framework,version,device,op_name,kernel_source,moe_dtype,num_tokens,hidden_size,inter_size,topk,num_experts,moe_tp_size,moe_ep_size,distribution,latency
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,8,balanced,0.01929066702723503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,8,balanced,0.0207893339296182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,8,balanced,0.019343999524911244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,8,balanced,0.019066666563351948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,8,balanced,0.020266667008399963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,8,balanced,0.02082666630546252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,8,balanced,0.020874666670958202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,8,balanced,0.023711999257405598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,8,balanced,0.02402666707833608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,8,balanced,0.02420799930890401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,8,balanced,0.03526933242877325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,8,balanced,0.03526400029659271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,8,balanced,0.03525333354870478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,8,balanced,0.03677866607904434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,8,balanced,0.03721066564321518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,8,balanced,0.03772266705830892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,8,balanced,0.040421334405740104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,8,balanced,0.04372799893220266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,8,balanced,0.04558933277924856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,8,balanced,0.05477866530418396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,8,balanced,0.06425599753856659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,8,balanced,0.08844266335169475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,8,balanced,0.10179733236630757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,8,balanced,0.13798399766286215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,8,balanced,0.17435733477274576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,8,power_law_1.2,0.024876800179481507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,8,balanced,0.24644800027211508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,8,power_law_1.2,0.022675199806690215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,8,power_law_1.2,0.02284799963235855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,8,power_law_1.2,0.025190401077270507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,8,power_law_1.2,0.027852800488471986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,8,power_law_1.2,0.029190400242805482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,128,balanced,0.062080000837643944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,8,power_law_1.2,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,128,balanced,0.05376000205675761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,128,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,128,balanced,0.050069332122802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,8,balanced,0.3078560034434001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,128,balanced,0.051701332132021584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,128,balanced,0.05401599903901418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,128,balanced,0.054511999090512596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,128,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,128,balanced,0.05611200133959452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,128,balanced,0.05782400071620941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,128,balanced,0.05607999861240387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,128,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,128,balanced,0.059877331058184304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,128,balanced,0.06195199986298879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,128,balanced,0.06607466439406078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,128,balanced,0.06644266843795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,128,balanced,0.0720960001150767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,128,balanced,0.07649066547552745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,128,balanced,0.08426133791605632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,8,power_law_1.2,0.030771198868751525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,128,balanced,0.10060800115267436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,128,balanced,0.11769599715868632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,8,power_law_1.2,0.030847999453544616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,128,balanced,0.14724266529083252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,128,balanced,0.17485332489013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,8,power_law_1.2,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,128,balanced,0.22854399681091309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,8,power_law_1.2,0.03839359879493713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,8,power_law_1.2,0.04230400025844574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,8,power_law_1.2,0.0440064013004303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,8,power_law_1.2,0.04321280121803284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,8,power_law_1.2,0.046937599778175354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,8,power_law_1.2,0.04928640127182007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,8,power_law_1.2,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,8,power_law_1.2,0.06249600052833557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,128,balanced,0.28516799211502075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,128,balanced,0.3888746500015259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,128,balanced,0.4976480007171631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,8,power_law_1.2,0.10053759813308716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,8,power_law_1.2,0.11121280193328857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,8,power_law_1.2,0.10240000486373901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,8,power_law_1.2,0.12276480197906495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,8,power_law_1.2,0.1435968041419983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,8,power_law_1.2,0.18984320163726806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,8,power_law_1.2,0.23493120670318604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,8,power_law_1.2,0.33722240924835206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,8,power_law_1.2,0.413478422164917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,4,balanced,0.04571733375390371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,4,balanced,0.052986666560173035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,4,balanced,0.06850133339564006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,4,balanced,0.09388800462086995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,4,balanced,0.14617600043614706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,4,balanced,0.17907732725143433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,4,balanced,0.1944426695505778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,4,balanced,0.19169066349665323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,4,balanced,0.19960000117619833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,4,balanced,0.20778133471806845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,4,balanced,0.23157866795857748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,4,balanced,0.25125332673390705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,4,balanced,0.2728213270505269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,4,balanced,0.24450665712356567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,4,balanced,0.2633066574732463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,4,balanced,0.28677332401275635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,4,balanced,0.3068053325017293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,4,balanced,0.3040213386217753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,1,balanced,0.07790400087833405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,1,balanced,0.10513066252072652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,1,balanced,0.16236266493797302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,4,balanced,0.4352586666742961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,1,balanced,0.1651893357435862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,1,balanced,0.16513599952061972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,4,balanced,0.47178133328755695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,1,balanced,0.16703999042510986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,1,balanced,0.16706132888793945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,4,balanced,0.5057760079701742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,4,balanced,0.559066653251648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,4,balanced,0.5908639828364054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,4,balanced,0.9971840381622314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,4,balanced,1.1281440258026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,1,balanced,0.1685439944267273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,4,balanced,1.6719733874003093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,1,balanced,0.17126933733622232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,1,balanced,0.1727679967880249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,1,balanced,0.17442133029301962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,1,balanced,0.17599467436472574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,1,balanced,0.18026133378346762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,4,balanced,2.159557342529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,1,balanced,0.18796267112096152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,1,balanced,0.1995840072631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,1,balanced,0.20356800158818564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,1,balanced,0.216538667678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,1,balanced,0.2691839933395386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,1,balanced,0.2897493243217468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,1,balanced,0.37279999256134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,1,balanced,0.4536159833272298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,1,balanced,0.6266719897588094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,1,balanced,0.8092373212178549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,1,balanced,1.135200023651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,1,balanced,1.4986666043599446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,1,balanced,2.1527199745178223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,1,balanced,2.874922752380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,32,power_law_1.2,0.08387200236320495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,32,power_law_1.2,0.08074880242347718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,32,power_law_1.2,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,32,power_law_1.2,0.08420479893684388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,32,power_law_1.2,0.08499199748039246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,32,power_law_1.2,0.07778559923171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,32,power_law_1.2,0.08436480164527893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,32,power_law_1.2,0.08549759984016418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,32,power_law_1.2,0.08570880293846131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,32,power_law_1.2,0.08641279935836792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,32,power_law_1.2,0.08855040073394775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,32,power_law_1.2,0.09016960263252258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,32,power_law_1.2,0.09175040125846863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,32,power_law_1.2,0.09216639995574952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,32,power_law_1.2,0.10488959550857543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,32,power_law_1.2,0.10867840051651001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,32,power_law_1.2,0.11025279760360718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,32,power_law_1.2,0.125273597240448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,32,power_law_1.2,0.1323456048965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,32,power_law_1.2,0.17479679584503174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,32,power_law_1.2,0.1941696047782898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,32,power_law_1.2,0.2491071939468384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,32,power_law_1.2,0.3399807929992676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,32,power_law_1.2,0.451039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,32,power_law_1.2,0.5978687763214111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,32,power_law_1.2,0.8746496200561523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,32,power_law_1.2,1.3198016166687012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,128,power_law_1.2,0.04629760086536407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,128,power_law_1.2,0.05429120063781738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,128,power_law_1.2,0.04310399889945984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,128,power_law_1.2,0.04311679899692535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,128,power_law_1.2,0.04289279878139496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,128,power_law_1.2,0.044710400700569156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,128,power_law_1.2,0.045023998618125914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,128,power_law_1.2,0.046003198623657225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,128,power_law_1.2,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,128,power_law_1.2,0.04802559912204742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,128,power_law_1.2,0.052857601642608644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,128,power_law_1.2,0.05413119792938233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,128,power_law_1.2,0.05565440058708191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,128,power_law_1.2,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,128,power_law_1.2,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,128,power_law_1.2,0.061414402723312375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,128,power_law_1.2,0.05669119954109192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,128,power_law_1.2,0.06417279839515685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,128,power_law_1.2,0.06948480010032654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,128,power_law_1.2,0.08572160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,128,power_law_1.2,0.0960319995880127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,128,power_law_1.2,0.12027519941329956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,128,power_law_1.2,0.13527040481567382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,128,power_law_1.2,0.1975808024406433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,128,power_law_1.2,0.24296960830688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,128,power_law_1.2,0.3530751943588257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,128,power_law_1.2,0.49560961723327634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,1,balanced,0.029135999580224354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,1,balanced,0.045791998505592346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,1,balanced,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,1,balanced,0.08638933300971985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,1,balanced,0.14266666769981384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,1,balanced,0.2574826677640279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,1,balanced,0.26522133747736615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,1,balanced,0.2717013359069824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,1,balanced,0.2793546716372172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,1,balanced,0.286901334921519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,1,balanced,0.28517333666483563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,1,balanced,0.29466134309768677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,1,balanced,0.30561065673828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,1,balanced,0.29940799872080487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,1,balanced,0.3099679946899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,1,balanced,0.32234134276707965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,1,balanced,0.3408106565475464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,1,balanced,0.3869973421096802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,1,balanced,0.39584000905354816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,1,balanced,0.48477331797281903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,1,balanced,0.5379039843877157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,1,balanced,0.7100533644358317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,1,balanced,0.6388800144195557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,1,balanced,0.8615360260009766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,1,balanced,1.0905120372772217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,1,balanced,1.5596960385640461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,1,balanced,2.062602678934733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,32,power_law_1.01,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,32,power_law_1.01,0.04031359851360321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,32,power_law_1.01,0.044819200038909913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,32,power_law_1.01,0.03852159976959228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,32,power_law_1.01,0.04113920032978058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,32,power_law_1.01,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,32,power_law_1.01,0.042828801274299624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,32,power_law_1.01,0.054150402545928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,32,power_law_1.01,0.0523967981338501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,32,power_law_1.01,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,32,power_law_1.01,0.08403199911117554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,32,power_law_1.01,0.08396160006523132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,32,power_law_1.01,0.08454399704933166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,32,power_law_1.01,0.08466560244560242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,32,power_law_1.01,0.09553920030593872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,32,power_law_1.01,0.11105920076370239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,32,power_law_1.01,0.11813759803771973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,32,power_law_1.01,0.14260480403900147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.01,0.17766400575637817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.01,0.22318720817565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.01,0.1256127953529358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.01,0.17052160501480101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.01,0.22024319171905518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.01,0.3023103952407837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.01,0.4347008228302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.01,0.6052544116973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.01,0.729529619216919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,8,power_law_1.01,0.04078719913959503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,8,power_law_1.01,0.041868799924850465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,8,power_law_1.01,0.0427839994430542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,8,power_law_1.01,0.045414400100708005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,8,power_law_1.01,0.045388799905776975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,8,power_law_1.01,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,8,power_law_1.01,0.049235200881958006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,8,power_law_1.01,0.0486847996711731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,8,power_law_1.01,0.04955520033836365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,8,power_law_1.01,0.050547200441360476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,8,power_law_1.01,0.05071359872817993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,8,power_law_1.01,0.05143679976463318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,8,power_law_1.01,0.05289599895477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,8,power_law_1.01,0.054022401571273804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,8,power_law_1.01,0.060166400671005246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,8,power_law_1.01,0.06158080101013184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,8,power_law_1.01,0.06967039704322815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,8,power_law_1.01,0.07326719760894776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,8,power_law_1.01,0.08160640001296997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,8,power_law_1.01,0.09440640211105347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,8,power_law_1.01,0.1055616021156311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,8,power_law_1.01,0.12177920341491699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,8,power_law_1.01,0.1527168035507202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,8,power_law_1.01,0.20079360008239747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,8,power_law_1.01,0.2728447914123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,8,power_law_1.01,0.3457855939865112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,8,power_law_1.01,0.43980798721313474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,1,power_law_1.2,0.02447360008955002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,1,power_law_1.2,0.03824000060558319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,1,power_law_1.2,0.054553598165512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,1,power_law_1.2,0.07667840123176575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,1,power_law_1.2,0.11110399961471558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,1,power_law_1.2,0.14789119958877564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,1,power_law_1.2,0.1955456018447876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,1,power_law_1.2,0.21278080940246583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,1,power_law_1.2,0.21874558925628662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,1,power_law_1.2,0.2302464008331299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,1,power_law_1.2,0.23843839168548583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,1,power_law_1.2,0.24591999053955077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,1,power_law_1.2,0.25785601139068604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,1,power_law_1.2,0.25774080753326417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,1,power_law_1.2,0.2627648115158081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,1,power_law_1.2,0.27733759880065917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,1,power_law_1.2,0.31067519187927245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,1,power_law_1.2,0.35778560638427737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,1,power_law_1.2,0.3566528081893921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,1,power_law_1.2,0.4386688232421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,1,power_law_1.2,0.46297597885131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,1,power_law_1.2,0.5902016162872314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,1,power_law_1.2,0.5610623836517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,1,power_law_1.2,0.7128767967224121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,1,power_law_1.2,0.8865407943725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,1,power_law_1.2,1.2225536346435546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,1,power_law_1.2,1.5531968116760253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,2,power_law_1.2,0.03880319893360138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,2,power_law_1.2,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,2,power_law_1.2,0.04158720076084137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,2,power_law_1.2,0.04147199988365173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,2,power_law_1.2,0.041875201463699344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,2,power_law_1.2,0.043859198689460754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,2,power_law_1.2,0.043628799915313723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,2,power_law_1.2,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,2,power_law_1.2,0.04561919867992401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,2,power_law_1.2,0.04842880070209503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,2,power_law_1.2,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,2,power_law_1.2,0.055718398094177245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,2,power_law_1.2,0.05676800012588501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,2,power_law_1.2,0.061222398281097413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,2,power_law_1.2,0.0694208025932312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,2,power_law_1.2,0.07277439832687378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,2,power_law_1.2,0.08235520124435425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,2,power_law_1.2,0.09553920030593872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.2,0.10080000162124633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.2,0.11242239475250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.2,0.15582079887390138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.2,0.1881152033805847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,16,balanced,0.05332266787687937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,16,balanced,0.044981335600217186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,16,balanced,0.04410133262475332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,16,balanced,0.04985066751639048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,16,balanced,0.05190933247407278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,16,balanced,0.07533866663773854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,16,balanced,0.09257599711418152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,16,balanced,0.09109333157539368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,16,balanced,0.09076266487439473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,16,balanced,0.09310932954152425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,16,balanced,0.09297066926956177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,16,balanced,0.09604799747467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,16,balanced,0.09620799620946248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,16,balanced,0.09921600421269734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,16,balanced,0.1032319962978363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,16,balanced,0.10608533024787903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.2,0.20205440521240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,16,balanced,0.10955733060836792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,16,balanced,0.11678399642308553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,16,balanced,0.12462932864824931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,16,balanced,0.14203733205795288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,16,balanced,0.1549066702524821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,16,balanced,0.18459200859069824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,16,balanced,0.22868265708287558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,16,balanced,0.28734399875005084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,16,balanced,0.3495573202768962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,16,balanced,0.4726453224817912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,16,balanced,0.5983999967575073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.2,0.3007551908493042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.2,0.4209280014038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,8,power_law_1.01,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,8,power_law_1.01,0.046598398685455324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,8,power_law_1.01,0.03891200125217438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,8,power_law_1.01,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,8,power_law_1.01,0.05140479803085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,8,power_law_1.01,0.051820802688598636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,8,power_law_1.01,0.05329279899597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,8,power_law_1.01,0.05383679866790771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,8,power_law_1.01,0.053337597846984865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,8,power_law_1.01,0.05495679974555969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,8,power_law_1.01,0.05482239723205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,8,power_law_1.01,0.05626239776611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,8,power_law_1.01,0.05786240100860596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,8,power_law_1.01,0.05867519974708557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,8,power_law_1.01,0.06574079990386963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,8,power_law_1.01,0.0669376015663147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,8,power_law_1.01,0.06502400040626526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,8,power_law_1.01,0.070278400182724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,8,power_law_1.01,0.07179520130157471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,8,power_law_1.01,0.08478720188140869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,8,power_law_1.01,0.0877951979637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,8,power_law_1.01,0.10746879577636718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,8,power_law_1.01,0.12328959703445434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,8,power_law_1.01,0.15697920322418213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,8,power_law_1.01,0.20004479885101317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,8,power_law_1.01,0.24188799858093263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,8,power_law_1.01,0.2846400022506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,16,8,power_law_1.2,0.05724160075187683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,16,8,power_law_1.2,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,16,8,power_law_1.2,0.06162559986114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,16,8,power_law_1.2,0.07002879977226258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,16,8,power_law_1.2,0.07369599938392639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,16,8,power_law_1.2,0.0797760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,16,8,power_law_1.2,0.08309760093688964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,16,8,power_law_1.2,0.07979519963264466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,16,8,power_law_1.2,0.08284800052642823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,16,8,power_law_1.2,0.0833407998085022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,16,8,power_law_1.2,0.08540800213813782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,16,8,power_law_1.2,0.08517119884490967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,16,8,power_law_1.2,0.08856319785118102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,16,8,power_law_1.2,0.0939136028289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,16,8,power_law_1.2,0.09789440035820007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,16,8,power_law_1.2,0.10021120309829712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,16,8,power_law_1.2,0.10560640096664428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,16,8,power_law_1.2,0.12350080013275147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,16,8,power_law_1.2,0.13557759523391724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,16,8,power_law_1.2,0.1621567964553833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,16,8,power_law_1.2,0.1903807997703552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,16,8,power_law_1.2,0.2527872085571289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,16,8,power_law_1.2,0.32385280132293703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,16,8,power_law_1.2,0.4792640209197998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,16,8,power_law_1.2,0.6044352054595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,16,8,power_law_1.2,0.8835840225219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,16,8,power_law_1.2,1.130784034729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.2,0.5760511875152587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.2,0.8315327644348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,4,power_law_1.2,0.06848000288009644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,4,power_law_1.2,0.08507519960403442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,4,power_law_1.2,0.0856768012046814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,4,power_law_1.2,0.09735680222511292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,4,power_law_1.2,0.11193599700927734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,4,power_law_1.2,0.12053120136260986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,4,power_law_1.2,0.13640960454940795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,4,power_law_1.2,0.13471360206604005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,4,power_law_1.2,0.1333567976951599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,4,power_law_1.2,0.1368191957473755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,4,power_law_1.2,0.13950719833374023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,4,power_law_1.2,0.1425536036491394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,4,power_law_1.2,0.14391039609909057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,4,power_law_1.2,0.1487488031387329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,4,power_law_1.2,0.15489920377731323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,4,power_law_1.2,0.15605759620666504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,4,power_law_1.2,0.1647487998008728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,4,power_law_1.2,0.18440959453582764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,4,power_law_1.2,0.20810239315032958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,4,power_law_1.2,0.25854079723358153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,4,power_law_1.2,0.2925695896148682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,4,power_law_1.2,0.38103680610656737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,4,power_law_1.2,0.5038335800170899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,4,power_law_1.2,0.6672063827514648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,4,power_law_1.2,0.8455615997314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,4,power_law_1.2,1.267308807373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,4,power_law_1.2,1.6408512115478515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,4,power_law_1.01,0.02011519968509674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,4,power_law_1.01,0.021030400693416596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,4,power_law_1.01,0.022988800704479218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,4,power_law_1.01,0.027059200406074523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,4,power_law_1.01,0.029504001140594482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,4,power_law_1.01,0.03224959969520569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,4,power_law_1.01,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,4,power_law_1.01,0.03446399867534637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,4,power_law_1.01,0.03596799969673157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,4,power_law_1.01,0.037222400307655334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,4,power_law_1.01,0.03859840035438537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,4,power_law_1.01,0.04382080137729645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,4,power_law_1.01,0.04909439980983734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,4,power_law_1.01,0.043558400869369504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,4,power_law_1.01,0.05432959794998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,4,power_law_1.01,0.05617920160293579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,4,power_law_1.01,0.059084802865982056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,4,power_law_1.01,0.0704576015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,4,power_law_1.01,0.07880319952964783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,4,power_law_1.01,0.09612799882888794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,4,power_law_1.01,0.12045439481735229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,4,power_law_1.01,0.1427135944366455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,4,power_law_1.01,0.16046719551086425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,4,power_law_1.01,0.20241920948028563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,4,power_law_1.01,0.2556864023208618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,4,power_law_1.01,0.3483839988708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,4,power_law_1.01,0.4541759967803955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,64,power_law_1.01,0.03244799971580505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,64,power_law_1.01,0.03163520097732544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,64,power_law_1.01,0.03086720108985901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,64,power_law_1.01,0.030406400561332703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,64,power_law_1.01,0.032076799869537355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,64,power_law_1.01,0.03215999901294708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,64,power_law_1.01,0.03275519907474518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,64,power_law_1.01,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,64,power_law_1.01,0.033657601475715636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,64,power_law_1.01,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,64,power_law_1.01,0.03511039912700653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,64,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,64,power_law_1.01,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,64,power_law_1.01,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,64,power_law_1.01,0.042508798837661746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,64,power_law_1.01,0.04344319999217987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,64,power_law_1.01,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,64,power_law_1.01,0.050918400287628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,64,power_law_1.01,0.057036799192428586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,64,power_law_1.01,0.0658623993396759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,64,power_law_1.01,0.0798143982887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,64,power_law_1.01,0.11016960144042968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,64,power_law_1.01,0.12995840311050416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,64,power_law_1.01,0.1784832000732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,64,power_law_1.01,0.2171839952468872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,64,power_law_1.01,0.3175935983657837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,64,power_law_1.01,0.39147520065307617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,1,power_law_1.01,0.0699072003364563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,1,power_law_1.01,0.09127039909362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,1,power_law_1.01,0.10488959550857543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,1,power_law_1.01,0.12384639978408814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,1,power_law_1.01,0.1351680040359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,1,power_law_1.01,0.1713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,1,power_law_1.01,0.18154879808425903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,1,power_law_1.01,0.18730239868164061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,1,power_law_1.01,0.1918272018432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,1,power_law_1.01,0.19690239429473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,1,power_law_1.01,0.19953919649124147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,1,power_law_1.01,0.20455679893493653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,1,power_law_1.01,0.20952959060668946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,1,power_law_1.01,0.21505920886993407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,1,power_law_1.01,0.21509120464324952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,1,power_law_1.01,0.22080640792846679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,1,power_law_1.01,0.23438720703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,1,power_law_1.01,0.2766848087310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.01,0.3180927991867065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.01,0.39883520603179934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.01,0.4702911853790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.01,0.6307136058807373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.01,0.793503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.01,1.1203968048095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.01,1.4408448219299317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.01,2.077382469177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.01,2.735647964477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,4,power_law_1.2,0.047891199588775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,4,power_law_1.2,0.0664192020893097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,4,power_law_1.2,0.07152000069618225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,4,power_law_1.2,0.08362240195274354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,4,power_law_1.2,0.1019968032836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,4,power_law_1.2,0.11139839887619019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,4,power_law_1.2,0.13036799430847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,4,power_law_1.2,0.1282431960105896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,4,power_law_1.2,0.12200959920883178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,4,power_law_1.2,0.12650879621505737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,4,power_law_1.2,0.13527040481567382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,4,power_law_1.2,0.14008320569992067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,4,power_law_1.2,0.14814079999923707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,4,power_law_1.2,0.15201280117034913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,4,power_law_1.2,0.16490880250930787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,4,power_law_1.2,0.18172160387039185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,1,balanced,0.04994133114814758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,1,balanced,0.06844800213972728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,1,balanced,0.10224533081054688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,1,balanced,0.1686720053354899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,1,balanced,0.2991466720898946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,1,balanced,0.39390401045481366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,1,balanced,0.39792001247406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,1,balanced,0.3973493178685506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,1,balanced,0.39767467975616455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,1,balanced,0.4017333189646403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,1,balanced,0.4013226826985677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,1,balanced,0.40515732765197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,4,power_law_1.2,0.17205120325088502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,4,power_law_1.2,0.19251840114593505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,4,power_law_1.2,0.20553600788116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,4,power_law_1.2,0.24634881019592286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,4,power_law_1.2,0.26536319255828855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,4,power_law_1.2,0.32252800464630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,1,balanced,0.40618666013081867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,1,balanced,0.4113653500874837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,4,power_law_1.2,0.404307222366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,1,balanced,0.41783467928568524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,1,balanced,0.4228479862213135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,4,power_law_1.2,0.5538879871368408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,4,power_law_1.2,0.6504511833190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,4,power_law_1.2,1.0
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,4,power_law_1.2,1.2572928428649903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,1,balanced,0.42522132396698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,1,balanced,0.4748053153355916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,1,balanced,0.455349326133728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,1,balanced,0.5612533489863077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,1,balanced,0.5164586702982584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,1,balanced,0.7155360380808512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,1,balanced,0.6463946501413981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,1,balanced,1.0037706693013508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,1,balanced,1.0275146961212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,1,balanced,1.587392012278239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,1,balanced,1.5302240053812664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,4,power_law_1.01,0.06247040033340454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,4,power_law_1.01,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,4,power_law_1.01,0.07862399816513062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,4,power_law_1.01,0.07693439722061157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,4,power_law_1.01,0.07746559977531434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,4,power_law_1.01,0.07878400087356567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,4,power_law_1.01,0.07922559976577759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,4,power_law_1.01,0.08502399921417236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,4,power_law_1.01,0.08829439878463745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,4,power_law_1.01,0.09055359959602356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,4,power_law_1.01,0.08848639726638793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,4,power_law_1.01,0.08956800103187561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,4,power_law_1.01,0.09423999786376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,4,power_law_1.01,0.10199040174484253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,4,power_law_1.01,0.1201024055480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,4,power_law_1.01,0.12430720329284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,4,power_law_1.01,0.13815040588378907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,4,power_law_1.01,0.1731328010559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,4,power_law_1.01,0.17694720029830932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,4,power_law_1.01,0.2070528030395508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,4,power_law_1.01,0.23876481056213378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,4,power_law_1.01,0.3719680070877075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,16,power_law_1.2,0.030323201417922975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,16,power_law_1.2,0.03209599852561951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,16,power_law_1.2,0.031462401151657104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,16,power_law_1.2,0.03166080117225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,16,power_law_1.2,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,16,power_law_1.2,0.03399679958820343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,16,power_law_1.2,0.03436160087585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,16,power_law_1.2,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,16,power_law_1.2,0.037196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,16,power_law_1.2,0.037561601400375365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,16,power_law_1.2,0.040659201145172116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,16,power_law_1.2,0.04482559859752655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,16,power_law_1.2,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,16,power_law_1.2,0.043673598766326906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,16,power_law_1.2,0.0484607994556427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,16,power_law_1.2,0.04967679977416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,16,power_law_1.2,0.052691197395324706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,16,power_law_1.2,0.05792639851570129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,16,power_law_1.2,0.06328960061073304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,16,power_law_1.2,0.07177600264549255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,16,power_law_1.2,0.08293759822845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,16,balanced,0.047797332207361855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,16,balanced,0.0499893327554067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,16,balanced,0.05376533170541128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,16,balanced,0.06208533545335134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,16,balanced,0.08038400113582611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,16,balanced,0.0872373382250468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,16,balanced,0.08714666962623596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,16,power_law_1.2,0.11109119653701782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,16,balanced,0.08730133374532063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,16,balanced,0.08769067128499348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,16,balanced,0.08855467041333516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,16,balanced,0.09026666482289632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,16,balanced,0.09297600388526917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,16,balanced,0.0934399962425232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,16,balanced,0.09511466821034749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,16,balanced,0.10084799925486247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,16,balanced,0.10291733344395955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,16,balanced,0.10841600100199382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,16,balanced,0.12370666861534119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,16,balanced,0.13194132844607034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,16,balanced,0.1644426683584849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,16,balanced,0.18916799624760947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,16,balanced,0.2564479907353719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,4,power_law_1.01,0.424345588684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,16,power_law_1.2,0.11510399580001832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,16,balanced,0.2467199961344401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,16,balanced,0.32790400584538776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,16,balanced,0.39792533715566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,16,balanced,0.5557493368784586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,16,balanced,0.676741361618042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,16,power_law_1.2,0.15041279792785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,4,power_law_1.01,0.7067967891693115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,16,power_law_1.2,0.18121600151062012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,2,power_law_1.2,0.043968001008033754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,2,power_law_1.2,0.050809597969055174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,2,power_law_1.2,0.05939840078353882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,2,power_law_1.2,0.0712768018245697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,2,power_law_1.2,0.07580159902572632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,2,power_law_1.2,0.10290559530258178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,2,power_law_1.2,0.10368000268936158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,2,power_law_1.2,0.10814080238342286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,2,power_law_1.2,0.11139839887619019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,2,power_law_1.2,0.11235840320587158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,2,power_law_1.2,0.11747200489044189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,2,power_law_1.2,0.1233024001121521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,2,power_law_1.2,0.12257920503616333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,2,power_law_1.2,0.13077759742736816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,2,power_law_1.2,0.140665602684021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,2,power_law_1.2,0.15005439519882202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,2,power_law_1.2,0.15213439464569092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,2,power_law_1.2,0.17669119834899902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.2,0.19086079597473143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.2,0.2351423978805542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.2,0.27820160388946535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.2,0.3566783905029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.2,0.45011200904846194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,16,power_law_1.2,0.2939840078353882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.2,0.5930880069732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,4,power_law_1.01,0.7926271915435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,4,power_law_1.01,1.2552191734313964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.2,0.8060480117797851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.2,1.0908160209655762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,16,power_law_1.01,0.05324800014495849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,16,power_law_1.01,0.06321280002593994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,16,power_law_1.01,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,16,power_law_1.01,0.05253120064735413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,16,power_law_1.01,0.051052802801132204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,16,power_law_1.01,0.05116159915924072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,16,power_law_1.01,0.054476797580718994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,16,power_law_1.01,0.056620800495147706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,16,power_law_1.01,0.05785599946975708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,16,power_law_1.01,0.05756800174713135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,16,power_law_1.01,0.05944960117340088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,16,power_law_1.01,0.060761600732803345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,16,power_law_1.01,0.0637440025806427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,16,power_law_1.01,0.06686080098152161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,16,power_law_1.01,0.07294080257415772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,16,power_law_1.01,0.08036479949951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,16,power_law_1.01,0.08415359854698182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,16,power_law_1.01,0.10065280199050904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,16,power_law_1.01,0.10581760406494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,16,power_law_1.01,0.1324671983718872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,16,power_law_1.01,0.15325440168380738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,16,power_law_1.01,0.20006399154663085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,16,power_law_1.01,0.2395008087158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,16,power_law_1.01,0.33743360042572024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,16,power_law_1.01,0.430617618560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,16,power_law_1.01,0.6416255950927734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,16,power_law_1.01,0.7877503871917725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,4,power_law_1.01,1.5868224143981933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,16,power_law_1.2,0.3362623929977417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.2,1.4088831901550294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,4,power_law_1.2,0.07395200133323669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,4,power_law_1.2,0.0984063982963562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,4,power_law_1.2,0.1019968032836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,4,power_law_1.2,0.13262079954147338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,4,power_law_1.2,0.16501760482788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,4,power_law_1.2,0.1745792031288147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,4,power_law_1.2,0.22262399196624755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,4,power_law_1.2,0.2227008104324341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,4,power_law_1.2,0.22417280673980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,4,power_law_1.2,0.2224639892578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,4,power_law_1.2,0.23082880973815917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,4,power_law_1.2,0.22903039455413818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,4,power_law_1.2,0.23291521072387694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,4,power_law_1.2,0.25050880908966067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,4,power_law_1.2,0.26026880741119385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,4,power_law_1.2,0.26531200408935546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,4,power_law_1.2,0.2934976100921631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,4,power_law_1.2,0.32702720165252686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,4,power_law_1.2,0.3433727979660034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,4,power_law_1.2,0.41052799224853515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,4,power_law_1.2,0.45674881935119627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,4,power_law_1.2,0.5561215877532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,4,power_law_1.2,0.6737984180450439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,4,power_law_1.2,0.8890624046325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,4,power_law_1.2,1.0674367904663087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,4,power_law_1.2,1.5174528121948243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,4,power_law_1.2,2.044595146179199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,64,balanced,0.0602400004863739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,64,balanced,0.05169600248336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,64,balanced,0.05192000170548757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,64,balanced,0.04985600213209788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,64,balanced,0.0539680023988088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,64,balanced,0.062234664956728615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,64,balanced,0.0796853353579839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,64,balanced,0.08021866778532664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,64,balanced,0.08121599753697713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,64,balanced,0.08168533444404602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,64,balanced,0.08226133386294048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,64,balanced,0.08375466863314311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,64,balanced,0.08364799618721008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,64,balanced,0.08604266246159871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,64,balanced,0.09194133679072063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,64,balanced,0.09301333626111348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,64,balanced,0.0985599954922994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,64,balanced,0.10479999581972758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,64,balanced,0.11130133271217346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,64,balanced,0.1258026659488678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,64,balanced,0.13863466183344522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,64,balanced,0.17060800393422446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,64,balanced,0.20212799310684204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,64,balanced,0.25838400920232135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,64,balanced,0.3189599911371867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,64,balanced,0.427621324857076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,64,balanced,0.5491413275400797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,1,balanced,0.11733333269755046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,1,balanced,0.14824533462524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,1,balanced,0.20825600624084473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,1,balanced,0.3288586735725403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,1,balanced,0.5706026554107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,1,balanced,1.0632373491923015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,1,balanced,1.557381312052409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,1,balanced,1.5530986785888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,1,balanced,1.5650506019592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,1,balanced,1.5640746752421062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,1,balanced,1.5737226804097493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,1,balanced,1.5818773905436199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,1,balanced,1.5915892918904622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,1,balanced,1.6032640139261882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,1,balanced,1.6141546567281086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,1,balanced,1.6273867289225261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,1,balanced,1.651242733001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,1,balanced,1.7012960116068523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,1,balanced,1.739664077758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,1,balanced,1.8278187115987141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,1,balanced,1.920138676961263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,1,balanced,2.1369120279947915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,1,balanced,2.330885410308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,1,balanced,2.7173598607381186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,1,balanced,3.507322629292806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,1,balanced,4.291440010070801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,1,balanced,5.791727701822917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,32,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,32,balanced,0.048021331429481506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,32,balanced,0.04701866706212362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,32,balanced,0.049679999550183616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,32,balanced,0.06797333558400472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,32,balanced,0.0922933320204417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,32,balanced,0.09091732899347942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,32,balanced,0.09264000256856282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,32,balanced,0.09315199653307597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,32,balanced,0.09310932954152425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,32,balanced,0.0944053332010905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,32,balanced,0.09708266456921895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,32,balanced,0.0950986643632253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,32,balanced,0.0974666674931844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,32,balanced,0.1027893324693044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,32,balanced,0.10252267122268677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,32,balanced,0.1051573355992635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,32,balanced,0.11071999867757161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,32,balanced,0.11346667011578877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,32,balanced,0.12469333410263062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,32,balanced,0.13541866342226663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,32,balanced,0.15494400262832642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,32,balanced,0.17626667022705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,32,balanced,0.21568532784779867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,32,balanced,0.25865066051483154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,32,balanced,0.343173344930013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,32,balanced,0.4341919819513957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,16,power_law_1.2,0.09364479780197144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,16,power_law_1.2,0.13295359611511232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,16,power_law_1.2,0.12828799486160278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,16,power_law_1.2,0.13050240278244019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,16,power_law_1.2,0.12777600288391114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,16,power_law_1.2,0.10956159830093384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,16,power_law_1.2,0.12717440128326415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,16,power_law_1.2,0.12732800245285034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,16,power_law_1.2,0.12047359943389893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,16,power_law_1.2,0.12934399843215943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,16,power_law_1.2,0.12449920177459717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,16,power_law_1.2,0.13003519773483277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,16,power_law_1.2,0.13964799642562867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,16,power_law_1.2,0.1531391978263855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,16,power_law_1.2,0.15857919454574584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,16,power_law_1.2,0.17237119674682616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,16,power_law_1.2,0.15601919889450072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,16,power_law_1.2,0.18265600204467775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,16,power_law_1.2,0.18786560297012328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,16,power_law_1.2,0.2286207914352417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,16,power_law_1.2,0.24072320461273194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,16,power_law_1.2,0.288703989982605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,16,power_law_1.2,0.3238271951675415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,16,power_law_1.2,0.43374080657958985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,16,power_law_1.2,0.5202303886413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,16,power_law_1.2,0.7699520111083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,16,power_law_1.2,0.9750975608825684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,4,power_law_1.2,0.023795199394226075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,4,power_law_1.2,0.027820798754692077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,4,power_law_1.2,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,4,power_law_1.2,0.043296000361442565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,4,power_law_1.2,0.04444159865379334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,4,power_law_1.2,0.05504639744758606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,4,power_law_1.2,0.0579584002494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,4,power_law_1.2,0.055257600545883176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,4,power_law_1.2,0.058905601501464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2688,1856,6,128,1,4,power_law_1.2,0.06071680188179016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2688,1856,6,128,1,4,power_law_1.2,0.06097279787063599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2688,1856,6,128,1,4,power_law_1.2,0.06302080154418946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2688,1856,6,128,1,4,power_law_1.2,0.0670527994632721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2688,1856,6,128,1,4,power_law_1.2,0.06846719980239868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2688,1856,6,128,1,4,power_law_1.2,0.06750079989433289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,4,balanced,0.043968002001444496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,4,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,4,balanced,0.04974933465321859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,4,balanced,0.05585599939028422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,4,balanced,0.07588799794514973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,4,balanced,0.10745066404342651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,4,balanced,0.10825066765149434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,4,balanced,0.10442133744557698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,4,balanced,0.10565867026646932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,4,balanced,0.10626666744550069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,4,balanced,0.10760533809661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,4,balanced,0.110042671362559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,4,balanced,0.11329600214958191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,4,balanced,0.11492266257603963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,4,balanced,0.12343466281890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,4,balanced,0.12577066818873087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,4,balanced,0.1376533309618632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,4,balanced,0.1562026639779409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,4,balanced,0.16663466890652975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,4,balanced,0.19565333922704062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,4,balanced,0.2278133432070414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,4,balanced,0.28653866052627563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,4,balanced,0.34146666526794434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,4,balanced,0.4920053482055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,4,balanced,0.6049013137817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,4,balanced,0.8655573527018229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,4,balanced,1.1262719631195068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,1,balanced,0.023120000958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,1,balanced,0.02537599951028824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,1,balanced,0.03170666595300039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,1,balanced,0.048250665267308555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,1,balanced,0.07291733225186665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,1,balanced,0.07623466849327087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,1,balanced,0.08017066617806752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,1,balanced,0.0844533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,1,balanced,0.08892800410588582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,1,balanced,0.09273067116737366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,1,balanced,0.08861333131790161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,1,balanced,0.0946613351504008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,1,balanced,0.10050132870674133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,1,balanced,0.09468799829483032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,1,balanced,0.10120532910029094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,1,balanced,0.10736533006032307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,1,balanced,0.11518399914105733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,1,balanced,0.13608533143997192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,1,balanced,0.1548960010210673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,1,balanced,0.20466667413711548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,1,balanced,0.18525334199269614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,1,balanced,0.2456159989039103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,1,balanced,0.22538133462270102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,1,balanced,0.38229866822560626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,1,balanced,0.38630934556325275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,1,balanced,0.560149351755778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,1,balanced,0.7349812984466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,8,power_law_1.01,0.06247040033340454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,8,power_law_1.01,0.08049280047416688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,8,power_law_1.01,0.0889087975025177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,8,power_law_1.01,0.09789440035820007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,8,power_law_1.01,0.10304000377655029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,8,power_law_1.01,0.11077120304107665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,8,power_law_1.01,0.0982208013534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,8,power_law_1.01,0.10824320316314698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,8,power_law_1.01,0.10917119979858399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,8,power_law_1.01,0.1047551989555359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,8,power_law_1.01,0.10785280466079712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,8,power_law_1.01,0.11080960035324097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,8,power_law_1.01,0.11191680431365966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,8,power_law_1.01,0.11953920125961304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,8,power_law_1.01,0.12407679557800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,8,power_law_1.01,0.12616319656372071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,8,power_law_1.01,0.1367616057395935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,8,power_law_1.01,0.15249279737472535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,8,power_law_1.01,0.1674880027770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,8,power_law_1.01,0.19474560022354126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,8,power_law_1.01,0.22430078983306884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,8,power_law_1.01,0.2755647897720337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,8,power_law_1.01,0.31893119812011717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,8,power_law_1.01,0.4295936107635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,8,power_law_1.01,0.5619455814361572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,8,power_law_1.01,0.7867136001586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,8,power_law_1.01,0.9347328186035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,2,power_law_1.2,0.05014399886131286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,2,power_law_1.2,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,2,power_law_1.2,0.1114240050315857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,2,power_law_1.2,0.14067200422286988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,2,power_law_1.2,0.1756160020828247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,2,power_law_1.2,0.2760832071304321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,2,power_law_1.2,0.289574408531189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,2,power_law_1.2,0.2856064081192017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,2,power_law_1.2,0.2928319931030273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,2,power_law_1.2,0.29991679191589354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,2,power_law_1.2,0.3123199939727783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,2,power_law_1.2,0.33530240058898925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,2,power_law_1.2,0.3530944108963013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,2,power_law_1.2,0.33810560703277587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,2,power_law_1.2,0.3481343984603882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,2,power_law_1.2,0.3646464109420776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,2,power_law_1.2,0.39123198986053465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,2,power_law_1.2,0.45638399124145507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.2,0.4310336112976074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.2,0.5319808006286622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.2,0.6238719940185546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.2,0.7010303974151612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.2,0.7864768028259277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.2,1.0205887794494628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.2,1.237343978881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.2,1.668377685546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,8,power_law_1.01,0.03377279937267304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,8,power_law_1.01,0.035571199655532834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,8,power_law_1.01,0.034643200039863584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,8,power_law_1.01,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,8,power_law_1.01,0.03543039858341217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,8,power_law_1.01,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,8,power_law_1.01,0.03896960020065308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,8,power_law_1.01,0.03900800049304962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,8,power_law_1.01,0.039552000164985654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,8,power_law_1.01,0.04145280122756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,8,power_law_1.01,0.04237439930438995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,8,power_law_1.01,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,8,power_law_1.01,0.04693120121955872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,8,power_law_1.01,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,8,power_law_1.01,0.053273600339889524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,8,power_law_1.01,0.05666559934616089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,8,power_law_1.01,0.056435197591781616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,8,power_law_1.01,0.06511359810829162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,8,power_law_1.01,0.0800704002380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,8,power_law_1.01,0.10312319993972778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,8,power_law_1.01,0.10558079481124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,8,power_law_1.01,0.1456063985824585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,8,power_law_1.01,0.1835584044456482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,8,power_law_1.01,0.25106561183929443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,8,power_law_1.01,0.35111680030822756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,8,power_law_1.01,0.499129581451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.2,2.1374080657958983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,8,power_law_1.01,0.5924928188323975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,4,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,4,balanced,0.05781333148479462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,4,balanced,0.06339199841022491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,4,balanced,0.08616000413894653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,4,balanced,0.11749333143234253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,4,balanced,0.16844799121220908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,4,balanced,0.17030932505925497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,4,balanced,0.1725013256072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,4,balanced,0.17375467220942178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,4,balanced,0.17506666978200278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,4,balanced,0.17863466342290243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,4,balanced,0.1785973310470581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,4,balanced,0.18126400311787924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,4,balanced,0.18407466014226279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,4,balanced,0.18986133734385172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,4,balanced,0.19187732537587485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,4,balanced,0.20082134008407593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,4,balanced,0.2142826716105143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,4,balanced,0.22936532894770303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,4,balanced,0.2585013310114543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,4,balanced,0.29480000336964923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,4,balanced,0.34299735228220624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,4,balanced,0.3941546678543091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,4,balanced,0.5487786531448364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,4,balanced,0.6526079972585043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,4,balanced,0.9148213068644205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,4,balanced,1.1768906911214192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,64,balanced,0.07443733513355255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,64,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,64,balanced,0.05003733436266581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,64,balanced,0.05387733379999796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,64,balanced,0.07396799822648366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,64,balanced,0.09975467125574748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,64,balanced,0.10165866216023763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,64,balanced,0.10097066561381023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,64,balanced,0.10099200407663982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,64,balanced,0.10372266173362732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,64,balanced,0.10260799527168274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,64,balanced,0.1039466659228007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,64,balanced,0.10389332969983418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,64,balanced,0.10538132985432942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,64,balanced,0.10837866862614949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,64,balanced,0.10940800110499065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,64,balanced,0.11327466368675232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,64,balanced,0.11727999647458394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,64,balanced,0.12187199791272481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,64,balanced,0.13179733355840048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,64,balanced,0.1402453382809957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,64,balanced,0.15923200050989786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,64,balanced,0.1811573306719462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,64,balanced,0.21875733137130737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,64,balanced,0.2596693237622579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,64,balanced,0.3431946833928426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,64,balanced,0.42165335019429523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,32,power_law_1.2,0.03158400058746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,32,power_law_1.2,0.03190400004386902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,32,power_law_1.2,0.030963200330734252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,32,power_law_1.2,0.029676800966262816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,32,power_law_1.2,0.03079040050506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,32,power_law_1.2,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,32,power_law_1.2,0.032518398761749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,32,power_law_1.2,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,32,power_law_1.2,0.03369599878787995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,32,power_law_1.2,0.033881598711013795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,32,power_law_1.2,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,32,power_law_1.2,0.03692159950733185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,32,power_law_1.2,0.03823359906673431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,32,power_law_1.2,0.03772799968719483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,32,power_law_1.2,0.04140160083770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,32,power_law_1.2,0.0434688001871109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,32,power_law_1.2,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,32,power_law_1.2,0.05160319805145264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,32,power_law_1.2,0.05655679702758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,32,power_law_1.2,0.06850559711456299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,32,power_law_1.2,0.08613759875297547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,32,power_law_1.2,0.1267583966255188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,32,power_law_1.2,0.12942719459533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,32,power_law_1.2,0.20515201091766358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,32,power_law_1.2,0.26369280815124513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,32,power_law_1.2,0.3963392019271851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,32,power_law_1.2,0.47601919174194335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,32,power_law_1.2,0.036339199542999266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,32,power_law_1.2,0.0355648010969162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,32,power_law_1.2,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,32,power_law_1.2,0.03875199854373932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,32,power_law_1.2,0.03971199989318848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,32,power_law_1.2,0.040115201473236085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,32,power_law_1.2,0.04163840115070343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,32,power_law_1.2,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,32,power_law_1.2,0.04588800072669983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,32,power_law_1.2,0.04538240134716034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,32,power_law_1.2,0.046463999152183535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,32,power_law_1.2,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,32,power_law_1.2,0.048307201266288756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,32,power_law_1.2,0.05370879769325256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,32,power_law_1.2,0.0630079984664917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,32,power_law_1.2,0.06380159854888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,32,power_law_1.2,0.06964480280876159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,32,power_law_1.2,0.07457919716835022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,32,power_law_1.2,0.09772160053253173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,32,power_law_1.2,0.10813440084457397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,32,power_law_1.2,0.1523200035095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,32,power_law_1.2,0.19883519411087036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,32,power_law_1.2,0.2533951997756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,32,power_law_1.2,0.4284031867980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,32,power_law_1.2,0.47096958160400393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,32,power_law_1.2,0.7201727867126465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,32,power_law_1.2,1.0170304298400878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,32,power_law_1.01,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,32,power_law_1.01,0.031033599376678468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,32,power_law_1.01,0.030687999725341798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,32,power_law_1.01,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,32,power_law_1.01,0.03372159898281098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,32,power_law_1.01,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,32,power_law_1.01,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,32,power_law_1.01,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,32,power_law_1.01,0.036083200573921205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,32,power_law_1.01,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,32,power_law_1.01,0.03758719861507416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,32,power_law_1.01,0.038022398948669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,32,power_law_1.01,0.03859840035438537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,32,power_law_1.01,0.039955198764801025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,32,power_law_1.01,0.04457600116729736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,32,power_law_1.01,0.045311999320983884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,32,power_law_1.01,0.05097600221633911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,32,power_law_1.01,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,32,power_law_1.01,0.06874240040779114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,32,power_law_1.01,0.077920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,32,power_law_1.01,0.09742720127105713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,32,power_law_1.01,0.12243839502334594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,32,power_law_1.01,0.14880000352859496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,32,power_law_1.01,0.21498239040374756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,32,power_law_1.01,0.2659712076187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,32,power_law_1.01,0.4243648052215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,32,power_law_1.01,0.5816703796386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,8,balanced,0.03499199946721395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,8,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,8,balanced,0.046480000019073486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,8,balanced,0.05542933444182078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,8,balanced,0.07761600116888683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,8,balanced,0.12643200159072876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,8,balanced,0.1360106666882833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,8,balanced,0.14664533734321594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,8,balanced,0.1584053337574005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,8,balanced,0.16746666034062704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,8,balanced,0.14577600359916687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,8,balanced,0.15169066190719604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,8,balanced,0.159770667552948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,8,balanced,0.1766293247540792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,8,balanced,0.19368533293406168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,8,balanced,0.20854399601618448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,8,balanced,0.22873600323994955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,8,balanced,0.24409600098927817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,8,balanced,0.44434134165445965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,8,balanced,0.5436266660690308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,256,power_law_1.2,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,8,balanced,0.3761813243230184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,256,power_law_1.2,0.03911679983139038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,8,balanced,0.4376586675643921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,8,balanced,0.4968053499857585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,8,balanced,0.6031733353932699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,8,balanced,0.7316479682922363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,8,balanced,1.1636906464894612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,8,balanced,1.3906133969624836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,256,power_law_1.2,0.039084801077842714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,256,power_law_1.2,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,256,power_law_1.2,0.042630401253700254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,256,power_law_1.2,0.044915199279785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,256,power_law_1.2,0.04537599980831146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,256,power_law_1.2,0.04595839977264404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,256,power_law_1.2,0.04761599898338318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,256,power_law_1.2,0.048172798752784726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,256,power_law_1.2,0.05232639908790589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,256,power_law_1.2,0.053939199447631835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,256,power_law_1.2,0.04978559911251068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,256,power_law_1.2,0.05488640069961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,256,power_law_1.2,0.06064640283584595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,256,power_law_1.2,0.061427199840545656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,256,power_law_1.2,0.06593279838562012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,256,power_law_1.2,0.07736960053443909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,256,power_law_1.2,0.08357120156288148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,256,power_law_1.2,0.10316799879074097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,256,power_law_1.2,0.12534400224685668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,256,power_law_1.2,0.16673280000686647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,256,power_law_1.2,0.22921600341796874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,256,power_law_1.2,0.33251841068267823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,256,power_law_1.2,0.42304000854492185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,256,power_law_1.2,0.6592895984649658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,256,power_law_1.2,0.9976960182189941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,8,power_law_1.01,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,8,power_law_1.01,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,8,power_law_1.01,0.034815999865531924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,8,power_law_1.01,0.044684800505638125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,8,power_law_1.01,0.042156800627708435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,8,power_law_1.01,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,8,power_law_1.01,0.04938240051269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,8,power_law_1.01,0.0514303982257843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,8,power_law_1.01,0.053478401899337766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,8,power_law_1.01,0.05251200199127197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,8,power_law_1.01,0.05338240265846252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,8,power_law_1.01,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,8,power_law_1.01,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,8,power_law_1.01,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,8,power_law_1.01,0.061894398927688596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,8,power_law_1.01,0.06585599780082703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,8,power_law_1.01,0.06702079772949218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,8,power_law_1.01,0.07185919880867005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,8,power_law_1.01,0.08529919981956482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,8,power_law_1.01,0.09407359957695008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,8,power_law_1.01,0.09688320159912109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,8,power_law_1.01,0.12588800191879274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,8,power_law_1.01,0.12806400060653686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,8,power_law_1.01,0.17677439451217652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,8,power_law_1.01,0.23840639591217042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,8,power_law_1.01,0.32057600021362304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,8,power_law_1.01,0.36075520515441895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,2,power_law_1.2,0.05048959851264954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,2,power_law_1.2,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,2,power_law_1.2,0.05973119735717773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,2,power_law_1.2,0.06773759722709656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,2,power_law_1.2,0.07838720083236694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,2,power_law_1.2,0.09094399809837342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,2,power_law_1.2,0.10908160209655762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,2,power_law_1.2,0.1097216010093689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,2,power_law_1.2,0.11360000371932984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,2,power_law_1.2,0.11934720277786255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,2,power_law_1.2,0.12306560277938842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,2,power_law_1.2,0.13155200481414794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,2,power_law_1.2,0.1372607946395874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,2,power_law_1.2,0.1512511968612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,2,power_law_1.2,0.16120320558547974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,2,power_law_1.2,0.1690111994743347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,2,power_law_1.2,0.1647487998008728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,2,power_law_1.2,0.2020927906036377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,2,power_law_1.2,0.21850240230560303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,2,power_law_1.2,0.26686720848083495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,2,power_law_1.2,0.3205951929092407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,2,power_law_1.2,0.4208320140838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,2,power_law_1.2,0.5640384197235108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,2,power_law_1.2,0.7375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,2,power_law_1.2,0.920742416381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,2,power_law_1.2,1.2807807922363281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,2,power_law_1.2,1.7542335510253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,1,power_law_1.01,0.05137919783592224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,1,power_law_1.01,0.055308800935745236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,1,power_law_1.01,0.06599040031433105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,1,power_law_1.01,0.08266239762306213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,1,power_law_1.01,0.1066431999206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,1,power_law_1.01,0.12697600126266478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,1,power_law_1.01,0.15701119899749755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,1,power_law_1.01,0.16359039545059204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,1,power_law_1.01,0.167302405834198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,1,power_law_1.01,0.17036800384521483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,1,power_law_1.01,0.17774720191955568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,1,power_law_1.01,0.18729599714279174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,1,power_law_1.01,0.18766080141067504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,1,power_law_1.01,0.19753600358963014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,1,power_law_1.01,0.20412158966064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,1,power_law_1.01,0.21427199840545655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,1,power_law_1.01,0.22609920501708985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,1,power_law_1.01,0.2543103933334351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,1,power_law_1.01,0.2798207998275757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,1,power_law_1.01,0.33482239246368406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,1,power_law_1.01,0.38235518932342527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,1,power_law_1.01,0.4812032222747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,1,power_law_1.01,0.5757567882537842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,1,power_law_1.01,0.7728384017944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,1,power_law_1.01,0.968876838684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,1,power_law_1.01,1.352524757385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,1,power_law_1.01,1.737433624267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,4,power_law_1.2,0.19084800481796266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,4,power_law_1.2,0.2617151975631714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,4,power_law_1.2,0.327129602432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,4,power_law_1.2,0.43804159164428713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,4,power_law_1.2,0.5445631980895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,4,power_law_1.2,0.7142848014831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,4,power_law_1.2,0.6897535800933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,4,power_law_1.2,0.7346367835998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,4,power_law_1.2,0.7528639793395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,4,power_law_1.2,0.7970623970031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,4,power_law_1.2,0.760640001296997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,4,power_law_1.2,0.7945280075073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,4,power_law_1.2,0.81594877243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,4,power_law_1.2,0.8508799552917481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,4,power_law_1.2,0.8359935760498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,4,power_law_1.2,0.8160127639770508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,4,power_law_1.2,0.9913215637207031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,4,power_law_1.2,1.108243179321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,4,power_law_1.2,1.0547136306762694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,4,power_law_1.2,1.3482560157775878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,4,power_law_1.2,1.3296064376831054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,4,power_law_1.2,1.584620761871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,4,power_law_1.2,1.6794879913330079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,4,power_law_1.2,2.088921546936035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,4,power_law_1.2,2.762713623046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,4,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,4,balanced,0.06914666791756947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,4,balanced,0.09222933650016785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,4,balanced,0.13572800159454346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,4,balanced,0.2216800053914388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,4,balanced,0.38837865988413495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,4,balanced,0.39189334710439044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,4,balanced,0.3927786747614543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,4,power_law_1.2,3.7916030883789062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,4,balanced,0.39587199687957764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,4,balanced,0.3969866832097371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,4,balanced,0.39905067284901935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,4,balanced,0.4030880133310954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,4,balanced,0.4057919979095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,4,balanced,0.409717321395874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,4,balanced,0.41420265038808185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,4,balanced,0.4174240032831828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,4,balanced,0.42663999398549396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,4,balanced,0.444485346476237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,4,balanced,0.4642080068588257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,4,balanced,0.49853332837422687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,4,balanced,0.5322453180948893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,4,balanced,0.5922346512476603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,4,balanced,0.6502079963684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,4,balanced,0.8512639999389648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,4,balanced,0.9353760083516439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,4,balanced,1.3920213381449382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,4,balanced,1.5581547419230144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,4,power_law_1.2,4.188902282714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,32,power_law_1.01,0.019392000138759614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,32,power_law_1.01,0.01923840045928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,32,power_law_1.01,0.018540799617767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,32,power_law_1.01,0.01809920072555542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,32,power_law_1.01,0.018777599930763243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,32,power_law_1.01,0.019398400187492372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,32,power_law_1.01,0.019526399672031403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,32,power_law_1.01,0.02040960043668747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,32,power_law_1.01,0.02096640020608902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,32,power_law_1.01,0.021087999641895293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,32,power_law_1.01,0.02465279996395111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,32,power_law_1.01,0.02512640058994293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,32,power_law_1.01,0.02476159930229187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,32,power_law_1.01,0.035071998834609985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,32,power_law_1.01,0.035392001271247864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,32,power_law_1.01,0.03696640133857727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,32,power_law_1.01,0.03731839954853058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,32,power_law_1.01,0.04016000032424927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,32,power_law_1.01,0.040550398826599124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,32,power_law_1.01,0.044607999920845035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,32,power_law_1.01,0.048902401328086854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,32,power_law_1.01,0.060063999891281125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,32,power_law_1.01,0.07539839744567871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,32,power_law_1.01,0.10062719583511352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,32,power_law_1.01,0.11715199947357177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,32,power_law_1.01,0.1671872019767761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,32,power_law_1.01,0.20694398880004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,32,power_law_1.2,0.021663999557495116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,32,power_law_1.2,0.020294399559497835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,32,power_law_1.2,0.02028159946203232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,32,power_law_1.2,0.023583999276161192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,32,power_law_1.2,0.0208064004778862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,32,power_law_1.2,0.02195200026035309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,32,power_law_1.2,0.02218240052461624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,32,power_law_1.2,0.025696000456809996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,32,power_law_1.2,0.02526719868183136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,32,power_law_1.2,0.026054400205612182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,32,power_law_1.2,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,32,power_law_1.2,0.040012800693511964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,32,power_law_1.2,0.03958399891853333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,32,power_law_1.2,0.04192639887332916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,32,power_law_1.2,0.0406143993139267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,32,power_law_1.2,0.04361599981784821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,32,power_law_1.2,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,32,power_law_1.2,0.046649599075317384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,32,power_law_1.2,0.0510591983795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,32,power_law_1.2,0.06245120167732239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,32,power_law_1.2,0.06992639899253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,32,power_law_1.2,0.09604480266571044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,32,power_law_1.2,0.11879680156707764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,32,power_law_1.2,0.16061439514160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,32,power_law_1.2,0.2058176040649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,32,power_law_1.2,0.30554239749908446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,32,power_law_1.2,0.37389440536499025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,1,power_law_1.01,0.08200960159301758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,1,power_law_1.01,0.10012799501419067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,1,power_law_1.01,0.1341248035430908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,1,power_law_1.01,0.176256000995636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,1,power_law_1.01,0.22400639057159424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,1,power_law_1.01,0.32933759689331055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,1,power_law_1.01,0.34876160621643065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,1,power_law_1.01,0.3674880027770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,1,power_law_1.01,0.37204480171203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,1,power_law_1.01,0.3797760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,1,power_law_1.01,0.3872704029083252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,1,power_law_1.01,0.39203200340270994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,1,power_law_1.01,0.40288639068603516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,1,power_law_1.01,0.416761589050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,1,power_law_1.01,0.44532480239868166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,1,power_law_1.01,0.4598720073699951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,1,power_law_1.01,0.4853055953979492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,1,power_law_1.01,0.5522047996520996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.01,0.6256832122802735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.01,0.7644032001495361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.01,0.8065792083740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.01,1.0271936416625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.01,1.2843392372131348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.01,1.695769691467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.01,2.117612838745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.01,2.930406379699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.01,3.7722816467285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,2,power_law_1.2,0.05009920001029968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,2,power_law_1.2,0.05608320236206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,2,power_law_1.2,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,2,power_law_1.2,0.07271040081977845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,2,power_law_1.2,0.07241600155830383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,2,power_law_1.2,0.07517439723014832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,2,power_law_1.2,0.07879040241241456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,2,power_law_1.2,0.08058879971504211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,2,power_law_1.2,0.08087040185928344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,2,power_law_1.2,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,2,power_law_1.2,0.08620160222053527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,2,power_law_1.2,0.08659200072288513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,2,power_law_1.2,0.08674560189247131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,2,power_law_1.2,0.09527040123939515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,2,power_law_1.2,0.10762879848480225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,2,power_law_1.2,0.10574719905853272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,2,power_law_1.2,0.12156800031661988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,2,power_law_1.2,0.13857920169830323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,2,power_law_1.2,0.1483839988708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,2,power_law_1.2,0.1729024052619934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,2,power_law_1.2,0.21525120735168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,2,power_law_1.2,0.29774720668792726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,2,power_law_1.2,0.30179200172424314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,2,power_law_1.2,0.4703104019165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,2,power_law_1.2,0.5863232135772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,2,power_law_1.2,0.8298815727233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,8,power_law_1.01,0.05564799904823303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,8,power_law_1.01,0.06213120222091675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,8,power_law_1.01,0.0657151997089386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,8,power_law_1.01,0.07442560195922851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,8,power_law_1.01,0.07434239983558655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,8,power_law_1.01,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,8,power_law_1.01,0.0804095983505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,8,power_law_1.01,0.08125439882278443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,8,power_law_1.01,0.08076159954071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,8,power_law_1.01,0.08074880242347718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,8,power_law_1.01,0.08343679904937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,8,power_law_1.01,0.08495360016822814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,8,power_law_1.01,0.08549119830131531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,8,power_law_1.01,0.0877120018005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,8,power_law_1.01,0.09105280041694641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,8,power_law_1.01,0.09262080192565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,8,power_law_1.01,0.09392639994621277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,8,power_law_1.01,0.10497280359268188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,8,power_law_1.01,0.11221120357513428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,8,power_law_1.01,0.1377087950706482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,8,power_law_1.01,0.15762560367584227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,8,power_law_1.01,0.19251840114593505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,8,power_law_1.01,0.22855041027069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,8,power_law_1.01,0.32665600776672366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,8,power_law_1.01,0.4148416042327881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,8,power_law_1.01,0.6082623958587646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,8,power_law_1.01,0.8095295906066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,2,power_law_1.2,1.1935104370117187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,64,power_law_1.2,0.05017600059509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,64,power_law_1.2,0.04779520034790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,64,power_law_1.2,0.04629760086536407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,64,power_law_1.2,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,64,power_law_1.2,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,64,power_law_1.2,0.04702079892158508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,64,power_law_1.2,0.04738560020923614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,64,power_law_1.2,0.04855040013790131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,64,power_law_1.2,0.04869759976863861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,64,power_law_1.2,0.04950399994850159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,64,power_law_1.2,0.05017600059509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,64,power_law_1.2,0.050969600677490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,64,power_law_1.2,0.051923197507858274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,64,power_law_1.2,0.05358719825744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,64,power_law_1.2,0.05756160020828247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,64,power_law_1.2,0.05809919834136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,64,power_law_1.2,0.061292797327041626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,64,power_law_1.2,0.06802560091018676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,64,power_law_1.2,0.07272959947586059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,64,power_law_1.2,0.08822399973869324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,64,power_law_1.2,0.09460480213165283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,64,power_law_1.2,0.12342400550842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,64,power_law_1.2,0.14832639694213867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,64,power_law_1.2,0.20337278842926027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,64,power_law_1.2,0.2671488046646118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,64,power_law_1.2,0.36421120166778564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,64,power_law_1.2,0.4819327831268311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,2,balanced,0.0421013335386912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,2,balanced,0.047983999053637184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,2,balanced,0.06150400141874949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,2,balanced,0.0653706689675649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,2,balanced,0.06671466430028279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,2,balanced,0.06637866795063019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,2,balanced,0.06704000135262807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,2,balanced,0.06840533514817555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,2,balanced,0.06789333124955495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,2,balanced,0.06753600140412648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,2,balanced,0.0680320014556249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,2,balanced,0.07025599976380666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,2,balanced,0.06966933111349742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,2,balanced,0.0695253312587738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,2,balanced,0.07658133407433827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,2,balanced,0.07836266855398814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,2,balanced,0.07829866806666057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,2,balanced,0.10652266939481099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,2,balanced,0.10671466588973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,2,balanced,0.13158399860064188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,2,balanced,0.16266133387883505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,2,balanced,0.2158613403638204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,2,balanced,0.2590986688931783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,2,balanced,0.36739198366800946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,2,balanced,0.4766453504562378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,2,balanced,0.6878986358642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,2,balanced,0.9010240236918131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,8,power_law_1.2,0.05000960230827332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,8,power_law_1.2,0.04937599897384644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,8,power_law_1.2,0.05196800231933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,8,power_law_1.2,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,8,power_law_1.2,0.05483520030975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,8,power_law_1.2,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,8,power_law_1.2,0.05585920214653015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,8,power_law_1.2,0.05635200142860412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,8,power_law_1.2,0.05689600110054016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,8,power_law_1.2,0.05799040198326111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,8,power_law_1.2,0.06064640283584595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,8,power_law_1.2,0.06291840076446534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,2,power_law_1.01,0.05619840025901794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,8,power_law_1.2,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,2,power_law_1.01,0.07304959893226623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,8,power_law_1.2,0.06866559982299805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,2,power_law_1.01,0.08542720079421998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,2,power_law_1.01,0.10933120250701904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,2,power_law_1.01,0.11743999719619751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,2,power_law_1.01,0.16759040355682372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,2,power_law_1.01,0.17165440320968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,2,power_law_1.01,0.17221120595932007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,8,power_law_1.2,0.0769919991493225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,8,power_law_1.2,0.08035200238227844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,8,power_law_1.2,0.08298879861831665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,8,power_law_1.2,0.09550719857215881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,8,power_law_1.2,0.11225600242614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,8,power_law_1.2,0.12836480140686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,8,power_law_1.2,0.16558079719543456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,8,power_law_1.2,0.20839040279388427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,8,power_law_1.2,0.25515520572662354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,2,power_law_1.01,0.17933440208435059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,8,power_law_1.2,0.3604288101196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,2,power_law_1.01,0.18515199422836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,2,power_law_1.01,0.18677120208740233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,2,power_law_1.01,0.19540480375289918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,2,power_law_1.01,0.19447679519653321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,2,power_law_1.01,0.20371840000152588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,2,power_law_1.01,0.21473278999328613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,2,power_law_1.01,0.21914238929748536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,2,power_law_1.01,0.2408128023147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,2,power_law_1.01,0.2818624019622803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,2,power_law_1.01,0.30468480587005614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,2,power_law_1.01,0.35761280059814454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,2,power_law_1.01,0.4384768009185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,2,power_law_1.01,0.5621376037597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,2,power_law_1.01,0.669868803024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,2,power_law_1.01,0.9139264106750489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,2,power_law_1.01,1.1368960380554198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,2,power_law_1.01,1.5208383560180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,2,power_law_1.01,2.038547134399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,8,power_law_1.2,0.43697280883789064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,8,power_law_1.2,0.6033152103424072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,8,power_law_1.2,0.832646369934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,8,power_law_1.2,0.016332800686359405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,8,power_law_1.2,0.016627199947834015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,8,power_law_1.2,0.016383999586105348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,8,power_law_1.2,0.017151999473571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,8,power_law_1.2,0.01802240014076233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,8,power_law_1.2,0.01796479970216751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,8,power_law_1.2,0.018163199722766876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,8,power_law_1.2,0.020403200387954713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,8,power_law_1.2,0.021190400421619415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,8,power_law_1.2,0.021279999613761903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,8,power_law_1.2,0.029548799991607665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,8,power_law_1.2,0.03108479976654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,8,power_law_1.2,0.03118079900741577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,8,power_law_1.2,0.030630400776863097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,8,power_law_1.2,0.031052801012992858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,8,power_law_1.2,0.03307519853115082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,8,power_law_1.2,0.032876798510551454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,8,power_law_1.2,0.03806079924106598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.2,0.040889599919319154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.2,0.05294719934463501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.2,0.05902720093727112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.2,0.07791360020637512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.2,0.09515519738197327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.2,0.12461440563201905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.2,0.15808639526367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.2,0.2207103967666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,4,power_law_1.2,0.041356799006462094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,4,power_law_1.2,0.04156160056591034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,4,power_law_1.2,0.043961599469184875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,4,power_law_1.2,0.045100799202919005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,4,power_law_1.2,0.04935039877891541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,4,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,4,power_law_1.2,0.05483520030975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,4,power_law_1.2,0.055270397663116456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,4,power_law_1.2,0.05619199872016907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,4,power_law_1.2,0.056959998607635495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,4,power_law_1.2,0.05952640175819397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,4,power_law_1.2,0.06114559769630432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,4,power_law_1.2,0.0639680027961731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,4,power_law_1.2,0.06949120163917541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,4,power_law_1.2,0.07505279779434204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,4,power_law_1.2,0.0791104018688202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,4,power_law_1.2,0.08652160167694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,4,power_law_1.2,0.10324480533599853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,4,power_law_1.2,0.10609279870986939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,4,power_law_1.2,0.12545280456542968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,4,power_law_1.2,0.15255680084228515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,4,power_law_1.2,0.1906623959541321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,4,power_law_1.2,0.25408639907836916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,4,power_law_1.2,0.32819840908050535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,4,power_law_1.2,0.4085824012756348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.2,0.28472959995269775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,4,power_law_1.2,0.6663616180419922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,4,power_law_1.2,0.8767487525939941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,4,balanced,0.04144533226887385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,4,balanced,0.03955200066169103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,4,balanced,0.062352001667022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,4,balanced,0.06186666587988535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,4,balanced,0.06364266574382782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,4,balanced,0.062218666076660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,4,balanced,0.06228800117969513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,4,balanced,0.0651146670182546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,4,balanced,0.06538666784763336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,4,balanced,0.06609599788983662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,4,balanced,0.06599466502666473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,4,balanced,0.06718933085600536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,4,balanced,0.06674666702747345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,4,balanced,0.07018666466077168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,4,balanced,0.0773173322280248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,4,balanced,0.07871466875076294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,4,balanced,0.08550933003425598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,4,balanced,0.09402666489283244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,4,balanced,0.1009226640065511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,4,balanced,0.12577600280443826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,4,balanced,0.14222932855288187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,4,balanced,0.1811306675275167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,4,balanced,0.2255893349647522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,4,balanced,0.3110613425572713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,4,balanced,0.3843413194020589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,4,balanced,0.5442826747894287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,4,balanced,0.6982293128967285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,1,power_law_1.01,0.05591679811477661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,1,power_law_1.01,0.05940480232238769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,1,power_law_1.01,0.06572160124778748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,1,power_law_1.01,0.08508800268173218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,1,power_law_1.01,0.11141760349273681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,1,power_law_1.01,0.12727680206298828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,1,power_law_1.01,0.15429120063781737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,1,power_law_1.01,0.16101759672164917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,1,power_law_1.01,0.1679103970527649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,1,power_law_1.01,0.17151999473571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,1,power_law_1.01,0.18161280155181886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,1,power_law_1.01,0.19664640426635743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,1,power_law_1.01,0.2035775899887085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,1,power_law_1.01,0.21945600509643554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,1,power_law_1.01,0.23805439472198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,1,power_law_1.01,0.2528640031814575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,1,power_law_1.01,0.2487488031387329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,1,power_law_1.01,0.2998080015182495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,1,power_law_1.01,0.3483328104019165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,1,power_law_1.01,0.43751039505004885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,4,power_law_1.2,0.032076799869537355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,1,power_law_1.01,0.5171584129333496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,4,power_law_1.2,0.03123840093612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,4,power_law_1.2,0.030950400233268737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,1,power_law_1.01,0.6982592105865478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,1,power_law_1.01,0.8749055862426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,1,power_law_1.01,1.2139967918395995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,4,power_law_1.2,0.03146879971027374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,1,power_law_1.01,1.5517888069152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,4,power_law_1.2,0.031948798894882204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,4,power_law_1.2,0.03326080143451691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,4,power_law_1.2,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,4,power_law_1.2,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,1,power_law_1.01,2.225587272644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,4,power_law_1.2,0.03699199855327606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,4,power_law_1.2,0.03814400136470795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,4,power_law_1.2,0.03777279853820801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,4,power_law_1.2,0.03973760008811951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,1,power_law_1.01,2.913644790649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,4,power_law_1.2,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,4,power_law_1.2,0.045465600490570066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,4,power_law_1.2,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,4,power_law_1.2,0.059680002927780154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,4,power_law_1.2,0.05451520085334778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,4,power_law_1.2,0.06577280163764954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,4,power_law_1.2,0.08001279830932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,4,power_law_1.2,0.10204800367355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,128,power_law_1.2,0.021875199675559998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,128,power_law_1.2,0.02173440009355545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,128,power_law_1.2,0.021555200219154358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,128,power_law_1.2,0.022655999660491942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,128,power_law_1.2,0.026841598749160766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,128,power_law_1.2,0.05316479802131653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,128,power_law_1.2,0.04007039964199066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,4,power_law_1.2,0.1183616042137146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,128,power_law_1.2,0.04028159976005554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,128,power_law_1.2,0.04044159948825836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,128,power_law_1.2,0.04028159976005554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,128,power_law_1.2,0.04005120098590851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,128,power_law_1.2,0.04076800048351288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,128,power_law_1.2,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,128,power_law_1.2,0.04110719859600067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,128,power_law_1.2,0.041356799006462094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,128,power_law_1.2,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,128,power_law_1.2,0.046316799521446225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,128,power_law_1.2,0.04838399887084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.2,0.056377601623535153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.2,0.06852480173110961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.2,0.0880832016468048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.2,0.1064255952835083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.2,0.13859200477600098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.2,0.1926591992378235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,4,power_law_1.2,0.15395840406417846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.2,0.24696319103240966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.2,0.3748800039291382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.2,0.4909823894500732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,4,power_law_1.2,0.18299520015716553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,4,power_law_1.2,0.23594241142272948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,1,power_law_1.01,0.10336639881134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,1,power_law_1.01,0.1385856032371521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,1,power_law_1.01,0.19351040124893187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,1,power_law_1.01,0.3041215896606445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,1,power_law_1.01,0.41629438400268554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,1,power_law_1.01,0.6544000148773194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,4,power_law_1.2,0.3095360040664673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,1,power_law_1.01,0.697049617767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,1,power_law_1.01,0.7315904140472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,1,power_law_1.01,0.741152000427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,1,power_law_1.01,0.7685887813568115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,1,power_law_1.01,0.7873151779174805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,1,power_law_1.01,0.8011648178100585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,1,power_law_1.01,0.8142848014831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,1,power_law_1.01,0.8360511779785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,1,power_law_1.01,0.8619584083557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,1,power_law_1.01,0.8826687812805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,1,power_law_1.01,0.9487551689147949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,1,power_law_1.01,1.0836607933044433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,1,power_law_1.01,1.1168767929077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,1,power_law_1.01,1.318899154663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,1,power_law_1.01,1.416806411743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,1,power_law_1.01,1.7579072952270507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,1,power_law_1.01,1.9610496520996095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,1,power_law_1.01,2.5479679107666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,1,power_law_1.01,3.4362369537353517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,1,power_law_1.01,4.612083053588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,1,power_law_1.01,5.587519836425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,4,power_law_1.2,0.4364607810974121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,4,power_law_1.2,0.6708928108215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,1,power_law_1.01,0.03945600092411041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,1,power_law_1.01,0.0571008026599884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,1,power_law_1.01,0.08039039969444275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,1,power_law_1.01,0.10517760515213012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,1,power_law_1.01,0.12958719730377197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,1,power_law_1.01,0.1909824013710022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,1,power_law_1.01,0.19140479564666749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,1,power_law_1.01,0.2011712074279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,1,power_law_1.01,0.20547199249267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,1,power_law_1.01,0.20791680812835694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,1,power_law_1.01,0.21816959381103515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,1,power_law_1.01,0.22425599098205568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,1,power_law_1.01,0.2311039924621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,1,power_law_1.01,0.23724160194396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,1,power_law_1.01,0.25048959255218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,1,power_law_1.01,0.25722880363464357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,1,power_law_1.01,0.2729599952697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,1,power_law_1.01,0.3114304065704346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,1,power_law_1.01,0.3403520107269287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,1,power_law_1.01,0.4216000080108643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,1,power_law_1.01,0.4837952136993408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,1,power_law_1.01,0.6350656032562256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,1,power_law_1.01,0.7948863983154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,1,power_law_1.01,1.0870335578918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,1,power_law_1.01,1.3493184089660644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,1,power_law_1.01,1.9147199630737304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,1,power_law_1.01,2.4717695236206056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,32,balanced,0.046015997727712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,32,balanced,0.045968001087506614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,32,balanced,0.046037331223487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,32,balanced,0.05035200218359629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,32,balanced,0.050101334849993386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,32,balanced,0.054101333022117615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,32,balanced,0.054245332876841225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,32,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,32,balanced,0.05610666672388712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,32,balanced,0.05527999997138977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,32,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,32,balanced,0.05861333509286245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,32,balanced,0.05985066791375478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,32,balanced,0.0617386649052302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,32,balanced,0.07049599786599477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,32,balanced,0.07060266534487407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,32,balanced,0.07633600135644276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,32,balanced,0.08939199646313985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,32,balanced,0.09092799822489421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,32,balanced,0.1090773344039917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,32,balanced,0.12417067090670268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,32,balanced,0.1554080049196879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,32,balanced,0.1811573306719462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,32,balanced,0.23493866125742593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,32,balanced,0.2869279980659485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,32,balanced,0.3946133454640706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,32,balanced,0.5033226807912191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,8,balanced,0.06003733476003011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,8,balanced,0.07863999903202057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,8,balanced,0.10422399640083313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,8,balanced,0.15956800182660422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,8,balanced,0.2611626585324605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,8,balanced,0.31519999106725055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,8,balanced,0.3195626735687256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,8,balanced,0.31890666484832764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,8,balanced,0.3205759922663371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,8,balanced,0.32548266649246216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,8,balanced,0.3196106751759847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,8,balanced,0.3203146656354268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,8,balanced,0.3206506570180257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,8,balanced,0.32473599910736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,8,balanced,0.3278613289197286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,8,balanced,0.32918399572372437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,8,balanced,0.3325440088907878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,8,balanced,0.3789066473642985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,8,balanced,0.35096534093221027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,8,balanced,0.41726934909820557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,8,balanced,0.3905973434448242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,8,balanced,0.6496853431065878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,8,balanced,0.48897067705790204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,8,balanced,0.7365439732869467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,8,balanced,0.7241919835408529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,8,balanced,1.283530632654826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,8,balanced,1.3313706715901692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,2,power_law_1.01,0.08798720240592957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,2,power_law_1.01,0.11607680320739747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,2,power_law_1.01,0.14176000356674195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,2,power_law_1.01,0.18847999572753907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,2,power_law_1.01,0.24179840087890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,2,power_law_1.01,0.29875199794769286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,2,power_law_1.01,0.38927359580993653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,2,power_law_1.01,0.3922111988067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,2,power_law_1.01,0.39590399265289306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,2,power_law_1.01,0.39831678867340087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,2,power_law_1.01,0.4234367847442627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,2,power_law_1.01,0.42479357719421384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,2,power_law_1.01,0.4277696132659912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,2,power_law_1.01,0.45399041175842286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,2,power_law_1.01,0.4652095794677734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,2,power_law_1.01,0.48175997734069825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,2,power_law_1.01,0.4871232032775879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,2,power_law_1.01,0.5331647872924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,2,power_law_1.01,0.5762112140655518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,2,power_law_1.01,0.6500544071197509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,2,power_law_1.01,0.7086783885955811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,2,power_law_1.01,0.8482751846313477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,2,power_law_1.01,0.9958399772644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,2,power_law_1.01,1.3063103675842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,2,power_law_1.01,1.5466879844665526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,2,power_law_1.01,2.1550912857055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,2,power_law_1.01,2.8471359252929687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,2,power_law_1.2,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,2,power_law_1.2,0.05487359762191772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,2,power_law_1.2,0.07043200135231018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,2,power_law_1.2,0.07869439721107482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,2,power_law_1.2,0.0871936023235321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,2,power_law_1.2,0.11568640470504761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,2,power_law_1.2,0.11892479658126831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,2,power_law_1.2,0.1268224000930786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,2,power_law_1.2,0.13144320249557495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,2,power_law_1.2,0.13582719564437867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,2,power_law_1.2,0.13863680362701417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,2,power_law_1.2,0.14997119903564454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,2,power_law_1.2,0.1552575945854187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,2,power_law_1.2,0.15312000513076782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,2,power_law_1.2,0.1627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,2,power_law_1.2,0.170905601978302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,2,power_law_1.2,0.18549760580062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,2,power_law_1.2,0.2162303924560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,2,power_law_1.2,0.23962879180908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,2,power_law_1.2,0.31537280082702634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,2,power_law_1.2,0.35050880908966064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,2,power_law_1.2,0.4678783893585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,2,power_law_1.2,0.6072512149810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,2,power_law_1.2,0.7954495906829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,2,power_law_1.2,1.1180352210998534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,2,power_law_1.2,1.4462271690368653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,4,power_law_1.01,0.05692800283432007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,4,power_law_1.01,0.06656000018119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,4,power_law_1.01,0.07230079770088196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,4,power_law_1.01,0.08636159896850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,4,power_law_1.01,0.08517760038375854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,4,power_law_1.01,0.10746879577636718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,4,power_law_1.01,0.10185600519180298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,4,power_law_1.01,0.10277119874954224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,4,power_law_1.01,0.1054527997970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,4,power_law_1.01,0.10687999725341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,4,power_law_1.01,0.1032256007194519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,4,power_law_1.01,0.10718719959259033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,4,power_law_1.01,0.10790400505065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,4,power_law_1.01,0.10776959657669068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,4,power_law_1.01,0.11396479606628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,4,power_law_1.01,0.11724799871444702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,4,power_law_1.01,0.12043520212173461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,4,power_law_1.01,0.12902400493621827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,4,power_law_1.01,0.1409343957901001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,4,power_law_1.01,0.16129280328750611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,4,power_law_1.01,0.18692480325698851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,4,power_law_1.01,0.24331519603729249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,4,power_law_1.01,0.2660288095474243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,4,power_law_1.01,0.36009600162506106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,4,power_law_1.01,0.43712639808654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,2,power_law_1.2,1.9415168762207031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,4,power_law_1.01,0.6708608150482178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,4,power_law_1.01,0.8600383758544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,8,power_law_1.2,0.04938879907131195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,8,power_law_1.2,0.05518720149993896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,8,power_law_1.2,0.0686847984790802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,8,power_law_1.2,0.07355520129203796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,8,power_law_1.2,0.08666239976882935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,8,power_law_1.2,0.09120640158653259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,8,power_law_1.2,0.0864512026309967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,8,power_law_1.2,0.09163519740104675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,8,power_law_1.2,0.09338240027427673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,8,power_law_1.2,0.09455999732017517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,8,power_law_1.2,0.09593600034713745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,8,power_law_1.2,0.09879680275917054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,8,power_law_1.2,0.10002559423446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,8,power_law_1.2,0.10692479610443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,8,power_law_1.2,0.11799039840698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,8,power_law_1.2,0.12320640087127685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,8,power_law_1.2,0.1350335955619812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,8,power_law_1.2,0.15400960445404052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,8,power_law_1.2,0.1622912049293518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,8,power_law_1.2,0.20026240348815919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,8,power_law_1.2,0.2347520112991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,8,power_law_1.2,0.32191359996795654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,8,power_law_1.2,0.38252160549163816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,8,power_law_1.2,0.5314367771148681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,8,power_law_1.2,0.6821951866149902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,8,power_law_1.2,1.0422080039978028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,8,power_law_1.2,1.2127615928649902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,8,balanced,0.04599999884764353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,8,balanced,0.0480373352766037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,8,balanced,0.048112000028292336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,8,balanced,0.05816533168156942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,8,balanced,0.07654400169849396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,8,balanced,0.10802666346232097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,8,balanced,0.13564266761144003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,8,balanced,0.1341333289941152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,8,balanced,0.13507733742396036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,8,balanced,0.13892799615859985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,8,balanced,0.13664533694585165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,8,balanced,0.14332266648610434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,8,balanced,0.1432319978872935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,8,balanced,0.14591466387112936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,8,balanced,0.15424000223477682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,8,balanced,0.15624533096949259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,8,balanced,0.16484799981117249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,8,balanced,0.17710934082667032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,8,balanced,0.1909280021985372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,8,balanced,0.21925334135691324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,8,balanced,0.24862933158874512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,8,balanced,0.299130658308665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,8,balanced,0.3666773239771525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,8,balanced,0.47792001565297443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,8,balanced,0.597599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,8,balanced,0.8214826583862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,8,balanced,1.056069294611613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,2,balanced,0.05043200155099233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,2,balanced,0.06401599943637848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,2,balanced,0.09267733494440715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,2,balanced,0.1455893317858378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,2,balanced,0.2547573248545329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,2,balanced,0.46744000911712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,2,balanced,0.4809173345565796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,2,balanced,0.49223466714223224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,2,balanced,0.5062613487243652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,2,balanced,0.5166293382644653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,2,balanced,0.5181706746419271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,2,balanced,0.538261334101359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,2,balanced,0.5554133256276449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,2,balanced,0.5289813280105591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,2,balanced,0.545578678448995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,2,balanced,0.5652106602986654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,2,balanced,0.5755146741867065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,2,balanced,0.6311466693878174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,2,balanced,0.6826559702555338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,2,balanced,0.8453439871470133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,2,balanced,0.7514932950337728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,2,balanced,1.053984006245931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,2,balanced,1.168453296025594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,2,balanced,1.2299359639485676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,2,balanced,1.3797173500061035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,2,balanced,2.3588266372680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,2,balanced,2.5232906341552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,2,balanced,0.031712000568707786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,2,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,2,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,2,balanced,0.06804266571998596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,2,balanced,0.1069493293762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,2,balanced,0.12372266252835591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,2,balanced,0.12543466687202454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,2,balanced,0.12769066294034323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,2,balanced,0.12736533085505167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,2,balanced,0.12717333436012268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,2,balanced,0.12983999649683634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,2,balanced,0.12981333335240683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,2,balanced,0.13040533661842346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,2,balanced,0.1320319970448812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,2,balanced,0.1341866652170817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,2,balanced,0.140773336092631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,2,balanced,0.14037866393725076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,2,balanced,0.16040533781051636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,2,balanced,0.15422399838765463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,2,balanced,0.18876266479492188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,2,balanced,0.17708265781402588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,2,balanced,0.2645866672197978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,2,balanced,0.24541334311167398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,2,balanced,0.3420693476994832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,2,balanced,0.37163734436035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,2,balanced,0.5249600013097128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,2,balanced,0.615013321240743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,2,power_law_1.01,0.07383679747581481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,2,power_law_1.01,0.08697599768638611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,2,power_law_1.01,0.11273599863052368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,2,power_law_1.01,0.12679680585861205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,2,power_law_1.01,0.16251519918441773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,2,power_law_1.01,0.1905727982521057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,2,power_law_1.01,0.21366400718688966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,2,power_law_1.01,0.2587007999420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,2,power_law_1.01,0.2689152002334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,2,power_law_1.01,0.2708672046661377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,2,power_law_1.01,0.26562559604644775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,2,power_law_1.01,0.28741118907928465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,2,power_law_1.01,0.29275519847869874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,2,power_law_1.01,0.30132479667663575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,2,power_law_1.01,0.30750720500946044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,2,power_law_1.01,0.3126015901565552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,2,power_law_1.01,0.32088959217071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,2,power_law_1.01,0.36548480987548826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,2,power_law_1.01,0.39950079917907716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,2,power_law_1.01,0.456876802444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,2,power_law_1.01,0.5285632133483886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,2,power_law_1.01,0.6645247936248779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,2,power_law_1.01,0.7513599872589112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,2,power_law_1.01,0.9905792236328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,2,power_law_1.01,1.2974783897399902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,2,power_law_1.01,1.8561471939086913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,2,power_law_1.01,2.2574783325195313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,128,power_law_1.01,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,128,power_law_1.01,0.06504960060119629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,128,power_law_1.01,0.045824000239372255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,128,power_law_1.01,0.04584319889545441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,128,power_law_1.01,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,128,power_law_1.01,0.047091200947761536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,128,power_law_1.01,0.04754559993743897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,128,power_law_1.01,0.04824320077896118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,128,power_law_1.01,0.04855040013790131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,128,power_law_1.01,0.04927999973297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,128,power_law_1.01,0.05262719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,128,power_law_1.01,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,128,power_law_1.01,0.05655679702758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,128,power_law_1.01,0.05991680026054382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,128,power_law_1.01,0.06542720198631287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,128,power_law_1.01,0.06718720197677612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,128,power_law_1.01,0.06350719928741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,128,power_law_1.01,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,128,power_law_1.01,0.08135039806365967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,128,power_law_1.01,0.09824640154838563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,128,power_law_1.01,0.10226559638977051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,128,power_law_1.01,0.1273344039916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,128,power_law_1.01,0.14803199768066405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,128,power_law_1.01,0.1969215989112854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,128,power_law_1.01,0.23754239082336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,128,power_law_1.01,0.32434558868408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,128,power_law_1.01,0.41680002212524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,1,balanced,0.05670933425426483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,1,balanced,0.062352001667022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,1,balanced,0.08348266283671062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,1,balanced,0.11692266662915547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,1,balanced,0.18209065993626913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,1,balanced,0.29449599981307983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,1,balanced,0.3006666700045268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,1,balanced,0.2938186724980672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,1,balanced,0.2961759964625041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,1,balanced,0.30033065875371295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,1,balanced,0.30248000224431354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,1,balanced,0.31011732419331867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,1,balanced,0.31567466259002686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,1,balanced,0.32238932450612384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,1,balanced,0.33164799213409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,1,balanced,0.34397868315378827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,1,balanced,0.371018648147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,1,balanced,0.42126933733622235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,1,balanced,0.4612746636072795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,1,balanced,0.5417706569035848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,1,balanced,0.6132160027821859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,1,balanced,0.8458720048268636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,1,balanced,0.9642240206400553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,1,balanced,1.383637269337972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,1,balanced,1.7165013949076335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,1,balanced,2.480858643849691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,1,balanced,3.1980107625325522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,4,power_law_1.2,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,4,power_law_1.2,0.07651200294494628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,4,power_law_1.2,0.09099519848823548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,4,power_law_1.2,0.10777599811553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,4,power_law_1.2,0.12265599966049194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,4,power_law_1.2,0.14754559993743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,4,power_law_1.2,0.14446719884872436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,4,power_law_1.2,0.17725440263748168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,4,power_law_1.2,0.18916480541229247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,4,power_law_1.2,0.17441920042037964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,4,power_law_1.2,0.17856639623641968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,4,power_law_1.2,0.19589760303497314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,4,power_law_1.2,0.18335360288619995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,4,power_law_1.2,0.20007040500640869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,4,power_law_1.2,0.21162879467010498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,4,power_law_1.2,0.2179327964782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,4,power_law_1.2,0.22885758876800538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,4,power_law_1.2,0.2564863920211792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,4,power_law_1.2,0.2901439905166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,4,power_law_1.2,0.34933760166168215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,4,power_law_1.2,0.38265600204467776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,4,power_law_1.2,0.5118207931518555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,4,power_law_1.2,0.6147007942199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,4,power_law_1.2,0.8265791893005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,4,power_law_1.2,0.9767680168151855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,4,power_law_1.2,1.3538047790527343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,4,power_law_1.2,1.7918207168579101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,128,power_law_1.2,0.016921600699424742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,128,power_law_1.2,0.01653759926557541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,128,power_law_1.2,0.017535999417304993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,128,power_law_1.2,0.01736319959163666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,128,power_law_1.2,0.018956799805164338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,128,power_law_1.2,0.025248000025749208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,128,power_law_1.2,0.02508159875869751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,128,power_law_1.2,0.025548800826072693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,128,power_law_1.2,0.025209599733352663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,128,power_law_1.2,0.024985599517822265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,128,power_law_1.2,0.025472000241279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,128,power_law_1.2,0.025382399559020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,128,power_law_1.2,0.026348799467086792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,128,power_law_1.2,0.027251198887825012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,128,power_law_1.2,0.028307199478149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,128,power_law_1.2,0.02831999957561493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,128,power_law_1.2,0.029676800966262816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,128,power_law_1.2,0.0324288010597229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,128,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,128,power_law_1.2,0.03703039884567261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,128,power_law_1.2,0.03887999951839447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,128,power_law_1.2,0.0452672004699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,128,power_law_1.2,0.05127679705619812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,128,power_law_1.2,0.06510720252990723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,128,power_law_1.2,0.08227840065956116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,128,power_law_1.2,0.11008640527725219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,128,power_law_1.2,0.13983999490737914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,8,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,8,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,8,balanced,0.07411199808120728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,8,balanced,0.09987200299898784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,8,balanced,0.14938666423161825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,8,balanced,0.24995199839274088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,8,balanced,0.34980801741282147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,8,balanced,0.34753068288167316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,8,balanced,0.35035733381907147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,8,balanced,0.3510773181915283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,8,balanced,0.3530079921086629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,8,balanced,0.3548053503036499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,8,balanced,0.35879464944203693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,8,balanced,0.3587520122528076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,8,balanced,0.37089065710703534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,8,balanced,0.37404266993204754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,8,balanced,0.3812853495279948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,8,balanced,0.3983413378397624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,8,balanced,0.40961066881815594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,8,balanced,0.44037334124247235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,8,balanced,0.4681280056635539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,8,balanced,0.523967981338501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,8,balanced,0.5857439835866293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,8,balanced,0.6947626272837321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,8,balanced,0.8436693350474039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,8,balanced,1.0574666659037273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,8,balanced,1.3882400194803874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,8,power_law_1.2,0.07075200080871583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,8,power_law_1.2,0.0921343982219696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,8,power_law_1.2,0.0926144003868103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,8,power_law_1.2,0.11192959547042847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,8,power_law_1.2,0.1191167950630188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,8,power_law_1.2,0.12637439966201783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,8,power_law_1.2,0.12267520427703857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,8,power_law_1.2,0.13278720378875733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,8,power_law_1.2,0.12528640031814575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,8,power_law_1.2,0.1325055956840515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,8,power_law_1.2,0.12997119426727294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,8,power_law_1.2,0.13141759634017944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,8,power_law_1.2,0.13159040212631226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,8,power_law_1.2,0.14326399564743042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,8,power_law_1.2,0.15262080430984498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,8,power_law_1.2,0.14762239456176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,8,power_law_1.2,0.16971520185470582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,8,power_law_1.2,0.18994560241699218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,8,power_law_1.2,0.20277760028839112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,8,power_law_1.2,0.2458048105239868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,8,power_law_1.2,0.2795072078704834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,8,power_law_1.2,0.3507456064224243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,8,power_law_1.2,0.4115903854370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,8,power_law_1.2,0.5594367980957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,8,power_law_1.2,0.710316801071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,8,power_law_1.2,0.9419903755187988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,8,power_law_1.2,1.179142379760742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,1,power_law_1.2,0.059935998916625974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,1,power_law_1.2,0.07094399929046631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,1,power_law_1.2,0.08506240248680115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,1,power_law_1.2,0.11259520053863525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,1,power_law_1.2,0.151910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,1,power_law_1.2,0.19623680114746095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,1,power_law_1.2,0.22854399681091309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,1,power_law_1.2,0.29011199474334715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,1,power_law_1.2,0.3071808099746704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,1,power_law_1.2,0.3166656017303467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,1,power_law_1.2,0.33033599853515627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,1,power_law_1.2,0.34215679168701174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,1,power_law_1.2,0.35355520248413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,1,power_law_1.2,0.3736000061035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,1,power_law_1.2,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,1,power_law_1.2,0.3848383903503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,1,power_law_1.2,0.05422719717025757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,1,power_law_1.2,0.0749888002872467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,1,power_law_1.2,0.09050880074501037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,1,power_law_1.2,0.09044479727745056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,1,power_law_1.2,0.09330559968948364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,1,power_law_1.2,0.09357439875602722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,1,power_law_1.2,0.09418879747390747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,1,power_law_1.2,0.0945855975151062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,1,power_law_1.2,0.3940927982330322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,1,power_law_1.2,0.09475200176239014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,1,power_law_1.2,0.42695040702819825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,1,power_law_1.2,0.09624959826469422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,1,power_law_1.2,0.09845759868621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,1,power_law_1.2,0.10502400398254394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,1,power_law_1.2,0.11348479986190796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,1,power_law_1.2,0.1260480046272278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,1,power_law_1.2,0.49497599601745607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,1,power_law_1.2,0.5395264148712158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,1,power_law_1.2,0.6438911914825439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,1,power_law_1.2,0.7226111888885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,1,power_law_1.2,0.9176896095275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,1,power_law_1.2,0.13041919469833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,1,power_law_1.2,1.1199999809265138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,1,power_law_1.2,0.13408000469207765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,1,power_law_1.2,0.1603008031845093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,4,power_law_1.01,0.06898559927940369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,4,power_law_1.01,0.08805760145187377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,4,power_law_1.01,0.10394879579544067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,4,power_law_1.01,0.1261888027191162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,4,power_law_1.01,0.1446720004081726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,1,power_law_1.2,1.47128324508667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.2,0.18913919925689698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,1,power_law_1.2,1.9555200576782226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,1,power_law_1.2,2.621958351135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,4,power_law_1.01,0.16506880521774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,4,power_law_1.01,0.1928704023361206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,4,power_law_1.01,0.20199038982391357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,4,power_law_1.01,0.19324159622192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,1,power_law_1.2,3.1442304611206056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,4,power_law_1.01,0.1911743998527527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,4,power_law_1.01,0.20055038928985597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,4,power_law_1.01,0.21067519187927247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,4,power_law_1.01,0.2104896068572998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,4,power_law_1.01,0.2049407958984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,4,power_law_1.01,0.22444159984588624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,4,power_law_1.01,0.220032000541687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,4,power_law_1.01,0.2383296012878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,4,power_law_1.01,0.2671360015869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.2,0.2607680082321167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,4,power_law_1.01,0.2662463903427124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.2,0.3129535913467407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,4,power_law_1.01,0.31554560661315917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,4,power_law_1.01,0.35790081024169923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,4,power_law_1.01,0.4345407962799072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,4,power_law_1.01,0.49069437980651853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,4,power_law_1.01,0.6085951805114747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,4,power_law_1.01,0.787059211730957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,4,power_law_1.01,1.0665663719177245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,4,power_law_1.01,1.2954815864562987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.2,0.42575998306274415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.2,0.5501760005950928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.2,0.7775040149688721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.2,1.0186559677124023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,2,power_law_1.01,0.03338240087032318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,2,power_law_1.01,0.04726400077342987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,2,power_law_1.01,0.05287039875984192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,2,power_law_1.01,0.06867200136184692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,2,power_law_1.01,0.07536640167236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,2,power_law_1.01,0.11284480094909669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,2,power_law_1.01,0.11201280355453491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,2,power_law_1.01,0.11237759590148926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,2,power_law_1.01,0.11365760564804077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,2,power_law_1.01,0.11650559902191163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,2,power_law_1.01,0.11947519779205322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,2,power_law_1.01,0.12392319440841675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,2,power_law_1.01,0.1254207968711853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,2,power_law_1.01,0.13003519773483277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,2,power_law_1.01,0.13386240005493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,2,power_law_1.01,0.14268800020217895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,2,power_law_1.01,0.1398911952972412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,2,power_law_1.01,0.1606335997581482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,2,power_law_1.01,0.16234240531921387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,2,power_law_1.01,0.1907520055770874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,2,power_law_1.01,0.21678080558776855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,2,power_law_1.01,0.2519615888595581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,2,power_law_1.01,0.279200005531311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,2,power_law_1.01,0.34547839164733884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,2,power_law_1.01,0.43344640731811523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,2,power_law_1.01,0.5537856101989747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,2,power_law_1.01,0.6858367919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.2,1.482534408569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.2,1.9461248397827149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,1,power_law_1.2,0.054630398750305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,1,power_law_1.2,0.05953279733657837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,1,power_law_1.2,0.06896640062332153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,1,power_law_1.2,0.08691840171813965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,1,power_law_1.2,0.10496640205383301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,1,power_law_1.2,0.12385280132293701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,1,power_law_1.2,0.16504960060119628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,1,power_law_1.2,0.16838400363922118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,1,power_law_1.2,0.17414400577545167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,1,power_law_1.2,0.17816319465637206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,1,power_law_1.2,0.18726400136947632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,1,power_law_1.2,0.1938431978225708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,1,power_law_1.2,0.2007040023803711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,1,power_law_1.2,0.20662400722503663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,1,power_law_1.2,0.21841280460357665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,1,power_law_1.2,0.23311998844146728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,1,power_law_1.2,0.24890880584716796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,1,power_law_1.2,0.2896895885467529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,1,power_law_1.2,0.3121279954910278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,1,power_law_1.2,0.3880831956863403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,1,power_law_1.2,0.4406720161437988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,1,power_law_1.2,0.5769023895263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,1,power_law_1.2,0.6837247848510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,1,power_law_1.2,0.9267071723937989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,1,power_law_1.2,1.1657600402832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,1,power_law_1.2,1.637868881225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,1,power_law_1.2,2.1075328826904296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,32,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,32,balanced,0.03562666724125544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,32,balanced,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,32,balanced,0.03740799923737844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,32,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,32,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,32,balanced,0.03915733347336451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,32,balanced,0.03941333293914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,32,balanced,0.03807999938726425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,32,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,32,balanced,0.03810133288304011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,32,balanced,0.04113066693147024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,32,balanced,0.03980266551176707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,32,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,32,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,32,balanced,0.045610666275024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,32,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,32,balanced,0.048026666045188904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,32,balanced,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,32,balanced,0.053861334919929504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,32,balanced,0.05593066910902659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,32,balanced,0.06919999917348225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,32,balanced,0.07857066889603932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,32,balanced,0.09783466657002766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,32,balanced,0.12035733461380005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,32,balanced,0.15398933490117392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,32,balanced,0.19369065761566162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,32,power_law_1.2,0.037452799081802365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,32,power_law_1.2,0.03384959995746613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,32,power_law_1.2,0.03290880024433136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,32,power_law_1.2,0.03209599852561951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,32,power_law_1.2,0.03377920091152191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,32,power_law_1.2,0.034041601419448855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,32,power_law_1.2,0.03421440124511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,32,power_law_1.2,0.034995201230049136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,32,power_law_1.2,0.03542400002479553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,32,power_law_1.2,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,32,power_law_1.2,0.036236798763275145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,32,power_law_1.2,0.03710080087184906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,32,power_law_1.2,0.03743360042572021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,32,power_law_1.2,0.03903360068798065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,32,power_law_1.2,0.04206080138683319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,32,power_law_1.2,0.0440064013004303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,32,power_law_1.2,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,32,power_law_1.2,0.05145599842071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,32,power_law_1.2,0.0546239972114563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,32,power_law_1.2,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,32,power_law_1.2,0.06823679804801941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,32,power_law_1.2,0.08233600258827209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,32,power_law_1.2,0.10876799821853637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,32,power_law_1.2,0.1418239951133728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,8,power_law_1.01,0.04567039906978607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,8,power_law_1.01,0.04959999918937683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,8,power_law_1.01,0.05246719717979431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,8,power_law_1.01,0.058143997192382814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,8,power_law_1.01,0.06537600159645081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,8,power_law_1.01,0.05825279951095581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,8,power_law_1.01,0.05775359869003296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,8,power_law_1.01,0.06341760158538819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,8,power_law_1.01,0.061292797327041626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,8,power_law_1.01,0.06122879981994629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,8,power_law_1.01,0.06640639901161194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,8,power_law_1.01,0.06756479740142822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,8,power_law_1.01,0.06796799898147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,8,power_law_1.01,0.06912000179290771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,8,power_law_1.01,0.07782400250434876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,8,power_law_1.01,0.08116480112075805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,8,power_law_1.01,0.08276479840278625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,32,power_law_1.2,0.1704767942428589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,8,power_law_1.01,0.0954367995262146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,8,power_law_1.01,0.09536640048027038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,8,power_law_1.01,0.11228799819946289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,8,power_law_1.01,0.12490240335464478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,8,power_law_1.01,0.15971200466156005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,8,power_law_1.01,0.1942080020904541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,8,power_law_1.01,0.24301440715789796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,8,power_law_1.01,0.303769588470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,8,power_law_1.01,0.4296383857727051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,8,power_law_1.01,0.5639039993286132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,32,power_law_1.2,0.26248960494995116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,1,power_law_1.2,0.05960959792137146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,1,power_law_1.2,0.07832959890365601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,1,power_law_1.2,0.10198400020599366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,1,power_law_1.2,0.14601600170135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,1,power_law_1.2,0.19020160436630248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,1,power_law_1.2,0.24118399620056152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,1,power_law_1.2,0.3395776033401489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,1,power_law_1.2,0.35983359813690186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,1,power_law_1.2,0.3766144037246704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,1,power_law_1.2,0.3809535980224609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,1,power_law_1.2,0.40805759429931643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,1,power_law_1.2,0.4186560153961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,1,power_law_1.2,0.4309247970581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,1,power_law_1.2,0.4516479969024658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,1,power_law_1.2,0.46586880683898924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,1,power_law_1.2,0.483622407913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,1,power_law_1.2,0.525881576538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,1,power_law_1.2,0.5844799995422363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,1,power_law_1.2,0.6324992179870605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,1,power_law_1.2,0.7332096099853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,1,power_law_1.2,0.8327360153198242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,1,power_law_1.2,1.0529855728149413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,1,power_law_1.2,1.2026752471923827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,1,power_law_1.2,1.518956756591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,1,power_law_1.2,1.8685247421264648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,1,power_law_1.2,2.5497791290283205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,1,power_law_1.2,3.259212875366211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,32,power_law_1.2,0.2938944101333618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,1,power_law_1.2,0.052665597200393675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,1,power_law_1.2,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,1,power_law_1.2,0.061913597583770755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,1,power_law_1.2,0.07137920260429383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,1,power_law_1.2,0.08209919929504395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,1,power_law_1.2,0.11149439811706544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,1,power_law_1.2,0.11779199838638306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,1,power_law_1.2,0.1199295997619629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,1,power_law_1.2,0.12288000583648681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,1,power_law_1.2,0.12416000366210937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,1,power_law_1.2,0.12873599529266358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,1,power_law_1.2,0.13437440395355224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,1,power_law_1.2,0.13763200044631957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,1,power_law_1.2,0.14333440065383912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,1,power_law_1.2,0.1515712022781372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,1,power_law_1.2,0.16011519432067872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,1,power_law_1.2,0.17169920206069947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,1,power_law_1.2,0.20473599433898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,1,power_law_1.2,0.22257919311523439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,1,power_law_1.2,0.27651200294494627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,1,power_law_1.2,0.3292608022689819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,1,power_law_1.2,0.44266881942749026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,1,power_law_1.2,0.5561535835266114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,1,power_law_1.2,0.7823103904724121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,1,power_law_1.2,0.9996928215026856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,1,power_law_1.2,1.434553623199463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,1,power_law_1.2,1.8575168609619142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,64,power_law_1.01,0.08768640160560608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,64,power_law_1.01,0.06295040249824524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,64,power_law_1.01,0.06111999750137329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,64,power_law_1.01,0.0603007972240448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,64,power_law_1.01,0.058143997192382814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,64,power_law_1.01,0.05866879820823669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,64,power_law_1.01,0.060889601707458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,64,power_law_1.01,0.06168320178985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,64,power_law_1.01,0.06292480230331421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,64,power_law_1.01,0.0644927978515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,64,power_law_1.01,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,64,power_law_1.01,0.0689791977405548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,64,power_law_1.01,0.06849920153617858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,64,power_law_1.01,0.06997119784355163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,64,power_law_1.01,0.07676799893379212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,64,power_law_1.01,0.08131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,64,power_law_1.01,0.08657280206680298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,64,power_law_1.01,0.09594879746437072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.01,0.1068608045578003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.01,0.12494080066680908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.01,0.1452288031578064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.01,0.17938560247421265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.01,0.21291520595550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.01,0.2697088003158569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.01,0.3359935998916626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.01,0.4817344188690186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.01,0.6410496234893799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,4,power_law_1.01,0.0692799985408783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,4,power_law_1.01,0.09173759818077087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,4,power_law_1.01,0.1077888011932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,4,power_law_1.01,0.11925760507583619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,4,power_law_1.01,0.14209920167922974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,4,power_law_1.01,0.1687872052192688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,4,power_law_1.01,0.20505599975585936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,4,power_law_1.01,0.20004479885101317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,4,power_law_1.01,0.17956479787826538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,4,power_law_1.01,0.19767040014266968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,4,power_law_1.01,0.21783039569854737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,4,power_law_1.01,0.197542405128479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,4,power_law_1.01,0.20289919376373292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,4,power_law_1.01,0.20723199844360352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,4,power_law_1.01,0.2297663927078247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,4,power_law_1.01,0.23635199069976806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,4,power_law_1.01,0.24801280498504638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,4,power_law_1.01,0.2751935958862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,4,power_law_1.01,0.2881727933883667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,4,power_law_1.01,0.3371072053909302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,4,power_law_1.01,0.38348801136016847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,4,power_law_1.01,0.4684351921081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,4,power_law_1.01,0.5473792076110839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,4,power_law_1.01,0.7587647914886475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,4,power_law_1.01,0.8975359916687011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,4,power_law_1.01,1.2508543968200683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,4,power_law_1.01,1.6373632431030274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,32,balanced,0.05443733433882395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,32,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,32,balanced,0.05392533540725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,32,balanced,0.0603359987338384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,32,balanced,0.0705973356962204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,32,balanced,0.08672533432642619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,32,balanced,0.08798933029174805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,32,balanced,0.08660800258318584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,32,balanced,0.0897653301556905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,32,balanced,0.08827733000119527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,32,balanced,0.08942400415738423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,32,balanced,0.09302933017412822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,32,balanced,0.09268800417582194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,32,balanced,0.09419199824333191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,32,balanced,0.10737066467603047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,32,balanced,0.10755733648935954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,32,balanced,0.10733333230018616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,32,balanced,0.11946666240692139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,32,balanced,0.12389333049456279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,32,balanced,0.13845866918563843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,32,balanced,0.15386666854222616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,32,balanced,0.18626133600870767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,32,balanced,0.2116746703783671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,32,balanced,0.287663996219635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,32,balanced,0.3399786551793416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,32,balanced,0.4594240188598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,32,balanced,0.5977973143259684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,64,balanced,0.017312000195185345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,64,balanced,0.01889066646496455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,64,balanced,0.01684800038735072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,64,balanced,0.018522666146357853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,64,balanced,0.018965333700180054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,64,balanced,0.021253332495689392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,64,balanced,0.020874666670958202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,64,balanced,0.025397333006064098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,64,balanced,0.025424001117547352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,64,balanced,0.02606400102376938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,64,balanced,0.0271519993742307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,64,balanced,0.027461332579453785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,64,balanced,0.027914665639400482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,64,balanced,0.029157333076000214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,64,balanced,0.029370665550231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,64,balanced,0.029440000653266907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,64,balanced,0.03128000100453695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,64,balanced,0.033258666594823204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,64,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,64,balanced,0.041402667760849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,64,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,64,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,64,balanced,0.05605333546797434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,64,balanced,0.0786293347676595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,64,balanced,0.09409600496292114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,64,balanced,0.12637866536776224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,64,balanced,0.16022400061289468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,2,balanced,0.03977066775163015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,2,balanced,0.04164266586303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,2,balanced,0.05198933184146881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,2,balanced,0.07583466668923695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,2,balanced,0.11143466830253601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,2,balanced,0.11349333326021831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,2,balanced,0.11333866914113362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,2,balanced,0.11525866389274597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,2,balanced,0.11524266997973125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,2,balanced,0.11964799960454305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,2,balanced,0.11957333485285442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,2,balanced,0.12000000476837158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,2,balanced,0.12388267119725545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,2,balanced,0.12752532958984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,2,balanced,0.13355200489362082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,2,balanced,0.13873066504796347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,2,balanced,0.14191466569900513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,2,balanced,0.16053332885106406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,2,balanced,0.17356799046198526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,2,balanced,0.20801067352294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,2,balanced,0.22955199082692465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,2,balanced,0.31538132826487225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,2,balanced,0.36322665214538574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,2,balanced,0.5309493144353231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,2,balanced,0.6642080148061117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,2,balanced,0.9620320002237955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,2,balanced,1.2555466492970784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,32,power_law_1.2,0.08052480220794678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,32,power_law_1.2,0.07756159901618957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,32,power_law_1.2,0.0785215973854065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,32,power_law_1.2,0.07990400195121765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,32,power_law_1.2,0.07831040024757385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,32,power_law_1.2,0.08090239763259888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,32,power_law_1.2,0.07831040024757385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,32,power_law_1.2,0.07642239928245545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,32,power_law_1.2,0.0793727993965149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,32,power_law_1.2,0.08268160223960877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,32,power_law_1.2,0.08371840119361877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,32,power_law_1.2,0.08555520176887513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,32,power_law_1.2,0.08746240139007569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,32,power_law_1.2,0.09157119989395142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,32,power_law_1.2,0.09919360280036926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,32,power_law_1.2,0.10384000539779663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,32,power_law_1.2,0.10823040008544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,32,power_law_1.2,0.12519680261611937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,32,power_law_1.2,0.13600000143051147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,32,power_law_1.2,0.16723840236663817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,32,power_law_1.2,0.204256010055542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,32,power_law_1.2,0.26971518993377686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,32,power_law_1.2,0.29051520824432375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,32,power_law_1.2,0.39549438953399657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,32,power_law_1.2,0.4887296199798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,32,power_law_1.2,0.6608511924743652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,32,power_law_1.2,0.8241408348083497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,16,power_law_1.01,0.05656319856643677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,16,power_law_1.01,0.04979200065135956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,16,power_law_1.01,0.04717440009117126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,16,power_law_1.01,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,16,power_law_1.01,0.04799999892711639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,16,power_law_1.01,0.04924159944057464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,16,power_law_1.01,0.050470399856567386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,16,power_law_1.01,0.056409597396850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,16,power_law_1.01,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,16,power_law_1.01,0.05741440057754517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,16,power_law_1.01,0.05736960172653198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,16,power_law_1.01,0.05745919942855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,16,power_law_1.01,0.05800960063934326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,16,power_law_1.01,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,16,power_law_1.01,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,16,power_law_1.01,0.05125120282173157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,16,power_law_1.01,0.05980160236358643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,16,power_law_1.01,0.051718401908874514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,16,power_law_1.01,0.05980799794197082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,16,power_law_1.01,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,16,power_law_1.01,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,16,power_law_1.01,0.06265599727630615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,16,power_law_1.01,0.06364799737930298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,16,power_law_1.01,0.05359359979629517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,16,power_law_1.01,0.06589440107345582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,16,power_law_1.01,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,16,power_law_1.01,0.07221760153770447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,16,power_law_1.01,0.07508479952812194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,16,power_law_1.01,0.07934079766273498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,16,power_law_1.01,0.09425280094146729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,16,power_law_1.01,0.10190720558166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,16,power_law_1.01,0.1248128056526184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,16,power_law_1.01,0.055641597509384154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,16,power_law_1.01,0.14240000247955323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,16,power_law_1.01,0.06183040142059326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,16,power_law_1.01,0.18228479623794555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,16,power_law_1.01,0.06538879871368408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,16,power_law_1.01,0.22286078929901124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,16,power_law_1.01,0.2840447902679443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,16,power_law_1.01,0.07914239764213563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,16,power_law_1.01,0.3740351915359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,16,power_law_1.01,0.10695680379867553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,16,power_law_1.01,0.5125951766967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,16,power_law_1.01,0.10412800312042236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,16,power_law_1.01,0.6792448043823243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,16,power_law_1.01,0.10764800310134888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,16,power_law_1.01,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,16,power_law_1.01,0.06918399930000305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,16,power_law_1.01,0.10700160264968872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,16,power_law_1.01,0.07963520288467407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,16,power_law_1.01,0.0852735996246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,16,power_law_1.01,0.10264320373535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,16,power_law_1.01,0.1107200026512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,16,power_law_1.01,0.10759040117263793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,16,power_law_1.01,0.10577280521392822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,16,power_law_1.01,0.10616960525512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,16,power_law_1.01,0.10952960252761841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,16,power_law_1.01,0.10822399854660034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,16,power_law_1.01,0.12059520483016968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,16,power_law_1.01,0.10919040441513062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,16,power_law_1.01,0.1590208053588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,16,power_law_1.01,0.1130944013595581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,16,power_law_1.01,0.20301439762115478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,16,power_law_1.01,0.11402239799499511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,16,power_law_1.01,0.2552128076553345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,16,power_law_1.01,0.12465280294418335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,16,power_law_1.01,0.30573439598083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,16,power_law_1.01,0.1257024049758911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,16,power_law_1.01,0.4276735782623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,16,power_law_1.01,0.13100800514221192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,16,power_law_1.01,0.5925439834594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,16,power_law_1.01,0.1511680006980896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,16,power_law_1.01,0.15489920377731323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,16,power_law_1.01,0.1733247995376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,16,power_law_1.01,0.2020479917526245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,16,power_law_1.01,0.26185600757598876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,16,power_law_1.01,0.30379519462585447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,16,power_law_1.01,0.4394815921783447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,16,power_law_1.01,0.5363647937774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,16,power_law_1.01,0.825318431854248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,16,power_law_1.01,1.0620800018310548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,1,power_law_1.01,0.04785279929637909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,1,power_law_1.01,0.06388480067253113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,1,power_law_1.01,0.09804159998893738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,1,power_law_1.01,0.14001280069351196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,1,power_law_1.01,0.19856640100479125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,1,power_law_1.01,0.3091264009475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,1,power_law_1.01,0.32227840423583987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,1,power_law_1.01,0.34408960342407224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,1,power_law_1.01,0.3363136053085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,1,power_law_1.01,0.35223679542541503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,1,power_law_1.01,0.3602816104888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,1,power_law_1.01,0.3641279935836792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,1,power_law_1.01,0.3701440095901489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,1,power_law_1.01,0.3836735963821411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,1,power_law_1.01,0.39457919597625735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,1,power_law_1.01,0.4018496036529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,1,power_law_1.01,0.42531838417053225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,1,power_law_1.01,0.4687488079071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,1,power_law_1.01,0.4811391830444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,1,power_law_1.01,0.5474944114685059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,1,power_law_1.01,0.5833663940429688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,1,power_law_1.01,0.68472318649292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,1,power_law_1.01,0.762336015701294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,1,power_law_1.01,0.9354880332946778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,1,power_law_1.01,1.2157247543334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,1,power_law_1.01,1.5404288291931152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,1,power_law_1.01,1.7746047973632812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,8,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,8,power_law_1.2,0.036320000886917114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,8,power_law_1.2,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,8,power_law_1.2,0.03570559918880463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,8,power_law_1.2,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,8,power_law_1.2,0.03989120125770569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,8,power_law_1.2,0.04695680141448975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,8,power_law_1.2,0.045542401075363156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,8,power_law_1.2,0.049593600630760196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,8,power_law_1.2,0.05002239942550659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,8,power_law_1.2,0.05001599788665771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,8,power_law_1.2,0.058310401439666745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,8,power_law_1.2,0.05914880037307739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,8,power_law_1.2,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,8,power_law_1.2,0.07555199861526489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,8,power_law_1.2,0.07541760206222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,8,power_law_1.2,0.08770560026168824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,8,power_law_1.2,0.10807679891586304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,8,power_law_1.2,0.1119871973991394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,8,power_law_1.2,0.14942079782485962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,8,power_law_1.2,0.18148479461669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,8,power_law_1.2,0.24121599197387694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,8,power_law_1.2,0.32042880058288575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,8,power_law_1.2,0.42940797805786135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,32,power_law_1.01,0.0494592010974884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,32,power_law_1.01,0.046105599403381346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,32,power_law_1.01,0.04482559859752655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,32,power_law_1.01,0.04733439981937408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,32,power_law_1.01,0.047654399275779726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,32,power_law_1.01,0.04903680086135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,32,power_law_1.01,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,32,power_law_1.01,0.05050240159034729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,32,power_law_1.01,0.0506496012210846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,32,power_law_1.01,0.0514303982257843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,32,power_law_1.01,0.051820802688598636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,32,power_law_1.01,0.05307520031929016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,32,power_law_1.01,0.053427201509475705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,32,power_law_1.01,0.05429760217666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,32,power_law_1.01,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,32,power_law_1.01,0.06041600108146668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,32,power_law_1.01,0.0648576021194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,32,power_law_1.01,0.0713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,32,power_law_1.01,0.07926399707794189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,32,power_law_1.01,0.09880959987640381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,32,power_law_1.01,0.1099392056465149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,32,power_law_1.01,0.1460927963256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,32,power_law_1.01,0.16809600591659546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,32,power_law_1.01,0.2361920118331909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,32,power_law_1.01,0.29697279930114745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,8,power_law_1.2,0.5289599895477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,32,power_law_1.01,0.4378367900848389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,32,power_law_1.01,0.5598464012145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,8,power_law_1.2,0.7759103775024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,1,power_law_1.2,0.041145598888397215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,1,power_law_1.2,0.043507200479507444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,1,power_law_1.2,0.04642559885978699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,1,power_law_1.2,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,1,power_law_1.2,0.061849600076675414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,1,power_law_1.2,0.06561920046806335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,8,power_law_1.2,1.0270976066589355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,1,power_law_1.2,0.06543359756469727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,1,power_law_1.2,0.0664255976676941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,1,power_law_1.2,0.06624640226364135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,1,power_law_1.2,0.06682239770889283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,1,power_law_1.2,0.0692351996898651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,1,power_law_1.2,0.07274240255355835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,1,power_law_1.2,0.0761023998260498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,1,power_law_1.2,0.0787775993347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,1,power_law_1.2,0.08840320110321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,1,power_law_1.2,0.08983039855957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,1,power_law_1.2,0.09557759761810303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,1,power_law_1.2,0.1124608039855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.2,0.12878079414367677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.2,0.1599552035331726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.2,0.18574719429016112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.2,0.24904320240020753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.2,0.30714240074157717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.2,0.4325695991516113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.2,0.5558911800384522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.2,0.7906239986419678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.2,1.0288191795349122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,64,balanced,0.019061333189407986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,64,balanced,0.019253333409627277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,64,balanced,0.018981333822011948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,64,balanced,0.018917333334684372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,64,balanced,0.01923199991385142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,64,balanced,0.019146667172511418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,64,balanced,0.024447999894618988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,64,balanced,0.023056000471115112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,64,balanced,0.029167999823888142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,64,balanced,0.027263998985290527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,64,balanced,0.034304000437259674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,64,balanced,0.03431999931732813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,64,balanced,0.031680000325044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,64,balanced,0.03192000091075897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,64,balanced,0.041477332512537636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,64,balanced,0.0406986673672994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,64,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,64,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,64,balanced,0.04980800052483877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,64,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,64,balanced,0.07041066884994507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,64,balanced,0.08981866637865703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,64,balanced,0.1127839982509613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,64,balanced,0.15069333712259927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,64,balanced,0.19266132513682047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,64,balanced,0.27507734298706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,64,balanced,0.3570079803466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,32,power_law_1.2,0.022252799570560457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,32,power_law_1.2,0.024403199553489685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,32,power_law_1.2,0.020768000185489653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,32,power_law_1.2,0.022316800057888032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,32,power_law_1.2,0.02232320010662079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,32,power_law_1.2,0.022643199563026427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,32,power_law_1.2,0.02396160066127777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,32,power_law_1.2,0.024684800207614897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,32,power_law_1.2,0.026080000400543212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,32,power_law_1.2,0.025183999538421632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,32,power_law_1.2,0.030995199084281923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,32,power_law_1.2,0.030124801397323608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,32,power_law_1.2,0.030137598514556885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,32,power_law_1.2,0.05240960121154785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,32,power_law_1.2,0.05403519868850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,32,power_law_1.2,0.056364798545837404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,32,power_law_1.2,0.05311359763145447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,32,power_law_1.2,0.05052800178527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,32,power_law_1.2,0.05960959792137146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,32,power_law_1.2,0.06584320068359376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,32,power_law_1.2,0.08426880240440368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,32,power_law_1.2,0.10734720230102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,32,power_law_1.2,0.12272000312805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,32,power_law_1.2,0.16193920373916626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,32,power_law_1.2,0.21787519454956056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,32,power_law_1.2,0.30796799659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,32,power_law_1.2,0.42691841125488283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,32,power_law_1.01,0.05711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,32,power_law_1.01,0.03902080059051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,32,power_law_1.01,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,32,power_law_1.01,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,32,power_law_1.01,0.03336319923400879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,32,power_law_1.01,0.034771201014518735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,32,power_law_1.01,0.035385599732398985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,32,power_law_1.01,0.03540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,32,power_law_1.01,0.036032000184059144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,32,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,32,power_law_1.01,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,32,power_law_1.01,0.040703999996185306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,32,power_law_1.01,0.042361599206924436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,32,power_law_1.01,0.04439679980278015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,32,power_law_1.01,0.050361597537994386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,32,power_law_1.01,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,32,power_law_1.01,0.05742719769477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,32,power_law_1.01,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,32,power_law_1.01,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,32,power_law_1.01,0.06878719925880432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,32,power_law_1.01,0.0765887975692749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,32,power_law_1.01,0.09020159840583801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,32,power_law_1.01,0.10286719799041748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,32,power_law_1.01,0.13556480407714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,32,power_law_1.01,0.15828479528427125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,32,power_law_1.01,0.22009599208831787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,32,power_law_1.01,0.2537152051925659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,4,power_law_1.2,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,4,power_law_1.2,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,4,power_law_1.2,0.03410559892654419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,4,power_law_1.2,0.037510401010513304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,4,power_law_1.2,0.038841599225997926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,4,power_law_1.2,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,4,power_law_1.2,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,4,power_law_1.2,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,4,power_law_1.2,0.04323840141296387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,4,power_law_1.2,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,4,power_law_1.2,0.04850560128688812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,4,power_law_1.2,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,4,power_law_1.2,0.05011839866638183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,4,power_law_1.2,0.0565887987613678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,4,power_law_1.2,0.06389120221138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,4,power_law_1.2,0.06588159799575806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,4,power_law_1.2,0.06766719818115234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,4,power_law_1.2,0.0775488018989563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,4,power_law_1.2,0.09654399752616882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,4,power_law_1.2,0.11425280570983887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,4,power_law_1.2,0.13890559673309327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,4,power_law_1.2,0.17662719488143921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,4,power_law_1.2,0.23116800785064698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,4,power_law_1.2,0.3183552026748657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,4,power_law_1.2,0.35308160781860354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,1,power_law_1.01,0.022809599339962006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,1,power_law_1.01,0.022380800545215608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,1,power_law_1.01,0.021855999529361726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,1,power_law_1.01,0.021459199488162994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,1,power_law_1.01,0.02263039946556091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,1,power_law_1.01,0.025971201062202454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,1,power_law_1.01,0.02924799919128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,1,power_law_1.01,0.030982398986816408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,1,power_law_1.01,0.038335999846458434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,1,power_law_1.01,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,1,power_law_1.01,0.037062400579452516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,1,power_law_1.01,0.041484799981117246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,1,power_law_1.01,0.04622719883918762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,1,power_law_1.01,0.048403200507164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,1,power_law_1.01,0.06568959951400757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,1,power_law_1.01,0.0568448007106781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,1,power_law_1.01,0.06375679969787598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,4,power_law_1.2,0.5477503776550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,1,power_law_1.01,0.08401920199394226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.01,0.09932799935340882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.01,0.13731839656829833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.01,0.17136000394821166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.01,0.24017279148101806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.01,0.29957120418548583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.01,0.43508481979370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,4,power_law_1.2,0.7045184135437011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.01,0.5636544227600098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.01,0.8286272048950195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,2,balanced,0.06403199831644694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,2,balanced,0.06630933284759521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,2,balanced,0.0728053351243337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,2,balanced,0.09869866569836934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,2,balanced,0.13900799552599588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,2,balanced,0.20910932620366415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,2,balanced,0.20964266856511435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,2,balanced,0.21161067485809326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,2,balanced,0.21209067106246948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,2,balanced,0.21330134073893228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,2,balanced,0.21636799971262613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,2,balanced,0.21897067626317343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,2,balanced,0.21954133113225302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,2,balanced,0.22601600488026938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,2,balanced,0.23176000515619913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,2,balanced,0.23405865828196207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,2,balanced,0.243450661500295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,2,balanced,0.2688373327255249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,2,balanced,0.2887626687685649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,2,balanced,0.33183467388153076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,2,balanced,0.36937065919240314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,2,balanced,0.45789865652720135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,2,balanced,0.539413332939148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,2,balanced,0.772266705830892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,2,balanced,0.9483093420664469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,2,balanced,1.3675840695699055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,2,balanced,1.764512062072754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.01,1.0954624176025392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,8,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,8,balanced,0.03930133332808813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,8,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,8,balanced,0.03938666731119156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,8,balanced,0.038506666819254555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,8,balanced,0.03779733429352442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,8,balanced,0.039621333281199135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,8,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,8,balanced,0.039503999054431915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,8,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,8,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,8,balanced,0.04080000023047129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,8,balanced,0.03959999978542328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,8,balanced,0.04196799794832865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,8,balanced,0.045642669002215065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,8,balanced,0.04626133541266123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,8,balanced,0.046037331223487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,8,balanced,0.05097066859404246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,8,balanced,0.05415999889373779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,8,balanced,0.062234664956728615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,8,balanced,0.06651199857393901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,8,balanced,0.08613333106040955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,8,balanced,0.0977226694424947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,8,balanced,0.12325867017110188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,8,balanced,0.14494933684666952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,8,balanced,0.18958399693171182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,8,balanced,0.23643734057744345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,8,power_law_1.2,0.12072960138320923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,8,power_law_1.2,0.12340480089187622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,8,power_law_1.2,0.12008320093154908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,8,power_law_1.2,0.1181056022644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,8,power_law_1.2,0.11846400499343872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,8,power_law_1.2,0.1267135977745056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,8,power_law_1.2,0.17225600481033326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,8,power_law_1.2,0.19912320375442505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,8,power_law_1.2,0.24791040420532226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,8,power_law_1.2,0.242195200920105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,8,power_law_1.2,0.1545024037361145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,8,power_law_1.2,0.17875839471817018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,8,power_law_1.2,0.18012800216674804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,8,power_law_1.2,0.22755839824676513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,8,power_law_1.2,0.3030911922454834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,8,power_law_1.2,0.30078721046447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,8,power_law_1.2,0.2780416011810303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,8,power_law_1.2,0.3380223989486694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,8,power_law_1.2,0.43932161331176756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,8,power_law_1.2,0.641868782043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,8,power_law_1.2,0.6724671840667724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,8,power_law_1.2,1.0610176086425782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,8,power_law_1.2,1.3426560401916503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,8,power_law_1.2,2.2055295944213866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,8,power_law_1.2,2.730246353149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,8,power_law_1.01,0.044921600818634035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,8,power_law_1.01,0.05392640233039856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,8,power_law_1.01,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,8,power_law_1.01,0.034483200311660765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,8,power_law_1.01,0.0356799989938736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,8,power_law_1.01,0.03580799996852875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,8,power_law_1.01,0.041356799006462094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,8,power_law_1.01,0.04227199852466583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,8,power_law_1.01,0.04459519982337952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,8,power_law_1.01,0.04950399994850159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,8,power_law_1.01,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,8,power_law_1.01,0.04991999864578247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,8,power_law_1.01,0.05121920108795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,8,power_law_1.01,0.0531391978263855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,8,power_law_1.01,0.06467840075492859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,8,power_law_1.01,0.0700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,8,power_law_1.01,0.07863039970397949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,8,power_law_1.01,0.09541119933128357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,8,power_law_1.01,0.10298880338668823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,8,power_law_1.01,0.1267840027809143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,8,power_law_1.01,0.1507904052734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,8,power_law_1.01,0.18396159410476684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,8,power_law_1.01,0.23627519607543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,8,power_law_1.2,4.025222396850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,8,balanced,0.049738665421803795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,8,balanced,0.051674668987592064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,8,balanced,0.054272000988324486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,8,balanced,0.06635199983914693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,8,balanced,0.08923199772834778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,8,balanced,0.09752000371615092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,8,balanced,0.0990826686223348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,8,balanced,0.09876799583435059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,8,balanced,0.10088533163070679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,8,balanced,0.10115733742713928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,8,balanced,0.10249599814414978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,8,balanced,0.10251200199127197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,8,balanced,0.10333866874376933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,8,balanced,0.10633599758148193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,8,balanced,0.11071999867757161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,8,balanced,0.11328533291816711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,8,balanced,0.11703999837239583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,8,balanced,0.12545599540074667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,8,balanced,0.134634663661321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,8,balanced,0.17011733849843344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,8,balanced,0.1672160029411316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,8,balanced,0.21763734022776285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,8,balanced,0.25100799401601154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,8,balanced,0.3497333526611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,8,power_law_1.01,0.34189438819885254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,8,balanced,0.414352019627889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,8,balanced,0.5597013235092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,8,balanced,0.726645310719808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,8,power_law_1.01,0.413753604888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,8,power_law_1.01,0.5692224025726318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,8,power_law_1.2,5.125267028808594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,8,power_law_1.01,0.6872320175170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,4,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,4,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,4,balanced,0.05585066477457682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,4,balanced,0.07509333391984303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,4,balanced,0.10316800077756245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,4,balanced,0.15507733821868896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,4,balanced,0.2042079965273539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,4,balanced,0.20528000593185425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,4,balanced,0.20833067099253336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,4,balanced,0.2099626660346985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,4,balanced,0.2097813288370768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,4,balanced,0.2137813369433085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,4,balanced,0.2175040046374003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,4,balanced,0.21619733174641928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,4,balanced,0.22712532679239908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,4,balanced,0.22816532850265503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,4,balanced,0.24294400215148926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,4,balanced,0.2573920090993245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,4,balanced,0.2714933355649312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,4,balanced,0.3003573417663574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,4,balanced,0.32868800560633343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,4,balanced,0.38497598965962726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,4,balanced,0.4556800127029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,4,balanced,0.5642666816711426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,4,balanced,0.7138613065083822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,4,balanced,0.9682133197784424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,4,balanced,1.208463986714681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,1,power_law_1.2,0.017824000120162962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,1,power_law_1.2,0.017561599612236023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,1,power_law_1.2,0.01791359931230545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,1,power_law_1.2,0.01761920005083084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,1,power_law_1.2,0.018380799889564516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,1,power_law_1.2,0.018911999464035035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,1,power_law_1.2,0.01996160000562668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,1,power_law_1.2,0.02046079933643341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,1,power_law_1.2,0.022969600558280946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,1,power_law_1.2,0.028646400570869444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,1,power_law_1.2,0.024588799476623534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,1,power_law_1.2,0.025190401077270507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,1,power_law_1.2,0.028652799129486085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,1,power_law_1.2,0.029049599170684816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,1,power_law_1.2,0.030803200602531434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,1,power_law_1.2,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,1,power_law_1.2,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,1,power_law_1.2,0.045612800121307376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.2,0.0484607994556427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.2,0.06395519971847534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.2,0.0755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.2,0.09615359902381897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.2,0.11967999935150146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.2,0.1645311951637268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,4,power_law_1.01,0.05422719717025757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,4,power_law_1.01,0.06316800117492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,4,power_law_1.01,0.05383679866790771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,4,power_law_1.01,0.06600959897041321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,4,power_law_1.01,0.06673280000686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,4,power_law_1.01,0.06616960167884826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,4,power_law_1.01,0.06691840291023254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,4,power_law_1.01,0.06924800276756286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,4,power_law_1.01,0.06976640224456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,4,power_law_1.01,0.07170559763908387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,4,power_law_1.01,0.07192959785461425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,4,power_law_1.01,0.0745855987071991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,4,power_law_1.01,0.07879040241241456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,4,power_law_1.01,0.08674560189247131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,4,power_law_1.01,0.09779199957847595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,4,power_law_1.01,0.10482560396194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,4,power_law_1.01,0.1066815972328186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,4,power_law_1.01,0.1350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.01,0.15955840349197387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.01,0.19434880018234252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.2,0.20651519298553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.01,0.2046207904815674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.01,0.28229761123657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.01,0.41068158149719236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.01,0.545689582824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.01,0.6550144195556641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.2,0.2949120044708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.01,1.0957311630249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.01,1.3507583618164063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.2,0.38122880458831787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,4,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,4,balanced,0.046341334780057274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,4,balanced,0.058261334896087646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,4,balanced,0.07851733267307281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,4,balanced,0.11157866319020589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,4,balanced,0.17938133080800375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,4,balanced,0.17836799224217734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,4,balanced,0.18007999658584595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,4,balanced,0.18134399255116782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,4,balanced,0.18067733446756998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,4,balanced,0.1823520064353943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,4,balanced,0.18678933382034302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,4,balanced,0.18807466824849448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,4,balanced,0.18780267238616943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,4,balanced,0.19813867410024008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,4,balanced,0.19681066274642944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,4,balanced,0.2045066754023234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,4,balanced,0.21969066063563028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,4,balanced,0.22707732518513998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,4,balanced,0.2502613266309102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,4,balanced,0.2760000030199687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,2,balanced,0.07384000221888225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,4,balanced,0.33268266916275024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,2,balanced,0.09986666838328044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,4,balanced,0.3635893265406291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,4,balanced,0.4843306541442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,4,balanced,0.5715680122375488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,4,balanced,0.8098453680674235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,4,balanced,0.9864053726196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,2,balanced,0.14961066842079163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,2,balanced,0.25091199080149335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,2,balanced,0.44658132394154865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,2,balanced,0.8398826917012533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,2,balanced,1.2338666915893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,2,balanced,1.2355893452962239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,2,balanced,1.2373546759287517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,2,balanced,1.2422880331675212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,2,balanced,1.2464426358540852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,2,balanced,1.2519466876983643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,2,balanced,1.2557706832885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,2,balanced,1.2671199639638264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,2,balanced,1.2792906761169434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,2,balanced,1.2887840270996094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,2,balanced,1.3015306790669758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,2,balanced,1.3343946139017742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,2,balanced,1.3683733940124512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,2,balanced,1.4296266237894695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,2,balanced,1.491967995961507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,2,balanced,1.6073333422342937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,2,balanced,1.7286879221598308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,2,balanced,1.9374027252197266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,2,balanced,2.2728427251180015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,2,balanced,2.6866238911946616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,2,balanced,3.659541447957357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,32,power_law_1.2,0.06736639738082886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,32,power_law_1.2,0.06590080261230469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,32,power_law_1.2,0.06466559767723083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,32,power_law_1.2,0.06455039978027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,32,power_law_1.2,0.06561279892921448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,32,power_law_1.2,0.07012479901313781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,32,power_law_1.2,0.07272959947586059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,32,power_law_1.2,0.07337599992752075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,32,power_law_1.2,0.07456640005111695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,32,power_law_1.2,0.08180480003356934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,32,power_law_1.2,0.08246399760246277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,32,power_law_1.2,0.08410239815711976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,32,power_law_1.2,0.07976959943771363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,32,power_law_1.2,0.08630399703979492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,32,power_law_1.2,0.09503359794616699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,32,power_law_1.2,0.091430401802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,32,power_law_1.2,0.1028864026069641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,32,power_law_1.2,0.11391359567642212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,32,power_law_1.2,0.12369279861450196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,32,power_law_1.2,0.16314239501953126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,32,power_law_1.2,0.1851263999938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,32,power_law_1.2,0.2622463941574097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,32,power_law_1.2,0.3393791913986206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,32,power_law_1.2,0.46260480880737304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,32,power_law_1.2,0.6557375907897949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,32,power_law_1.2,1.0303423881530762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,32,power_law_1.2,1.430412769317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,4,balanced,0.03182933231194814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,4,balanced,0.031136001149813335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,4,balanced,0.029711998999118805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,4,balanced,0.033471999069054924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,4,balanced,0.03421866645415624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,4,balanced,0.03937600056330363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,4,balanced,0.06060799956321716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,4,balanced,0.04637866715590159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,4,balanced,0.06414933502674103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,4,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,4,balanced,0.05760000149408976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,4,balanced,0.0702400008837382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,4,balanced,0.07687999804814656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,4,balanced,0.07439466814200084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,4,balanced,0.10754666725794475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,4,balanced,0.08498666683832805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,4,balanced,0.0990559955437978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,4,balanced,0.14016000429789224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,4,balanced,0.16480533281962076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,4,balanced,0.23286932706832886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,4,balanced,0.2953866720199585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,4,balanced,0.41974933942159015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,4,balanced,0.5299946864446005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,4,balanced,0.7813333670298258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,4,balanced,1.0223146279652913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,4,balanced,1.5100159645080566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,4,balanced,2.003472010294596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,1,power_law_1.01,0.03518719971179962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,1,power_law_1.01,0.03824000060558319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,1,power_law_1.01,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,1,power_law_1.01,0.06662399768829345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,1,power_law_1.01,0.07904639840126038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,1,power_law_1.01,0.11068160533905029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,1,power_law_1.01,0.11578240394592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,1,power_law_1.01,0.12163840532302857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,1,power_law_1.01,0.12259839773178101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,1,power_law_1.01,0.1255552053451538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,1,power_law_1.01,0.1257024049758911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,1,power_law_1.01,0.12703360319137574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,1,power_law_1.01,0.12828160524368287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,1,power_law_1.01,0.1343808054924011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,1,power_law_1.01,0.14172799587249757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,1,power_law_1.01,0.1441151976585388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,1,power_law_1.01,0.1522752046585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,1,power_law_1.01,0.16839040517807008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,1,power_law_1.01,0.1853119969367981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,1,power_law_1.01,0.2197119951248169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,1,power_law_1.01,0.24875519275665284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,1,power_law_1.01,0.3068928003311157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,1,power_law_1.01,0.360863995552063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,1,power_law_1.01,0.4737855911254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,1,power_law_1.01,0.5864575862884521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,1,power_law_1.01,0.8155391693115235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,1,power_law_1.01,1.0481087684631347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,8,power_law_1.2,0.12627840042114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,8,power_law_1.2,0.16810879707336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,8,power_law_1.2,0.16022399663925171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,8,power_law_1.2,0.20285439491271973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,8,power_law_1.2,0.23601920604705812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,8,power_law_1.2,0.2619328022003174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,8,power_law_1.2,0.2583168029785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,8,power_law_1.2,0.23787519931793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,8,power_law_1.2,0.23636479377746583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,8,power_law_1.2,0.2577471971511841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,8,power_law_1.2,0.2554368019104004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,8,power_law_1.2,0.26119680404663087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,8,power_law_1.2,0.25927679538726806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,8,power_law_1.2,0.25861759185791017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,8,power_law_1.2,0.28012158870697024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,8,power_law_1.2,0.2852031946182251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,8,power_law_1.2,0.291974401473999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,8,power_law_1.2,0.3118016004562378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,8,power_law_1.2,0.3319616079330444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,32,power_law_1.2,0.05268480181694031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,32,power_law_1.2,0.052249598503112796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,4,power_law_1.01,0.12726399898529053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,32,power_law_1.2,0.05192959904670715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,4,power_law_1.01,0.1589568018913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,4,power_law_1.01,0.17956479787826538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,4,power_law_1.01,0.21072640419006347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,4,power_law_1.01,0.2110464096069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,4,power_law_1.01,0.2172991991043091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,4,power_law_1.01,0.22622718811035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,4,power_law_1.01,0.22215039730072023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,4,power_law_1.01,0.2485503911972046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,32,power_law_1.2,0.05255680084228516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,4,power_law_1.01,0.25459840297698977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,32,power_law_1.2,0.0513152003288269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,32,power_law_1.2,0.05098239779472351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,32,power_law_1.2,0.05493119955062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,32,power_law_1.2,0.05548160076141358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,32,power_law_1.2,0.05512319803237915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,32,power_law_1.2,0.0562175989151001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,32,power_law_1.2,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,4,power_law_1.01,0.25167360305786135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,32,power_law_1.2,0.056377601623535153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,4,power_law_1.01,0.2650943994522095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,32,power_law_1.2,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,4,power_law_1.01,0.27844479084014895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,32,power_law_1.2,0.05978879928588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,4,power_law_1.01,0.2732032060623169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,32,power_law_1.2,0.06855040192604064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,4,power_law_1.01,0.3439296007156372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,32,power_law_1.2,0.06903679966926575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,4,power_law_1.01,0.3463615894317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,4,power_law_1.01,0.40787200927734374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,4,power_law_1.01,0.5260799884796142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,32,power_law_1.2,0.07434239983558655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,32,power_law_1.2,0.08625919818878174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,4,power_law_1.01,0.5223360061645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,32,power_law_1.2,0.09283199906349182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,32,power_law_1.2,0.10960639715194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,32,power_law_1.2,0.13277440071105956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,32,power_law_1.2,0.16325119733810425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,32,power_law_1.2,0.20389120578765868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,32,power_law_1.2,0.253766393661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,32,power_law_1.2,0.3098112106323242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,32,power_law_1.2,0.4370751857757568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,32,power_law_1.2,0.5826240062713623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,4,power_law_1.01,0.7039231777191162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,4,power_law_1.01,0.984556770324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,4,power_law_1.01,1.1942015647888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,4,power_law_1.01,1.6203264236450194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,4,power_law_1.01,2.603558349609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,4,power_law_1.01,3.6942657470703124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,4,power_law_1.01,5.17004165649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,4,power_law_1.01,5.951244735717774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,8,balanced,0.04427200059096018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,8,balanced,0.05193066596984863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,8,balanced,0.04964800179004669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,8,balanced,0.04959466556708018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,8,balanced,0.050399998823801674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,8,balanced,0.0509493350982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,8,balanced,0.05052266518274943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,8,balanced,0.035631999373435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,8,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,8,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,8,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,8,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,8,balanced,0.03941333293914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,8,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,8,balanced,0.04966933528582255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,8,balanced,0.04959466556708018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,8,balanced,0.05412266651789347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,8,balanced,0.061903998255729675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,8,balanced,0.06182399888833364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,8,balanced,0.07938133180141449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,8,balanced,0.08036266764005025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,8,balanced,0.10125866532325745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,8,balanced,0.11183466513951619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,8,balanced,0.150026669104894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,8,balanced,0.1744640072186788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,8,balanced,0.23527467250823975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,8,balanced,0.29874134063720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,32,power_law_1.2,0.051609599590301515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,32,power_law_1.2,0.05056639909744263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,32,power_law_1.2,0.04935039877891541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,32,power_law_1.2,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,32,power_law_1.2,0.05575680136680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,32,power_law_1.2,0.05642240047454834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,32,power_law_1.2,0.057715201377868654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,32,power_law_1.2,0.05875840187072754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,32,power_law_1.2,0.058956801891326904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,32,power_law_1.2,0.058899199962615965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,32,power_law_1.2,0.06112639904022217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,32,power_law_1.2,0.060736000537872314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,32,power_law_1.2,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,32,power_law_1.2,0.07187839746475219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,32,power_law_1.2,0.08171520233154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,32,power_law_1.2,0.07873280048370361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,32,power_law_1.2,0.08737919926643371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,32,power_law_1.2,0.0988864004611969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,32,power_law_1.2,0.11301759481430054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,32,power_law_1.2,0.13805439472198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,32,power_law_1.2,0.16231679916381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,32,power_law_1.2,0.2415679931640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,32,power_law_1.2,0.26191999912261965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,64,balanced,0.07585066556930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,64,balanced,0.0761653333902359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,64,balanced,0.07472000022729237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,64,balanced,0.07853333155314128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,64,balanced,0.0900320013364156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,64,balanced,0.11886933445930481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,64,balanced,0.12784000237782797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,64,balanced,0.13168000181516012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,64,balanced,0.11409067114194234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,64,balanced,0.11422933141390483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,64,balanced,0.13311466574668884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,64,balanced,0.11345066626866658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,64,balanced,0.11900800466537476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,64,balanced,0.11970133582750957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,64,balanced,0.1325813333193461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,64,balanced,0.13025599718093872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,64,balanced,0.1299626628557841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,64,balanced,0.12811199824015299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,64,balanced,0.12956266601880392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,64,balanced,0.1469013293584188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,64,balanced,0.1488053301970164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,64,balanced,0.1702400048573812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,64,balanced,0.19099199771881104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,64,balanced,0.25683732827504474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,32,power_law_1.2,0.4534463882446289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,64,balanced,0.2898719906806946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,64,balanced,0.39417600631713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,64,balanced,0.5007199843724569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,32,power_law_1.2,0.5785600185394287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,32,power_law_1.2,0.9236479759216308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,32,power_law_1.2,1.1325183868408204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,4,power_law_1.01,0.018681600689888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,4,power_law_1.01,0.01842560023069382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,4,power_law_1.01,0.018783999979496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,4,power_law_1.01,0.018617600202560425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,4,power_law_1.01,0.01963520050048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,4,power_law_1.01,0.02022400051355362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,4,power_law_1.01,0.024953599274158477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,4,power_law_1.01,0.02308479994535446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,4,power_law_1.01,0.028486400842666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,4,power_law_1.01,0.026700800657272337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,4,power_law_1.01,0.027564799785614012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,4,power_law_1.01,0.030611199140548707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,4,power_law_1.01,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,4,power_law_1.01,0.03404799997806549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,4,power_law_1.01,0.04423039853572845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,4,power_law_1.01,0.04237439930438995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,4,power_law_1.01,0.044889599084854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,4,power_law_1.01,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.01,0.06592000126838685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.01,0.08532480001449586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.01,0.10677759647369385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,4,power_law_1.2,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,4,power_law_1.2,0.0310591995716095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,4,power_law_1.2,0.03116160035133362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,4,power_law_1.2,0.03345920145511627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,4,power_law_1.2,0.03468799889087677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,4,power_law_1.2,0.035180801153182985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,4,power_law_1.2,0.035334399342536925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,4,power_law_1.2,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,4,power_law_1.2,0.03614720106124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,4,power_law_1.2,0.038361600041389464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,4,power_law_1.2,0.03865599930286408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,4,power_law_1.2,0.04312959909439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,4,power_law_1.2,0.04475519955158234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,4,power_law_1.2,0.04652160108089447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,4,power_law_1.2,0.057126402854919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.01,0.14592000246047973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,4,power_law_1.2,0.06101760268211365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,4,power_law_1.2,0.0710207998752594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,4,power_law_1.2,0.0916159987449646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,4,power_law_1.2,0.09520639777183533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.01,0.17859840393066406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,4,power_law_1.2,0.1449280023574829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.01,0.2534336090087891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,4,power_law_1.2,0.17368320226669312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,4,power_law_1.2,0.2368000030517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,4,power_law_1.2,0.3101183891296387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.01,0.33082239627838134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,4,power_law_1.2,0.4571072101593018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,4,power_law_1.2,0.6427008152008057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.01,0.4726272106170654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.01,0.6209983825683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,4,power_law_1.2,0.9605440139770508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,4,power_law_1.2,1.3203007698059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,2,power_law_1.01,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,2,power_law_1.01,0.05472000241279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,2,power_law_1.01,0.06119679808616638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,2,power_law_1.01,0.07541760206222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,2,power_law_1.01,0.08617600202560424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,2,power_law_1.01,0.10161279439926148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,2,power_law_1.01,0.12460160255432129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,2,power_law_1.01,0.12008960247039795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,2,power_law_1.01,0.12634880542755128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,2,power_law_1.01,0.12725759744644166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,2,power_law_1.01,0.12996480464935303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,2,power_law_1.01,0.13124480247497558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,2,power_law_1.01,0.13963520526885986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,2,power_law_1.01,0.14143999814987182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,2,power_law_1.01,0.1498304009437561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,2,power_law_1.01,0.15256320238113402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,2,power_law_1.01,0.1689471960067749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,2,power_law_1.01,0.195251202583313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,2,power_law_1.01,0.2172096014022827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,2,power_law_1.01,0.2708415985107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,2,power_law_1.01,0.3358400106430054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,2,power_law_1.01,0.42351360321044923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,2,power_law_1.01,0.5338496208190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,2,power_law_1.01,0.6983359813690185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,2,power_law_1.01,0.9236927986145019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,2,power_law_1.01,1.228281593322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,2,power_law_1.01,1.5901951789855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,128,power_law_1.01,0.060166400671005246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,128,power_law_1.01,0.08707839846611024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,128,power_law_1.01,0.052716797590255736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,128,power_law_1.01,0.05001599788665771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,128,power_law_1.01,0.0523967981338501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,128,power_law_1.01,0.05170559883117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,128,power_law_1.01,0.052223998308181765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,128,power_law_1.01,0.05340800285339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,128,power_law_1.01,0.05348479747772217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,128,power_law_1.01,0.053913599252700804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,128,power_law_1.01,0.05432959794998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,128,power_law_1.01,0.05603839755058289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,128,power_law_1.01,0.05665919780731201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,128,power_law_1.01,0.05708799958229065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,128,power_law_1.01,0.06257280111312866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,128,power_law_1.01,0.0645695984363556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,128,power_law_1.01,0.07144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,128,power_law_1.01,0.08488320112228394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.01,0.09311360120773315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.01,0.1109503984451294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.01,0.12844799757003783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.01,0.16506880521774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.01,0.187718403339386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.01,0.2537152051925659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.01,0.31055359840393065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.01,0.44087681770324705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.01,0.5497856140136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,8,power_law_1.01,0.032441601157188416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,8,power_law_1.01,0.0348800003528595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,8,power_law_1.01,0.03023360073566437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,8,power_law_1.01,0.029824000597000123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,8,power_law_1.01,0.030239999294281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,8,power_law_1.01,0.030956798791885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,8,power_law_1.01,0.03076480031013489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,8,power_law_1.01,0.03275519907474518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,8,power_law_1.01,0.03452160060405731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,8,power_law_1.01,0.03488639891147614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,8,power_law_1.01,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,8,power_law_1.01,0.034918400645256045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,8,power_law_1.01,0.03463039994239807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,8,power_law_1.01,0.03694080114364624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,8,power_law_1.01,0.042559999227523806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,8,power_law_1.01,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,8,power_law_1.01,0.04798719882965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,8,power_law_1.01,0.053881597518920896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,8,power_law_1.01,0.05811200141906738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,8,power_law_1.01,0.07402880191802978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,8,power_law_1.01,0.08221439719200134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,8,power_law_1.01,0.10697599649429321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,8,power_law_1.2,0.04732159972190857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,8,power_law_1.2,0.04643200039863586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,8,power_law_1.2,0.045023998618125914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,8,power_law_1.2,0.046598398685455324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,8,power_law_1.2,0.04806399941444397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,8,power_law_1.2,0.05226879715919495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,8,power_law_1.2,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,8,power_law_1.2,0.06588159799575806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,8,power_law_1.2,0.08849279880523682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,8,power_law_1.2,0.07958400249481201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,8,power_law_1.2,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,8,power_law_1.2,0.13432320356369018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,8,power_law_1.2,0.12861440181732178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,8,power_law_1.2,0.10021120309829712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,8,power_law_1.2,0.14425599575042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,8,power_law_1.2,0.12486399412155151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,8,power_law_1.01,0.12683520317077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,8,power_law_1.2,0.16179200410842895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,8,power_law_1.2,0.21132800579071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.2,0.27324159145355226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.2,0.3878976106643677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.2,0.519814395904541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.2,0.7427264213562011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,8,power_law_1.01,0.17331199645996093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.2,0.9860544204711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,8,power_law_1.01,0.2027967929840088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.2,1.4532671928405763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,8,power_law_1.01,0.29434239864349365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.2,1.914566421508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,8,power_law_1.01,0.35034880638122556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.2,2.870303916931152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,64,power_law_1.2,0.06749439835548401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,64,power_law_1.2,0.06049280166625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,64,power_law_1.2,0.05294079780578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,64,power_law_1.2,0.0537280023097992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,64,power_law_1.2,0.053446400165557864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,64,power_law_1.2,0.054611200094223024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,64,power_law_1.2,0.05448960065841675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,64,power_law_1.2,0.05624319911003113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,64,power_law_1.2,0.05539199709892273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,16,balanced,0.018885333091020584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,16,balanced,0.017173333714405697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,16,balanced,0.017114666601022083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,16,balanced,0.01720000058412552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,64,power_law_1.2,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,16,balanced,0.01878400022784869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,16,balanced,0.01924266666173935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,16,balanced,0.01911466692884763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,16,balanced,0.021018666525681812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,64,power_law_1.2,0.057030397653579715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,16,balanced,0.021018666525681812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,16,balanced,0.021205333371957142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,16,balanced,0.0233599990606308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,64,power_law_1.2,0.05873280167579651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,64,power_law_1.2,0.06082559823989868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,64,power_law_1.2,0.061689597368240354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,64,power_law_1.2,0.06683520078659058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,64,power_law_1.2,0.06777600049972535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,64,power_law_1.2,0.07482879757881164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,16,balanced,0.02493866781393687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,16,balanced,0.02502399931351344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,16,balanced,0.03046400099992752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,16,balanced,0.034815999368826546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,16,balanced,0.03495466709136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,16,balanced,0.0349440003434817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,16,balanced,0.0356480007370313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,16,balanced,0.04144533226887385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,16,balanced,0.04584000011285146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,16,balanced,0.05202133456865946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,16,balanced,0.061887999375661217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,16,balanced,0.07300266623497009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,16,balanced,0.09829333424568176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,16,balanced,0.1218826671441396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,16,balanced,0.16931732495625815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,16,balanced,0.21760000785191855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.2,3.800454330444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,128,power_law_1.01,0.05554559826850891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,128,power_law_1.01,0.05914880037307739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,128,power_law_1.01,0.04681600034236908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,128,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,64,power_law_1.2,0.0852735996246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,128,power_law_1.01,0.047193598747253415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,64,power_law_1.2,0.09518719911575317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,128,power_law_1.01,0.04762240052223206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,64,power_law_1.2,0.12206720113754273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,128,power_law_1.01,0.04837760031223297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,128,power_law_1.01,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,128,power_law_1.01,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,128,power_law_1.01,0.05065600275993347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,128,power_law_1.01,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,128,power_law_1.01,0.05251200199127197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,128,power_law_1.01,0.0538752019405365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,64,power_law_1.2,0.13857920169830323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,128,power_law_1.01,0.055379199981689456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,64,power_law_1.2,0.18865920305252076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,64,power_law_1.2,0.21887359619140626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,64,power_law_1.2,0.30690560340881345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,64,power_law_1.2,0.3796159982681274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,64,power_law_1.2,0.5476160049438477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,128,power_law_1.01,0.06042879819869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,64,power_law_1.2,0.7237055778503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,128,power_law_1.01,0.06318079829216003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,128,power_law_1.01,0.06977919936180114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,128,power_law_1.01,0.088646399974823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,128,power_law_1.01,0.09465600252151489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,128,power_law_1.01,0.12117120027542114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,128,power_law_1.01,0.13893760442733766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,128,power_law_1.01,0.1818943977355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,128,power_law_1.01,0.22239999771118163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,128,power_law_1.01,0.32006399631500243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,128,power_law_1.01,0.3889087915420532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,128,power_law_1.01,0.553439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,128,power_law_1.01,0.6898240089416504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,64,power_law_1.01,0.09296640157699584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,64,power_law_1.01,0.0626688003540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,64,power_law_1.01,0.05924479961395264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,64,power_law_1.01,0.05866879820823669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,64,power_law_1.01,0.056006401777267456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,64,power_law_1.01,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,64,power_law_1.01,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,64,power_law_1.01,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,64,power_law_1.01,0.06275839805603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,64,power_law_1.01,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,64,power_law_1.01,0.06405119895935059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,64,power_law_1.01,0.06775040030479432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,64,power_law_1.01,0.06495360136032105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,64,power_law_1.01,0.07164160013198853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,64,power_law_1.01,0.0759872019290924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,64,power_law_1.01,0.07934719920158387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,64,power_law_1.01,0.07898880243301391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,64,power_law_1.01,0.08362879753112792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,64,power_law_1.01,0.08798720240592957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,64,power_law_1.01,0.1010815978050232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,64,power_law_1.01,0.11203199625015259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,64,power_law_1.01,0.1391167998313904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,64,power_law_1.01,0.15269119739532472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,64,power_law_1.01,0.20467839241027833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,64,power_law_1.01,0.2528127908706665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,64,power_law_1.01,0.3347968101501465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,64,power_law_1.01,0.4716032028198242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,2,power_law_1.2,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,2,power_law_1.2,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,2,power_law_1.2,0.03735679984092712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,2,power_law_1.2,0.042361599206924436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,2,power_law_1.2,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,2,power_law_1.2,0.06688640117645264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,2,power_law_1.2,0.0676800012588501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,2,power_law_1.2,0.07217919826507568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,2,power_law_1.2,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,2,power_law_1.2,0.07266560196876526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,2,power_law_1.2,0.07422720193862915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,2,power_law_1.2,0.07837439775466919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,2,power_law_1.2,0.08037760257720947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,2,power_law_1.2,0.08652160167694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,2,power_law_1.2,0.0941760003566742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,2,power_law_1.2,0.09752960205078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,2,power_law_1.2,0.09521920084953309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,2,power_law_1.2,0.11253119707107544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,2,power_law_1.2,0.1291584014892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,2,power_law_1.2,0.15863679647445678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,2,power_law_1.2,0.19124480485916137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,2,power_law_1.2,0.2503232002258301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,2,power_law_1.2,0.3111936092376709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,2,power_law_1.2,0.4360447883605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,2,power_law_1.2,0.5512959957122803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,2,power_law_1.2,0.7086207866668701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,64,power_law_1.01,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,64,power_law_1.01,0.052236801385879515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,64,power_law_1.01,0.05139200091361999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,64,power_law_1.01,0.05426560044288635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,64,power_law_1.01,0.0526528000831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,64,power_law_1.01,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,64,power_law_1.01,0.05500800013542175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,64,power_law_1.01,0.05568000078201294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,64,power_law_1.01,0.055251199007034305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,64,power_law_1.01,0.054585599899291994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,64,power_law_1.01,0.05651199817657471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,64,power_law_1.01,0.057062399387359616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,64,power_law_1.01,0.05942400097846985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,64,power_law_1.01,0.058278399705886844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,64,power_law_1.01,0.06166399717330932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,64,power_law_1.01,0.06373119950294495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,64,power_law_1.01,0.06570879817008972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,64,power_law_1.01,0.07337599992752075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,64,power_law_1.01,0.07772799730300903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,64,power_law_1.01,0.09134719967842102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,64,power_law_1.01,0.10375679731369018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,64,power_law_1.01,0.12917120456695558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,64,power_law_1.01,0.1525887966156006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,64,power_law_1.01,0.20158720016479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,64,power_law_1.01,0.24833920001983642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,64,power_law_1.01,0.3442879915237427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,64,power_law_1.01,0.44401278495788576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,2,power_law_1.2,0.9763072013854981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,8,power_law_1.2,0.057920002937316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,8,power_law_1.2,0.061350399255752565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,8,power_law_1.2,0.060601598024368285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,8,power_law_1.2,0.06401280164718628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,8,power_law_1.2,0.06636160016059875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,8,power_law_1.2,0.06759679913520814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,8,power_law_1.2,0.0716863989830017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,8,power_law_1.2,0.07306879758834839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,8,power_law_1.2,0.07571200132369996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,8,power_law_1.2,0.07471359968185425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,8,power_law_1.2,0.07522559762001038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,8,power_law_1.2,0.0778880000114441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,8,power_law_1.2,0.07752959728240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,8,power_law_1.2,0.08086400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,8,power_law_1.2,0.09102079868316651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,8,power_law_1.2,0.09435520172119141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,8,power_law_1.2,0.10427520275115967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,8,power_law_1.2,0.12496639490127563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.2,0.13962240219116212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.2,0.17331199645996093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.2,0.17927680015563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.2,0.22554240226745606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.2,0.30553600788116453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.2,0.4186431884765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.2,0.546067190170288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.2,0.7779200077056885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.2,1.0056832313537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,4,balanced,0.023120000958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,4,balanced,0.024005333582560223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,4,balanced,0.027274665733178455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,4,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,4,balanced,0.057189335425694786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,4,balanced,0.08348266283671062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,4,balanced,0.08594133456548055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,4,balanced,0.08845866719881694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,4,balanced,0.090229332447052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,4,balanced,0.09198400378227234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,4,balanced,0.0967733363310496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,4,balanced,0.09644800424575806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,4,balanced,0.09917866190274556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,4,balanced,0.111135999361674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,4,balanced,0.11223999659220378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,4,balanced,0.11946133772532146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,4,balanced,0.13900799552599588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,4,balanced,0.14691733320554098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,4,balanced,0.1551040013631185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,4,balanced,0.17475199699401855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,4,balanced,0.25278399387995404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,4,balanced,0.2800053358078003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,4,balanced,0.31522132953008014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,4,balanced,0.3834880193074544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,4,balanced,0.44882134596506756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,4,balanced,0.7070720195770264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,4,balanced,0.8539679845174154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,2,balanced,0.07022400200366974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,2,balanced,0.08980266253153484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,2,balanced,0.13432533542315164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,2,balanced,0.22195200125376383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,2,balanced,0.3904373248418172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,2,balanced,0.7314186890920004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,2,balanced,0.7347520192464193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,2,balanced,0.7310986518859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,2,balanced,0.7354026635487875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,2,balanced,0.7371093432108561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,8,power_law_1.2,0.01764480024576187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,2,balanced,0.7391786575317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,8,power_law_1.2,0.01765120029449463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,2,balanced,0.7474773724873861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,8,power_law_1.2,0.01777919977903366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,2,balanced,0.7481333414713541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,2,balanced,0.7572213013966879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,2,balanced,0.770255963007609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,2,balanced,0.7735733191172282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,2,balanced,0.7838453451792399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,2,balanced,0.8176693121592203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,2,balanced,0.8518719673156738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,8,power_law_1.2,0.018406400084495546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,2,balanced,0.9052800337473551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,8,power_law_1.2,0.01883520036935806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,8,power_law_1.2,0.01912959963083267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,2,balanced,0.9524532953898112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,8,power_law_1.2,0.020076799392700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,8,power_law_1.2,0.021081599593162536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,8,power_law_1.2,0.023366400599479677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,8,power_law_1.2,0.022284799814224245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,8,power_law_1.2,0.029971200227737426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,8,power_law_1.2,0.034272000193595886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,8,power_law_1.2,0.03372800052165985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,8,power_law_1.2,0.030937600135803222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,2,balanced,1.04475736618042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,8,power_law_1.2,0.03412480056285858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,2,balanced,1.1786080201466878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,8,power_law_1.2,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,2,balanced,1.4752052625020344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,2,balanced,1.5676320393880208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,8,power_law_1.2,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,2,balanced,2.449061393737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,8,power_law_1.2,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,2,balanced,2.6606879234313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.2,0.048019200563430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.2,0.062483197450637816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.2,0.0744704008102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.2,0.0964352011680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.2,0.12009600400924683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.2,0.16416000127792357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.2,0.20629119873046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.2,0.2927615880966187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.2,0.3802367925643921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,32,power_law_1.01,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,32,power_law_1.01,0.029977598786354066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,32,power_law_1.01,0.029951998591423036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,32,power_law_1.01,0.029043200612068176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,32,power_law_1.01,0.029824000597000123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,32,power_law_1.01,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,32,power_law_1.01,0.031027200818061828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,32,power_law_1.01,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,32,power_law_1.01,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,32,power_law_1.01,0.0319680005311966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,32,power_law_1.01,0.032716798782348636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,32,power_law_1.01,0.033523198962211606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,32,power_law_1.01,0.033958399295806886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,32,power_law_1.01,0.0358271986246109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,32,power_law_1.01,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,32,power_law_1.01,0.03985919952392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,32,power_law_1.01,0.04169600009918213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,32,power_law_1.01,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,32,power_law_1.01,0.05022720098495483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,32,power_law_1.01,0.05447040200233459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,32,power_law_1.01,0.05964159965515137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,32,power_law_1.01,0.07290239930152893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,32,power_law_1.01,0.08671360015869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,32,power_law_1.01,0.12570879459381104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,32,power_law_1.01,0.15781760215759277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,32,power_law_1.01,0.21557118892669677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,32,power_law_1.01,0.2552191972732544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,8,power_law_1.2,0.020550400018692017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,8,power_law_1.2,0.023827199637889863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,8,power_law_1.2,0.025561600923538208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,8,power_law_1.2,0.025913599133491515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,8,power_law_1.2,0.024774399399757386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,8,power_law_1.2,0.026719999313354493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,8,power_law_1.2,0.030035200715065002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,8,power_law_1.2,0.029862400889396668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,8,power_law_1.2,0.03172479867935181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,8,power_law_1.2,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,8,power_law_1.2,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,8,power_law_1.2,0.03439359962940216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,8,power_law_1.2,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,8,power_law_1.2,0.04850560128688812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,8,power_law_1.2,0.05261440277099609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,8,power_law_1.2,0.05738239884376526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,8,power_law_1.2,0.08197759985923767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,8,power_law_1.2,0.09125760197639465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,8,power_law_1.2,0.09880319833755494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,8,power_law_1.2,0.1276800036430359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,8,power_law_1.2,0.09683200120925903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,8,power_law_1.2,0.13059840202331544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,8,power_law_1.2,0.15185920000076295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,8,power_law_1.2,0.20032639503479005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,8,power_law_1.2,0.24757759571075438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,8,power_law_1.2,0.35834240913391113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,8,power_law_1.2,0.4591616153717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,1,power_law_1.01,0.0379967987537384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,1,power_law_1.01,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,1,power_law_1.01,0.06033279895782471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,1,power_law_1.01,0.07468159794807434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,1,power_law_1.01,0.08845440149307252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,1,power_law_1.01,0.12190079689025879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,1,power_law_1.01,0.12739839553833007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,1,power_law_1.01,0.13615360260009765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,1,power_law_1.01,0.1405184030532837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,1,power_law_1.01,0.1450368046760559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,1,power_law_1.01,0.1493631958961487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,1,power_law_1.01,0.1527168035507202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,1,power_law_1.01,0.1589568018913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,1,power_law_1.01,0.16265599727630614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,1,power_law_1.01,0.17293440103530883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,1,power_law_1.01,0.18303359746932985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,1,power_law_1.01,0.19788800477981566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,1,power_law_1.01,0.22855041027069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,1,power_law_1.01,0.262009596824646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,1,power_law_1.01,0.32992000579833985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,1,power_law_1.01,0.3819583892822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,1,power_law_1.01,0.5131584167480469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,1,power_law_1.01,0.6369664192199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,1,power_law_1.01,0.8814271926879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,1,power_law_1.01,1.121548843383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,1,power_law_1.01,1.607276725769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,1,power_law_1.01,2.1137983322143556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,16,power_law_1.2,0.08008319735527039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,16,power_law_1.2,0.08376960158348083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,16,power_law_1.2,0.08067839741706848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,16,power_law_1.2,0.08236160278320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,16,power_law_1.2,0.06985599994659424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,16,power_law_1.2,0.08145279884338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,16,power_law_1.2,0.0800320029258728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,16,power_law_1.2,0.0812991976737976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,16,power_law_1.2,0.08162559866905213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,16,power_law_1.2,0.08350719809532166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,16,power_law_1.2,0.08810880184173583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,16,power_law_1.2,0.09035519957542419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,16,power_law_1.2,0.09351680278778077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,16,power_law_1.2,0.09490560293197632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,16,power_law_1.2,0.10526080131530761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,16,power_law_1.2,0.10585600137710571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,16,power_law_1.2,0.12145919799804687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,16,power_law_1.2,0.14625279903411864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,16,power_law_1.2,0.1404863953590393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,16,power_law_1.2,0.1761664032936096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,16,power_law_1.2,0.2085632085800171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,16,power_law_1.2,0.279040002822876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,16,power_law_1.2,0.3713344097137451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,16,power_law_1.2,0.4931072235107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,16,power_law_1.2,0.5928127765655518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,16,power_law_1.2,0.8278464317321778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,8,balanced,0.04473066826661428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,8,balanced,0.04562133550643921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,8,balanced,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,8,balanced,0.04972266654173533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,8,balanced,0.07123733560244243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,8,balanced,0.09913600484530131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,8,balanced,0.09942400455474854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,8,balanced,0.09985066453615825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,8,balanced,0.10187733173370361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,8,balanced,0.10148266951243083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,8,balanced,0.10265599687894185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,8,balanced,0.10665599505106609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,8,balanced,0.10492799679438274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,8,balanced,0.1090880036354065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,8,balanced,0.11693333586057027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,8,balanced,0.11851200461387634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,8,balanced,0.12169599533081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,8,balanced,0.13503467043240866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,8,balanced,0.14285332957903543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,8,balanced,0.16755199432373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,8,balanced,0.1893440087636312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,8,balanced,0.23706666628519693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,8,balanced,0.27875733375549316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,8,balanced,0.3797279993693034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,8,balanced,0.4647359848022461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,8,balanced,0.6615466674168905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,8,balanced,0.8353866736094157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,16,power_law_1.2,1.0287487983703614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,2,power_law_1.2,0.04754559993743897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,2,power_law_1.2,0.05237759947776795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,2,power_law_1.2,0.05671679973602295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,2,power_law_1.2,0.070169597864151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,2,power_law_1.2,0.08558080196380616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,2,power_law_1.2,0.10103039741516114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,2,power_law_1.2,0.11897599697113037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,2,power_law_1.2,0.12260479927062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,2,power_law_1.2,0.12535040378570556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,2,power_law_1.2,0.12437119483947753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,2,power_law_1.2,0.13297280073165893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,2,power_law_1.2,0.13284480571746826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,2,power_law_1.2,0.14024319648742675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,2,power_law_1.2,0.14521600008010865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,2,power_law_1.2,0.15344640016555786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,2,power_law_1.2,0.15580159425735474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,2,power_law_1.2,0.17240320444107055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,2,power_law_1.2,0.19589120149612427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,2,power_law_1.2,0.22762880325317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,2,power_law_1.2,0.2851072072982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,2,power_law_1.2,0.3406208038330078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,2,power_law_1.2,0.42256641387939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,2,power_law_1.2,0.5593152046203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,2,power_law_1.2,0.71943039894104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,2,power_law_1.2,0.9960895538330078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,2,power_law_1.2,1.294809627532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,2,power_law_1.2,1.622060775756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,4,power_law_1.2,0.056985598802566526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,4,power_law_1.2,0.07145599722862243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,4,power_law_1.2,0.09488000273704529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,4,power_law_1.2,0.1097216010093689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,4,power_law_1.2,0.13239680528640746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,4,power_law_1.2,0.14476799964904785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,4,power_law_1.2,0.1649791955947876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,4,power_law_1.2,0.18069759607315064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,4,power_law_1.2,0.18305920362472533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,4,power_law_1.2,0.19249919652938843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,4,power_law_1.2,0.18695679903030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,4,power_law_1.2,0.1806272029876709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,4,power_law_1.2,0.18403199911117554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,4,power_law_1.2,0.2002943992614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,4,power_law_1.2,0.2047935962677002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,4,power_law_1.2,0.20721919536590577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,4,power_law_1.2,0.22055039405822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,4,power_law_1.2,0.2441920042037964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,4,power_law_1.2,0.2700608015060425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,4,power_law_1.2,0.3152895927429199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,4,power_law_1.2,0.36293759346008303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,4,power_law_1.2,0.44623360633850095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,4,power_law_1.2,0.5379712104797363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,4,power_law_1.2,0.7013887882232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,4,power_law_1.2,0.8438912391662597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,4,power_law_1.2,1.1378751754760743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,4,power_law_1.2,1.4125951766967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,4,balanced,0.06246933341026306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,4,balanced,0.06637866795063019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,4,balanced,0.07569066683451335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,4,balanced,0.10672000050544739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,4,balanced,0.151829332113266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,4,balanced,0.23825067281723022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,4,balanced,0.22743467489878336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,4,balanced,0.21808532873789468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,4,balanced,0.22045334180196127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,4,balanced,0.20864532391230264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,4,balanced,0.2069759964942932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,4,balanced,0.2082080046335856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,4,balanced,0.21314134200414023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,4,balanced,0.20747200647989908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,4,balanced,0.21483200788497925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,4,balanced,0.22714134057362875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,4,balanced,0.21724265813827515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,4,balanced,0.23306665817896524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,4,balanced,0.24737600485483804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,4,balanced,0.2667093276977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,4,balanced,0.2911520004272461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,4,balanced,0.3399306535720825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,4,balanced,0.39321601390838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,4,balanced,0.5865279833475748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,4,balanced,0.6864213148752848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,4,balanced,0.9663093090057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,4,balanced,1.2611786524454753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,2,power_law_1.2,0.06974719762802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,2,power_law_1.2,0.08895999789237977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,2,power_law_1.2,0.10928000211715698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,2,power_law_1.2,0.1446272015571594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,2,power_law_1.2,0.1925503969192505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,2,power_law_1.2,0.2218048095703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,2,power_law_1.2,0.30188798904418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,2,power_law_1.2,0.30604801177978513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,2,power_law_1.2,0.32001280784606934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,2,power_law_1.2,0.32563199996948244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,2,power_law_1.2,0.350816011428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,2,power_law_1.2,0.34250879287719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,2,power_law_1.2,0.3496511936187744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,2,power_law_1.2,0.3635839939117432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,2,power_law_1.2,0.39434239864349363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,2,power_law_1.2,0.3957632064819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,2,power_law_1.2,0.4147776126861572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,2,power_law_1.2,0.4608640193939209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,2,power_law_1.2,0.4805056095123291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,2,power_law_1.2,0.5538559913635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,2,power_law_1.2,0.5960768222808838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,2,power_law_1.2,0.7190080165863038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,2,power_law_1.2,0.8324928283691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,2,power_law_1.2,1.0647104263305665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,2,power_law_1.2,1.2598655700683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,2,power_law_1.2,1.7755071640014648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,2,power_law_1.2,2.1673471450805666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,128,balanced,0.03614933292071024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,128,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,128,balanced,0.03618133316437403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,128,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,128,balanced,0.039503999054431915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,128,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,128,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,128,balanced,0.04114133367935816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,128,balanced,0.041178666055202484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,128,balanced,0.03955733279387156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,128,balanced,0.04013866682847341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,128,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,128,balanced,0.04186133543650309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,128,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,128,balanced,0.04372266431649526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,128,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,128,balanced,0.045567999283472695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,128,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,128,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,128,balanced,0.051856001218159996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,128,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,128,balanced,0.0614879975716273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,128,balanced,0.06418666740258534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,128,balanced,0.08328533172607422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,128,balanced,0.09573866923650105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,128,balanced,0.125082661708196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,128,balanced,0.1511679987112681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,32,balanced,0.025472000241279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,32,balanced,0.027242665489514668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,32,balanced,0.023391999304294586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,32,balanced,0.023546665906906128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,32,balanced,0.031471999982992806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,32,balanced,0.036943999429543815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,32,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,32,balanced,0.04177066683769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,32,balanced,0.044954667488733925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,32,balanced,0.04390400151411692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,32,balanced,0.05017599960168203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,32,balanced,0.053157334526379905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,32,balanced,0.05158400038878123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,32,balanced,0.08170133332411449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,32,balanced,0.08486933509508769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,32,balanced,0.08543466528256734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,32,balanced,0.11122133334477742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,32,balanced,0.14266666769981384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,32,balanced,0.15507733821868896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,32,balanced,0.21851199865341187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,32,balanced,0.14034133156140646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,32,balanced,0.18474666277567545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,32,balanced,0.23249600330988565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,32,balanced,0.3192426760991414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,32,balanced,0.4009386698404948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,32,balanced,0.6014613310496012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,32,balanced,0.7813440163930258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,2,power_law_1.2,0.07203840017318726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,2,power_law_1.2,0.07768959999084472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,2,power_law_1.2,0.0893567979335785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,2,power_law_1.2,0.12194559574127198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,2,power_law_1.2,0.12374399900436402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,2,power_law_1.2,0.12501120567321777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,2,power_law_1.2,0.12778879404067994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,2,power_law_1.2,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,2,power_law_1.2,0.13605120182037353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,2,power_law_1.2,0.13679360151290892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,2,power_law_1.2,0.1385472059249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,2,power_law_1.2,0.14504319429397583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,2,power_law_1.2,0.14455679655075074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,2,power_law_1.2,0.15646079778671265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,2,power_law_1.2,0.17874560356140137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,2,power_law_1.2,0.1781440019607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,2,power_law_1.2,0.19352960586547852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,2,power_law_1.2,0.2416383981704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,2,power_law_1.2,0.2679935932159424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,2,power_law_1.2,0.3197824001312256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,2,power_law_1.2,0.3960191965103149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,2,power_law_1.2,0.5369152069091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,32,balanced,0.019333332777023315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,32,balanced,0.017050666113694508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,32,balanced,0.017114666601022083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,32,balanced,0.019120000302791595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,32,balanced,0.020773333807786305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,32,balanced,0.023103999594847362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,32,balanced,0.033471999069054924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,32,balanced,0.03401600072781245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,32,balanced,0.03299200038115183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,32,balanced,0.033002667129039764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,32,balanced,0.031957333286603294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,32,balanced,0.03124266614516576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,32,balanced,0.031248000760873158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,32,balanced,0.02961066613594691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,32,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,32,balanced,0.035360001027584076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,32,balanced,0.03309333324432373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,32,balanced,0.03722133239110311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,32,balanced,0.044581333796183266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,32,balanced,0.0513866643110911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,32,balanced,0.06006399790445963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,32,balanced,0.07674666742483775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,32,balanced,0.0960640013217926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,32,balanced,0.12617599964141846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,32,balanced,0.16031466921170553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,32,balanced,0.22585066159566244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,32,balanced,0.2929439942042033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,2,power_law_1.2,0.6970047950744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,2,power_law_1.2,0.9743167877197265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,16,power_law_1.01,0.06083199977874756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,16,power_law_1.01,0.07063680291175842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,16,power_law_1.01,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,16,power_law_1.01,0.06843519806861878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,16,power_law_1.01,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,2,power_law_1.2,1.2003456115722657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,16,power_law_1.01,0.06475520133972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,16,power_law_1.01,0.06765440106391907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,16,power_law_1.01,0.06846719980239868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,16,power_law_1.01,0.06877440214157104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,16,power_law_1.01,0.07093120217323304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,16,power_law_1.01,0.07131519913673401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,16,power_law_1.01,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,16,power_law_1.01,0.0733568012714386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,16,power_law_1.01,0.07893120050430298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,16,power_law_1.01,0.08487679958343505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,16,power_law_1.01,0.08568959832191467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,16,power_law_1.01,0.0899071991443634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,16,power_law_1.01,0.09966080188751221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,16,power_law_1.01,0.1050495982170105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,16,power_law_1.01,0.12009600400924683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,16,power_law_1.01,0.1435263991355896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,16,power_law_1.01,0.17225600481033326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,16,power_law_1.01,0.20683519840240477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,16,power_law_1.01,0.286297607421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,16,power_law_1.01,0.35221760272979735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,16,power_law_1.01,0.48088321685791013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,16,power_law_1.01,0.576032018661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,2,power_law_1.2,2.0782079696655273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,32,power_law_1.2,0.07088000178337098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,32,power_law_1.2,0.07973759770393371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,32,power_law_1.2,0.07672960162162781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,32,power_law_1.2,0.08017920255661011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,32,power_law_1.2,0.0807424008846283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,32,power_law_1.2,0.08005759716033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,32,power_law_1.2,0.07916799783706666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,32,power_law_1.2,0.08013439774513245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,32,power_law_1.2,0.0800000011920929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,32,power_law_1.2,0.08372480273246766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,32,power_law_1.2,0.08361600041389465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,32,power_law_1.2,0.08294399976730346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,32,power_law_1.2,0.08401280045509338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,32,power_law_1.2,0.09420160055160523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,32,power_law_1.2,0.10468480587005616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,32,power_law_1.2,0.10713599920272827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,32,power_law_1.2,0.10529919862747192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,32,power_law_1.2,0.12567039728164672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,32,power_law_1.2,0.1309183955192566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,32,power_law_1.2,0.15761280059814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,32,power_law_1.2,0.18255360126495362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,32,power_law_1.2,0.235532808303833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,32,power_law_1.2,0.2988672018051147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,32,power_law_1.2,0.4655424118041992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,32,power_law_1.2,0.6078015804290772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,32,power_law_1.2,0.794649600982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,32,power_law_1.2,1.1172863960266113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,2,power_law_1.2,2.581119918823242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,4,balanced,0.07389866809050243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,4,balanced,0.09883200128873189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,4,balanced,0.1471946636835734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,4,balanced,0.2471733291943868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,4,balanced,0.4437066713968913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,4,balanced,0.8327626387278239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,4,balanced,0.8364106814066569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,4,balanced,0.8419466813405355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,4,balanced,0.8439626693725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,4,balanced,0.8470293680826823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,4,balanced,0.84933869043986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,4,balanced,0.8484960397084554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,4,balanced,0.851855993270874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,4,balanced,0.854576031366984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,4,balanced,0.8633386294047037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,4,balanced,0.86516801516215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,4,balanced,0.8767146269480387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,4,balanced,0.8874026934305826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,4,balanced,0.9155466556549072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,4,balanced,0.951695998509725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,4,balanced,0.9833226998647054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,4,balanced,1.0327733357747395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,4,balanced,1.1590080261230469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,4,balanced,1.2655413150787354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,4,balanced,1.4388160705566406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,128,balanced,0.05171733101209005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,128,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,128,balanced,0.05407999952634176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,4,balanced,1.9114665985107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,128,balanced,0.05580266813437144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,128,balanced,0.057904000083605446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,128,balanced,0.05774400134881338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,128,balanced,0.05825600028038025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,128,balanced,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,4,balanced,2.1739253997802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,128,balanced,0.05804799993832906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,128,balanced,0.05807466804981232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,128,balanced,0.05990933378537496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,128,balanced,0.062277331948280334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,128,balanced,0.06198933223883311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,128,balanced,0.06404800216356914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,128,balanced,0.0683786670366923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,128,balanced,0.06879466772079468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,128,balanced,0.0738506664832433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,128,balanced,0.08250133196512859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,128,balanced,0.0885813335577647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,128,balanced,0.10946133732795715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,128,balanced,0.12168533603350322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,128,balanced,0.16760534048080444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,128,balanced,0.15771200259526572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,128,balanced,0.20963199933369955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,128,balanced,0.2331999937693278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,128,balanced,0.3078826665878296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,128,balanced,0.37204798062642414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,8,balanced,0.051925331354141235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,8,balanced,0.054234668612480164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,8,balanced,0.06426133215427399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,8,balanced,0.08794666330019633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,8,balanced,0.12038399775822957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,8,balanced,0.13775466879208884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,8,balanced,0.1381333371003469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,8,balanced,0.1395840048789978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,8,balanced,0.1389173368612925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,8,balanced,0.14009066422780356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,8,balanced,0.14109333356221518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,8,balanced,0.14320533474286398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,8,balanced,0.14407466848691305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,8,balanced,0.1467519998550415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,8,balanced,0.15212800105412802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,8,balanced,0.1534933348496755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,8,balanced,0.15983999768892923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,8,balanced,0.17011733849843344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,8,balanced,0.17690134048461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,8,balanced,0.2076479991277059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,8,balanced,0.21266667048136392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,8,balanced,0.2752266724904378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,8,balanced,0.30932267506917316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,8,balanced,0.43034664789835614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,8,balanced,0.5063946644465128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,8,balanced,0.7043146292368571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,8,balanced,0.880021333694458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,32,8,power_law_1.2,0.01940480023622513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,32,8,power_law_1.2,0.021158400177955627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,32,8,power_law_1.2,0.02227199971675873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,32,8,power_law_1.2,0.02255360037088394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,32,8,power_law_1.2,0.024582399427890776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,32,8,power_law_1.2,0.02595840096473694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,32,8,power_law_1.2,0.027904000878334046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,32,8,power_law_1.2,0.02837119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,32,8,power_law_1.2,0.029363200068473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,32,8,power_law_1.2,0.030502399802207945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,32,8,power_law_1.2,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,2,power_law_1.01,0.09841279983520508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,32,8,power_law_1.2,0.03282560110092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,2,power_law_1.01,0.14754559993743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,2,power_law_1.01,0.19527039527893067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,2,power_law_1.01,0.2767359972000122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,2,power_law_1.01,0.3825472116470337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,2,power_law_1.01,0.4451263904571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,32,8,power_law_1.2,0.03407360017299652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,32,8,power_law_1.2,0.03560959994792938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,2,power_law_1.01,0.65414400100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,32,8,power_law_1.2,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,32,8,power_law_1.2,0.04154880046844482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,2,power_law_1.01,0.6527872085571289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,32,8,power_law_1.2,0.04456959962844849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,2,power_law_1.01,0.6790847778320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,2,power_law_1.01,0.6715392112731934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,2,power_law_1.01,0.7111231803894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,2,power_law_1.01,0.7356544017791748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,2,power_law_1.01,0.7127359867095947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,32,8,power_law_1.2,0.05286399722099304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,32,8,power_law_1.2,0.07767680287361145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,32,8,power_law_1.2,0.09145600199699402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,32,8,power_law_1.2,0.08929920196533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,32,8,power_law_1.2,0.1074112057685852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,32,8,power_law_1.2,0.1323904037475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,32,8,power_law_1.2,0.18378880023956298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,2,power_law_1.01,0.7810111999511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,32,8,power_law_1.2,0.22495999336242675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,2,power_law_1.01,0.7947968006134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,32,8,power_law_1.2,0.319648003578186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,2,power_law_1.01,0.7951871871948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,32,8,power_law_1.2,0.3861759901046753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,2,power_law_1.01,0.8211199760437011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,2,power_law_1.01,0.9003071784973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,2,power_law_1.01,0.9390912055969238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,2,power_law_1.01,1.0194111824035645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,2,power_law_1.01,1.0519488334655762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,2,power_law_1.01,1.18472318649292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,2,power_law_1.01,1.392627239227295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,8,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,8,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,8,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,8,balanced,0.041509332756201424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,2,power_law_1.01,1.6958784103393554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,8,balanced,0.049738665421803795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,8,balanced,0.04984533290068308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,8,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,8,balanced,0.05194133520126343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,8,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,8,balanced,0.051882664362589516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,8,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,8,balanced,0.05393599967161814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,8,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,8,balanced,0.05811200042565664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,8,balanced,0.062090665102005005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,8,balanced,0.06316799918810527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,8,balanced,0.0644053320089976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,8,balanced,0.07227733234564464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,8,balanced,0.08724799752235413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,8,balanced,0.09668266773223877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,8,balanced,0.10055999954541524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,8,balanced,0.12692800164222717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,8,balanced,0.1483786702156067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,8,balanced,0.18966400623321533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,8,balanced,0.23336533705393472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,8,balanced,0.3208213249842326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,8,balanced,0.4059786796569824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,2,power_law_1.01,2.0487295150756837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,2,power_law_1.01,2.6056896209716798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,2,power_law_1.01,3.2840831756591795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,8,power_law_1.01,0.02094080001115799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,8,power_law_1.01,0.025196799635887147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,8,power_law_1.01,0.02494720071554184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,8,power_law_1.01,0.02584959864616394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,8,power_law_1.01,0.0239424005150795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,8,power_law_1.01,0.026815998554229736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,8,power_law_1.01,0.028812798857688903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,8,power_law_1.01,0.029164800047874452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,8,power_law_1.01,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,8,power_law_1.01,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,8,power_law_1.01,0.032902398705482484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,8,power_law_1.01,0.03282560110092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,8,power_law_1.01,0.034860798716545106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,8,power_law_1.01,0.04390400052070618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,8,power_law_1.01,0.05119360089302063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,8,power_law_1.01,0.05631359815597534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,8,power_law_1.01,0.08554880023002624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,8,power_law_1.01,0.09118080139160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,8,power_law_1.01,0.09783040285110474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,8,power_law_1.01,0.10935039520263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,8,power_law_1.01,0.09530879855155945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,8,power_law_1.01,0.1223423957824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,8,power_law_1.01,0.14236799478530884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,8,power_law_1.01,0.19134080410003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,8,power_law_1.01,0.24039039611816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,8,power_law_1.01,0.3237695932388306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,8,power_law_1.01,0.4384960174560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,1,power_law_1.2,0.033657601475715636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,1,power_law_1.2,0.03824639916419983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,1,power_law_1.2,0.04922240078449249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,1,power_law_1.2,0.060710400342941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,1,power_law_1.2,0.0700543999671936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,1,power_law_1.2,0.10694400072097779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,1,power_law_1.2,0.11304960250854493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,1,power_law_1.2,0.11437439918518066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,1,power_law_1.2,0.11713279485702514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,1,power_law_1.2,0.12161920070648194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,1,power_law_1.2,0.12531839609146117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,1,power_law_1.2,0.12672640085220338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,1,power_law_1.2,0.12982399463653566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,1,power_law_1.2,0.13279999494552613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,1,power_law_1.2,0.1427199959754944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,1,power_law_1.2,0.14471039772033692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,1,power_law_1.2,0.15852799415588378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,1,power_law_1.2,0.1845952033996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,1,power_law_1.2,0.1773375988006592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,1,power_law_1.2,0.21143040657043458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,1,power_law_1.2,0.25224320888519286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,1,power_law_1.2,0.31052799224853517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,1,power_law_1.2,0.385152006149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,1,power_law_1.2,0.49874558448791506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,1,power_law_1.2,0.560262393951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,1,power_law_1.2,0.783135986328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,1,power_law_1.2,1.0141887664794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,64,balanced,0.03158933420976003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,64,balanced,0.03092266619205475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,64,balanced,0.031514666974544525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,64,balanced,0.03166399896144867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,64,balanced,0.03190933416287104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,64,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,64,balanced,0.03431999931732813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,64,balanced,0.03495466709136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,64,balanced,0.03526933242877325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,64,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,64,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,64,balanced,0.03732266773780187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,64,balanced,0.03632533301909765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,64,balanced,0.03729599962631861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,64,balanced,0.03941333293914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,64,balanced,0.04196266829967499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,64,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,64,balanced,0.0424586683511734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,64,balanced,0.04554666578769684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,64,balanced,0.04964800179004669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,64,balanced,0.05197866757710775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,64,balanced,0.06434133152167003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,64,balanced,0.08611733714739482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,64,balanced,0.1165706713994344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,64,balanced,0.141375998655955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,64,balanced,0.18951465686162314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,64,balanced,0.2420426607131958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,128,power_law_1.01,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,128,power_law_1.01,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,128,power_law_1.01,0.0322816014289856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,128,power_law_1.01,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,128,power_law_1.01,0.035667198896408084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,128,power_law_1.01,0.03170560002326965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,128,power_law_1.01,0.031667199730873105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,128,power_law_1.01,0.03113600015640259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,128,power_law_1.01,0.03159680068492889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,128,power_law_1.01,0.031673601269721983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,128,power_law_1.01,0.03283199965953827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,128,power_law_1.01,0.03361279964447021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,128,power_law_1.01,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,128,power_law_1.01,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,128,power_law_1.01,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,128,power_law_1.01,0.03891839981079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,128,power_law_1.01,0.04046080112457275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,128,power_law_1.01,0.04542079865932465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,128,power_law_1.01,0.05065600275993347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,128,power_law_1.01,0.05911039710044861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,128,power_law_1.01,0.0648639976978302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,128,power_law_1.01,0.0832256019115448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,128,power_law_1.01,0.09708160161972046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,128,power_law_1.01,0.12674560546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,128,power_law_1.01,0.15467519760131837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,128,power_law_1.01,0.18978559970855713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,128,power_law_1.01,0.2719680070877075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,4,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,4,balanced,0.04329599936803182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,4,balanced,0.045738667249679565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,4,balanced,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,4,balanced,0.05199466645717621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,4,balanced,0.07428266604741414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,4,balanced,0.07482133309046428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,4,balanced,0.07295999924341838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,4,balanced,0.07441066702206929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,4,balanced,0.07468266785144806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,4,balanced,0.07449066638946533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,4,balanced,0.07938666641712189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,4,balanced,0.07810133198897044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,4,balanced,0.080485333998998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,4,balanced,0.0876693328221639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,4,balanced,0.08757332960764568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,4,balanced,0.09690133730570476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,4,balanced,0.10900800426801045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,4,balanced,0.11388799548149109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,4,balanced,0.1379146675268809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,4,balanced,0.15388799707094827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,4,balanced,0.2016213337580363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,4,balanced,0.2411253253618876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,4,balanced,0.33105067412058514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,4,balanced,0.4217706521352132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,4,balanced,0.6079839865366617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,4,balanced,0.7848693529764811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,4,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,4,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,4,balanced,0.03349866718053818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,4,balanced,0.05002133548259735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,4,balanced,0.07160000006357829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,4,balanced,0.07912533481915791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,4,balanced,0.07831466694672902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,4,balanced,0.07902400195598602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,4,balanced,0.07937600215276082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,4,balanced,0.07824533184369405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,4,balanced,0.08042133351167043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,4,balanced,0.08078399797280629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,4,balanced,0.08062933385372162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,4,balanced,0.08099199831485748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,4,balanced,0.08684800068537395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,4,balanced,0.08666132887204488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,4,balanced,0.08991466959317525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,4,balanced,0.10745599865913391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,4,balanced,0.09642133116722107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,4,balanced,0.13185600439707437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,4,balanced,0.1109386682510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,4,balanced,0.14331733187039694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,4,balanced,0.1485919952392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,4,balanced,0.2050666610399882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,4,balanced,0.23031999667485556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,4,balanced,0.32819199562072754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,4,balanced,0.35124798615773517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,1,power_law_1.2,0.059545600414276124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,1,power_law_1.2,0.0786624014377594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,1,power_law_1.2,0.09710080027580262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,1,power_law_1.2,0.13301759958267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,1,power_law_1.2,0.17772799730300903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,1,power_law_1.2,0.21865599155426024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,1,power_law_1.2,0.30649600028991697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,1,power_law_1.2,0.32785921096801757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,1,power_law_1.2,0.34314239025115967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,1,power_law_1.2,0.34206719398498536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,1,power_law_1.2,0.3582144021987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,1,power_law_1.2,0.3824575901031494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,1,power_law_1.2,0.388102388381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,1,power_law_1.2,0.3936255931854248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,1,power_law_1.2,0.4199359893798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,1,power_law_1.2,0.436902379989624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,1,power_law_1.2,0.4689151763916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,1,power_law_1.2,0.5353727817535401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,1,power_law_1.2,0.57292799949646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,1,power_law_1.2,0.6827263832092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,1,power_law_1.2,0.7705152034759521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,1,power_law_1.2,0.9600383758544921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,1,power_law_1.2,1.1323455810546874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,1,power_law_1.2,1.4916416168212892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,1,power_law_1.2,1.8567615509033204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,1,power_law_1.2,2.5541248321533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,1,power_law_1.2,3.2716480255126954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,32,balanced,0.037733333806196846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,32,balanced,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,32,balanced,0.03164266546567281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,32,balanced,0.03436266630887985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,32,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,32,balanced,0.03532266616821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,32,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,32,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,32,balanced,0.03580799947182337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,32,balanced,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,32,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,32,balanced,0.03651199986537298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,32,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,32,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,32,balanced,0.03982933362325033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,32,balanced,0.04365866879622141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,32,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,32,balanced,0.04671466847260793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,32,balanced,0.04489600161711375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,32,balanced,0.0539680023988088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,32,balanced,0.052101333936055504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,32,balanced,0.07064533233642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,32,balanced,0.08089600006739299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,32,balanced,0.10781332850456238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,32,balanced,0.12165333827336629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,32,balanced,0.16432000199953714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,32,balanced,0.19100266695022583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,4,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,4,balanced,0.06201066573460897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,4,balanced,0.07037866612275441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,4,balanced,0.09402133027712505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,4,balanced,0.12849600116411844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,4,balanced,0.1909866730372111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,4,balanced,0.2555359999338786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,4,balanced,0.2555946707725525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,4,balanced,0.2582933306694031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,4,balanced,0.2593013246854146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,4,balanced,0.26136000951131183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,4,balanced,0.26498132944107056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,4,balanced,0.27100799481074017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,4,balanced,0.27185599009195965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,4,balanced,0.2797813415527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,4,balanced,0.2834666570027669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,4,balanced,0.29478933413823444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,4,balanced,0.3094080090522766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,4,balanced,0.3289066751797994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,4,balanced,0.36790398756663006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,4,balanced,0.39798398812611896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,4,balanced,0.477509339650472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,4,balanced,0.5353226661682129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,4,balanced,0.6584746837615967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,4,balanced,0.8582986990610758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,4,balanced,1.1175039609273274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,4,balanced,1.4792052904764812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,8,power_law_1.01,0.04742400050163269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,8,power_law_1.01,0.06452479958534241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,8,power_law_1.01,0.06980479955673217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,8,power_law_1.01,0.08164479732513427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,8,power_law_1.01,0.09191039800643921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,8,power_law_1.01,0.08650879859924317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,8,power_law_1.01,0.09402880072593689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,8,power_law_1.01,0.09232640266418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,8,power_law_1.01,0.09210879802703857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,8,power_law_1.01,0.09592959880828858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,8,power_law_1.01,0.09511039853096008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,8,power_law_1.01,0.0992959976196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,8,power_law_1.01,0.10213760137557984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,8,power_law_1.01,0.10588159561157226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,8,power_law_1.01,0.11619839668273926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,8,power_law_1.01,0.11809279918670654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,8,power_law_1.01,0.12346880435943604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,8,power_law_1.01,0.14072320461273194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,8,power_law_1.01,0.1545024037361145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,8,power_law_1.01,0.18176640272140504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,8,power_law_1.01,0.2061824083328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,8,power_law_1.01,0.26924159526824953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,8,power_law_1.01,0.32586240768432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,8,power_law_1.01,0.45464320182800294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,8,power_law_1.01,0.5562880039215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,8,power_law_1.01,0.793446397781372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,8,power_law_1.01,1.0119808197021485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,16,2,power_law_1.2,0.0844543993473053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,16,2,power_law_1.2,0.0930176019668579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,16,2,power_law_1.2,0.11153279542922974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,16,2,power_law_1.2,0.12439039945602418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,16,2,power_law_1.2,0.13428479433059692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,16,2,power_law_1.2,0.15193599462509155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,16,2,power_law_1.2,0.17136000394821166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,16,2,power_law_1.2,0.17645440101623536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,16,2,power_law_1.2,0.17838079929351808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,16,2,power_law_1.2,0.17916799783706666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,16,2,power_law_1.2,0.18154879808425903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,16,2,power_law_1.2,0.1945088028907776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,16,2,power_law_1.2,0.19427839517593384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,16,2,power_law_1.2,0.2020927906036377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,16,2,power_law_1.2,0.20208001136779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,16,2,power_law_1.2,0.21215999126434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,16,2,power_law_1.2,0.22094080448150635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,16,2,power_law_1.2,0.2551487922668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,16,2,power_law_1.2,0.28551039695739744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,16,2,power_law_1.2,0.359987211227417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,16,2,power_law_1.2,0.4365248203277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,16,2,power_law_1.2,0.5419072151184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,16,2,power_law_1.2,0.6954944133758545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,16,2,power_law_1.2,1.0127424240112304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,16,2,power_law_1.2,1.252672004699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,16,2,power_law_1.2,1.921625518798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,16,2,power_law_1.2,2.509619140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,8,power_law_1.2,0.05790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,8,power_law_1.2,0.05870720148086548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,8,power_law_1.2,0.06322559714317322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,8,power_law_1.2,0.06607999801635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,8,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,8,power_law_1.2,0.07637760043144226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,8,power_law_1.2,0.07591040134429931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,8,power_law_1.2,0.08079360127449035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,8,power_law_1.2,0.07964800000190735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,8,power_law_1.2,0.0815999984741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,8,power_law_1.2,0.08275200128555298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,8,power_law_1.2,0.08513280153274536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,8,power_law_1.2,0.08533120155334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,8,power_law_1.2,0.09045760035514831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,8,power_law_1.2,0.09781119823455811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,8,power_law_1.2,0.10204160213470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,8,power_law_1.2,0.10720640420913696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,8,power_law_1.2,0.12355200052261353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.2,0.13690240383148194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.2,0.1677567958831787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.2,0.18301440477371217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.2,0.24916479587554932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.2,0.32994558811187746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.2,0.4171008110046387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.2,0.6285632133483887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.2,0.7769919872283936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.2,1.1192768096923829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,4,balanced,0.04775466521581014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,4,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,4,balanced,0.06995200117429097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,4,balanced,0.09624000390370686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,4,balanced,0.1482080022493998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,4,balanced,0.2458826700846354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,4,balanced,0.24634132782618204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,4,balanced,0.24832000335057577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,4,balanced,0.2497119903564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,4,balanced,0.2494879961013794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,4,balanced,0.2529333432515462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,4,balanced,0.2564479907353719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,4,balanced,0.26105600595474243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,4,balanced,0.2589866717656453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,4,balanced,0.2675519982973735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,4,balanced,0.27002133925755817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,4,balanced,0.27535466353098553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,4,balanced,0.29102933406829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,4,balanced,0.3030986587206523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,4,balanced,0.3280319968859355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,4,balanced,0.36618133385976154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,4,balanced,0.41285868485768634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,4,balanced,0.4617226521174113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,4,balanced,0.5930933157602946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,4,balanced,0.6871999899546305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,4,balanced,0.9519253571828207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,4,balanced,1.1654773553212483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,16,power_law_1.01,0.021932800114154816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,16,power_law_1.01,0.022176000475883483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,16,power_law_1.01,0.02248319983482361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,16,power_law_1.01,0.02245119959115982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,16,power_law_1.01,0.02263039946556091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,16,power_law_1.01,0.024332800507545473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,16,power_law_1.01,0.025536000728607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,16,power_law_1.01,0.025248000025749208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,16,power_law_1.01,0.02565760016441345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,16,power_law_1.01,0.0263808012008667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,16,power_law_1.01,0.02725760042667389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,16,power_law_1.01,0.02885119915008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,16,power_law_1.01,0.030355200171470642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,16,power_law_1.01,0.03379839956760407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,16,power_law_1.01,0.03521920144557953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,16,power_law_1.01,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,16,power_law_1.01,0.052723199129104614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,16,power_law_1.01,0.058355200290679934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.01,0.06327040195465088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.01,0.07439360022544861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.01,0.08180480003356934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.01,0.10319360494613647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.01,0.12215679883956909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.01,0.1571903944015503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.01,0.22011520862579345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.01,0.2907327890396118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.01,0.38191359043121337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,2,balanced,0.05637866755326589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,2,balanced,0.07926400005817413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,2,balanced,0.12148800492286682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,2,balanced,0.1255466639995575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,2,balanced,0.12574399511019388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,2,balanced,0.12718400359153748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,2,balanced,0.12621866663297018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,2,balanced,0.12756266196568808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,2,balanced,0.12731200456619263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,2,balanced,0.12754133343696594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,2,balanced,0.13398399949073792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,2,balanced,0.1309386690457662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,2,balanced,0.13398399949073792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,2,balanced,0.1425333321094513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,2,balanced,0.1555519998073578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,2,balanced,0.15866133570671082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,2,balanced,0.1630400021870931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,2,balanced,0.19681066274642944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,2,balanced,0.2012373407681783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,2,balanced,0.2569653391838074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,2,balanced,0.29214932521184284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,2,balanced,0.40116266409556073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,2,balanced,0.5014079809188843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,2,balanced,0.7137227058410645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,2,balanced,0.9431306521097819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,2,balanced,1.3603253364562988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,2,balanced,1.758090655008952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,16,power_law_1.2,0.04910080134868622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,16,power_law_1.2,0.07907840013504028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,16,power_law_1.2,0.0700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,16,power_law_1.2,0.07872639894485474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,16,power_law_1.2,0.0776639997959137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,16,power_law_1.2,0.06444159746170045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,16,power_law_1.2,0.08279039859771728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,16,power_law_1.2,0.07470719814300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,16,power_law_1.2,0.07612159848213196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,16,power_law_1.2,0.08220800161361694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,16,power_law_1.2,0.08366079926490784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,16,power_law_1.2,0.08176000118255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,16,power_law_1.2,0.08723199963569642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,16,power_law_1.2,0.09387519955635071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,16,power_law_1.2,0.10209920406341552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,16,power_law_1.2,0.1156991958618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,16,power_law_1.2,0.19293440580368043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,16,power_law_1.2,0.2237312078475952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.2,0.16346240043640137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.2,0.17181440591812133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.2,0.19507839679718017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.2,0.24332160949707032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.2,0.30962560176849363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.2,0.42108798027038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.2,0.566809606552124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.2,0.788102388381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.2,1.021504020690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,4,power_law_1.2,0.05364480018615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,4,power_law_1.2,0.07507839798927307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,4,power_law_1.2,0.07913600206375122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,4,power_law_1.2,0.09259520173072815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,4,power_law_1.2,0.11250560283660889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,4,power_law_1.2,0.12056959867477417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,4,power_law_1.2,0.13793920278549193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,4,power_law_1.2,0.1471295952796936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,4,power_law_1.2,0.14482560157775878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,4,power_law_1.2,0.13919999599456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,4,power_law_1.2,0.14731520414352417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,4,power_law_1.2,0.1581760048866272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,4,power_law_1.2,0.16613119840621948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,4,power_law_1.2,0.16747519969940186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,4,power_law_1.2,0.1796223998069763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,4,power_law_1.2,0.17846399545669556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,4,power_law_1.2,0.1946239948272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,4,power_law_1.2,0.2268224000930786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,4,power_law_1.2,0.2406464099884033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,4,power_law_1.2,0.29668478965759276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,4,power_law_1.2,0.332204794883728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,4,power_law_1.2,0.4481664180755615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,4,power_law_1.2,0.5279808044433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,4,power_law_1.2,0.7079552173614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,4,power_law_1.2,0.9351360321044921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,4,power_law_1.2,1.360211181640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,4,power_law_1.2,1.6585344314575194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,32,4,power_law_1.2,0.02025599926710129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,32,4,power_law_1.2,0.023680000007152556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,32,4,power_law_1.2,0.02346239984035492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,32,4,power_law_1.2,0.0281792014837265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,32,4,power_law_1.2,0.02930560111999512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,32,4,power_law_1.2,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,32,4,power_law_1.2,0.03500159978866577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,32,4,power_law_1.2,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,32,4,power_law_1.2,0.037555199861526486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,32,4,power_law_1.2,0.039443200826644896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,32,4,power_law_1.2,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,32,4,power_law_1.2,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,32,4,power_law_1.2,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,32,4,power_law_1.2,0.045151999592781066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,32,4,power_law_1.2,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,32,4,power_law_1.2,0.05144960284233093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,32,4,power_law_1.2,0.057120001316070555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,32,4,power_law_1.2,0.07180160284042358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,32,4,power_law_1.2,0.08684800267219543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,32,4,power_law_1.2,0.11082240343093872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,32,4,power_law_1.2,0.13763200044631957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,32,4,power_law_1.2,0.1589568018913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,32,4,power_law_1.2,0.18255360126495362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,32,4,power_law_1.2,0.24069759845733643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,32,4,power_law_1.2,0.2971456050872803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,32,4,power_law_1.2,0.41201281547546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,32,4,power_law_1.2,0.5096831798553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,16,power_law_1.2,0.06101120114326477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,16,power_law_1.2,0.06851199865341187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,16,power_law_1.2,0.07550719976425171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,16,power_law_1.2,0.06766719818115234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,16,power_law_1.2,0.06730239987373351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,16,power_law_1.2,0.06510720252990723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,16,power_law_1.2,0.06727679967880248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,16,power_law_1.2,0.06925439834594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,16,power_law_1.2,0.0688256025314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,16,power_law_1.2,0.06959999799728393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,16,power_law_1.2,0.06897280216217042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,16,power_law_1.2,0.07262719869613647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,16,power_law_1.2,0.07284479737281799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,16,power_law_1.2,0.07430400252342224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,16,power_law_1.2,0.08109440207481385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,16,power_law_1.2,0.08339840173721313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,16,power_law_1.2,0.09291520118713378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,16,power_law_1.2,0.10899840593338013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,16,power_law_1.2,0.12351360321044921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,16,power_law_1.2,0.14585599899291993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,16,power_law_1.2,0.17906559705734254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,16,power_law_1.2,0.22410240173339843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,16,power_law_1.2,0.28023040294647217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,16,power_law_1.2,0.3703039884567261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,16,power_law_1.2,0.4770559787750244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,16,power_law_1.2,0.6676479816436768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,16,power_law_1.2,0.91146879196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,1,balanced,0.019039999693632126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,1,balanced,0.020549333343903225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,1,balanced,0.020410666863123577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,1,balanced,0.023120000958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,1,balanced,0.030981334547201794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,1,balanced,0.05072533090909322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,1,balanced,0.0517493337392807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,1,balanced,0.0537120004494985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,1,balanced,0.055455997586250305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,1,balanced,0.05610666672388712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,1,balanced,0.057631999254226685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,1,balanced,0.05890666445096334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,1,balanced,0.06018666426340739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,1,balanced,0.06192533175150553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,1,balanced,0.0644053320089976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,1,balanced,0.06504000226656596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,1,balanced,0.06804266571998596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,1,balanced,0.07673066854476929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,1,balanced,0.08515733480453491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,1,balanced,0.11415466666221619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,1,balanced,0.12800533572832742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,1,balanced,0.1717066764831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,1,balanced,0.17695466677347818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,1,balanced,0.24420267343521118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,1,balanced,0.3225546677907308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,1,balanced,0.4510186513264974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,1,balanced,0.5909013350804647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,32,power_law_1.2,0.052211201190948485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,32,power_law_1.2,0.047712001204490664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,32,power_law_1.2,0.04758400022983551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,32,power_law_1.2,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,32,power_law_1.2,0.04837760031223297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,32,power_law_1.2,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,32,power_law_1.2,0.050521600246429446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,32,power_law_1.2,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,32,power_law_1.2,0.05162879824638367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,32,power_law_1.2,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,32,power_law_1.2,0.05470719933509827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,32,power_law_1.2,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,32,power_law_1.2,0.056992000341415404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,32,power_law_1.2,0.06020479798316956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,32,power_law_1.2,0.06672000288963317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,32,power_law_1.2,0.07018880248069763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,32,power_law_1.2,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,32,power_law_1.2,0.09057279825210571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,32,power_law_1.2,0.09987199902534485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,32,power_law_1.2,0.12748160362243652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,32,power_law_1.2,0.15160319805145264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,32,power_law_1.2,0.2087104082107544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,32,power_law_1.2,0.252128005027771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,32,power_law_1.2,0.34101119041442873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,32,power_law_1.2,0.4439551830291748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,32,power_law_1.2,0.6645567893981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,32,power_law_1.2,0.8446528434753418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,16,power_law_1.2,0.07500159740447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,16,power_law_1.2,0.07304319739341736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,16,power_law_1.2,0.07577599883079529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,16,power_law_1.2,0.07297919988632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,16,power_law_1.2,0.06926720142364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,16,power_law_1.2,0.07441279888153077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,16,power_law_1.2,0.07735679745674133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,16,power_law_1.2,0.07374079823493958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,16,power_law_1.2,0.08149759769439698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,16,power_law_1.2,0.07908480167388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,16,power_law_1.2,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,16,power_law_1.2,0.08370559811592101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,16,power_law_1.2,0.08704000115394592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,16,power_law_1.2,0.08792319893836975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,16,power_law_1.2,0.10302720069885254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,16,power_law_1.2,0.10547840595245361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,16,power_law_1.2,0.10115840435028076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,16,power_law_1.2,0.11553280353546143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,16,power_law_1.2,0.12872320413589478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,16,power_law_1.2,0.15137920379638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,16,power_law_1.2,0.18559999465942384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,16,power_law_1.2,0.2004863977432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,16,power_law_1.2,0.22439041137695312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,16,power_law_1.2,0.30583040714263915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,16,power_law_1.2,0.4202752113342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,16,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,16,balanced,0.04985600213209788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,16,balanced,0.050437331199645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,16,balanced,0.05171200136343638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,16,balanced,0.06737066805362701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,16,balanced,0.07523733377456665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,16,balanced,0.07638933261235555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,16,balanced,0.07761600116888683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,16,balanced,0.07788800199826558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,16,balanced,0.0795360008875529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,16,balanced,0.07854933540026347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,16,balanced,0.07929599781831105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,16,balanced,0.08060266574223836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,16,balanced,0.08272000153859456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,16,balanced,0.08692266543706258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,16,balanced,0.08859733740488689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,16,balanced,0.0909440020720164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,16,balanced,0.09915733337402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,16,balanced,0.10310399532318115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,16,balanced,0.12378666798273723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,16,balanced,0.13581333557764688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,16,balanced,0.16844266653060913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,16,balanced,0.1792853275934855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,16,balanced,0.25616000096003216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,16,balanced,0.2940853238105774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,16,balanced,0.39973334471384686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,16,balanced,0.5009066661198934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,16,power_law_1.2,0.5670783996582032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,16,power_law_1.2,0.7653439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,4,power_law_1.01,0.061900800466537474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,4,power_law_1.01,0.07859200239181519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,4,power_law_1.01,0.09162240028381348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,4,power_law_1.01,0.10544639825820923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,4,power_law_1.01,0.12509440183639525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,4,power_law_1.01,0.13425920009613038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,4,power_law_1.01,0.14264960289001466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,4,power_law_1.01,0.15266560316085814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,4,power_law_1.01,0.1526080012321472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,4,power_law_1.01,0.15088000297546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,4,power_law_1.01,0.16720000505447388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,4,power_law_1.01,0.1641983985900879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,4,power_law_1.01,0.16314879655838013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,4,power_law_1.01,0.17150720357894897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,4,power_law_1.01,0.17300479412078856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,4,power_law_1.01,0.17713279724121095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,4,power_law_1.01,0.18518400192260742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,4,power_law_1.01,0.20920960903167723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,4,power_law_1.01,0.2086400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,4,power_law_1.01,0.23785600662231446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,4,power_law_1.01,0.27308800220489504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,4,power_law_1.01,0.33464319705963136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,4,power_law_1.01,0.3618240118026733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,4,power_law_1.01,0.45027837753295896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,4,power_law_1.01,0.5596672058105469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,4,power_law_1.01,0.748038387298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,4,power_law_1.01,0.9276032447814941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,16,balanced,0.04794133206208547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,16,balanced,0.04996266464392344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,16,balanced,0.07213866710662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,16,balanced,0.09500799576441447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,16,balanced,0.14621333281199136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,16,balanced,0.2459519902865092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,16,balanced,0.24688533941904703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,16,balanced,0.24515734116236368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,16,balanced,0.2462773323059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,16,balanced,0.2482773264249166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,16,balanced,0.2505653301874797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,16,balanced,0.25041067600250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,16,balanced,0.25314666827519733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,16,balanced,0.2512906591097514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,16,balanced,0.2585279941558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,16,balanced,0.2592586676279704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,16,balanced,0.2632960081100464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,16,balanced,0.2691253423690796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,16,balanced,0.2791999975840251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,16,balanced,0.2902346650759379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,16,balanced,0.3155679901440938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,16,balanced,0.3421333233515422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,16,balanced,0.36507201194763184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,16,balanced,0.4513546625773112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,16,balanced,0.48873066902160645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,16,balanced,0.6641600131988525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,16,balanced,0.7527413368225098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,32,power_law_1.01,0.060550397634506224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,32,power_law_1.01,0.06238080263137817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,32,power_law_1.01,0.058278399705886844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,32,power_law_1.01,0.06062080264091492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,32,power_law_1.01,0.06154239773750305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,32,power_law_1.01,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,32,power_law_1.01,0.06262400150299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,32,power_law_1.01,0.06190720200538635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,32,power_law_1.01,0.06362879872322083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,32,power_law_1.01,0.06307200193405152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,32,power_law_1.01,0.06362239718437195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,32,power_law_1.01,0.06677759885787964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,32,power_law_1.01,0.06826879978179931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,32,power_law_1.01,0.07102720141410827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,32,power_law_1.01,0.07780479788780212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,32,power_law_1.01,0.07938560247421264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,32,power_law_1.01,0.08644480109214783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,32,power_law_1.01,0.10623999834060668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,32,power_law_1.01,0.11144959926605225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,32,power_law_1.01,0.13354239463806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,32,power_law_1.01,0.16805119514465333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,32,power_law_1.01,0.21233279705047609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,32,power_law_1.01,0.21584639549255372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,32,power_law_1.01,0.28097920417785643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,32,power_law_1.01,0.33269760608673093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,32,power_law_1.01,0.4967807769775391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,32,power_law_1.01,0.5791872024536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,1,power_law_1.2,0.03223679959774017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,1,power_law_1.2,0.03253119885921478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,1,power_law_1.2,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,1,power_law_1.2,0.0365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,1,power_law_1.2,0.03696640133857727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,1,power_law_1.2,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,1,power_law_1.2,0.042668798565864564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,1,power_law_1.2,0.04273279905319214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,1,power_law_1.2,0.045414400100708005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,1,power_law_1.2,0.04682239890098572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,1,power_law_1.2,0.045075199007987975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,1,power_law_1.2,0.04816640019416809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,1,power_law_1.2,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,1,power_law_1.2,0.0562175989151001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,1,power_law_1.2,0.0643775999546051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,1,power_law_1.2,0.06814720034599304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,1,power_law_1.2,0.07491199970245362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,1,power_law_1.2,0.09089279770851136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,1,power_law_1.2,0.1044927954673767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,1,power_law_1.2,0.13146239519119263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,1,power_law_1.2,0.15436160564422607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,1,power_law_1.2,0.21442561149597167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,1,power_law_1.2,0.26064000129699705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,1,power_law_1.2,0.3609920024871826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,1,power_law_1.2,0.4678976058959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,2,balanced,0.025445332129796345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,2,balanced,0.02719466636578242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,2,balanced,0.02569066733121872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,2,balanced,0.03358400116364161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,2,balanced,0.06471999982992808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,2,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,2,balanced,0.0547626664241155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,2,balanced,0.05574400226275126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,2,balanced,0.05609600245952606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,2,balanced,0.05495466788609823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,2,balanced,0.05602133274078369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,2,balanced,0.05783466498057047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,2,balanced,0.05782400071620941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,2,balanced,0.06700266897678375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,2,balanced,0.07249600191911061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,2,balanced,0.07039466500282288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,2,balanced,0.07166933516661327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,2,balanced,0.08582400282224019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,2,balanced,0.09085333347320557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,2,balanced,0.10593600074450175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,2,balanced,0.1148533324400584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,2,balanced,0.1511840025583903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,2,balanced,0.17302932341893515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,2,balanced,0.2566933234532674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,2,balanced,0.3301279942194621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,2,balanced,0.4645013411839803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,2,balanced,0.595413327217102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,1,power_law_1.2,0.6739776134490967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,32,power_law_1.01,0.07622399926185608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,32,power_law_1.01,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,32,power_law_1.01,0.06743040084838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,32,power_law_1.01,0.06695039868354798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,32,power_law_1.01,0.06696320176124573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,32,power_law_1.01,0.06450560092926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,32,power_law_1.01,0.06676480174064636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,32,power_law_1.01,0.06451200246810913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,32,power_law_1.01,0.06773120164871216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,32,power_law_1.01,0.06787840127944947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,32,power_law_1.01,0.06968320012092591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,32,power_law_1.01,0.07107840180397033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,32,power_law_1.01,0.0717311978340149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,32,power_law_1.01,0.07665280103683472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,32,power_law_1.01,0.08153600096702576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,32,power_law_1.01,0.08320639729499817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,32,power_law_1.01,0.0870527982711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,32,power_law_1.01,0.09649919867515563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,32,power_law_1.01,0.09934080243110657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,32,power_law_1.01,0.12205439805984497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,1,power_law_1.2,0.8836671829223632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,32,power_law_1.01,0.13134080171585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,32,power_law_1.01,0.16513279676437378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,32,power_law_1.01,0.19468159675598146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,32,power_law_1.01,0.2646143913269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,32,power_law_1.01,0.31390719413757323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,32,power_law_1.01,0.4513855934143066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,32,power_law_1.01,0.5393343925476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,256,balanced,0.060266668597857155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,256,balanced,0.06366933385531108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,256,balanced,0.06001066664854685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,256,balanced,0.06645333270231883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,256,balanced,0.06192533175150553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,256,balanced,0.06820799907048543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,256,balanced,0.06663466493288676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,256,balanced,0.06838933130105336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,256,balanced,0.06811200082302094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,256,balanced,0.06835733354091644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,256,balanced,0.06846400101979573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,256,balanced,0.06830400228500366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,256,balanced,0.07146666447321574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,256,balanced,0.06950933237870534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,256,balanced,0.07447466750939687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,256,balanced,0.0766293356815974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,256,balanced,0.07800533374150594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,256,balanced,0.08472533027331035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,256,balanced,0.08986133337020874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,256,balanced,0.10160000125567119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,256,balanced,0.11109866698582967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,256,balanced,0.14206399520238241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,256,balanced,0.16564800341924033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,256,balanced,0.21782400210698447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,256,balanced,0.2666613260904948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,256,balanced,0.3668533166249593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,256,balanced,0.46191465854644775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,1,balanced,0.05039466420809428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,1,balanced,0.055770665407180786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,1,balanced,0.070783997575442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,1,balanced,0.09433066844940186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,1,balanced,0.1388213336467743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,1,balanced,0.15890666842460632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,1,balanced,0.1604586640993754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,1,balanced,0.16337066888809204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,1,balanced,0.16513599952061972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,1,balanced,0.16705065965652466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,1,balanced,0.17165867487589517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,1,balanced,0.17603200674057007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,1,balanced,0.18100800116856894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,1,balanced,0.19371734062830606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,1,balanced,0.2055786649386088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,1,balanced,0.2118720014890035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,1,balanced,0.22489066918691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,1,balanced,0.26603732506434125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,1,balanced,0.29364800453186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,1,balanced,0.418559988339742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,1,balanced,0.4458560148874919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,1,balanced,0.6680959860483805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,1,balanced,0.7721439997355143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,1,balanced,1.1913706461588542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,1,balanced,1.4369440078735352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,1,balanced,2.2466932932535806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,1,balanced,2.747056007385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,8,power_law_1.01,0.05340800285339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,8,power_law_1.01,0.07099519968032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,8,power_law_1.01,0.07425280213356018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,8,power_law_1.01,0.08024960160255432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,8,power_law_1.01,0.08870400190353393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,8,power_law_1.01,0.08598399758338929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,8,power_law_1.01,0.09103360176086425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,8,power_law_1.01,0.08972160220146179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,8,power_law_1.01,0.09303039908409119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,8,power_law_1.01,0.09085440039634704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,8,power_law_1.01,0.09312000274658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,8,power_law_1.01,0.09824000000953674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,8,power_law_1.01,0.09694719910621644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,8,power_law_1.01,0.10083199739456176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,8,power_law_1.01,0.10775680541992187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,8,power_law_1.01,0.1082751989364624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,8,power_law_1.01,0.1150272011756897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,8,power_law_1.01,0.12911360263824462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,8,power_law_1.01,0.13637759685516357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,8,power_law_1.01,0.15697920322418213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,8,power_law_1.01,0.18496639728546144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,8,power_law_1.01,0.22050559520721436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,8,power_law_1.01,0.27402238845825194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,8,power_law_1.01,0.351091194152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,8,power_law_1.01,0.45500798225402833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,8,power_law_1.01,0.6181568145751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,8,power_law_1.01,0.75065598487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,2,power_law_1.2,0.04734080135822296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,2,power_law_1.2,0.0516543984413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,2,power_law_1.2,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,2,power_law_1.2,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,2,power_law_1.2,0.072953599691391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,2,power_law_1.2,0.09578239917755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,2,power_law_1.2,0.09797120094299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,2,power_law_1.2,0.1024832010269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,2,power_law_1.2,0.10335999727249146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,2,power_law_1.2,0.10493439435958862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,2,power_law_1.2,0.10799360275268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,2,power_law_1.2,0.11201920509338378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,2,power_law_1.2,0.11249279975891113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,2,power_law_1.2,0.1172287940979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,2,power_law_1.2,0.13000320196151732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,2,power_law_1.2,0.13332480192184448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,2,power_law_1.2,0.14311039447784424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,2,power_law_1.2,0.1644863963127136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,2,power_law_1.2,0.1926911950111389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,2,power_law_1.2,0.250598406791687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,2,power_law_1.2,0.2934335947036743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,2,power_law_1.2,0.3905663967132568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,2,power_law_1.2,0.4920447826385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,2,power_law_1.2,0.6643968105316163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,2,power_law_1.2,0.9092672348022461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,2,power_law_1.2,1.2255040168762208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,4,balanced,0.03369066615899404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,4,balanced,0.035029334326585136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,4,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,4,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,4,balanced,0.03638399889071783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,4,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,4,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,4,balanced,0.03738666574160258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,4,balanced,0.036933332681655884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,4,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,4,balanced,0.03746666759252548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,4,balanced,0.037802666425704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,4,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,4,balanced,0.04050666590531667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,4,balanced,0.04554133117198944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,4,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,4,balanced,0.04558399816354116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,4,balanced,0.052416001756985985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,4,balanced,0.05308799942334493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,4,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,4,balanced,0.07112533350785573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,4,balanced,0.09268266956011455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,4,balanced,0.11105599999427795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,4,balanced,0.14494400223096213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,4,balanced,0.17430933316548666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,4,balanced,0.2313279906908671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,4,balanced,0.2849973241488139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,2,power_law_1.2,1.5308480262756348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,1,balanced,0.08932800094286601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,1,balanced,0.09478933612505595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,1,balanced,0.10729066530863444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,1,balanced,0.13547733426094055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,1,balanced,0.1822026570638021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,1,balanced,0.268613338470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,1,balanced,0.2664373318354289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,1,balanced,0.266159991423289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,1,balanced,0.26452799638112384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,1,balanced,0.26684800783793133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,1,balanced,0.269594669342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,1,balanced,0.27530133724212646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,1,balanced,0.278437336285909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,1,balanced,0.2797440091768901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,1,balanced,0.29037866989771527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,1,balanced,0.2919626633326213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,1,balanced,0.30718932549158734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,1,balanced,0.3386026620864868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,1,balanced,0.36613865693410236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,1,balanced,0.4246986707051595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,1,balanced,0.48099199930826825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,1,balanced,0.6008373498916626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,1,balanced,0.7203306357065836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,1,balanced,1.0895466804504395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,1,balanced,1.3392106691996257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,1,balanced,1.9501439730326335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,1,balanced,2.5414932568868003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,2,power_law_1.2,0.046649599075317384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,2,power_law_1.2,0.04648320078849792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,2,power_law_1.2,0.0455487996339798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,2,power_law_1.2,0.047193598747253415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,2,power_law_1.2,0.047942399978637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,2,power_law_1.2,0.05311359763145447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,2,power_law_1.2,0.07422080039978027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,2,power_law_1.2,0.06606079936027527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,2,power_law_1.2,0.08784639835357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,2,power_law_1.2,0.07955200076103211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,2,power_law_1.2,0.07553279995918274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,2,power_law_1.2,0.13565440177917482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,2,power_law_1.2,0.12471679449081421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,2,power_law_1.2,0.10042239427566528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,2,power_law_1.2,0.14646400213241578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,2,power_law_1.2,0.12392959594726563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,2,power_law_1.2,0.1624575972557068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,2,power_law_1.2,0.21080319881439208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.2,0.27422080039978025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.2,0.38883841037750244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.2,0.5198207855224609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.2,0.7430975914001465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,32,power_law_1.01,0.10145920515060425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,32,power_law_1.01,0.07923200130462646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,32,power_law_1.01,0.07858560085296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,32,power_law_1.01,0.07052800059318542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,32,power_law_1.01,0.07169920206069946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,32,power_law_1.01,0.07991039752960205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,32,power_law_1.01,0.07976959943771363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,32,power_law_1.01,0.08030719757080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,32,power_law_1.01,0.07555199861526489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,32,power_law_1.01,0.07917439937591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,32,power_law_1.01,0.08232319951057435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,32,power_law_1.01,0.082259202003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,32,power_law_1.01,0.08317440152168273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,32,power_law_1.01,0.08688639998435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,32,power_law_1.01,0.09899520277976989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,32,power_law_1.01,0.09879680275917054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,32,power_law_1.01,0.10077439546585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,32,power_law_1.01,0.11043200492858887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,32,power_law_1.01,0.11920000314712524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,32,power_law_1.01,0.14295680522918702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,32,power_law_1.01,0.1613759994506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,32,power_law_1.01,0.2016767978668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,32,power_law_1.01,0.23644800186157228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,32,power_law_1.01,0.3025536060333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,32,power_law_1.01,0.38360960483551027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,32,power_law_1.01,0.6213247776031494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,32,power_law_1.01,0.6681407928466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.2,0.9875007629394531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.2,1.4575167655944825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.2,1.9168319702148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.2,2.8724544525146483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.2,3.7987201690673826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,1,power_law_1.01,0.13909759521484374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,1,power_law_1.01,0.20277121067047119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,1,power_law_1.01,0.2708928108215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,1,power_law_1.01,0.4144576072692871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,1,power_law_1.01,0.4321023941040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,1,power_law_1.01,0.44714879989624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,1,power_law_1.01,0.448038387298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,1,power_law_1.01,0.45525121688842773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,1,power_law_1.01,0.4632575988769531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,1,power_law_1.01,0.47436161041259767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,1,power_law_1.01,0.4835008144378662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,1,power_law_1.01,0.5009664058685303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,1,power_law_1.01,0.5033472061157227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,1,power_law_1.01,0.5049727916717529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,1,power_law_1.01,0.5624192237854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,1,power_law_1.01,0.5583487987518311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,1,power_law_1.01,0.6160384178161621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,1,power_law_1.01,0.6931839942932129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,1,power_law_1.01,0.7644544124603272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,1,power_law_1.01,0.9509823799133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,1,power_law_1.01,1.1625791549682618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,1,power_law_1.01,1.6204736709594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,1,power_law_1.01,1.8867263793945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,8,power_law_1.01,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,8,power_law_1.01,0.0417279988527298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,8,power_law_1.01,0.042208001017570496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,8,power_law_1.01,0.043859198689460754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,8,power_law_1.01,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,8,power_law_1.01,0.044352000951766966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,8,power_law_1.01,0.04454399943351746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,8,power_law_1.01,0.04457600116729736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,8,power_law_1.01,0.0444927990436554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,8,power_law_1.01,0.04441600143909454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,8,power_law_1.01,0.04450559914112091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,8,power_law_1.01,0.04563199877738953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,8,power_law_1.01,0.04625920057296753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,8,power_law_1.01,0.046419200301170346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,8,power_law_1.01,0.05114240050315857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,1,power_law_1.01,2.7225088119506835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,8,power_law_1.01,0.05121279954910278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,8,power_law_1.01,0.05432959794998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,8,power_law_1.01,0.0585919976234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,8,power_law_1.01,0.06632320284843445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,8,power_law_1.01,0.07841280102729797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,8,power_law_1.01,0.0904640018939972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,1,power_law_1.01,3.802912139892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,8,power_law_1.01,0.12403199672698975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,8,power_law_1.01,0.14783999919891358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,8,power_law_1.01,0.20896639823913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,1,power_law_1.01,5.379923248291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,8,power_law_1.01,0.2599936008453369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,1,power_law_1.2,0.060115200281143186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,1,power_law_1.2,0.07323520183563233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,1,power_law_1.2,0.10305279493331909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,1,power_law_1.2,0.13203840255737304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,1,power_law_1.2,0.13157119750976562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,1,power_law_1.2,0.13987840414047242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,1,power_law_1.2,0.1415295958518982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,1,power_law_1.2,0.14127999544143677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,1,power_law_1.2,0.14528000354766846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,1,power_law_1.2,0.14923520088195802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,1,power_law_1.2,0.14556800127029418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,1,power_law_1.2,0.15599360466003417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,1,power_law_1.2,0.15824639797210693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,1,power_law_1.2,0.16095999479293824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,1,power_law_1.2,0.1820736050605774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,1,power_law_1.2,0.18984320163726806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,1,power_law_1.2,0.19761279821395875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,1,power_law_1.2,0.23077120780944824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,1,power_law_1.2,0.2636735916137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,1,power_law_1.2,0.3241280078887939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,8,power_law_1.01,0.3502271890640259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,1,power_law_1.2,0.38853759765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,1,power_law_1.01,6.8808128356933596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,1,power_law_1.2,0.5209727764129639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,1,power_law_1.2,0.6640255928039551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,16,power_law_1.2,0.06734079718589783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,16,power_law_1.2,0.08221439719200134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,16,power_law_1.2,0.060185599327087405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,16,power_law_1.2,0.06232960224151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,16,power_law_1.2,0.0615231990814209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,16,power_law_1.2,0.06111999750137329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,16,power_law_1.2,0.062041598558425906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,16,power_law_1.2,0.06282879710197449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,16,power_law_1.2,0.06356480121612548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,16,power_law_1.2,0.06483839750289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,16,power_law_1.2,0.06448000073432922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,16,power_law_1.2,0.06736639738082886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,16,power_law_1.2,0.06755200028419495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,16,power_law_1.2,0.07015680074691773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,16,power_law_1.2,0.07621759772300721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,16,power_law_1.2,0.08042240142822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,16,power_law_1.2,0.08615679740905761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,16,power_law_1.2,0.09945600032806397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.2,0.11275520324707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.2,0.1328703999519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.2,0.16125439405441283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.2,0.2048703908920288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.2,0.24159998893737794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.2,0.34117119312286376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.2,0.44591360092163085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.2,0.6075007915496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.2,0.7871808052062989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,1,power_law_1.2,0.8982272148132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,8,power_law_1.01,0.45952639579772947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,1,power_law_1.2,1.1589632034301758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,1,power_law_1.2,1.6569023132324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,1,power_law_1.2,2.1593984603881835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,8,power_law_1.01,0.0883840024471283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,8,power_law_1.01,0.11527680158615113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,8,power_law_1.01,0.10440959930419921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,8,power_law_1.01,0.11222399473190307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,8,power_law_1.01,0.12446719408035278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,8,power_law_1.01,0.12532479763031007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,8,power_law_1.01,0.12641279697418212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,8,power_law_1.01,0.12519680261611937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,8,power_law_1.01,0.131660795211792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,8,power_law_1.01,0.13564800024032592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,8,power_law_1.01,0.1351744055747986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,8,power_law_1.01,0.13798400163650512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,8,power_law_1.01,0.13939839601516724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,8,power_law_1.01,0.1425920009613037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,8,power_law_1.01,0.1504320025444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,8,power_law_1.01,0.15297919511795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,8,power_law_1.01,0.1640128016471863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,8,power_law_1.01,0.1807360053062439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.01,0.20861439704895018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.01,0.24824960231781007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.01,0.28025600910186765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.01,0.32956159114837646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.01,0.400710391998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.01,0.5368703842163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.01,0.6514304161071778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.01,0.9725760459899903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.01,1.3110655784606933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,32,power_law_1.01,0.06322559714317322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,32,power_law_1.01,0.05667200088500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,32,power_law_1.01,0.05326719880104065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,32,power_law_1.01,0.04224640130996704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,32,power_law_1.01,0.04878720045089722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,32,power_law_1.01,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,32,power_law_1.01,0.05253120064735413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,32,power_law_1.01,0.055827200412750244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,32,power_law_1.01,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,32,power_law_1.01,0.05621119737625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,32,power_law_1.01,0.05796480178833008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,32,power_law_1.01,0.05913599729537964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,32,power_law_1.01,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,32,power_law_1.01,0.06281599998474122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,32,power_law_1.01,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,32,power_law_1.01,0.06812160015106201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,32,power_law_1.01,0.0717311978340149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,32,power_law_1.01,0.07631999850273133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,32,power_law_1.01,0.0864960014820099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,32,power_law_1.01,0.09976959824562073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,32,power_law_1.01,0.11508480310440064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,32,power_law_1.01,0.14549119472503663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,32,power_law_1.01,0.1509119987487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,32,power_law_1.01,0.21658880710601808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,32,power_law_1.01,0.28869760036468506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,32,power_law_1.01,0.3634687900543213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,32,power_law_1.01,0.4611968040466309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,4,power_law_1.2,0.08492799997329711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,4,power_law_1.2,0.09168639779090881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,4,power_law_1.2,0.09479039907455444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,4,power_law_1.2,0.11393280029296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,4,power_law_1.2,0.11512960195541382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,4,power_law_1.2,0.14891519546508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,4,power_law_1.2,0.145414400100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,4,power_law_1.2,0.15033600330352784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,4,power_law_1.2,0.15210239887237548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,4,power_law_1.2,0.15175039768218995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,4,power_law_1.2,0.15714559555053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,4,power_law_1.2,0.15889279842376708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,4,power_law_1.2,0.16284159421920777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,4,power_law_1.2,0.16520960330963136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,4,power_law_1.2,0.17160320281982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,4,power_law_1.2,0.17407360076904296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,4,balanced,0.039450667798519135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,4,power_law_1.2,0.19215359687805175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,4,power_law_1.2,0.20956799983978272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,4,power_law_1.2,0.23663361072540284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,4,power_law_1.2,0.2829952001571655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,4,power_law_1.2,0.35265920162200926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,4,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,4,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,4,balanced,0.0353973334034284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,4,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,4,power_law_1.2,0.4587967872619629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,4,balanced,0.037263999382654824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,4,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,4,balanced,0.0355679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,4,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,4,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,4,power_law_1.2,0.6010560035705567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,4,balanced,0.039818666875362396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,4,balanced,0.03818666686614355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,4,balanced,0.03880000114440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,4,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,4,balanced,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,4,balanced,0.05481599768002828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,4,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,4,balanced,0.06840000053246816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,4,balanced,0.0660159985224406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,4,balanced,0.07839466631412506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,4,power_law_1.2,0.8920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,4,balanced,0.0835093359152476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,4,balanced,0.1011786659558614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,4,balanced,0.12330666184425354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,4,balanced,0.15715733170509338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,4,balanced,0.192848006884257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,4,balanced,0.2548533280690511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,4,balanced,0.30905065933863324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,4,power_law_1.2,1.051091194152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,4,power_law_1.2,1.499180793762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,4,power_law_1.2,2.1697919845581053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,8,power_law_1.01,0.020160000026226043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,8,power_law_1.01,0.019782400131225585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,8,power_law_1.01,0.019519999623298645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,8,power_law_1.01,0.019993600249290467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,8,power_law_1.01,0.01930239945650101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,8,power_law_1.01,0.022111999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,8,power_law_1.01,0.03963519930839539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,8,power_law_1.01,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,8,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,8,power_law_1.01,0.03447679877281189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,8,power_law_1.01,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,8,power_law_1.01,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,8,power_law_1.01,0.04191359877586365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,8,power_law_1.01,0.03415040075778961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,8,power_law_1.01,0.04405120015144348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,8,power_law_1.01,0.04252800047397613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,8,power_law_1.01,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,8,power_law_1.01,0.05745279788970947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.01,0.06563839912414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.01,0.08595200181007386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.01,0.10721280574798583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.01,0.14551680088043212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.01,0.1786239981651306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.01,0.2523711919784546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.01,0.32575359344482424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.01,0.4722623825073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,1,balanced,0.05825066566467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,1,balanced,0.08454933762550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,1,balanced,0.13269333044687906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,1,balanced,0.13780799508094788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,1,balanced,0.1390720009803772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,1,balanced,0.13851733009020487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,1,balanced,0.13885333140691122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,1,balanced,0.1421173314253489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,1,balanced,0.14503467082977295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,1,balanced,0.14427733421325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,1,balanced,0.14717866977055868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,1,balanced,0.14961066842079163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,1,balanced,0.15506666898727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,1,balanced,0.1576586663722992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,1,balanced,0.17327467600504556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,1,balanced,0.178330659866333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,1,balanced,0.18186134099960327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,1,balanced,0.22338134050369263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,1,balanced,0.22920000553131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,1,balanced,0.31117866436640423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,1,balanced,0.3449546496073405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,1,balanced,0.4737813472747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,1,balanced,0.5894879897435507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,1,balanced,0.8582293192545573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,1,balanced,1.0939146677652996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,1,balanced,1.6055679321289062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,1,balanced,2.1096906661987305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.01,0.6191616058349609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,2,power_law_1.2,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,2,power_law_1.2,0.05660799741744995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,2,power_law_1.2,0.06711680293083191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,2,power_law_1.2,0.07850880026817322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,2,power_law_1.2,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,2,power_law_1.2,0.11442559957504272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,2,power_law_1.2,0.1261952042579651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,2,power_law_1.2,0.1450943946838379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,2,power_law_1.2,0.15121920108795167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,2,power_law_1.2,0.15447039604187013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,2,power_law_1.2,0.15377919673919677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,2,power_law_1.2,0.16561280488967894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,2,power_law_1.2,0.17247999906539918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,2,power_law_1.2,0.18059519529342652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,2,power_law_1.2,0.19594240188598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,2,power_law_1.2,0.1979840040206909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,2,power_law_1.2,0.2163520097732544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,2,power_law_1.2,0.24465279579162597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,2,power_law_1.2,0.2807039976119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,2,power_law_1.2,0.34197120666503905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,2,power_law_1.2,0.4019328117370605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,2,power_law_1.2,0.535807991027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,2,power_law_1.2,0.6618559837341309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,2,power_law_1.2,0.9189951896667481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,2,power_law_1.2,1.1026240348815919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,2,power_law_1.2,1.6748672485351563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,2,power_law_1.2,2.057542419433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,1,power_law_1.01,0.10328320264816285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,1,power_law_1.01,0.13315199613571166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,1,power_law_1.01,0.196396803855896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,1,power_law_1.01,0.28286080360412597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,1,power_law_1.01,0.37011840343475344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,1,power_law_1.01,0.5630335807800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,1,power_law_1.01,0.6053184032440185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,1,power_law_1.01,0.6420224189758301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,1,power_law_1.01,0.6377408027648925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,1,power_law_1.01,0.6560959815979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,1,power_law_1.01,0.6780032157897949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,1,power_law_1.01,0.6861504077911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,1,power_law_1.01,0.6849599838256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,1,power_law_1.01,0.7030848026275635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,1,power_law_1.01,0.7425983905792236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,1,power_law_1.01,0.7577343940734863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,1,power_law_1.01,0.7930560111999512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,1,power_law_1.01,0.8577280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.01,0.9451199531555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.01,1.1286399841308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.01,1.1724160194396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.01,1.4714624404907226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.01,1.8339136123657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.01,2.401875114440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.01,2.9687936782836912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.01,4.1343742370605465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.01,5.203091049194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,8,balanced,0.04228266576925913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,8,balanced,0.043525333205858864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,8,balanced,0.04603200157483419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,8,balanced,0.07020799815654755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,8,balanced,0.09282666444778442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,8,balanced,0.13885333140691122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,8,balanced,0.13743999600410461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,8,balanced,0.13876799742380777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,8,balanced,0.1402239998181661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,8,balanced,0.141893337170283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,8,balanced,0.14010666807492575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,8,balanced,0.1418719987074534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,8,balanced,0.14664533734321594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,8,balanced,0.1469119985898336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,8,balanced,0.15545599659283957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,8,balanced,0.15624533096949259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,8,balanced,0.1599146624406179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,8,balanced,0.17283199230829874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,8,balanced,0.1819253365198771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,8,balanced,0.2076693375905355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,8,balanced,0.23227733373641968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,8,balanced,0.2844746708869934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,8,balanced,0.32211732864379883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,8,balanced,0.42774399121602374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,8,balanced,0.5221173365910848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,8,balanced,0.7215733528137207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,8,balanced,0.9028106530507406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,4,balanced,0.030970667799313862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,4,balanced,0.029690665503342945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,4,balanced,0.029605334003766377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,4,balanced,0.03325333446264267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,4,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,4,balanced,0.03377600014209747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,4,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,4,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,4,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,4,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,4,balanced,0.03530666728814443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,4,balanced,0.03757333258787791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,4,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,4,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,4,balanced,0.04359466830889384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,4,balanced,0.04590400060017904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,4,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,4,balanced,0.05349866549173991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,4,balanced,0.06005333364009857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,4,balanced,0.06645866731802623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,4,balanced,0.08073066671689351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,4,balanced,0.1025279959042867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,4,balanced,0.124399999777476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,4,balanced,0.17642132441202799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,4,balanced,0.21385065714518228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,4,balanced,0.3020640015602112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,4,balanced,0.41067198912302655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,8,power_law_1.2,0.023104000091552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,8,power_law_1.2,0.026240000128746034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,8,power_law_1.2,0.030214399099349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,8,power_law_1.2,0.03343360126018524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,8,power_law_1.2,0.036774399876594546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,8,power_law_1.2,0.04421760141849518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,8,power_law_1.2,0.050367999076843264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,8,power_law_1.2,0.049753600358963014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,8,power_law_1.2,0.049132800102233885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2688,1856,6,128,1,8,power_law_1.2,0.06769919991493226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2688,1856,6,128,1,8,power_law_1.2,0.06566399931907654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2688,1856,6,128,1,8,power_law_1.2,0.0623744010925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,8,balanced,0.044639999667803444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,8,balanced,0.04772266745567322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,8,balanced,0.047082667549451195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,8,balanced,0.04970133304595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,8,balanced,0.05422399938106537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,8,balanced,0.06301866471767426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,8,balanced,0.07029333213965099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,8,balanced,0.0722453345855077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,8,balanced,0.07259200016657512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,8,balanced,0.07327466706434886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,8,balanced,0.07628799974918365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,8,balanced,0.07646400233109792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,8,balanced,0.07976533472537994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,8,balanced,0.08116800089677174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,8,balanced,0.08773333827654521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,8,balanced,0.08841066559155782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,8,balanced,0.09544000029563904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,8,balanced,0.1162506639957428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,8,balanced,0.1274186670780182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,8,balanced,0.1562986671924591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,8,balanced,0.17965332667032877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,8,balanced,0.22949333985646567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,8,balanced,0.29651200771331787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,8,balanced,0.398357351620992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,8,balanced,0.5212746858596802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,8,balanced,0.7451466719309489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,8,balanced,0.9547359943389893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,64,power_law_1.2,0.06728320121765137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,64,power_law_1.2,0.06595839858055115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,64,power_law_1.2,0.04893440008163452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,64,power_law_1.2,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,64,power_law_1.2,0.04890879988670349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,64,power_law_1.2,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,64,power_law_1.2,0.0526528000831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,64,power_law_1.2,0.05383679866790771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,64,power_law_1.2,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,64,power_law_1.2,0.05299199819564819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,64,power_law_1.2,0.05459840297698974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,64,power_law_1.2,0.05668479800224304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,64,power_law_1.2,0.05913599729537964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,64,power_law_1.2,0.060192000865936277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,64,power_law_1.2,0.06807680130004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,64,power_law_1.2,0.06887680292129517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,64,power_law_1.2,0.07516160011291503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,64,power_law_1.2,0.08541439771652222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,64,power_law_1.2,0.09275519847869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,64,power_law_1.2,0.11352319717407226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,64,power_law_1.2,0.13373440504074097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,64,power_law_1.2,0.17636480331420898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,64,power_law_1.2,0.19924479722976685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,64,power_law_1.2,0.27099521160125734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,64,power_law_1.2,0.35810561180114747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,64,power_law_1.2,0.535423994064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,64,power_law_1.2,0.6604928016662598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,1,power_law_1.01,0.0976639986038208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,1,power_law_1.01,0.10992000102996827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,1,power_law_1.01,0.13931519985198976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,1,power_law_1.01,0.17286399602890015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,1,power_law_1.01,0.23919999599456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,1,power_law_1.01,0.31141760349273684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,1,power_law_1.01,0.35376639366149903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,1,power_law_1.01,0.4335040092468262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,1,power_law_1.01,0.44552321434020997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,1,power_law_1.01,0.4553664207458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,1,power_law_1.01,0.4677120208740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,1,balanced,0.03153600047032038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,1,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,1,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,1,balanced,0.03726933399836222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,1,balanced,0.037818667789300285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,1,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,1,balanced,0.038618666430314384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,1,balanced,0.040037333965301514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,1,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,1,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,1,balanced,0.04247466723124186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,1,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,1,balanced,0.04557866851488749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,1,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,1,balanced,0.0621919979651769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,1,balanced,0.063360000650088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,1,balanced,0.07277866701285045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,1,power_law_1.01,0.4916351795196533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,1,balanced,0.09157333771387736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,1,balanced,0.09397866328557332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,1,balanced,0.12333866953849792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,1,balanced,0.13637333114941916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,1,balanced,0.1792373259862264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,1,balanced,0.2206773360570272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,1,balanced,0.3118559916814168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,1,balanced,0.4041173458099365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,1,balanced,0.58023468653361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,1,power_law_1.01,0.4991744041442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,1,balanced,0.7514986991882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,1,power_law_1.01,0.5188032150268554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,1,power_law_1.01,0.5296383857727051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,1,power_law_1.01,0.5404928207397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,1,power_law_1.01,0.5696703910827636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,1,power_law_1.01,0.6386496067047119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.01,0.7012095928192139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.01,0.8065279960632324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.01,0.9151679992675781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.01,1.1404735565185546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.01,1.3195008277893066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.01,1.739468765258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,1,power_law_1.01,2.262553596496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,1,power_law_1.01,3.075276756286621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,1,power_law_1.01,4.051667022705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,4,power_law_1.01,0.07256320118904114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,4,power_law_1.01,0.08082559704780579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,4,power_law_1.01,0.08507519960403442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,4,power_law_1.01,0.09733120203018189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,4,power_law_1.01,0.11288319826126099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,4,power_law_1.01,0.11720319986343383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,4,power_law_1.01,0.13581440448760987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,4,power_law_1.01,0.13496960401535035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,4,power_law_1.01,0.1342144012451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,4,power_law_1.01,0.13021440505981446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,4,power_law_1.01,0.14234880208969117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,4,power_law_1.01,0.13619840145111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,4,power_law_1.01,0.13969279527664186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,4,power_law_1.01,0.13899519443511962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,4,power_law_1.01,0.14684799909591675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,4,power_law_1.01,0.148089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,4,power_law_1.01,0.15665919780731202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,4,power_law_1.01,0.1699136018753052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,4,power_law_1.01,0.18836480379104614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,4,power_law_1.01,0.21940479278564454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,4,power_law_1.01,0.2517695903778076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,4,power_law_1.01,0.3213248014450073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,4,power_law_1.01,0.38719360828399657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,4,power_law_1.01,0.5252416133880615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,4,power_law_1.01,0.6742335796356201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,4,power_law_1.01,0.9741056442260743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,4,power_law_1.01,1.2690367698669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,1,balanced,0.08266666531562805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,1,balanced,0.09289600451787312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,1,balanced,0.11955733100573222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,1,balanced,0.16391467054684958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,1,balanced,0.24761066834131876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,1,balanced,0.40882666905721027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,1,balanced,0.4172906478246053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,1,balanced,0.41433600584665936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,1,balanced,0.4184639851252238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,1,balanced,0.41811732451121014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,1,balanced,0.42869865894317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,1,balanced,0.43270933628082275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,1,balanced,0.4365546703338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,1,balanced,0.44122131665547687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,1,balanced,0.45322132110595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,1,balanced,0.45688533782958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,1,balanced,0.4752800067265828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,1,balanced,0.5226986805597941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,1,balanced,0.5549386739730835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,1,balanced,0.633349339167277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,1,balanced,0.718010663986206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,1,balanced,0.8870293299357096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,1,balanced,1.0581920146942139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,1,balanced,1.5472906430562336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,1,balanced,1.8565813700358074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,1,balanced,2.7864160537719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,1,balanced,3.503391901652018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,128,power_law_1.01,0.08087040185928344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,128,power_law_1.01,0.08851199746131896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,128,power_law_1.01,0.06563839912414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,128,power_law_1.01,0.06588799953460693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,128,power_law_1.01,0.060121601819992064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,128,power_law_1.01,0.06095359921455383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,128,power_law_1.01,0.06554880142211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,128,power_law_1.01,0.06636800169944763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,128,power_law_1.01,0.06727679967880248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,128,power_law_1.01,0.06736639738082886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,128,power_law_1.01,0.06862080097198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,128,power_law_1.01,0.06683520078659058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,128,power_law_1.01,0.06963840126991272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,128,power_law_1.01,0.06983680129051209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,128,power_law_1.01,0.07589759826660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,128,power_law_1.01,0.07919359803199769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,128,power_law_1.01,0.08843520283699036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,128,power_law_1.01,0.09756799936294555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,128,power_law_1.01,0.10698239803314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,128,power_law_1.01,0.12985600233078004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,128,power_law_1.01,0.13822720050811768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,128,power_law_1.01,0.18165760040283202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,128,power_law_1.01,0.20298879146575927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,128,power_law_1.01,0.2644160032272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,128,power_law_1.01,0.36659839153289797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,128,power_law_1.01,0.5098368167877197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,128,power_law_1.01,0.5825920104980469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,1,power_law_1.01,0.03411200046539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,1,power_law_1.01,0.03813759982585907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,1,power_law_1.01,0.04382719993591309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,1,power_law_1.01,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,1,power_law_1.01,0.05573760271072388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,1,power_law_1.01,0.05699840188026428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,1,power_law_1.01,0.060684800148010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,1,power_law_1.01,0.05927039980888367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,1,power_law_1.01,0.0606656014919281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,1,power_law_1.01,0.06321920156478882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,1,power_law_1.01,0.06221439838409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,1,power_law_1.01,0.06433280110359192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,1,power_law_1.01,0.06684160232543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,1,power_law_1.01,0.07216640114784241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,1,power_law_1.01,0.08273280262947083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,1,power_law_1.01,0.08608639836311341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,1,power_law_1.01,0.09293439984321594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,1,power_law_1.01,0.11564799547195434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,1,power_law_1.01,0.12396800518035889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,1,power_law_1.01,0.15569280385971068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,1,power_law_1.01,0.18153599500656128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,1,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,1,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,1,balanced,0.0753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,1,balanced,0.07123733560244243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,1,balanced,0.06949333349863689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,1,balanced,0.07849066456158955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,1,balanced,0.06593599915504456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,1,balanced,0.07014399766921997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,1,balanced,0.06810666620731354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,1,balanced,0.0643093337615331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,1,balanced,0.07007466753323872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,1,balanced,0.06677333513895671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,1,power_law_1.01,0.2493567943572998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,1,power_law_1.01,0.3034303903579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,1,balanced,0.07513066629568736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,1,balanced,0.07436266541481018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,1,balanced,0.075013334552447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,1,balanced,0.07655466596285503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,1,balanced,0.08045866588751475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,1,balanced,0.11052800218264262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,1,balanced,0.11148266990979512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,1,balanced,0.14178666472434998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,1,balanced,0.16723199685414633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,1,balanced,0.22606933116912842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,1,balanced,0.28174932797749835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,1,balanced,0.39637335141499835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,1,balanced,0.5114773511886597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,1,balanced,0.7334026495615641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,1,power_law_1.01,0.42629117965698243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,1,balanced,0.9672586917877197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,1,power_law_1.01,0.5500224113464356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,16,power_law_1.01,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,1,power_law_1.01,0.7844223976135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,16,power_law_1.01,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,16,power_law_1.01,0.05335680246353149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,16,power_law_1.01,0.05015040040016174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,16,power_law_1.01,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,16,power_law_1.01,0.054073601961135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,16,power_law_1.01,0.05541120171546936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,16,power_law_1.01,0.05459200143814087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,16,power_law_1.01,0.05496320128440857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,16,power_law_1.01,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,16,power_law_1.01,0.05618559718132019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,16,power_law_1.01,0.05619199872016907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,16,power_law_1.01,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,16,power_law_1.01,0.0599232017993927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,16,power_law_1.01,0.06554239988327026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,16,power_law_1.01,0.06509439945220948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,16,power_law_1.01,0.06810879707336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,16,power_law_1.01,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,16,power_law_1.01,0.0828224003314972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,16,power_law_1.01,0.09467520117759705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,16,power_law_1.01,0.10851199626922607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,16,power_law_1.01,0.13319040536880494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,16,power_law_1.01,0.15028480291366578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,16,power_law_1.01,0.21118719577789308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,16,power_law_1.01,0.2291584014892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,16,power_law_1.01,0.338483190536499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,16,power_law_1.01,0.44112000465393064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,1,power_law_1.01,1.0282112121582032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,256,power_law_1.01,0.09084799885749817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,256,power_law_1.01,0.061286401748657224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,256,power_law_1.01,0.0536575973033905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,256,power_law_1.01,0.05377920269966126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,256,power_law_1.01,0.0531391978263855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,256,power_law_1.01,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,256,power_law_1.01,0.05449600219726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,256,power_law_1.01,0.05457280278205871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,256,power_law_1.01,0.054662400484085084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,256,power_law_1.01,0.05578240156173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,256,power_law_1.01,0.056992000341415404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,256,power_law_1.01,0.0611840009689331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,256,power_law_1.01,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,256,power_law_1.01,0.06565759778022766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,256,power_law_1.01,0.07402880191802978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,256,power_law_1.01,0.07571200132369996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,256,power_law_1.01,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,256,power_law_1.01,0.09609599709510804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,256,power_law_1.01,0.10449919700622559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,256,power_law_1.01,0.13363200426101685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,256,power_law_1.01,0.1570688009262085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,256,power_law_1.01,0.20965120792388917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,256,power_law_1.01,0.20892798900604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,256,power_law_1.01,0.2802623987197876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,256,power_law_1.01,0.3456000089645386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,256,power_law_1.01,0.49312639236450195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,256,power_law_1.01,0.5831999778747559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,2,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,2,balanced,0.05011733373006185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,2,balanced,0.07030933101971944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,2,balanced,0.10197333494822185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,2,balanced,0.15668800473213196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,2,balanced,0.15747732917467752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,2,balanced,0.15928533673286438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,2,balanced,0.16060800353686014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,2,balanced,0.16024532914161682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,2,balanced,0.16541332999865213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,2,balanced,0.1643786629041036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,2,balanced,0.16850666205088297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,2,balanced,0.17123732964197794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,2,balanced,0.17229332526524863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,2,balanced,0.18073066075642905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,2,balanced,0.18440532684326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,2,balanced,0.19074134031931558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,2,balanced,0.2102773388226827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,2,balanced,0.23772799968719482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,2,balanced,0.2678133249282837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,2,balanced,0.2986560066541036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,2,balanced,0.40512533982594806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,2,balanced,0.46639466285705566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,2,balanced,0.6543146769205729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,2,balanced,0.826645294825236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,2,balanced,1.1887093385060628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,2,balanced,1.5160214106241863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,16,power_law_1.01,0.025043201446533204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,16,power_law_1.01,0.04545280039310455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,16,power_law_1.01,0.045177599787712096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,16,power_law_1.01,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,16,power_law_1.01,0.04472959935665131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,16,power_law_1.01,0.044038400053977966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,16,power_law_1.01,0.04737280011177063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,16,power_law_1.01,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,16,power_law_1.01,0.047270399332046506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,16,power_law_1.01,0.05182719826698303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,16,power_law_1.01,0.0500544011592865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,16,power_law_1.01,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,16,power_law_1.01,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,16,power_law_1.01,0.059910398721694944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,16,power_law_1.01,0.06332799792289734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,16,power_law_1.01,0.06474239826202392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,16,power_law_1.01,0.1093951940536499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,16,power_law_1.01,0.11900800466537476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,16,power_law_1.01,0.1391487956047058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,16,power_law_1.01,0.1602944016456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,16,power_law_1.01,0.11736960411071777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,16,power_law_1.01,0.13791999816894532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,16,power_law_1.01,0.16735999584197997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,16,power_law_1.01,0.22910079956054688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,16,power_law_1.01,0.28453760147094725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,16,power_law_1.01,0.41672320365905763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,16,power_law_1.01,0.5074687957763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,16,balanced,0.019061333189407986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,16,balanced,0.018960000326236088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,16,balanced,0.018976000448067982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,16,balanced,0.0191040001809597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,16,balanced,0.020213333268960316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,16,balanced,0.02125866711139679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,16,balanced,0.02329600105683009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,16,balanced,0.023386667172114056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,16,balanced,0.025216000775496166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,16,balanced,0.02712533374627431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,16,balanced,0.025034666061401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,16,balanced,0.027109332382678986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,16,balanced,0.026943999032179516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,16,balanced,0.03170666595300039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,16,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,16,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,16,balanced,0.045370668172836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,16,balanced,0.049365331729253135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,16,balanced,0.05137600004673004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,16,balanced,0.05927466849486033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,16,balanced,0.06955199937025706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,16,balanced,0.0844586690266927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,16,balanced,0.10054933031400044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,16,balanced,0.12822399536768594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,16,balanced,0.1542080044746399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,16,balanced,0.2256960074106852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,16,balanced,0.2808000048001607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,8,balanced,0.04600533346335093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,8,balanced,0.04994133114814758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,8,balanced,0.05060799916585287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,8,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,8,balanced,0.05564799904823303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,8,balanced,0.06425066788991292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,8,balanced,0.06434133152167003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,8,balanced,0.06542400022347768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,8,balanced,0.06621333460013072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,8,balanced,0.06739200154940288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,8,balanced,0.0660159985224406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,8,balanced,0.06689600149790446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,8,balanced,0.06833600004514058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,8,balanced,0.07016533116499583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,8,balanced,0.07389866809050243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,8,balanced,0.07513066629568736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,8,balanced,0.07896533111731212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,8,balanced,0.08612799644470215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,8,balanced,0.09266666571299235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,8,balanced,0.10558933019638062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,8,balanced,0.12105066577593486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,8,balanced,0.149317334095637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,8,balanced,0.1814240018526713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,8,balanced,0.25244800249735516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,8,balanced,0.3062346577644348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,8,balanced,0.4198506673177083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,8,balanced,0.5423253377278646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,2,power_law_1.01,0.09142400026321411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,2,power_law_1.01,0.1348863959312439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,2,power_law_1.01,0.17034239768981935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,2,power_law_1.01,0.2620800018310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,2,power_law_1.01,0.35260159969329835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,2,power_law_1.01,0.44635519981384275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,2,power_law_1.01,0.5901567935943604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,2,power_law_1.01,0.5840640068054199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,2,power_law_1.01,0.5756288051605225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,2,power_law_1.01,0.6117119789123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,2,power_law_1.01,0.5989696025848389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,2,power_law_1.01,0.6386752128601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,2,power_law_1.01,0.6565055847167969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,2,power_law_1.01,0.6729728221893311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,2,power_law_1.01,0.7119232177734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,2,power_law_1.01,0.7043456077575684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,2,power_law_1.01,0.7608448028564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,2,power_law_1.01,0.8309951782226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,2,power_law_1.01,0.8354111671447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,2,power_law_1.01,0.9736767768859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,2,power_law_1.01,0.963584041595459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,2,power_law_1.01,1.108198356628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,2,power_law_1.01,1.3248640060424806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,2,power_law_1.01,1.6541696548461915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,2,power_law_1.01,1.9696256637573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,2,power_law_1.01,2.5753536224365234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,2,power_law_1.01,3.212748718261719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,64,power_law_1.2,0.05999360084533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,64,power_law_1.2,0.0581055998802185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,64,power_law_1.2,0.05717759728431702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,64,power_law_1.2,0.058873599767684935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,64,power_law_1.2,0.05702400207519531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,64,power_law_1.2,0.061427199840545656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,64,power_law_1.2,0.06098560094833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,64,power_law_1.2,0.06352000236511231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,64,power_law_1.2,0.06254720091819763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,64,power_law_1.2,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,64,power_law_1.2,0.0649728000164032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,64,power_law_1.2,0.06380800008773804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,64,power_law_1.2,0.06666240096092224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,64,power_law_1.2,0.06622080206871032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,64,power_law_1.2,0.07214080095291138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,64,power_law_1.2,0.07352319955825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,64,power_law_1.2,0.07805439829826355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,64,power_law_1.2,0.09171199798583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,64,power_law_1.2,0.10341119766235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,64,power_law_1.2,0.13374719619750977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,64,power_law_1.2,0.15607680082321168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,64,power_law_1.2,0.20674560070037842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,64,power_law_1.2,0.2534976005554199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,64,power_law_1.2,0.3533375978469849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,64,power_law_1.2,0.4613952159881592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,64,power_law_1.2,0.6829311847686768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,64,power_law_1.2,0.9413439750671386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,8,balanced,0.03173866619666418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,8,balanced,0.029285334050655365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,8,balanced,0.03125333289305369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,8,balanced,0.03173866619666418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,8,balanced,0.03523733218510946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,8,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,8,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,8,balanced,0.03587199995915095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,8,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,8,balanced,0.03764266769091288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,8,balanced,0.0373333344856898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,8,balanced,0.03792533278465271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,8,balanced,0.037871999045213066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,8,balanced,0.03995733211437861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,8,balanced,0.04371733466784159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,8,balanced,0.041984001795450844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,8,balanced,0.045850664377212524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,8,balanced,0.048565333088239036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,8,balanced,0.05618133147557577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,8,balanced,0.058464000622431435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,8,balanced,0.06824000179767609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,8,balanced,0.09115200241406758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,8,balanced,0.10502933462460835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,8,balanced,0.1362506647904714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,8,balanced,0.16156799594561258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,8,balanced,0.23098133007685342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,8,balanced,0.27929067611694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,8,power_law_1.2,0.019340799748897554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,8,power_law_1.2,0.020979200303554536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,8,power_law_1.2,0.021817600727081297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,8,power_law_1.2,0.02343679964542389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,8,power_law_1.2,0.025248000025749208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,8,power_law_1.2,0.025772801041603087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,8,power_law_1.2,0.026675200462341307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,8,power_law_1.2,0.027699199318885804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,8,power_law_1.2,0.028428798913955687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,8,power_law_1.2,0.029235199093818665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,8,power_law_1.2,0.029094401001930236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,8,power_law_1.2,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,8,power_law_1.2,0.03224959969520569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,8,power_law_1.2,0.03356800079345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,8,power_law_1.2,0.03640320003032684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,8,power_law_1.2,0.0395904004573822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,8,power_law_1.2,0.04413439929485321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,8,power_law_1.2,0.05169919729232788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,8,power_law_1.2,0.0810368001461029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,8,power_law_1.2,0.09285119771957398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,8,power_law_1.2,0.08896639943122864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,8,power_law_1.2,0.11054719686508178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,8,power_law_1.2,0.13453439474105836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,8,power_law_1.2,0.18036479949951173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,8,power_law_1.2,0.21731839179992676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,8,power_law_1.2,0.2996927976608276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,8,power_law_1.2,0.3935296058654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,2,power_law_1.2,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,2,power_law_1.2,0.07683200240135193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,2,power_law_1.2,0.08718079924583436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,2,power_law_1.2,0.10514559745788574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,2,power_law_1.2,0.11083519458770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,2,power_law_1.2,0.14973440170288085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,2,power_law_1.2,0.15538560152053832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,2,power_law_1.2,0.1627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,2,power_law_1.2,0.1611199975013733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,2,power_law_1.2,0.16805119514465333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,2,power_law_1.2,0.17221120595932007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,2,power_law_1.2,0.1754431962966919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,2,power_law_1.2,0.17697919607162477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,2,power_law_1.2,0.18204159736633302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,2,power_law_1.2,0.18558080196380616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,2,power_law_1.2,0.19088640213012695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,2,power_law_1.2,0.19771519899368287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,2,power_law_1.2,0.21797120571136475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,2,power_law_1.2,0.2437119960784912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,2,power_law_1.2,0.2974208116531372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,2,power_law_1.2,0.3683903932571411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,2,power_law_1.2,0.4576831817626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,2,power_law_1.2,0.6391615867614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,2,power_law_1.2,0.8419136047363281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,2,power_law_1.2,1.149235153198242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,4,power_law_1.2,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,4,power_law_1.2,0.03857919871807099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,4,power_law_1.2,0.03790720105171204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,4,power_law_1.2,0.04079999923706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,4,power_law_1.2,0.041171199083328246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,4,power_law_1.2,0.042752000689506534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,4,power_law_1.2,0.04154239892959595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,4,power_law_1.2,0.04549759924411774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,4,power_law_1.2,0.045286399126052854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,4,power_law_1.2,0.04505600035190582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,4,power_law_1.2,0.04715520143508911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,4,power_law_1.2,0.05503360033035278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,4,power_law_1.2,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,4,power_law_1.2,0.06430720090866089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,4,power_law_1.2,0.07919999957084656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,4,power_law_1.2,0.07978879809379577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,4,power_law_1.2,0.09701759815216064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,4,power_law_1.2,0.1211840033531189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,4,power_law_1.2,0.1323583960533142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,4,power_law_1.2,0.18887679576873778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,4,power_law_1.2,0.22135040760040284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,2,power_law_1.2,1.6637632369995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,4,power_law_1.2,0.31900160312652587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,4,power_law_1.2,0.408512020111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,4,power_law_1.2,0.6433472156524658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,2,power_law_1.2,2.261004829406738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,4,power_law_1.2,0.8138815879821777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,4,power_law_1.2,1.2201727867126464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,4,power_law_1.2,1.6402751922607421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,1,power_law_1.2,0.016575999557971954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,1,power_law_1.2,0.016812799870967864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,1,power_law_1.2,0.018163199722766876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,1,power_law_1.2,0.020627200603485107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,1,power_law_1.2,0.02417919933795929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,1,power_law_1.2,0.027609598636627198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,1,power_law_1.2,0.029440000653266907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,1,power_law_1.2,0.02990719974040985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,1,power_law_1.2,0.031481599807739256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,1,power_law_1.2,0.03301759958267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,1,power_law_1.2,0.03391999900341034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,1,power_law_1.2,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,1,power_law_1.2,0.03769600093364715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,1,power_law_1.2,0.03727999925613403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,1,power_law_1.2,0.039680001139640805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,1,power_law_1.2,0.04085119962692261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,1,power_law_1.2,0.04469760060310364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,1,power_law_1.2,0.05400320291519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,1,power_law_1.2,0.06064000129699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,1,power_law_1.2,0.07775999903678894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,1,power_law_1.2,0.08559359908103943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,1,power_law_1.2,0.11058559417724609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,1,power_law_1.2,0.12353279590606689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,1,power_law_1.2,0.15820159912109374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,1,power_law_1.2,0.1916159987449646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,1,power_law_1.2,0.259769606590271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,8,power_law_1.01,0.055731201171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,8,power_law_1.01,0.07196159958839417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,8,power_law_1.01,0.08346880078315735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,8,power_law_1.01,0.08515840172767639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,8,power_law_1.01,0.08980479836463928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,8,power_law_1.01,0.0913919985294342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,8,power_law_1.01,0.09907839894294738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,8,power_law_1.01,0.09278079867362976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,8,power_law_1.01,0.09738240242004395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,8,power_law_1.01,0.10209280252456665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,8,power_law_1.01,0.10311039686203002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,8,power_law_1.01,0.10166399478912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,8,power_law_1.01,0.10531200170516967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,8,power_law_1.01,0.1086527943611145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,8,power_law_1.01,0.11441919803619385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,8,power_law_1.01,0.11624959707260132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,8,power_law_1.01,0.12378879785537719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,8,power_law_1.01,0.13420159816741944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,8,power_law_1.01,0.15576959848403932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,8,power_law_1.01,0.17722879648208617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,8,power_law_1.01,0.20440320968627929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,8,power_law_1.01,0.24703359603881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,8,power_law_1.01,0.29932799339294436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,8,power_law_1.01,0.3946559906005859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,8,power_law_1.01,0.4885439872741699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,8,power_law_1.01,0.6780672073364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,8,power_law_1.01,0.8703488349914551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,1,power_law_1.01,0.04892799854278564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,1,power_law_1.01,0.0754368007183075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,1,power_law_1.01,0.13162879943847655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,1,power_law_1.01,0.20876801013946533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,1,power_law_1.01,0.31101438999176023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,1,power_law_1.01,0.4213888168334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,1,power_law_1.01,0.5961088180541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,1,power_law_1.01,0.6207935810089111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,1,power_law_1.01,0.6478400230407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,1,power_law_1.01,0.6768191814422607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,1,power_law_1.01,0.6927296161651612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,1,power_law_1.01,0.7026368141174316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,1,power_law_1.01,0.7448128223419189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,1,power_law_1.01,0.7414656162261963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,1,power_law_1.01,0.7466303825378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,1,power_law_1.01,0.7965760231018066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,1,power_law_1.2,0.329638409614563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,1,power_law_1.01,0.7974271774291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,1,power_law_1.01,0.9057279586791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.01,0.9451711654663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.01,1.1043328285217284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.01,1.286073589324951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.01,1.6253887176513673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.01,1.4964159965515136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.01,1.9296127319335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.01,2.3719295501708983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.01,3.2138046264648437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.01,4.09447021484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,64,power_law_1.2,0.023846399784088135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,64,power_law_1.2,0.02248319983482361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,64,power_law_1.2,0.022176000475883483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,64,power_law_1.2,0.021049599349498748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,64,power_law_1.2,0.021132799983024596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,64,power_law_1.2,0.022201600670814513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,64,power_law_1.2,0.02273920029401779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,64,power_law_1.2,0.0262719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,64,power_law_1.2,0.026681599020957947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,64,power_law_1.2,0.026342400908470155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,64,power_law_1.2,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,64,power_law_1.2,0.04437119960784912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,64,power_law_1.2,0.04349440038204193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,64,power_law_1.2,0.04332799911499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,64,power_law_1.2,0.043833601474761966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,64,power_law_1.2,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,64,power_law_1.2,0.050470399856567386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,4,power_law_1.2,0.0775168001651764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,64,power_law_1.2,0.0500927984714508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.2,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.2,0.06769919991493226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.2,0.07675520181655884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.2,0.11011840105056762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.2,0.1313024044036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,4,power_law_1.2,0.08356480002403259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.2,0.17592959403991698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,4,power_law_1.2,0.11344000101089477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.2,0.2241919994354248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.2,0.3341759920120239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.2,0.4441215991973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,4,power_law_1.2,0.11646720170974731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,4,power_law_1.2,0.14216959476470947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,4,power_law_1.2,0.1775488018989563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,4,power_law_1.2,0.19216639995574952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,4,power_law_1.2,0.19107199907302858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,4,power_law_1.2,0.19512959718704223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,4,power_law_1.2,0.20328960418701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,4,power_law_1.2,0.20453760623931885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,4,power_law_1.2,0.21305599212646484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,4,power_law_1.2,0.21543679237365723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,4,power_law_1.2,0.22985599040985108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,4,power_law_1.2,0.23512959480285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,4,power_law_1.2,0.2506943941116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,4,power_law_1.2,0.25151360034942627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,4,power_law_1.2,0.2745599985122681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,4,power_law_1.2,0.3012991905212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,4,power_law_1.2,0.3501760005950928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,4,power_law_1.2,0.3405312061309814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,4,power_law_1.2,0.4824384212493896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,4,power_law_1.2,0.5463551998138427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,4,power_law_1.2,0.6678592205047608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,4,power_law_1.2,0.8043135643005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,4,power_law_1.2,1.0994303703308106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,4,power_law_1.2,1.4977279663085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,128,power_law_1.2,0.03519999980926514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,128,power_law_1.2,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,128,power_law_1.2,0.034534400701522826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,128,power_law_1.2,0.037350401282310486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,128,power_law_1.2,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,128,power_law_1.2,0.03823359906673431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,128,power_law_1.2,0.03877759873867035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,128,power_law_1.2,0.03858560025691986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,128,power_law_1.2,0.03860479891300202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,128,power_law_1.2,0.0384768009185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,128,power_law_1.2,0.039103999733924866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,128,power_law_1.2,0.039129599928855896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,128,power_law_1.2,0.03946239948272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,128,power_law_1.2,0.039878401160240176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,128,power_law_1.2,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,128,power_law_1.2,0.04359680116176605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,128,power_law_1.2,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,128,power_law_1.2,0.04545280039310455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,128,power_law_1.2,0.04764159917831421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,128,power_law_1.2,0.05144960284233093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,128,power_law_1.2,0.058719998598098753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,128,power_law_1.2,0.07941759824752807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,128,power_law_1.2,0.09045119881629944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,128,power_law_1.2,0.12049280405044556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,128,power_law_1.2,0.14130560159683228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,128,power_law_1.2,0.218176007270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,128,power_law_1.2,0.2806976079940796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,16,power_law_1.01,0.07427840232849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,16,power_law_1.01,0.09927039742469787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,8,power_law_1.01,0.07588480114936828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,16,power_law_1.01,0.08858879804611205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,8,power_law_1.01,0.0906175971031189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,16,power_law_1.01,0.09059839844703674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,8,power_law_1.01,0.08705919981002808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,16,power_law_1.01,0.08994560241699219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,8,power_law_1.01,0.09994879961013795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,16,power_law_1.01,0.08526719808578491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,8,power_law_1.01,0.08555520176887513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,8,power_law_1.01,0.10517760515213012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,8,power_law_1.01,0.10951039791107178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,8,power_law_1.01,0.11640959978103638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,8,power_law_1.01,0.11393920183181763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,8,power_law_1.01,0.1167296051979065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,8,power_law_1.01,0.11708799600601197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,8,power_law_1.01,0.12163840532302857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,16,power_law_1.01,0.0889024019241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,8,power_law_1.01,0.12131199836730958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,8,power_law_1.01,0.1264256000518799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,16,power_law_1.01,0.09233279824256897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,8,power_law_1.01,0.13809280395507811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,8,power_law_1.01,0.13867520093917846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,16,power_law_1.01,0.09182720184326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,8,power_law_1.01,0.15010559558868408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,16,power_law_1.01,0.0881663978099823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,8,power_law_1.01,0.17055360078811646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,16,power_law_1.01,0.08935040235519409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,8,power_law_1.01,0.19061119556427003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,16,power_law_1.01,0.09255679845809936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,8,power_law_1.01,0.23085439205169678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,16,power_law_1.01,0.0980351984500885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,8,power_law_1.01,0.2537087917327881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,16,power_law_1.01,0.10362880229949951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,16,power_law_1.01,0.10878080129623413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,16,power_law_1.01,0.11471999883651733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,16,power_law_1.01,0.11768319606781005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,16,power_law_1.01,0.13439359664916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,16,power_law_1.01,0.14024319648742675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,16,power_law_1.01,0.16803840398788453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,16,power_law_1.01,0.18929280042648317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,8,power_law_1.01,0.3123136043548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,16,power_law_1.01,0.2505408048629761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,8,power_law_1.01,0.3878079891204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,16,power_law_1.01,0.2964799880981445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,8,power_law_1.01,0.5034111976623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,16,power_law_1.01,0.39498879909515383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,8,power_law_1.01,0.630790376663208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,8,power_law_1.01,0.9208576202392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,8,power_law_1.01,1.24017276763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,16,power_law_1.01,0.5300928115844726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,16,power_law_1.01,0.6922751903533936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,16,power_law_1.01,0.9023039817810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,8,power_law_1.01,0.017548799514770508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,8,power_law_1.01,0.018374399840831758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,8,power_law_1.01,0.019020800292491914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,8,power_law_1.01,0.019539199769496918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,8,power_law_1.01,0.021196800470352172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,8,power_law_1.01,0.022067199647426605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,8,power_law_1.01,0.024345600605010988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,8,power_law_1.01,0.025414401292800905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,8,power_law_1.01,0.026393601298332216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,8,power_law_1.01,0.026841598749160766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,8,power_law_1.01,0.026822400093078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,8,power_law_1.01,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,8,power_law_1.01,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,8,power_law_1.01,0.033766400814056394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,8,power_law_1.01,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,8,power_law_1.01,0.046937599778175354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,8,power_law_1.01,0.04755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,8,power_law_1.01,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,8,power_law_1.01,0.05384960174560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,8,power_law_1.01,0.06162559986114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,8,power_law_1.01,0.06718080043792725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,8,power_law_1.01,0.07763839960098266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,8,power_law_1.01,0.08671360015869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,8,power_law_1.01,0.11132160425186158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,8,power_law_1.01,0.13477760553359985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,8,power_law_1.01,0.1877120018005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,8,power_law_1.01,0.23487999439239501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,4,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,4,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,4,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,4,balanced,0.035599999129772186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,4,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,4,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,4,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,4,balanced,0.03737066686153412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,4,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,4,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,4,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,4,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,4,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,4,balanced,0.039733332892258964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,4,balanced,0.046623999873797096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,4,balanced,0.04953599969546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,4,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,4,balanced,0.056218668818473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,4,balanced,0.06229333579540253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,4,balanced,0.08069866895675659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,4,balanced,0.09913600484530131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,4,balanced,0.13297067085901895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,4,balanced,0.16054399808247885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,4,balanced,0.22159467140833536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,4,balanced,0.2660800019900004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,4,balanced,0.37244268258412677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,4,balanced,0.4793013334274292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,64,power_law_1.2,0.07614079713821412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,64,power_law_1.2,0.0750976026058197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,64,power_law_1.2,0.07397760152816772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,64,power_law_1.2,0.07679359912872315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,64,power_law_1.2,0.07600640058517456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,64,power_law_1.2,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,64,power_law_1.2,0.08294399976730346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,64,power_law_1.2,0.08234879970550538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,64,power_law_1.2,0.08310400247573853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,64,power_law_1.2,0.08532480001449586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,64,power_law_1.2,0.08386560082435608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,64,power_law_1.2,0.09159680008888245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,64,power_law_1.2,0.0934719979763031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,64,power_law_1.2,0.09125120043754578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,64,power_law_1.2,0.1037824034690857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,64,power_law_1.2,0.10515199899673462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,64,power_law_1.2,0.10872960090637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,64,power_law_1.2,0.12662400007247926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,64,power_law_1.2,0.13764480352401734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,64,power_law_1.2,0.1746880054473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,64,power_law_1.2,0.19687680006027222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,64,power_law_1.2,0.2745599985122681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,64,power_law_1.2,0.33882880210876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,64,power_law_1.2,0.4922175884246826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,64,power_law_1.2,0.6577280044555665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,64,power_law_1.2,0.9439359664916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,64,power_law_1.2,1.2852543830871581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,4,balanced,0.05481599768002828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,4,balanced,0.057061334451039634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,4,balanced,0.0689279983441035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,4,balanced,0.09245866537094116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,4,balanced,0.1262506643931071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,4,balanced,0.14401066303253174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,4,balanced,0.1439520021279653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,4,balanced,0.14670399824778238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,4,balanced,0.1466506620248159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,4,balanced,0.14734933773676553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,4,balanced,0.14823466539382935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,4,balanced,0.14866133530934653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,4,balanced,0.15340266625086466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,4,balanced,0.1548479994138082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,4,balanced,0.16145066420237222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,4,balanced,0.16525866587956747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,4,balanced,0.17147199312845865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,4,balanced,0.18479466438293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,4,balanced,0.19659199317296347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,4,balanced,0.26571200291315716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,4,balanced,0.2502506573994954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,4,balanced,0.33713066577911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,4,balanced,0.3777279853820801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,4,balanced,0.5328960021336874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,4,balanced,0.6555413405100504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,4,balanced,0.8976106643676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,4,balanced,1.1748586495717366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,1,power_law_1.01,0.023212799429893495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,1,power_law_1.01,0.03601279854774475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,1,power_law_1.01,0.04679040014743805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,1,power_law_1.01,0.06678400039672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,1,power_law_1.01,0.08850560188293458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,1,power_law_1.01,0.12866560220718384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,1,power_law_1.01,0.13893120288848876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,1,power_law_1.01,0.14181120395660402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,1,power_law_1.01,0.14933760166168214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,1,power_law_1.01,0.14698879718780516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,1,power_law_1.01,0.1598847985267639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,1,power_law_1.01,0.16131839752197266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,1,power_law_1.01,0.16750719547271728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,1,power_law_1.01,0.16971520185470582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,1,power_law_1.01,0.17405439615249635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,1,power_law_1.01,0.1808127999305725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,1,power_law_1.01,0.2183680057525635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,1,power_law_1.01,0.24338560104370116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.01,0.24505600929260254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.01,0.31089279651641843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.01,0.30853118896484377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.01,0.38530559539794923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.01,0.4468224048614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.01,0.5312320232391358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.01,0.6576704025268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.01,0.8955904006958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.01,1.1220224380493165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,32,power_law_1.2,0.07463039755821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,32,power_law_1.2,0.09059200286865235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,32,power_law_1.2,0.0856000006198883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,32,power_law_1.2,0.08461440205574036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,32,power_law_1.2,0.08520960211753845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,32,power_law_1.2,0.09012479782104492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,32,power_law_1.2,0.07444480061531067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,32,power_law_1.2,0.07581440210342408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,32,power_law_1.2,0.08732799887657165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,32,power_law_1.2,0.08618879914283753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,32,power_law_1.2,0.08977280259132385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,32,power_law_1.2,0.08952320218086243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,32,power_law_1.2,0.09350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,32,power_law_1.2,0.08873599767684937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,32,power_law_1.2,0.10063999891281128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,32,power_law_1.2,0.10614399909973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,32,power_law_1.2,0.11425280570983887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,32,power_law_1.2,0.13051520586013793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,32,power_law_1.2,0.14663679599761964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,32,power_law_1.2,0.18272639513015748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,32,power_law_1.2,0.1990847945213318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,32,power_law_1.2,0.26474239826202395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,32,power_law_1.2,0.3267584085464478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,32,power_law_1.2,0.4533184051513672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,32,power_law_1.2,0.5613887786865235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,32,power_law_1.2,0.8257535934448242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,32,power_law_1.2,1.0036607742309571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,256,power_law_1.2,0.02210559993982315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,256,power_law_1.2,0.021753600239753722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,256,power_law_1.2,0.023801599442958832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,256,power_law_1.2,0.02643199861049652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,256,power_law_1.2,0.052236801385879515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,256,power_law_1.2,0.04007039964199066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,256,power_law_1.2,0.03973119854927063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,256,power_law_1.2,0.04073599874973297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,256,power_law_1.2,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,256,power_law_1.2,0.04059520065784454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,256,power_law_1.2,0.04156799912452698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,256,power_law_1.2,0.041017600893974306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,256,power_law_1.2,0.0406143993139267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,256,power_law_1.2,0.03953279852867127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,256,power_law_1.2,0.042771199345588685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,256,power_law_1.2,0.04784640073776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,256,power_law_1.2,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,256,power_law_1.2,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.2,0.053324800729751584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.2,0.0723904013633728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.2,0.08515840172767639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.2,0.11047040224075318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.2,0.13877760171890258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.2,0.20353279113769532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.2,0.25464959144592286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.2,0.3890687942504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.2,0.49224319458007815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,1,power_law_1.01,0.039552000164985654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,1,power_law_1.01,0.04933120012283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,1,power_law_1.01,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,1,power_law_1.01,0.07582719922065735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,1,power_law_1.01,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,1,power_law_1.01,0.07870079874992371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,1,power_law_1.01,0.08041599988937378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,1,power_law_1.01,0.07948799729347229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,1,power_law_1.01,0.0812928020954132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,1,power_law_1.01,0.08423680067062378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,1,power_law_1.01,0.08335999846458435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,1,power_law_1.01,0.09068160057067871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,1,power_law_1.01,0.09166719913482665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,1,power_law_1.01,0.09605759978294373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,1,power_law_1.01,0.10788480043411255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,1,power_law_1.01,0.11113599538803101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,1,power_law_1.01,0.11488640308380127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,1,power_law_1.01,0.1380735993385315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,1,power_law_1.01,0.15777280330657958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,1,power_law_1.01,0.20042879581451417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,1,power_law_1.01,0.23765120506286622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,1,power_law_1.01,0.32261760234832765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,1,power_law_1.01,0.3890496015548706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,4,power_law_1.01,0.04780159890651703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,4,power_law_1.01,0.06204800009727478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,4,power_law_1.01,0.0673919975757599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,4,power_law_1.01,0.07448319792747497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,4,power_law_1.01,0.08195199966430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,4,power_law_1.01,0.09370239973068237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,4,power_law_1.01,0.10415359735488891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,4,power_law_1.01,0.10841599702835084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,4,power_law_1.01,0.11300480365753174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,4,power_law_1.01,0.10844800472259522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,4,power_law_1.01,0.10599679946899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,4,power_law_1.01,0.10926079750061035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,4,power_law_1.01,0.11436159610748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,4,power_law_1.01,0.11754879951477051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,4,power_law_1.01,0.12165759801864624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,4,power_law_1.01,0.12331520318984986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,4,power_law_1.01,0.1320639967918396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,4,power_law_1.01,0.14030079841613768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,4,power_law_1.01,0.15337599515914918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,4,power_law_1.01,0.1839743971824646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,4,power_law_1.01,0.1945024013519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,4,power_law_1.01,0.2527807950973511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,4,power_law_1.01,0.2807744026184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,4,power_law_1.01,0.36847360134124757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,4,power_law_1.01,0.45122561454772947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,4,power_law_1.01,0.6090047836303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,4,power_law_1.01,0.7966784000396728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,1,power_law_1.01,0.5443583965301514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,1,power_law_1.01,0.6853119850158691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,1,power_law_1.01,0.9777088165283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,1,power_law_1.01,1.2801983833312989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,4,balanced,0.05829866727193197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,4,balanced,0.06224533418814341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,4,balanced,0.06591466565926869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,4,balanced,0.07657066484292348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,4,balanced,0.10318932930628459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,4,balanced,0.1469066639741262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,4,balanced,0.18923733631769815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,4,balanced,0.19024533033370972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,4,balanced,0.19218132893244425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,4,balanced,0.1938506762186686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,4,balanced,0.1936533252398173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,4,balanced,0.19498666127522787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,4,balanced,0.19770665963490805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,4,balanced,0.20083733399709067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,4,balanced,0.20696000258127847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,4,balanced,0.2078933318456014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,4,balanced,0.22025599082310995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,4,balanced,0.2336533268292745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,4,balanced,0.25036799907684326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,4,balanced,0.2815093398094177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,4,balanced,0.30846933523813885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,4,balanced,0.3729066848754883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,4,balanced,0.43771199385325116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,4,balanced,0.5661386648813883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,4,balanced,0.7474079926808676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,4,balanced,1.0012266635894775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,4,balanced,1.3411146799723308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,8,power_law_1.01,0.048403200507164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,8,power_law_1.01,0.05459200143814087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,8,power_law_1.01,0.06230400204658508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,8,power_law_1.01,0.06629120111465454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,8,power_law_1.01,0.07659519910812378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,8,power_law_1.01,0.0743552029132843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,8,power_law_1.01,0.07680000066757202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,8,power_law_1.01,0.07627519965171814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,8,power_law_1.01,0.07867519855499268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,8,power_law_1.01,0.08028799891471863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,8,power_law_1.01,0.08335999846458435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,8,power_law_1.01,0.08625919818878174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,8,power_law_1.01,0.0864575982093811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,8,power_law_1.01,0.09349120259284974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,8,power_law_1.01,0.10414719581604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,8,power_law_1.01,0.10821759700775146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,8,power_law_1.01,0.11008640527725219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,8,power_law_1.01,0.12748160362243652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,8,power_law_1.01,0.14341119527816773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,8,power_law_1.01,0.17389440536499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,8,power_law_1.01,0.19837440252304078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,8,power_law_1.01,0.25731201171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,8,power_law_1.01,0.3197119951248169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,8,power_law_1.01,0.4525760173797607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,8,power_law_1.01,0.5651519775390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,8,power_law_1.01,0.8542400360107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,8,power_law_1.01,1.0395392417907714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,1,balanced,0.037392000357309975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,1,balanced,0.041178666055202484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,1,balanced,0.045653333266576133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,1,balanced,0.07010133564472198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,1,balanced,0.0990826686223348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,1,balanced,0.09938133756319682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,1,balanced,0.10387200117111206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,1,balanced,0.10393066207567851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,1,balanced,0.10743467013041179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,1,balanced,0.10732266306877136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,1,balanced,0.11136000355084737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,1,balanced,0.11607999602953593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,1,balanced,0.12025066216786702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,1,balanced,0.12571199735005698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,1,balanced,0.13593066732088724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,1,balanced,0.14146666725476584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,1,balanced,0.15195733308792114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,1,balanced,0.18093333641688028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,1,balanced,0.20018666982650757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,1,balanced,0.28405332565307617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,1,balanced,0.30980799595514935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,1,balanced,0.456773320833842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,1,balanced,0.5539466540018717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,1,balanced,0.8128853638966879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,1,balanced,1.0237600008646648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,1,balanced,1.5067626635233562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,1,balanced,1.9841920534769695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,64,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,64,balanced,0.029658667743206024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,64,balanced,0.02945599953333537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,64,balanced,0.03126933425664902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,64,balanced,0.03181333343187968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,64,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,64,balanced,0.03165333221356074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,64,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,64,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,64,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,64,balanced,0.034287999073664345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,64,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,64,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,64,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,64,balanced,0.03957333415746689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,64,balanced,0.03914133210976919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,64,balanced,0.039450667798519135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,64,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,64,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,64,balanced,0.04350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,64,balanced,0.04399466514587402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,64,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,64,balanced,0.05421333511670431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,64,balanced,0.07044800122578938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,64,balanced,0.08921066919962566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,64,balanced,0.12365333239237468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,64,balanced,0.15447466572125754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,128,power_law_1.2,0.045177599787712096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,128,power_law_1.2,0.043910399079322815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,128,power_law_1.2,0.04332799911499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,128,power_law_1.2,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,128,power_law_1.2,0.04838399887084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,128,power_law_1.2,0.05004159808158874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,128,power_law_1.2,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,128,power_law_1.2,0.05358719825744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,128,power_law_1.2,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,128,power_law_1.2,0.05374720096588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,128,power_law_1.2,0.0577023983001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,128,power_law_1.2,0.059564799070358276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,128,power_law_1.2,0.06292480230331421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,128,power_law_1.2,0.059468799829483034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,128,power_law_1.2,0.07257599830627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,128,power_law_1.2,0.07582719922065735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,128,power_law_1.2,0.07925760149955749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,128,power_law_1.2,0.09229440093040467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,128,power_law_1.2,0.10217599868774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,128,power_law_1.2,0.13153280019760133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,128,power_law_1.2,0.15551359653472902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,128,power_law_1.2,0.22001919746398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,128,power_law_1.2,0.28032000064849855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,128,power_law_1.2,0.37043840885162355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,128,power_law_1.2,0.5243328094482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,128,power_law_1.2,0.799071979522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,128,power_law_1.2,1.2259391784667968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,2,power_law_1.01,0.018323199450969697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,2,power_law_1.01,0.02106879949569702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,2,power_law_1.01,0.025088000297546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,2,power_law_1.01,0.030015999078750612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,2,power_law_1.01,0.0315775990486145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,2,power_law_1.01,0.040870401263237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,2,power_law_1.01,0.043609601259231565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,2,power_law_1.01,0.04402559995651245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,2,power_law_1.01,0.044012799859046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,2,power_law_1.01,0.04627839922904968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,2,power_law_1.01,0.04594559967517853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,2,power_law_1.01,0.047628799080848695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,2,power_law_1.01,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,2,power_law_1.01,0.04952319860458374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,2,power_law_1.01,0.05133439898490906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,2,power_law_1.01,0.05367680191993714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,2,power_law_1.01,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,2,power_law_1.01,0.06158080101013184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,2,power_law_1.01,0.08815360069274902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,2,power_law_1.01,0.11700479984283448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,2,power_law_1.01,0.10343040227890014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,2,power_law_1.01,0.11818239688873292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,2,power_law_1.01,0.14063359498977662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,2,power_law_1.01,0.18101119995117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,2,power_law_1.01,0.22097280025482177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,2,power_law_1.01,0.3161535978317261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,2,power_law_1.01,0.3981695890426636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,32,power_law_1.01,0.04580479860305786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,32,power_law_1.01,0.03681280016899109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,32,power_law_1.01,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,32,power_law_1.01,0.034329599142074584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,32,power_law_1.01,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,32,power_law_1.01,0.0369024008512497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,32,power_law_1.01,0.037215998768806456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,32,power_law_1.01,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,32,power_law_1.01,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,32,power_law_1.01,0.0416128009557724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,32,power_law_1.01,0.044710400700569156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,32,power_law_1.01,0.047366398572921756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,32,power_law_1.01,0.04930559992790222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,32,power_law_1.01,0.049747198820114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,32,power_law_1.01,0.05605760216712952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,32,power_law_1.01,0.05751680135726929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,32,power_law_1.01,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,32,power_law_1.01,0.069760000705719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,32,power_law_1.01,0.08004480004310607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,32,power_law_1.01,0.09664000272750854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,32,power_law_1.01,0.10980479717254639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,32,power_law_1.01,0.140447998046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,32,power_law_1.01,0.16517119407653807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,32,power_law_1.01,0.21742079257965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,32,power_law_1.01,0.293503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,32,power_law_1.01,0.4116032123565674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,32,power_law_1.01,0.5021632194519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,2,power_law_1.01,0.03954559862613678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,2,power_law_1.01,0.052211201190948485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,2,power_law_1.01,0.060140800476074216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,2,power_law_1.01,0.0753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,2,power_law_1.01,0.09267839789390564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,2,power_law_1.01,0.11562880277633666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,2,power_law_1.01,0.1189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,2,power_law_1.01,0.11967999935150146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,2,power_law_1.01,0.1243008017539978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,2,power_law_1.01,0.12565759420394898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,2,power_law_1.01,0.13147519826889037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,2,power_law_1.01,0.13472000360488892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,2,power_law_1.01,0.13596800565719605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,2,power_law_1.01,0.14241280555725097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,2,power_law_1.01,0.15358079671859742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,2,power_law_1.01,0.15905920267105103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,2,power_law_1.01,0.16386560201644898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,2,power_law_1.01,0.18416639566421508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,2,power_law_1.01,0.2073280096054077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,2,power_law_1.01,0.2578943967819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,2,power_law_1.01,0.2883008003234863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,2,power_law_1.01,0.39254400730133054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,2,power_law_1.01,0.4566080093383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,2,power_law_1.01,0.6795263767242432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,2,power_law_1.01,0.8318911552429199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,2,power_law_1.01,1.1728960037231446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,2,power_law_1.01,1.4370944023132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,2,power_law_1.2,0.02590720057487488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,2,power_law_1.2,0.026630398631095887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,2,power_law_1.2,0.024684800207614897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,2,power_law_1.2,0.02507520020008087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,2,power_law_1.2,0.026240000128746034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,2,power_law_1.2,0.031206399202346802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,2,power_law_1.2,0.04341759979724884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,2,power_law_1.2,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,2,power_law_1.2,0.05076479911804199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,2,power_law_1.2,0.0442111998796463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,2,power_law_1.2,0.04629760086536407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,2,power_law_1.2,0.0827455997467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,2,power_law_1.2,0.07742080092430115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,2,power_law_1.2,0.060550397634506224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,2,power_law_1.2,0.09329919815063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,2,power_law_1.2,0.07393280267715455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,2,power_law_1.2,0.09435520172119141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,2,power_law_1.2,0.11752320528030395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.2,0.14291839599609374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.2,0.1994879961013794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.2,0.2554624080657959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.2,0.3570175886154175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.2,0.46799359321594236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,16,power_law_1.01,0.05177599787712097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,16,power_law_1.01,0.06970239877700805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,16,power_law_1.01,0.05571200251579285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,16,power_law_1.01,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,16,power_law_1.01,0.05459200143814087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,16,power_law_1.01,0.05690240263938904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,16,power_law_1.01,0.05730559825897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,16,power_law_1.01,0.058361601829528806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,16,power_law_1.01,0.05902720093727112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,16,power_law_1.01,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,16,power_law_1.01,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,16,power_law_1.01,0.06459519863128663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,16,power_law_1.01,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,16,power_law_1.01,0.07300480008125305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,16,power_law_1.01,0.07901440262794494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,16,power_law_1.01,0.08542720079421998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,16,power_law_1.01,0.09439359903335572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,16,power_law_1.01,0.11017600297927857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,16,power_law_1.01,0.11760640144348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,16,power_law_1.01,0.1434880018234253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,16,power_law_1.01,0.16985599994659423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,16,power_law_1.01,0.20773119926452638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,16,power_law_1.01,0.2609791994094849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,16,power_law_1.01,0.3782272100448608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,16,power_law_1.01,0.4660543918609619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,16,power_law_1.01,0.6742656230926514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,16,power_law_1.01,0.8553983688354492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.2,0.675219202041626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,128,power_law_1.2,0.018111999332904815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,128,power_law_1.2,0.018700799345970152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,128,power_law_1.2,0.019411200284957887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,128,power_law_1.2,0.02195200026035309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.2,0.8841792106628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.2,1.3064512252807616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.2,1.7280704498291015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,2,balanced,0.052101333936055504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,2,balanced,0.06807999809583028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,2,balanced,0.09805867075920105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,2,balanced,0.1530933380126953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,2,balanced,0.2622879942258199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,2,balanced,0.32547734181086224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,2,balanced,0.33875731627146405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,2,balanced,0.3356320063273112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,2,balanced,0.34537065029144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,2,balanced,0.353770653406779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,2,balanced,0.3726773262023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,2,balanced,0.39050134023030597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,2,balanced,0.4110399881998698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,2,balanced,0.3869599898656209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,2,balanced,0.40460264682769775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,2,balanced,0.42562135060628253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,2,balanced,0.4314560095469157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,2,balanced,0.5929333368937174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,2,balanced,0.47501333554585773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,2,balanced,0.8354026476542155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,2,balanced,0.8167466322580973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,2,balanced,0.8441867033640543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,2,balanced,0.8243626753489176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,2,balanced,1.4909013112386067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,2,balanced,1.5719626744588215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,2,balanced,2.345285256703695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,2,balanced,2.9871946970621743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,8,power_law_1.01,0.08515200018882751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,8,power_law_1.01,0.11626880168914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,8,power_law_1.01,0.12377599477767945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,8,power_law_1.01,0.13653119802474975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,8,power_law_1.01,0.1377408027648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,8,power_law_1.01,0.1525887966156006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,8,power_law_1.01,0.15893759727478027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,8,power_law_1.01,0.1542464017868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,8,power_law_1.01,0.15999360084533693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,8,power_law_1.01,0.16441600322723388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,8,power_law_1.01,0.1587839961051941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,8,power_law_1.01,0.1657920002937317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,8,power_law_1.01,0.16406400203704835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,8,power_law_1.01,0.1657088041305542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,8,power_law_1.01,0.1777791976928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,8,power_law_1.01,0.17805440425872804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,8,power_law_1.01,0.18074239492416383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,8,power_law_1.01,0.20620160102844237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,8,power_law_1.01,0.21100161075592042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,8,power_law_1.01,0.24465279579162597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,8,power_law_1.01,0.2951488018035889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,8,power_law_1.01,0.3401535987854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,8,power_law_1.01,0.4296000003814697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,8,power_law_1.01,0.5879744052886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,8,power_law_1.01,0.7289535999298096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,8,power_law_1.01,1.1551808357238769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,8,power_law_1.01,1.3912447929382323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,4,power_law_1.2,0.04258559942245484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,4,power_law_1.2,0.041017600893974306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,4,power_law_1.2,0.04162560105323791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,4,power_law_1.2,0.04267520010471344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,4,power_law_1.2,0.04310399889945984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,4,power_law_1.2,0.04431999921798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,4,power_law_1.2,0.044614401459693906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,4,power_law_1.2,0.04425599873065948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,4,power_law_1.2,0.05014399886131286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,4,power_law_1.2,0.052767997980117796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,4,power_law_1.2,0.05237759947776795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,4,power_law_1.2,0.05621119737625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,4,power_law_1.2,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,4,power_law_1.2,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,4,power_law_1.2,0.07292159795761108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,4,power_law_1.2,0.07423359751701356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,4,power_law_1.2,0.08117759823799134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,4,power_law_1.2,0.09426559805870056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.2,0.10947200059890747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.2,0.12513920068740844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.2,0.15530240535736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.2,0.19745919704437256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.2,0.23759360313415528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,32,power_law_1.2,0.08678399920463561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,32,power_law_1.2,0.06087679862976074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,32,power_law_1.2,0.060761600732803345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,32,power_law_1.2,0.059468799829483034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,32,power_law_1.2,0.054502397775650024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,32,power_law_1.2,0.060550397634506224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,32,power_law_1.2,0.06149119734764099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,32,power_law_1.2,0.061587202548980716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,32,power_law_1.2,0.06283519864082336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,32,power_law_1.2,0.06394879817962647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,32,power_law_1.2,0.06709120273590088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,32,power_law_1.2,0.06681600213050842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,32,power_law_1.2,0.07079039812088013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,32,power_law_1.2,0.0757695972919464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,32,power_law_1.2,0.08085759878158569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,32,power_law_1.2,0.07998719811439514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,32,power_law_1.2,0.08755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,32,power_law_1.2,0.10090240240097045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,32,power_law_1.2,0.10451840162277222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,32,power_law_1.2,0.12732160091400146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,32,power_law_1.2,0.13349119424819947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,32,power_law_1.2,0.17619839906692505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,32,power_law_1.2,0.22576639652252198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,32,power_law_1.2,0.3195519924163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,32,power_law_1.2,0.3798079967498779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.2,0.3212991952896118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,32,power_law_1.2,0.5897664070129395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,32,power_law_1.2,0.6914303779602051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.2,0.4628352165222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,2,power_law_1.01,0.03648000061511993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,2,power_law_1.01,0.036320000886917114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,2,power_law_1.01,0.03656960129737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,2,power_law_1.01,0.04126079976558685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,2,power_law_1.01,0.041536000370979306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,2,power_law_1.01,0.04281600117683411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,2,power_law_1.01,0.04306559860706329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,2,power_law_1.01,0.04376960098743439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,2,power_law_1.01,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,2,power_law_1.01,0.05075200200080872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,2,power_law_1.01,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,16,4,power_law_1.01,0.0685375988483429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,2,power_law_1.01,0.05500800013542175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,16,4,power_law_1.01,0.0765887975692749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,16,4,power_law_1.01,0.08446720242500305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,16,4,power_law_1.01,0.09889280200004577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,16,4,power_law_1.01,0.1028864026069641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,16,4,power_law_1.01,0.10795520544052124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,16,4,power_law_1.01,0.12213120460510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,16,4,power_law_1.01,0.12247680425643921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,16,4,power_law_1.01,0.12011519670486451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,2,power_law_1.01,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,2,power_law_1.01,0.062431997060775755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,2,power_law_1.01,0.06821759939193725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,2,power_law_1.01,0.07162879705429077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,2,power_law_1.01,0.07827200293540955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,16,4,power_law_1.01,0.12288639545440674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,2,power_law_1.01,0.09352959990501404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,16,4,power_law_1.01,0.12249599695205689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,16,4,power_law_1.01,0.12572159767150878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,16,4,power_law_1.01,0.1283136010169983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.01,0.10154240131378174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,16,4,power_law_1.01,0.13401600122451782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,16,4,power_law_1.01,0.13942400217056275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,16,4,power_law_1.01,0.1415552020072937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,16,4,power_law_1.01,0.1478271961212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,16,4,power_law_1.01,0.1615615963935852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,16,4,power_law_1.01,0.17828480005264283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,16,4,power_law_1.01,0.22278399467468263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.01,0.1297727942466736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,16,4,power_law_1.01,0.26893439292907717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,16,4,power_law_1.01,0.3614784002304077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,16,4,power_law_1.01,0.42693119049072265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,16,4,power_law_1.01,0.5977344036102294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,16,4,power_law_1.01,0.7795328140258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.01,0.14175360202789306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,16,4,power_law_1.01,1.1438976287841798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,16,4,power_law_1.01,1.4802495956420898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.01,0.1894912004470825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.01,0.22430078983306884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.2,0.701151990890503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.01,0.3022336006164551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.01,0.4238719940185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.01,0.5324031829833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.01,0.7501247882843017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.2,0.8017600059509278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,8,power_law_1.2,0.030342400074005127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,8,power_law_1.2,0.02974080145359039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,8,power_law_1.2,0.02961919903755188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,8,power_law_1.2,0.032102400064468385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,8,power_law_1.2,0.03246079981327057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,8,power_law_1.2,0.03384959995746613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,8,power_law_1.2,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,8,power_law_1.2,0.03460479974746704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,8,power_law_1.2,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,8,power_law_1.2,0.03654400110244751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,8,power_law_1.2,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,8,power_law_1.2,0.04033919870853424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,8,power_law_1.2,0.0404992014169693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,8,power_law_1.2,0.04421760141849518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,8,power_law_1.2,0.050809597969055174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,8,power_law_1.2,0.05605760216712952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,8,power_law_1.2,0.06355839967727661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,8,power_law_1.2,0.07678719758987426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,8,power_law_1.2,0.09134079813957215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,8,power_law_1.2,0.1136512041091919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,8,power_law_1.2,0.152019202709198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,8,power_law_1.2,0.22620160579681398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,8,power_law_1.2,0.2884416103363037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,8,power_law_1.01,0.03253760039806366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,8,power_law_1.01,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,8,power_law_1.01,0.03349120020866394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,8,power_law_1.01,0.03543039858341217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,8,power_law_1.01,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,8,power_law_1.01,0.03678719997406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,8,power_law_1.01,0.03744640052318573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,8,power_law_1.01,0.0382207989692688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,8,power_law_1.01,0.04017919898033142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,8,power_law_1.01,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,8,power_law_1.01,0.04106239974498749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,8,power_law_1.01,0.04412800073623657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,8,power_law_1.01,0.04575360119342804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,8,power_law_1.01,0.052908802032470705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,8,power_law_1.01,0.06162559986114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,8,power_law_1.01,0.06401280164718628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,8,power_law_1.01,0.07074559926986694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,8,power_law_1.01,0.09426559805870056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,8,power_law_1.01,0.1128000020980835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,8,power_law_1.01,0.15412479639053345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,8,power_law_1.2,0.4261375904083252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,8,power_law_1.01,0.20467839241027833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,8,power_law_1.01,0.25781118869781494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,1,power_law_1.2,0.05644800066947937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,1,power_law_1.2,0.06650879979133606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,1,power_law_1.2,0.07249919772148132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,1,power_law_1.2,0.09473919868469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,1,power_law_1.2,0.09253119826316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,1,power_law_1.2,0.09397760033607483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,1,power_law_1.2,0.09224960207939148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,8,power_law_1.01,0.3457535982131958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,1,power_law_1.2,0.09553920030593872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,1,power_law_1.2,0.09182080030441284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,1,power_law_1.2,0.09360640048980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,1,power_law_1.2,0.09388160109519958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,1,power_law_1.2,0.09360640048980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,1,power_law_1.2,0.1032256007194519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,1,power_law_1.2,0.10799360275268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,1,power_law_1.2,0.12403839826583862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,1,power_law_1.2,0.121343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,8,power_law_1.2,0.5988927841186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,1,power_law_1.2,0.12547199726104735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,1,power_law_1.2,0.1397760033607483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,1,power_law_1.2,0.16810879707336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,1,power_law_1.2,0.21288321018218995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,1,power_law_1.2,0.23352959156036376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,1,power_law_1.2,0.3098304033279419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,8,power_law_1.01,0.5648960113525391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,1,power_law_1.2,0.3985599994659424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,8,power_law_1.2,0.9421759605407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,8,power_law_1.01,0.7964159965515136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,16,power_law_1.2,0.05069440007209778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,16,power_law_1.2,0.06865280270576476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,16,power_law_1.2,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,16,power_law_1.2,0.05505920052528381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,16,power_law_1.2,0.05681279897689819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,16,power_law_1.2,0.05754240155220032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,1,power_law_1.2,0.5449408054351806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,16,power_law_1.2,0.05754240155220032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,16,power_law_1.2,0.058739197254180905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,16,power_law_1.2,0.058873599767684935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,16,power_law_1.2,0.060070401430130003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,16,power_law_1.2,0.0637503981590271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,16,power_law_1.2,0.06870399713516236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,16,power_law_1.2,0.06888319849967957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,16,power_law_1.2,0.07392640113830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,16,power_law_1.2,0.08486400246620178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,16,power_law_1.2,0.08567039966583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,16,power_law_1.2,0.09440000057220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,16,power_law_1.2,0.11327999830245972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,16,power_law_1.2,0.11860480308532714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,16,power_law_1.2,0.1431040048599243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,16,power_law_1.2,0.1755136013031006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,16,power_law_1.2,0.22783360481262208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,16,power_law_1.2,0.2836352109909058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,16,power_law_1.2,0.3958591938018799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,16,power_law_1.2,0.5264063835144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,16,power_law_1.2,0.7058176040649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,16,power_law_1.2,0.9688896179199219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,8,power_law_1.01,1.0976448059082031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,8,power_law_1.01,1.5051712036132812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,1,power_law_1.2,0.7021312236785888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,8,power_law_1.2,1.2736703872680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,1,power_law_1.2,1.012441635131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,1,power_law_1.2,1.3070079803466796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,16,power_law_1.2,0.058911997079849246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,16,power_law_1.2,0.05867519974708557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,16,power_law_1.2,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,16,power_law_1.2,0.055315202474594115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,16,power_law_1.2,0.05235840082168579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,16,power_law_1.2,0.05576320290565491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,16,power_law_1.2,0.0586624026298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,16,power_law_1.2,0.06439039707183838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,16,power_law_1.2,0.05906559824943543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,16,power_law_1.2,0.0642304003238678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,16,power_law_1.2,0.07128319740295411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,16,power_law_1.2,0.07652480006217957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,16,power_law_1.2,0.0760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,16,power_law_1.2,0.12208000421524048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,16,power_law_1.2,0.13047679662704467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,16,power_law_1.2,0.14124159812927245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,16,power_law_1.2,0.16428799629211427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,16,power_law_1.2,0.19629440307617188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.2,0.2327104091644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.2,0.3450239896774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.2,0.17933440208435059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.2,0.21833600997924804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.2,0.29701120853424073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.2,0.4031871795654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.2,0.550387191772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.2,0.7644991874694824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.2,1.0020544052124023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,16,2,power_law_1.01,0.08457599878311158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,16,2,power_law_1.01,0.09224320054054261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,16,2,power_law_1.01,0.1140544056892395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,16,2,power_law_1.01,0.13054720163345337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,16,2,power_law_1.01,0.141484797000885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,16,2,power_law_1.01,0.14629119634628296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,16,2,power_law_1.01,0.16901760101318358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,16,2,power_law_1.01,0.17137919664382933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,16,2,power_law_1.01,0.17446399927139283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,16,2,power_law_1.01,0.1820799946784973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,16,2,power_law_1.01,0.18451839685440063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,16,2,power_law_1.01,0.1890496015548706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,16,2,power_law_1.01,0.19216639995574952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,16,2,power_law_1.01,0.2026815891265869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,16,2,power_law_1.01,0.20117759704589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,16,2,power_law_1.01,0.20787839889526366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,16,2,power_law_1.01,0.2202239990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,16,2,power_law_1.01,0.24447999000549317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,16,2,power_law_1.01,0.28107519149780275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,16,2,power_law_1.01,0.3441407918930054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,16,2,power_law_1.01,0.41353597640991213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,16,2,power_law_1.01,0.5288127899169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,16,2,power_law_1.01,0.6647744178771973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,16,2,power_law_1.01,0.9341567993164063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,16,2,power_law_1.01,1.1947263717651366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,16,2,power_law_1.01,1.7182720184326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,16,2,power_law_1.01,2.227654457092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,4,balanced,0.062122667829195656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,4,balanced,0.06825600067774455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,4,balanced,0.08989866574605306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,4,balanced,0.12339199582735698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,4,balanced,0.1822026570638021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,4,balanced,0.30131200949350995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,4,balanced,0.42253867785135907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,4,balanced,0.4289120038350423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,4,balanced,0.42641599973042804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,4,balanced,0.42901333173116046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,4,balanced,0.4329226811726888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,4,balanced,0.43563199043273926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,4,balanced,0.4389226833979289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,4,balanced,0.4424479802449544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,4,balanced,0.45085867245992023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,4,balanced,0.4553653399149577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,4,balanced,0.4656960169474284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,4,balanced,0.47869332631429035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,4,balanced,0.49397865931193036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,4,balanced,0.5264053344726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,4,balanced,0.55676798025767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,4,balanced,0.6315999825795492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,4,balanced,0.7076266606648763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,4,balanced,0.8394773006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,4,balanced,1.0702346960703533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,4,balanced,1.3253493309020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,4,balanced,1.7730132738749187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,4,power_law_1.01,0.01767680048942566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,4,power_law_1.01,0.01831679940223694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,4,power_law_1.01,0.0176704004406929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,4,power_law_1.01,0.018464000523090364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,4,power_law_1.01,0.01895039975643158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,4,power_law_1.01,0.020601600408554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,4,power_law_1.01,0.020339199900627138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,4,power_law_1.01,0.022540800273418427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,4,power_law_1.01,0.024607999622821806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,4,power_law_1.01,0.025747200846672057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,4,power_law_1.01,0.022284799814224245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,4,power_law_1.01,0.024153600633144378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,4,power_law_1.01,0.023904000222682954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,4,power_law_1.01,0.027980801463127137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,4,power_law_1.01,0.02993279993534088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,4,power_law_1.01,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,4,power_law_1.01,0.032441601157188416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,4,power_law_1.01,0.037062400579452516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,4,power_law_1.01,0.0517632007598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,4,power_law_1.01,0.05934079885482788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,4,power_law_1.01,0.05383679866790771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,4,power_law_1.01,0.06564480066299438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,4,power_law_1.01,0.0788096010684967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,4,power_law_1.01,0.10215679407119752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,4,power_law_1.01,0.12369920015335083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,4,power_law_1.01,0.16507519483566285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,4,power_law_1.01,0.2095871925354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,2,power_law_1.2,0.06603519916534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,2,power_law_1.2,0.0815999984741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,2,power_law_1.2,0.09134719967842102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,2,power_law_1.2,0.10990719795227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,2,power_law_1.2,0.12243200540542602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,2,power_law_1.2,0.1511680006980896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,2,power_law_1.2,0.19044480323791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,2,power_law_1.2,0.1975551962852478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,2,power_law_1.2,0.1994879961013794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,2,power_law_1.2,0.21089279651641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,2,power_law_1.2,0.20999679565429688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,2,power_law_1.2,0.2237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,2,power_law_1.2,0.22578558921813965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,2,power_law_1.2,0.22891519069671631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,2,power_law_1.2,0.24430079460144044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,2,power_law_1.2,0.2471679925918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,2,power_law_1.2,0.26959359645843506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,2,power_law_1.2,0.29577600955963135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,2,power_law_1.2,0.3225663900375366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,2,power_law_1.2,0.398144006729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,2,power_law_1.2,0.4291264057159424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,2,power_law_1.2,0.5232704162597657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,2,power_law_1.2,0.6328896045684814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,2,power_law_1.2,0.8767231941223145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,2,power_law_1.2,1.1018495559692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,2,power_law_1.2,1.4574784278869628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,2,power_law_1.2,1.9431808471679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,4,power_law_1.01,0.0843455970287323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,4,power_law_1.01,0.10489599704742432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,4,power_law_1.01,0.1200063943862915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,4,power_law_1.01,0.14043519496917725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,4,power_law_1.01,0.1624127984046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,4,power_law_1.01,0.18069759607315064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,4,power_law_1.01,0.20567679405212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,4,power_law_1.01,0.18706560134887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,4,power_law_1.01,0.19783040285110473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,4,power_law_1.01,0.194758403301239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,4,power_law_1.01,0.20062720775604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,4,power_law_1.01,0.21283841133117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,4,power_law_1.01,0.22049920558929442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,4,power_law_1.01,0.21484160423278809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,4,power_law_1.01,0.21604480743408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,2,balanced,0.04072533299525579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,2,balanced,0.055626665552457176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,2,balanced,0.07788800199826558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,2,balanced,0.1220693290233612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,2,balanced,0.20805867513020834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,2,balanced,0.21167999505996704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,2,balanced,0.21137599150339761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,2,balanced,0.2114773392677307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,2,balanced,0.2137333353360494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,2,balanced,0.21363200743993124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,2,balanced,0.21649066607157388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,2,balanced,0.21836266915003458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,2,balanced,0.2221013307571411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,2,balanced,0.22673600912094116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,2,balanced,0.23286932706832886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,2,balanced,0.23676800727844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,2,balanced,0.24382932980855307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,2,balanced,0.26447999477386475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,2,balanced,0.2816266616185506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,2,balanced,0.31648532549540204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,2,balanced,0.3476053476333618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,2,balanced,0.4458933273951213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,2,balanced,0.5080960194269816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,2,balanced,0.7186400095621744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,2,balanced,0.8546826839447021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,2,balanced,1.2117066383361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,2,balanced,1.59225066502889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,32,power_law_1.01,0.05633919835090637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,32,power_law_1.01,0.055379199981689456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,32,power_law_1.01,0.05551360249519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,32,power_law_1.01,0.05511040091514587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,32,power_law_1.01,0.05554559826850891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,32,power_law_1.01,0.05556480288505554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,32,power_law_1.01,0.05619840025901794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,32,power_law_1.01,0.057715201377868654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,32,power_law_1.01,0.057094401121139525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,32,power_law_1.01,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,32,power_law_1.01,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,32,power_law_1.01,0.0604095995426178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,32,power_law_1.01,0.06116480231285095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,32,power_law_1.01,0.062457597255706786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,32,power_law_1.01,0.0673919975757599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,32,power_law_1.01,0.06972799897193908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,32,power_law_1.01,0.07468799948692321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,8,power_law_1.01,0.10319360494613647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,32,power_law_1.01,0.08280320167541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,8,power_law_1.01,0.15156480073928832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,8,power_law_1.01,0.13237119913101197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,8,power_law_1.01,0.1475391983985901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,8,power_law_1.01,0.1512768030166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,8,power_law_1.01,0.15539200305938722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,8,power_law_1.01,0.16498559713363647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,8,power_law_1.01,0.17894400358200074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.01,0.09174399971961975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.01,0.11237759590148926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.01,0.12382080554962158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.01,0.1557760000228882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.01,0.1988479971885681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,8,power_law_1.01,0.17278720140457154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.01,0.24951679706573487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,8,power_law_1.01,0.1735551953315735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.01,0.3203200101852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,8,power_law_1.01,0.1769215941429138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,8,power_law_1.01,0.178220796585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.01,0.4701568126678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,8,power_law_1.01,0.18833279609680176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,8,power_law_1.01,0.19680000543594361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,8,power_law_1.01,0.20490880012512208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,8,power_law_1.01,0.21577599048614501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,8,power_law_1.01,0.22736639976501466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,8,power_law_1.01,0.2652928113937378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,8,power_law_1.01,0.2761087894439697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,8,power_law_1.01,0.32816638946533205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.01,0.5849855899810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,8,power_law_1.01,0.3759552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,8,power_law_1.01,0.47832322120666504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,8,power_law_1.01,0.49436798095703127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,8,power_law_1.01,0.6497983932495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,8,power_law_1.01,0.8782527923583985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,8,power_law_1.01,1.2143487930297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,8,power_law_1.01,1.4917183876037599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,4,power_law_1.2,0.09340159893035889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,4,power_law_1.2,0.10398720502853394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,4,power_law_1.2,0.10881919860839843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,4,power_law_1.2,0.1374079942703247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,4,power_law_1.2,0.14342399835586547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,4,power_law_1.2,0.16499840021133422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,4,power_law_1.2,0.18883199691772462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,4,power_law_1.2,0.19900159835815429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,4,power_law_1.2,0.2030656099319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,4,power_law_1.2,0.19714560508728027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,4,power_law_1.2,0.18743040561676025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,4,power_law_1.2,0.20704638957977295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,4,power_law_1.2,0.21230719089508057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,4,power_law_1.2,0.21132800579071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,4,power_law_1.2,0.2248768091201782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,4,power_law_1.2,0.22421119213104249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,4,power_law_1.2,0.24157440662384033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,4,power_law_1.2,0.2504256010055542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,4,power_law_1.2,0.2778752088546753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,4,power_law_1.2,0.33765759468078616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,4,power_law_1.2,0.3950655937194824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,4,power_law_1.2,0.4992703914642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,4,power_law_1.2,0.63024001121521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,4,power_law_1.2,0.9142848014831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,4,power_law_1.2,1.1488832473754882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,4,power_law_1.2,1.6050687789916993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,4,power_law_1.2,2.1710975646972654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,64,balanced,0.03583466758330663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,64,balanced,0.03806400050719579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,64,balanced,0.03427733232577642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,64,balanced,0.03409600009520849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,64,balanced,0.03842666745185852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,64,balanced,0.03790933390458425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,64,balanced,0.03734933336575826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,64,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,64,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,64,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,64,balanced,0.03824000060558319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,64,balanced,0.03737599899371465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,64,balanced,0.03810133288304011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,64,balanced,0.03989866624275843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,64,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,64,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,64,balanced,0.04394666850566864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,64,balanced,0.044218664367993675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,64,balanced,0.048154667019844055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,64,balanced,0.05190933247407278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,64,balanced,0.0637600024541219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,64,balanced,0.06711466610431671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,64,balanced,0.06923733154932658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,64,balanced,0.07977599898974101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,64,balanced,0.09275199969609578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,64,balanced,0.10744532942771912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,64,balanced,0.14056533575057983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,2,power_law_1.2,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,2,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,2,power_law_1.2,0.05421440005302429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,2,power_law_1.2,0.06475520133972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,2,power_law_1.2,0.06361600160598754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,2,power_law_1.2,0.06682239770889283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,2,power_law_1.2,0.06764159798622131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,2,power_law_1.2,0.0677183985710144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,2,power_law_1.2,0.06824319958686828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,2,power_law_1.2,0.07000960111618042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,2,power_law_1.2,0.07107840180397033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,2,power_law_1.2,0.07100160121917724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,2,power_law_1.2,0.07575680017471313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,2,power_law_1.2,0.07701119780540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,2,power_law_1.2,0.08698880076408386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,2,power_law_1.2,0.08907520174980163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,2,power_law_1.2,0.10741759538650512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,2,power_law_1.2,0.11976319551467896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.2,0.1276927947998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.2,0.1601855993270874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.2,0.17920000553131105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.2,0.2504319906234741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.2,0.30098559856414797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.2,0.4443840026855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.2,0.6413375854492187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.2,0.831123161315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,8,balanced,0.02696000039577484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,8,balanced,0.027477333943049114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,8,balanced,0.03257066756486893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,8,balanced,0.04391466577847799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,8,balanced,0.05659733215967814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,8,balanced,0.06515199939409892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,8,balanced,0.07502933343251546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.2,0.9886655807495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,8,balanced,0.07229333122571309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,8,balanced,0.07823466757933299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,8,balanced,0.10427199800809224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,8,balanced,0.09528000156084697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,8,balanced,0.1029919981956482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,8,balanced,0.1388320028781891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,8,balanced,0.1402186652024587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,8,balanced,0.14620266358057657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,8,balanced,0.14323733250300089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,8,balanced,0.13948266704877219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,8,balanced,0.15426133076349893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,8,balanced,0.16402133305867514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,8,balanced,0.19149333238601685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,8,balanced,0.21128533283869425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,8,balanced,0.2586453358332316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,8,balanced,0.29476267099380493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,8,balanced,0.4413813352584839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,8,balanced,0.5278826554616293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,8,balanced,0.7767039934794108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,8,balanced,1.0164533456166585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,8,power_law_1.01,0.05848960280418396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,8,power_law_1.01,0.06389759778976441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,8,power_law_1.01,0.06821119785308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,8,power_law_1.01,0.06842880249023438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,8,power_law_1.01,0.07018880248069763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,8,power_law_1.01,0.07159039974212647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,8,power_law_1.01,0.07247359752655029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,8,power_law_1.01,0.07406719923019409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,8,power_law_1.01,0.0732479989528656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,8,power_law_1.01,0.07630720138549804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,8,power_law_1.01,0.07685760259628296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,8,power_law_1.01,0.07919999957084656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,8,power_law_1.01,0.08223360180854797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,8,power_law_1.01,0.08330879807472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,8,power_law_1.01,0.08852480053901672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,8,power_law_1.01,0.09127039909362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,8,power_law_1.01,0.10105600357055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,8,power_law_1.01,0.11534080505371094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,8,power_law_1.01,0.1344256043434143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,8,power_law_1.01,0.1646656036376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,8,power_law_1.01,0.1729472041130066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,8,power_law_1.01,0.21262080669403077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,8,power_law_1.01,0.27781119346618655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,8,power_law_1.01,0.37502079010009765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,8,power_law_1.01,0.48850560188293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,8,power_law_1.01,0.7410560131072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,8,power_law_1.01,0.9217472076416016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,32,2,balanced,0.04762133459250132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,32,2,balanced,0.04804799954096476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,32,2,balanced,0.05199466645717621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,32,2,balanced,0.05807466804981232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,32,2,balanced,0.08649599552154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,32,2,balanced,0.12687466541926065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,32,2,balanced,0.13179199894269308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,32,2,balanced,0.12217066685358684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,32,2,balanced,0.12523733576138815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,32,2,balanced,0.12297067046165466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,32,2,balanced,0.12638399998346964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,32,2,balanced,0.1295253336429596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,32,2,balanced,0.1312320033709208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,32,2,balanced,0.13403733571370444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,32,2,balanced,0.1399893363316854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,32,2,balanced,0.1441439986228943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,32,2,balanced,0.15371200442314148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,32,2,balanced,0.17148266235987344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,32,2,balanced,0.18489599227905273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,32,2,balanced,0.22260266542434692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,32,2,balanced,0.25413332382837933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,32,2,balanced,0.35492801666259766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,32,2,balanced,0.40743998686472577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,32,2,balanced,0.595301349957784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,32,2,balanced,0.7181066672007242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,32,2,balanced,1.0500106811523438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,32,2,balanced,1.365664005279541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,64,power_law_1.01,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,64,power_law_1.01,0.0403903990983963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,64,power_law_1.01,0.03948799967765808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,64,power_law_1.01,0.03903360068798065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,64,power_law_1.01,0.039552000164985654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,64,power_law_1.01,0.04196479916572571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,64,power_law_1.01,0.043136000633239746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,64,power_law_1.01,0.043007999658584595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,64,power_law_1.01,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,64,power_law_1.01,0.04394879937171936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,64,power_law_1.01,0.047577598690986635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,64,power_law_1.01,0.050963199138641356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,64,power_law_1.01,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,64,power_law_1.01,0.04798080027103424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,64,power_law_1.01,0.052211201190948485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,64,power_law_1.01,0.05214080214500427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,64,power_law_1.01,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,64,power_law_1.01,0.06328319907188415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,64,power_law_1.01,0.06615039706230164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,64,power_law_1.01,0.07786880135536194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,64,power_law_1.01,0.08912640213966369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,64,power_law_1.01,0.1086016058921814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,64,power_law_1.01,0.12428159713745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,64,power_law_1.01,0.16286079883575438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,64,power_law_1.01,0.2010432004928589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,64,power_law_1.01,0.2863744020462036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,64,power_law_1.01,0.3729664087295532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,128,power_law_1.01,0.057868802547454835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,128,power_law_1.01,0.05591040253639221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,128,power_law_1.01,0.056595200300216676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,128,power_law_1.01,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,128,power_law_1.01,0.058329600095748904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,128,power_law_1.01,0.05802239775657654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,128,power_law_1.01,0.058534401655197146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,128,power_law_1.01,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,128,power_law_1.01,0.06107519865036011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,128,power_law_1.01,0.06430720090866089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,128,power_law_1.01,0.06638079881668091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,128,power_law_1.01,0.07088639736175537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,128,power_law_1.01,0.07491199970245362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,128,power_law_1.01,0.0760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,128,power_law_1.01,0.08503680229187012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,128,power_law_1.01,0.08789759874343872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,128,power_law_1.01,0.09390079975128174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,128,power_law_1.01,0.11006720066070556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,128,power_law_1.01,0.11891839504241944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,128,power_law_1.01,0.14794880151748657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,128,power_law_1.01,0.16503679752349854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,128,power_law_1.01,0.21754879951477052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,128,power_law_1.01,0.2235647916793823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,128,power_law_1.01,0.29783039093017577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,128,power_law_1.01,0.35078399181365966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,128,power_law_1.01,0.5076288223266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,128,power_law_1.01,0.6176640033721924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,1,power_law_1.01,0.027692800760269164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,1,power_law_1.01,0.04862079918384552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,1,power_law_1.01,0.0754368007183075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,1,power_law_1.01,0.11907839775085449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,1,power_law_1.01,0.17123199701309205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,1,power_law_1.01,0.23343360424041748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,1,power_law_1.01,0.3093120098114014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,1,power_law_1.01,0.32644479274749755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,1,power_law_1.01,0.3529088020324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,1,power_law_1.01,0.34399359226226806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,1,power_law_1.01,0.36135680675506593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,1,power_law_1.01,0.3804415941238403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,1,power_law_1.01,0.38646399974823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,1,power_law_1.01,0.3883455991744995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,1,power_law_1.01,0.4026175975799561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,1,power_law_1.01,0.41364479064941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,1,power_law_1.01,0.4340991973876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,1,power_law_1.01,0.4786687850952148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.01,0.49740800857543943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.01,0.5988927841186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.01,0.6639423847198487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.01,0.87838716506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.01,0.813702392578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.01,1.0450112342834472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.01,1.302131175994873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.01,1.7787519454956056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.01,2.23624324798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,128,power_law_1.2,0.025996801257133485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,128,power_law_1.2,0.025798401236534117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,128,power_law_1.2,0.026495999097824095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,128,power_law_1.2,0.028275200724601747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,128,power_law_1.2,0.028863999247550964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,128,power_law_1.2,0.02885119915008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,128,power_law_1.2,0.029286399483680725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,128,power_law_1.2,0.029811200499534608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,128,power_law_1.2,0.030124801397323608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,128,power_law_1.2,0.03086079955101013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,128,power_law_1.2,0.03160960078239441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,128,power_law_1.2,0.035667198896408084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,128,power_law_1.2,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,128,power_law_1.2,0.03798399865627289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,128,power_law_1.2,0.04166400134563446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,128,power_law_1.2,0.04289279878139496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,128,power_law_1.2,0.049107199907302855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,128,power_law_1.2,0.057132798433303836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,128,power_law_1.2,0.06828799843788147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,128,power_law_1.2,0.09031680226325989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,128,power_law_1.2,0.10958720445632934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,128,power_law_1.2,0.14819200038909913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,128,power_law_1.2,0.20103039741516113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,128,power_law_1.2,0.30448639392852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,128,power_law_1.2,0.40446081161499026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,128,power_law_1.2,0.6022463798522949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,128,power_law_1.2,0.8567680358886719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,2,power_law_1.01,0.06830719709396363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,2,power_law_1.01,0.07420160174369812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,2,power_law_1.01,0.08058239817619324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,2,power_law_1.01,0.09378560185432434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,2,power_law_1.01,0.10748159885406494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,2,power_law_1.01,0.11626880168914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,2,power_law_1.01,0.14033279418945313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,2,power_law_1.01,0.14095360040664673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,2,power_law_1.01,0.14936959743499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,2,power_law_1.01,0.1459264039993286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,2,power_law_1.01,0.15007359981536866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,2,power_law_1.01,0.15648640394210817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,2,power_law_1.01,0.15987839698791503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,2,power_law_1.01,0.16487679481506348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,2,power_law_1.01,0.17311359643936158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,2,power_law_1.01,0.17630720138549805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,2,power_law_1.01,0.18569600582122803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,2,power_law_1.01,0.2082047939300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,2,power_law_1.01,0.23712000846862794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,2,power_law_1.01,0.29027841091156004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,2,power_law_1.01,0.34910080432891843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,2,power_law_1.01,0.4573184013366699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,2,power_law_1.01,0.5673408031463623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,2,power_law_1.01,0.7705599784851074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,2,power_law_1.01,0.977996826171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,2,power_law_1.01,1.4281791687011718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,2,power_law_1.01,1.9417024612426759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,2,power_law_1.2,0.09507840275764465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,2,power_law_1.2,0.11043200492858887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,2,power_law_1.2,0.1103935956954956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,2,power_law_1.2,0.15301120281219482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,2,power_law_1.2,0.15868799686431884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,2,power_law_1.2,0.15940480232238768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,2,power_law_1.2,0.15491199493408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,2,power_law_1.2,0.1607807993888855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,2,power_law_1.2,0.16038399934768677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,2,power_law_1.2,0.16136959791183472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,2,power_law_1.2,0.16170239448547363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,2,power_law_1.2,0.1767359972000122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,2,power_law_1.2,0.18691200017929077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,2,power_law_1.2,0.20072319507598876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,2,power_law_1.2,0.21436800956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,2,power_law_1.2,0.2240000009536743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,2,power_law_1.2,0.23764479160308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,2,power_law_1.2,0.3156991958618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,2,power_law_1.2,0.3613055944442749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,2,power_law_1.2,0.4562943935394287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,2,power_law_1.2,0.6049536228179931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,2,power_law_1.2,0.8189760208129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,2,power_law_1.2,1.1562175750732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,2,power_law_1.2,1.5087295532226563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,2,power_law_1.2,1.9643072128295898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,2,balanced,0.06655466556549072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,2,balanced,0.07443733513355255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,2,balanced,0.08749866485595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,2,balanced,0.11281067132949829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,2,balanced,0.16739199558893839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,2,balanced,0.16916799545288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,2,balanced,0.1697546641031901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,2,balanced,0.16915200153986612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,2,balanced,0.17230933904647827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,2,balanced,0.1734559933344523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,2,balanced,0.17218667268753052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,2,balanced,0.18093333641688028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,2,balanced,0.17946134010950723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,2,balanced,0.18242667118708292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,2,balanced,0.18618667125701904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,2,balanced,0.18929600715637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,2,balanced,0.1946880022684733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,2,balanced,0.20907199382781982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,2,balanced,0.22263999780019125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,2,balanced,0.2548426588376363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,2,balanced,0.2851840058962504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,2,balanced,0.4045013189315796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,2,balanced,0.4702186584472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,4,power_law_1.01,0.09966719746589661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,2,balanced,0.6563839912414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,4,power_law_1.01,0.15032960176467897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,2,balanced,0.8586826324462891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,4,power_law_1.01,0.2531840085983276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,2,balanced,1.234218676884969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,4,power_law_1.01,0.24739840030670165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,2,balanced,1.6282347043355305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,4,power_law_1.01,0.3462591886520386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,4,power_law_1.01,0.421676778793335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,4,power_law_1.01,0.44544639587402346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,4,power_law_1.01,0.5164671897888183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,4,power_law_1.01,0.5076799869537354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,4,power_law_1.01,0.4814784049987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,4,power_law_1.01,0.5521664142608642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,4,power_law_1.01,0.5734272003173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,4,power_law_1.01,0.6007040023803711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,4,power_law_1.01,0.5500864028930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,4,power_law_1.01,0.5607232093811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,4,power_law_1.01,0.5637375831604003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,4,power_law_1.01,0.6344064235687256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,4,power_law_1.01,0.6617663860321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,4,power_law_1.01,0.7274943828582764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,2,power_law_1.2,2.597433662414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,4,power_law_1.01,0.8125439643859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,4,power_law_1.01,0.8172224044799805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,4,power_law_1.01,1.005337619781494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,4,power_law_1.01,1.0614208221435546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,4,power_law_1.01,1.2677824020385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,4,power_law_1.01,1.5550399780273438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,4,power_law_1.01,2.0037439346313475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,4,power_law_1.01,2.573075294494629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,8,power_law_1.2,0.08067839741706848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,8,power_law_1.2,0.08332800269126892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,8,balanced,0.04390933116277059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,8,balanced,0.04242133100827535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,8,balanced,0.04373333354791006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,8,balanced,0.04599999884764353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,8,balanced,0.04789866507053375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,8,balanced,0.048309331138928734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,8,balanced,0.04811733464399973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,8,balanced,0.04868799944718679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,8,balanced,0.054511999090512596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,8,balanced,0.05605333546797434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,8,balanced,0.056287998954455055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,8,balanced,0.0521919975678126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,8,balanced,0.05121066669623057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,8,power_law_1.2,0.08480640053749085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,8,balanced,0.058304001887639366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,8,balanced,0.06844266752401988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,8,balanced,0.07677866518497467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,8,balanced,0.0683840016523997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,8,balanced,0.08301866551240285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,8,balanced,0.08171733220418294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,8,balanced,0.10531199971834819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,8,balanced,0.11141332983970642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,8,balanced,0.14006933569908142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,8,power_law_1.2,0.07859839797019959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,8,balanced,0.18626666069030762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,8,balanced,0.24860799312591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,8,balanced,0.3135733405749003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,8,balanced,0.4597546656926473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,8,balanced,0.5939893325169882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,8,power_law_1.2,0.08152959942817688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,8,power_law_1.2,0.07955840229988098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,8,power_law_1.2,0.08823680281639099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,8,power_law_1.2,0.08979200124740601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,8,power_law_1.2,0.0933568000793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,8,power_law_1.2,0.09657599925994872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,8,power_law_1.2,0.0944703996181488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,8,power_law_1.2,0.10461440086364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,8,power_law_1.2,0.10359679460525513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,8,power_law_1.2,0.1062656044960022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,8,power_law_1.2,0.14803199768066405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,8,power_law_1.2,0.14439040422439575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,8,power_law_1.2,0.1586624026298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,8,power_law_1.2,0.21615359783172608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,8,power_law_1.2,0.2610752105712891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,8,power_law_1.2,0.3587071895599365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,2,power_law_1.2,3.6613185882568358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,8,power_law_1.2,0.3637183904647827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,8,power_law_1.2,0.5736576080322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,8,power_law_1.2,0.6749311923980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,8,power_law_1.2,1.069536018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,8,power_law_1.2,1.4113792419433593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,2,balanced,0.027552001178264618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,2,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,2,balanced,0.05099200208981832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,2,balanced,0.0763626645008723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,2,balanced,0.11892799536387126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,2,balanced,0.1209333340326945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,2,balanced,0.12056000034014384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,2,balanced,0.1223520040512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,2,balanced,0.12244266271591187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,2,balanced,0.1216373344262441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,2,balanced,0.12164266904195149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,2,balanced,0.12468266487121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,2,balanced,0.12617066502571106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,2,balanced,0.12800533572832742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,2,balanced,0.133242666721344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,2,balanced,0.13369066516558328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,2,balanced,0.14005866646766663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,2,balanced,0.1469386617342631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,2,balanced,0.15642133355140686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,2,balanced,0.18023999532063803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,2,balanced,0.19071465730667114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,2,balanced,0.23531200488408408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,2,balanced,0.2681173284848531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,2,balanced,0.3685866594314575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,2,balanced,0.43296531836191815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,2,balanced,0.6024853388468424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,8,power_law_1.2,2.190662384033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,2,balanced,0.7813546657562256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,128,balanced,0.05377600093682607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,128,balanced,0.06011199951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,128,balanced,0.048101335763931274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,128,balanced,0.044031997521718345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,128,balanced,0.0467199981212616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,128,balanced,0.04985600213209788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,128,balanced,0.05046399931112925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,128,balanced,0.051882664362589516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,128,balanced,0.05208533505598704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,128,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,128,balanced,0.05420800050099691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,128,balanced,0.05568000177542368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,128,balanced,0.05407999952634176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,128,balanced,0.056976000467936196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,128,balanced,0.062074666221936546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,128,balanced,0.06241600215435028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,128,balanced,0.0672106643517812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,128,balanced,0.0790880024433136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,128,balanced,0.08868799606959026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,128,balanced,0.10780800382296245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,128,balanced,0.12034666538238525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,128,balanced,0.15107199549674988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,128,balanced,0.17988266547520956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,128,balanced,0.23503466447194418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,128,balanced,0.2927466630935669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,128,balanced,0.40668265024820965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,128,balanced,0.5137866735458374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,8,power_law_1.2,2.81341438293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,32,power_law_1.2,0.07488639950752259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,32,power_law_1.2,0.07027199864387512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,32,power_law_1.2,0.0686079978942871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,32,power_law_1.2,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,32,power_law_1.2,0.07153919935226441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,32,power_law_1.2,0.06568319797515869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,32,power_law_1.2,0.06939520239830017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,32,power_law_1.2,0.07064319849014282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,32,power_law_1.2,0.07041919827461243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,32,power_law_1.2,0.07260800004005433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,32,power_law_1.2,0.07124480009078979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,32,power_law_1.2,0.07228800058364868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,32,power_law_1.2,0.07276160120964051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,32,power_law_1.2,0.07195519804954528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,32,power_law_1.2,0.07763199806213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,32,power_law_1.2,0.078847998380661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,32,power_law_1.2,0.08540800213813782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,32,power_law_1.2,0.09829760193824769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,32,power_law_1.2,0.10736000537872314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,32,power_law_1.2,0.1268671989440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,32,power_law_1.2,0.1424512028694153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,32,power_law_1.2,0.18094079494476317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,32,power_law_1.2,0.2205631971359253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,32,power_law_1.2,0.3017215967178345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,32,power_law_1.2,0.38188159465789795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,32,power_law_1.2,0.5694655895233154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,32,power_law_1.2,0.7313151836395264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,2,power_law_1.2,0.06185600161552429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,2,power_law_1.2,0.06846079826354981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,2,power_law_1.2,0.07657600045204163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,2,power_law_1.2,0.09159680008888245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,2,power_law_1.2,0.10309760570526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,2,power_law_1.2,0.11432960033416747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,2,power_law_1.2,0.14052480459213257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,2,power_law_1.2,0.14587520360946654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,2,power_law_1.2,0.14826879501342774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,2,power_law_1.2,0.15386879444122314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,2,power_law_1.2,0.15243519544601442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,2,power_law_1.2,0.16035840511322022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,2,power_law_1.2,0.16378240585327147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,2,power_law_1.2,0.17189760208129884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,2,power_law_1.2,0.1851456046104431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,2,power_law_1.2,0.1896064043045044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,2,power_law_1.2,0.19838080406188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,2,power_law_1.2,0.22021760940551757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,2,power_law_1.2,0.2403007984161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,2,power_law_1.2,0.29358720779418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,2,power_law_1.2,0.31601920127868655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,2,power_law_1.2,0.41957759857177734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,2,power_law_1.2,0.5018752098083497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,2,power_law_1.2,0.6671040058135986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,2,power_law_1.2,0.8960384368896485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,2,power_law_1.2,1.2741312026977538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,2,power_law_1.2,1.545695972442627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,4,power_law_1.2,0.08782079815864563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,4,power_law_1.2,0.12844159603118896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,4,power_law_1.2,0.15491199493408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,4,power_law_1.2,0.17861759662628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,4,power_law_1.2,0.23678081035614013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,4,power_law_1.2,0.2659199953079224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,4,power_law_1.2,0.3125695943832397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,4,power_law_1.2,0.3680448055267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,4,power_law_1.2,0.3083199977874756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,4,power_law_1.2,0.30989439487457277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,4,power_law_1.2,0.3351999998092651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,4,power_law_1.2,0.3451904058456421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,4,power_law_1.2,0.354803204536438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,4,power_law_1.2,0.36258559226989745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,4,power_law_1.2,0.39232640266418456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,4,power_law_1.2,0.39054079055786134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,4,power_law_1.2,0.4379583835601807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,4,power_law_1.2,0.482374382019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,4,power_law_1.2,0.48818559646606446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,4,power_law_1.2,0.5661695957183838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,4,power_law_1.2,0.5979135990142822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,4,power_law_1.2,0.6970111846923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,4,power_law_1.2,0.8398143768310546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,4,power_law_1.2,1.0597951889038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,4,power_law_1.2,1.2624128341674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,4,power_law_1.2,1.769536018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,4,power_law_1.2,2.046873664855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,32,4,balanced,0.021173333128293354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,32,4,balanced,0.021269333859284718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,32,4,balanced,0.023077333966890972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,32,4,balanced,0.027141332626342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,32,4,balanced,0.029215998947620392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,32,4,balanced,0.03262399882078171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,32,4,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,32,4,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,32,4,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,32,4,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,32,4,balanced,0.03949866692225138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,32,4,balanced,0.04081599911053976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,32,4,balanced,0.0425600012143453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,32,4,balanced,0.04571733375390371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,32,4,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,32,4,balanced,0.050698667764663696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,32,4,balanced,0.06878399848937988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,32,4,balanced,0.07955733438332875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,32,4,balanced,0.11370133360226949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,32,4,balanced,0.1250933309396108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,32,4,balanced,0.14289066195487976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,32,4,balanced,0.17357333501180014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,32,4,balanced,0.19566933314005533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,32,4,balanced,0.24683733781178793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,32,4,balanced,0.2887199918429057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,32,4,balanced,0.4440319935480754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,32,4,balanced,0.5431679884592692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,1,balanced,0.058133333921432495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,1,balanced,0.07746133208274841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,1,balanced,0.10905067125956218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,1,balanced,0.16113600134849548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,1,balanced,0.26910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,16,balanced,0.06286400059858958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,16,balanced,0.06227200229962667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,16,balanced,0.06410133341948192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,16,balanced,0.07231466472148895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,1,balanced,0.3243839939435323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,16,balanced,0.08495466907819112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,16,balanced,0.10085333387056987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,16,balanced,0.1146506667137146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,16,balanced,0.11515200138092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,16,balanced,0.11774933338165283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,16,balanced,0.11555733283360799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,16,balanced,0.11762133240699768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,16,balanced,0.12032000223795573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,16,balanced,0.12039466698964436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,16,balanced,0.12185066938400269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,1,balanced,0.3348906834920247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,16,balanced,0.1322719951470693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,1,balanced,0.3366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,16,balanced,0.13242666920026144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,1,balanced,0.33860798676808673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,16,balanced,0.13809600472450256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,1,balanced,0.34008534749348956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,1,balanced,0.3424533208211263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,16,power_law_1.2,0.03617280125617981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,1,balanced,0.349178671836853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,16,power_law_1.2,0.03246079981327057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,1,balanced,0.3535573482513428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,1,balanced,0.36353600025177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,16,power_law_1.2,0.03134720027446747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,1,balanced,0.369429349899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,16,power_law_1.2,0.029363200068473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,16,power_law_1.2,0.030399999022483824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,16,power_law_1.2,0.03287039995193482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,16,balanced,0.14908799529075623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,16,power_law_1.2,0.032627201080322264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,16,balanced,0.15599999825159708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,16,balanced,0.1705013314882914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,16,balanced,0.1904319922129313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,16,balanced,0.22931732734044394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,16,balanced,0.2664213379224141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,16,balanced,0.3463840087254842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,16,balanced,0.4543946584065755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,16,balanced,0.6228053172429403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,16,power_law_1.2,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,16,balanced,0.8197440306345621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,16,power_law_1.2,0.03269760012626648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,16,power_law_1.2,0.03326080143451691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,1,balanced,0.3792266845703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,1,balanced,0.3962666591008504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,16,power_law_1.2,0.03489919900894165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,1,balanced,0.46087467670440674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,16,power_law_1.2,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,16,power_law_1.2,0.03624959886074066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,16,power_law_1.2,0.03766399919986725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,16,power_law_1.2,0.04069760143756866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,16,power_law_1.2,0.042828801274299624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,1,balanced,0.46426133314768475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,1,balanced,0.6074239810307821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,1,balanced,0.6404373248418173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,1,balanced,0.9931413332621256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,1,balanced,0.9868906339009603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,1,balanced,1.4746400515238445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,16,power_law_1.2,0.04405759871006012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,16,power_law_1.2,0.04909439980983734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,1,balanced,1.7415733337402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,16,power_law_1.2,0.054662400484085084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,16,power_law_1.2,0.06447359919548035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,1,balanced,2.611738681793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,16,power_law_1.2,0.07400320172309875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,1,balanced,3.2854719161987305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,16,power_law_1.2,0.08777599930763244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,16,power_law_1.2,0.11762559413909912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,16,power_law_1.2,0.17138559818267823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,2,power_law_1.01,0.03736959993839264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,2,power_law_1.01,0.03711360096931458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,2,power_law_1.01,0.03976320028305054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,2,power_law_1.01,0.040582400560379026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,2,power_law_1.01,0.0416703999042511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,2,power_law_1.01,0.043680000305175784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,2,power_law_1.01,0.04456959962844849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,2,power_law_1.01,0.045721599459648134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,2,power_law_1.01,0.04498560130596161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,2,power_law_1.01,0.045151999592781066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,2,power_law_1.01,0.046239998936653134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,2,power_law_1.01,0.051123201847076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,2,power_law_1.01,0.06220800280570984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,2,power_law_1.01,0.06291840076446534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,2,power_law_1.01,0.07086719870567322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,2,power_law_1.01,0.07031679749488831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,2,power_law_1.01,0.08214399814605713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,2,power_law_1.01,0.09599999785423279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,2,power_law_1.01,0.11390080451965331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,16,power_law_1.2,0.1962496042251587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,2,power_law_1.01,0.15086079835891725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,2,power_law_1.01,0.18213119506835937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,16,power_law_1.2,0.2928191900253296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,2,power_law_1.01,0.2458240032196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,2,power_law_1.01,0.3150399923324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,2,power_law_1.01,0.4243775844573975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,32,power_law_1.01,0.03505280017852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,32,power_law_1.01,0.032601600885391234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,32,power_law_1.01,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,32,power_law_1.01,0.03270399868488312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,32,power_law_1.01,0.03367680013179779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,32,power_law_1.01,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,32,power_law_1.01,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,32,power_law_1.01,0.035462400317192076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,32,power_law_1.01,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,32,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,32,power_law_1.01,0.03612160086631775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,32,power_law_1.01,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,32,power_law_1.01,0.03790079951286316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,32,power_law_1.01,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,32,power_law_1.01,0.04234879910945892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,32,power_law_1.01,0.042847999930381776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,32,power_law_1.01,0.0435263991355896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,32,power_law_1.01,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,32,power_law_1.01,0.05182719826698303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,32,power_law_1.01,0.05756800174713135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,32,power_law_1.01,0.0651968002319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,32,power_law_1.01,0.07313920259475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,32,power_law_1.01,0.08538240194320679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,32,power_law_1.01,0.11982079744338989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,32,power_law_1.01,0.13899519443511962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,32,power_law_1.01,0.20218238830566407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,32,power_law_1.01,0.256166410446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,16,power_law_1.2,0.34637439250946045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,2,power_law_1.01,0.584281587600708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,2,power_law_1.01,0.9455936431884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,2,power_law_1.01,1.2300992012023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,32,4,power_law_1.01,0.020601600408554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,32,4,power_law_1.01,0.023078399896621703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,32,4,power_law_1.01,0.02351360023021698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,32,4,power_law_1.01,0.02871679961681366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,32,4,power_law_1.01,0.02959359884262085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,32,4,power_law_1.01,0.0324288010597229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,32,4,power_law_1.01,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,32,4,power_law_1.01,0.03699199855327606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,32,4,power_law_1.01,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,32,4,power_law_1.01,0.03856000006198883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,32,4,power_law_1.01,0.03912320137023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,32,4,power_law_1.01,0.040703999996185306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,32,4,power_law_1.01,0.04217599928379059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,32,4,power_law_1.01,0.044659200310707095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,32,4,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,32,4,power_law_1.01,0.050918400287628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,32,4,power_law_1.01,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,32,4,power_law_1.01,0.07211520075798035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,32,4,power_law_1.01,0.0859391987323761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,32,4,power_law_1.01,0.11047040224075318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,32,4,power_law_1.01,0.13122559785842897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,32,4,power_law_1.01,0.15735039710998536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,32,4,power_law_1.01,0.18127360343933105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,32,4,power_law_1.01,0.23527040481567382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,32,4,power_law_1.01,0.2912384033203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,32,4,power_law_1.01,0.39643518924713134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,32,4,power_law_1.01,0.501087999343872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,32,power_law_1.01,0.10421760082244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,32,power_law_1.01,0.10146559476852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,32,power_law_1.01,0.10012160539627075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,32,power_law_1.01,0.10389759540557861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,32,power_law_1.01,0.10199040174484253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,32,power_law_1.01,0.1054144024848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,32,power_law_1.01,0.10458879470825196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,32,power_law_1.01,0.10368000268936158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,32,power_law_1.01,0.10707839727401733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,32,power_law_1.01,0.1077888011932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,32,power_law_1.01,0.10830719470977783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,32,power_law_1.01,0.11308799982070923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,32,power_law_1.01,0.11342719793319703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,32,power_law_1.01,0.11449600458145141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,32,power_law_1.01,0.13330559730529784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,32,power_law_1.01,0.13422080278396606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,32,power_law_1.01,0.13489919900894165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,32,power_law_1.01,0.13938560485839843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,32,power_law_1.01,0.15745279788970948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,32,power_law_1.01,0.17477760314941407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,32,power_law_1.01,0.19806079864501952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,32,power_law_1.01,0.25975039005279543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,32,power_law_1.01,0.2942784070968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,32,power_law_1.01,0.41565442085266113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,32,power_law_1.01,0.5549439907073974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,32,power_law_1.01,0.7947711944580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,32,power_law_1.01,1.0348799705505372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,8,power_law_1.2,0.048716801404953006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,8,power_law_1.2,0.07414399981498718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,8,power_law_1.2,0.07311360239982605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,8,power_law_1.2,0.10338560342788697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,8,power_law_1.2,0.1148800015449524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,8,power_law_1.2,0.1123263955116272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,8,power_law_1.2,0.12869759798049926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,8,power_law_1.2,0.12452479600906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,8,power_law_1.2,0.12338559627532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,8,power_law_1.2,0.11909760236740112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,8,power_law_1.2,0.13288960456848145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,8,power_law_1.2,0.1439039945602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,8,power_law_1.2,0.15178240537643434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,8,power_law_1.2,0.15494400262832642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,8,power_law_1.2,0.15694719552993774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,8,power_law_1.2,0.17404160499572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,8,power_law_1.2,0.16803200244903566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,8,power_law_1.2,0.18483840227127074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.2,0.246995210647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.2,0.25703039169311526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.2,0.30612480640411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.2,0.33408639430999754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.2,0.40755200386047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.2,0.5320511817932129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.2,0.6618688106536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.2,0.9750592231750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.2,1.2486783981323242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,128,power_law_1.01,0.016659200191497803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,128,power_law_1.01,0.01655679941177368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,128,power_law_1.01,0.017158399522304534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,128,power_law_1.01,0.01918720006942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,128,power_law_1.01,0.030483201146125793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,128,power_law_1.01,0.025811201333999632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,128,power_law_1.01,0.025945600867271424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,128,power_law_1.01,0.025900799036026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,128,power_law_1.01,0.02632960081100464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,128,power_law_1.01,0.026649600267410277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,128,power_law_1.01,0.02645759880542755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,128,power_law_1.01,0.026739200949668883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,128,power_law_1.01,0.026579201221466064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,128,power_law_1.01,0.02744320034980774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,128,power_law_1.01,0.028998398780822755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,128,power_law_1.01,0.0293503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,128,power_law_1.01,0.0302592009305954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,128,power_law_1.01,0.032716798782348636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,128,power_law_1.01,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,128,power_law_1.01,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,128,power_law_1.01,0.039680001139640805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,128,power_law_1.01,0.046316799521446225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,128,power_law_1.01,0.0559935986995697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,128,power_law_1.01,0.06986879706382751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,128,power_law_1.01,0.08775039911270141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,128,power_law_1.01,0.12497919797897339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,128,power_law_1.01,0.15484800338745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,4,power_law_1.2,0.05101439952850342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,4,power_law_1.2,0.05666559934616089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,4,power_law_1.2,0.06120319962501526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,4,power_law_1.2,0.06527360081672669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,4,power_law_1.2,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,4,power_law_1.2,0.06588159799575806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,4,power_law_1.2,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,4,power_law_1.2,0.06686720252037048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,4,power_law_1.2,0.06727039813995361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,4,power_law_1.2,0.06815999746322632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,4,power_law_1.2,0.06941440105438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,4,power_law_1.2,0.07838720083236694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,4,power_law_1.2,0.0828927993774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,4,power_law_1.2,0.08797439932823181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,4,power_law_1.2,0.10064640045166015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,4,power_law_1.2,0.10833280086517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,4,power_law_1.2,0.11690880060195923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,4,power_law_1.2,0.15714559555053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,4,power_law_1.2,0.18124159574508666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,4,power_law_1.2,0.25430400371551515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,4,power_law_1.2,0.3083775997161865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,8,power_law_1.01,0.05842559933662415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,8,power_law_1.01,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,8,power_law_1.01,0.0573311984539032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,8,power_law_1.01,0.06222079992294312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,8,power_law_1.01,0.06803200244903565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,8,power_law_1.01,0.07533439993858337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,8,power_law_1.01,0.07479040026664734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,8,power_law_1.01,0.07534719705581665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,8,power_law_1.01,0.07794560194015503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,8,power_law_1.01,0.07958400249481201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,8,power_law_1.01,0.07871360182762147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,8,power_law_1.01,0.07871360182762147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,8,power_law_1.01,0.07951359748840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,8,power_law_1.01,0.08293759822845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,8,power_law_1.01,0.09030399918556213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,8,power_law_1.01,0.0920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,4,power_law_1.2,0.4455103874206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,8,power_law_1.01,0.09559040069580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,8,power_law_1.01,0.10454399585723877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,8,power_law_1.01,0.1147968053817749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,8,power_law_1.01,0.13527040481567382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,8,power_law_1.01,0.14424320459365844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,8,power_law_1.01,0.17861119508743287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,8,power_law_1.01,0.21697280406951905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,8,power_law_1.01,0.3002943992614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,8,power_law_1.01,0.3657599925994873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,8,power_law_1.01,0.529529619216919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,2,power_law_1.01,0.03973119854927063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,2,power_law_1.01,0.039417600631713866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,8,power_law_1.01,0.6343808174133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,2,power_law_1.01,0.04490239918231964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,2,power_law_1.01,0.053964799642562865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,2,power_law_1.01,0.059654402732849124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,2,power_law_1.01,0.07686399817466735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,2,power_law_1.01,0.07740160226821899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,2,power_law_1.01,0.07901440262794494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,2,power_law_1.01,0.08254079818725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,2,power_law_1.01,0.08429440259933471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,2,power_law_1.01,0.08641279935836792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,2,power_law_1.01,0.09052159786224365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,2,power_law_1.01,0.09171199798583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,2,power_law_1.01,0.09637119770050048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,2,power_law_1.01,0.10529279708862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,2,power_law_1.01,0.10759040117263793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,4,power_law_1.2,0.5645760059356689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,2,power_law_1.01,0.11535999774932862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,2,power_law_1.01,0.13099520206451415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,2,power_law_1.01,0.14164479970932006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,2,power_law_1.01,0.1718783974647522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,2,power_law_1.01,0.19701759815216063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,2,power_law_1.01,0.25099520683288573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,2,power_law_1.01,0.3333631992340088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,2,power_law_1.01,0.46940159797668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,2,power_law_1.01,0.5738687992095948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,2,power_law_1.01,0.8120960235595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,2,power_law_1.01,1.090118408203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,4,power_law_1.2,0.8209919929504395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,256,power_law_1.2,0.055238401889801024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,4,power_law_1.2,1.063929557800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,256,power_law_1.2,0.056364798545837404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,256,power_law_1.2,0.055692797899246214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,256,power_law_1.2,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,256,power_law_1.2,0.059494400024414064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,256,power_law_1.2,0.06047359704971313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,256,power_law_1.2,0.060761600732803345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,256,power_law_1.2,0.05990399718284607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,256,power_law_1.2,0.06121600270271301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,256,power_law_1.2,0.06079360246658325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,256,power_law_1.2,0.06114559769630432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,256,power_law_1.2,0.06164479851722717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,256,power_law_1.2,0.06224640011787415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,256,power_law_1.2,0.062118399143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,256,power_law_1.2,0.06528639793395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,256,power_law_1.2,0.06429439783096313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,256,power_law_1.2,0.0741823971271515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,256,power_law_1.2,0.08241279721260071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,256,power_law_1.2,0.09164159893989562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,256,power_law_1.2,0.11588480472564697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,256,power_law_1.2,0.1324288010597229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,256,power_law_1.2,0.17701760530471802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,256,power_law_1.2,0.21687040328979493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,256,power_law_1.2,0.30453760623931886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,256,power_law_1.2,0.39870080947875974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,256,power_law_1.2,0.5982656002044677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,256,power_law_1.2,0.7428031921386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,4,power_law_1.01,0.03716480135917664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,4,power_law_1.01,0.04523519873619079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,4,power_law_1.01,0.046988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,4,power_law_1.01,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,4,power_law_1.01,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,4,power_law_1.01,0.07539839744567871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,4,power_law_1.01,0.07385600209236146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,4,power_law_1.01,0.07759360074996949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,4,power_law_1.01,0.07752320170402527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,4,power_law_1.01,0.0767359972000122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,4,power_law_1.01,0.0794368028640747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,4,power_law_1.01,0.082259202003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,4,power_law_1.01,0.07991039752960205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,4,power_law_1.01,0.08324480056762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,4,power_law_1.01,0.08799999952316284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,4,power_law_1.01,0.09204480051994324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,4,power_law_1.01,0.09659519791603088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,4,power_law_1.01,0.10228480100631714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,4,power_law_1.01,0.10758399963378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,4,power_law_1.01,0.131167995929718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,4,power_law_1.01,0.14839680194854737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,4,power_law_1.01,0.1733247995376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,4,power_law_1.01,0.1942911982536316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,4,power_law_1.01,0.23955199718475342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,4,power_law_1.01,0.32254719734191895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,4,power_law_1.01,0.4159552097320557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,4,power_law_1.01,0.5595263957977294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,4,power_law_1.2,1.6110143661499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,4,power_law_1.2,2.048147201538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,2,balanced,0.052341332038243614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,2,balanced,0.055813332398732506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,2,balanced,0.07733866572380066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,2,balanced,0.10667733351389568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,2,balanced,0.157151997089386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,2,balanced,0.26014933983484906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,2,balanced,0.359279990196228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,2,balanced,0.36690131823221844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,2,balanced,0.3724000056584676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,2,balanced,0.3676053285598755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,2,balanced,0.37487467130025226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,2,balanced,0.38599467277526855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,2,balanced,0.3879306713740031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,2,balanced,0.39195199807484943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,2,balanced,0.401311993598938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,2,balanced,0.4044693311055501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,2,balanced,0.41998934745788574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,2,balanced,0.4458346764246623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,2,balanced,0.46933865547180176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,2,balanced,0.524453322092692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,2,balanced,0.573082685470581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,2,balanced,0.6790826320648193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,2,balanced,0.8201493422190348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,2,balanced,1.0295360088348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,2,balanced,1.3152746359507244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,2,balanced,1.7336427370707195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,2,balanced,2.321354707082113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,8,power_law_1.2,0.01961600035429001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,8,power_law_1.2,0.01934719979763031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,8,power_law_1.2,0.019148799777030944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,8,power_law_1.2,0.018585599958896637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,8,power_law_1.2,0.019513599574565887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,8,power_law_1.2,0.022188800573349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,8,power_law_1.2,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,8,power_law_1.2,0.039705601334571836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,8,power_law_1.2,0.03511039912700653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,8,power_law_1.2,0.03443840146064758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,8,power_law_1.2,0.03272320032119751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,8,power_law_1.2,0.04302720129489899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,8,power_law_1.2,0.04103679955005646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,8,power_law_1.2,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,8,power_law_1.2,0.044064000248909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,8,power_law_1.2,0.04270080029964447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,8,power_law_1.2,0.04450559914112091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,8,power_law_1.2,0.05745279788970947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.2,0.06559360027313232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.2,0.08607360124588012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.2,0.1078719973564148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.2,0.14684159755706788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.2,0.18102400302886962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,8,power_law_1.01,0.1382143974304199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,8,power_law_1.01,0.211027193069458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,8,power_law_1.01,0.22259199619293213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,8,power_law_1.01,0.2871999979019165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,8,power_law_1.01,0.26305279731750486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,8,power_law_1.01,0.2625344038009644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,8,power_law_1.01,0.297158408164978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,8,power_law_1.01,0.28149120807647704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,8,power_law_1.01,0.2999423980712891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,8,power_law_1.01,0.28104960918426514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,8,power_law_1.01,0.2827647924423218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,8,power_law_1.01,0.29117441177368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,8,power_law_1.01,0.3147648096084595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,8,power_law_1.01,0.309004807472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,8,power_law_1.01,0.2991679906845093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.2,0.25419518947601316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,8,power_law_1.01,0.30614399909973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,8,power_law_1.01,0.31938560009002687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,8,power_law_1.01,0.3285952091217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.01,0.3655616044998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.01,0.41098880767822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.01,0.43951997756958006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.01,0.5549888134002685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.01,0.6323647975921631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.01,0.8352959632873536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.01,0.9814911842346191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.01,1.458073616027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.01,1.780294418334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.2,0.32700159549713137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.2,0.4741568088531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.2,0.6206655979156495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,1,power_law_1.01,0.12306560277938842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,1,power_law_1.01,0.15520000457763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,1,power_law_1.01,0.2124608039855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,1,power_law_1.01,0.32868480682373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,1,power_law_1.01,0.5198783874511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,1,power_law_1.01,0.7294591903686524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,1,power_law_1.01,0.8318079948425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,1,power_law_1.01,1.0732352256774902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,1,power_law_1.01,1.119539165496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,1,power_law_1.01,1.1526335716247558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,1,power_law_1.01,1.1959232330322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,1,power_law_1.01,1.2498496055603028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,1,power_law_1.01,1.2565183639526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,1,power_law_1.01,1.2844799995422362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,1,power_law_1.01,1.3532352447509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,1,power_law_1.01,1.3733183860778808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,1,power_law_1.01,1.395423984527588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,1,power_law_1.01,1.4951359748840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.01,1.5791935920715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.01,1.727872085571289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.01,1.935807991027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.01,2.2708480834960936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.01,2.4099775314331056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.01,2.9912384033203123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,1,power_law_1.01,3.7976192474365233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,1,power_law_1.01,5.015315246582031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,1,power_law_1.01,6.079257583618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,16,power_law_1.2,0.07998080253601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,16,power_law_1.2,0.11766400337219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,16,power_law_1.2,0.09373440146446228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,16,power_law_1.2,0.10545920133590699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,16,power_law_1.2,0.10046720504760742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,16,power_law_1.2,0.09413120150566101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,16,power_law_1.2,0.09654399752616882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,16,power_law_1.2,0.09710720181465149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,16,power_law_1.2,0.0981760025024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,16,power_law_1.2,0.10253440141677857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,16,power_law_1.2,0.10577280521392822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,16,power_law_1.2,0.10673279762268066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,16,power_law_1.2,0.10869760513305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,16,power_law_1.2,0.11227519512176513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,16,power_law_1.2,0.11593600511550903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,16,power_law_1.2,0.12225919961929321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,16,power_law_1.2,0.12791039943695068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,16,power_law_1.2,0.14265600442886353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,16,power_law_1.2,0.16355199813842775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,16,power_law_1.2,0.18579200506210328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,16,power_law_1.2,0.20695679187774657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,16,power_law_1.2,0.26182401180267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,16,power_law_1.2,0.3150784015655518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,16,power_law_1.2,0.4361152172088623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,16,power_law_1.2,0.5749248027801513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,16,power_law_1.2,0.7274687767028809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,16,power_law_1.2,0.9514495849609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,16,power_law_1.2,0.05259519815444946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,16,power_law_1.2,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,16,power_law_1.2,0.05006719827651977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,16,power_law_1.2,0.049593600630760196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,16,power_law_1.2,0.04955520033836365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,16,power_law_1.2,0.05134720206260681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,16,power_law_1.2,0.052076798677444455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,16,power_law_1.2,0.05308799743652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,16,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,16,power_law_1.2,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,16,power_law_1.2,0.05686399936676025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,16,power_law_1.2,0.05947520136833191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,16,power_law_1.2,0.06120319962501526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,16,power_law_1.2,0.06645119786262513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,16,power_law_1.2,0.07455360293388366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,16,power_law_1.2,0.07752959728240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,16,power_law_1.2,0.08025599718093872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,16,power_law_1.2,0.0951807975769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,16,power_law_1.2,0.10741759538650512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,16,power_law_1.2,0.1331071972846985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,16,power_law_1.2,0.16102399826049804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,16,power_law_1.2,0.20500481128692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,16,power_law_1.2,0.2742144107818604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,16,power_law_1.2,0.3506367921829224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,32,balanced,0.04224533339341482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,32,balanced,0.04170133173465729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,32,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,32,balanced,0.043791999419530235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,32,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,32,balanced,0.0499893327554067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,32,balanced,0.05061866839726766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,32,balanced,0.05162666738033295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,32,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,32,balanced,0.05236266553401947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,32,balanced,0.05369600156943003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,32,balanced,0.05629866818586985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,32,balanced,0.054485330979029335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,32,balanced,0.056133334835370384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,32,balanced,0.06196799874305725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,32,balanced,0.06218666831652323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,32,balanced,0.06596266726652782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,32,balanced,0.07229866584142049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,32,balanced,0.0762613316377004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,32,balanced,0.09364266196886699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,32,balanced,0.11409067114194234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,32,balanced,0.15362667044003805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,16,power_law_1.2,0.4505216121673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,32,balanced,0.18427733580271402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,32,balanced,0.2474773327509562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,32,balanced,0.3121386567751567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,32,balanced,0.4347573518753052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,32,balanced,0.5593493382136027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,16,power_law_1.2,0.6664000034332276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,16,power_law_1.2,0.9656767845153809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,2,power_law_1.2,0.09077759981155395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,2,power_law_1.2,0.10598399639129638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,2,power_law_1.2,0.12151039838790893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,2,power_law_1.2,0.1397760033607483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,2,power_law_1.2,0.152128005027771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,2,power_law_1.2,0.17520639896392823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,2,power_law_1.2,0.1983423948287964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,2,power_law_1.2,0.19582079648971557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,2,power_law_1.2,0.21042559146881104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,2,power_law_1.2,0.2128511905670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,2,power_law_1.2,0.2205631971359253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,2,power_law_1.2,0.2181312084197998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,2,power_law_1.2,0.2311039924621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,2,power_law_1.2,0.2350719928741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,2,power_law_1.2,0.243174409866333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,2,power_law_1.2,0.24341120719909667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,2,power_law_1.2,0.255513596534729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,2,power_law_1.2,0.28935039043426514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,2,power_law_1.2,0.32581119537353515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,2,power_law_1.2,0.4011263847351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,2,power_law_1.2,0.46065921783447267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,2,power_law_1.2,0.6258944034576416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,2,power_law_1.2,0.7752768039703369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,2,power_law_1.2,1.092972755432129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,2,power_law_1.2,1.3775296211242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,2,power_law_1.2,2.058803176879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,2,power_law_1.2,2.4108352661132812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,64,power_law_1.01,0.060889601707458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,64,power_law_1.01,0.05930240154266357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,64,power_law_1.01,0.05738880038261414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,64,power_law_1.01,0.06249600052833557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,64,power_law_1.01,0.06251519918441772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,64,power_law_1.01,0.06677759885787964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,64,power_law_1.01,0.06820480227470398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,64,power_law_1.01,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,64,power_law_1.01,0.07299200296401978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,64,power_law_1.01,0.07236480116844177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,64,power_law_1.01,0.07505919933319091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,64,power_law_1.01,0.08154879808425904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,64,power_law_1.01,0.07987200021743775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,64,power_law_1.01,0.07585279941558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,64,power_law_1.01,0.08654080033302307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,64,power_law_1.01,0.09347839951515198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,64,power_law_1.01,0.09207040071487427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,64,power_law_1.01,0.10831999778747559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,64,power_law_1.01,0.11658240556716919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,64,power_law_1.01,0.13527040481567382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,64,power_law_1.01,0.15248639583587648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,64,power_law_1.01,0.20073599815368653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,64,power_law_1.01,0.24094080924987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,64,power_law_1.01,0.3289151906967163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,64,power_law_1.01,0.4299839973449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,64,power_law_1.01,0.6360703945159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,64,power_law_1.01,0.8043071746826171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,64,power_law_1.01,0.13985919952392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,64,power_law_1.01,0.04913919866085052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,64,power_law_1.01,0.044819200038909913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,64,power_law_1.01,0.04563199877738953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,64,power_law_1.01,0.045177599787712096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,64,power_law_1.01,0.046393600106239316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,64,power_law_1.01,0.04730879962444305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,64,power_law_1.01,0.04824959933757782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,64,power_law_1.01,0.04917759895324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,64,power_law_1.01,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,64,power_law_1.01,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,64,power_law_1.01,0.052160000801086424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,64,power_law_1.01,0.053651201725006106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,64,power_law_1.01,0.055667197704315184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,64,power_law_1.01,0.06085119843482971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,64,power_law_1.01,0.06161919832229614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,64,power_law_1.01,0.06988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,64,power_law_1.01,0.07927680015563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,64,power_law_1.01,0.08464639782905578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,64,power_law_1.01,0.10826239585876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,64,power_law_1.01,0.12371200323104858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,64,power_law_1.01,0.17134079933166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,64,power_law_1.01,0.2023103952407837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,64,power_law_1.01,0.276857590675354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,64,power_law_1.01,0.34362239837646485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,64,power_law_1.01,0.502668809890747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,64,power_law_1.01,0.6609024047851563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,4,power_law_1.2,0.04930559992790222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,4,power_law_1.2,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,4,power_law_1.2,0.05633919835090637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,4,power_law_1.2,0.06458879709243774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,4,power_law_1.2,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,4,power_law_1.2,0.07463039755821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,4,power_law_1.2,0.07643520236015319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,4,power_law_1.2,0.07785599827766418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,4,power_law_1.2,0.07848960161209106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,4,power_law_1.2,0.08004480004310607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,1,power_law_1.01,0.0584447979927063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,4,power_law_1.2,0.08117759823799134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,1,power_law_1.01,0.07793279886245727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,4,power_law_1.2,0.08333439826965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,1,power_law_1.01,0.1184831976890564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,4,power_law_1.2,0.08312960267066956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,1,power_law_1.01,0.16508159637451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,1,power_law_1.01,0.20860800743103028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,4,power_law_1.2,0.08897280097007751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,1,power_law_1.01,0.31496319770812986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,4,power_law_1.2,0.09314560294151306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,1,power_law_1.01,0.33805439472198484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,4,power_law_1.2,0.09481599926948547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,4,power_law_1.2,0.10161919593811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,4,power_law_1.2,0.12049920558929443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,4,power_law_1.2,0.13568639755249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,4,power_law_1.2,0.1615872025489807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,4,power_law_1.2,0.1904255986213684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,1,power_law_1.01,0.35075199604034424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,1,power_law_1.01,0.3568000078201294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,4,power_law_1.2,0.2734463930130005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,1,power_law_1.01,0.35600640773773196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,1,power_law_1.01,0.37202560901641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,1,power_law_1.01,0.38163199424743655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,1,power_law_1.01,0.3845952033996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,1,power_law_1.01,0.403718376159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,1,power_law_1.01,0.42201600074768064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,1,power_law_1.01,0.43633279800415037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,4,power_law_1.2,0.3611327886581421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,1,power_law_1.01,0.4376383781433105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,1,power_law_1.01,0.48862719535827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,1,power_law_1.01,0.5095935821533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,1,power_law_1.01,0.5950911998748779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,1,power_law_1.01,0.051910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,1,power_law_1.01,0.05348479747772217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,1,power_law_1.01,0.05667200088500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,1,power_law_1.01,0.06495360136032105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,1,power_law_1.01,0.07710080146789551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,1,power_law_1.01,0.0889087975025177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,1,power_law_1.01,0.10504319667816162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,1,power_law_1.01,0.6846015930175782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,1,power_law_1.01,0.10812159776687622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,1,power_law_1.01,0.11114879846572875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,1,power_law_1.01,0.11383680105209351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,1,power_law_1.01,0.11724159717559815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,1,power_law_1.01,0.12349439859390259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,1,power_law_1.01,0.12616319656372071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,1,power_law_1.01,0.12575360536575317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,1,power_law_1.01,0.13092479705810547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,1,power_law_1.01,0.8547648429870606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,1,power_law_1.01,0.13477120399475098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,1,power_law_1.01,0.1530176043510437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,1,power_law_1.01,1.0414400100708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,1,power_law_1.01,0.1828287959098816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,1,power_law_1.01,0.2016832113265991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,1,power_law_1.01,1.390617561340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,1,power_law_1.01,0.24995839595794678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,1,power_law_1.01,0.29560959339141846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,1,power_law_1.01,1.7718271255493163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,1,power_law_1.01,0.3760384082794189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,1,power_law_1.01,0.46145920753479003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,1,power_law_1.01,2.437977600097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,4,power_law_1.2,0.5458943843841553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,1,power_law_1.01,3.1266752243041993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,1,power_law_1.01,0.6330495834350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,1,power_law_1.01,0.8055232048034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,1,power_law_1.01,1.1445247650146484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,1,power_law_1.01,1.482259178161621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,4,power_law_1.2,0.7120063781738282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,4,power_law_1.2,0.9820287704467774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,4,power_law_1.2,1.3846143722534179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,1,power_law_1.01,0.017836800217628478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,1,power_law_1.01,0.017811200022697447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,1,power_law_1.01,0.01910399943590164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,1,power_law_1.01,0.022060799598693847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,1,power_law_1.01,0.026335999369621277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,1,power_law_1.01,0.027910399436950683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,1,power_law_1.01,0.028627198934555054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,1,power_law_1.01,0.029625600576400755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,1,power_law_1.01,0.031174400448799135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,1,power_law_1.01,0.0321727991104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,1,power_law_1.01,0.03359360098838806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,1,power_law_1.01,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,1,power_law_1.01,0.03681919872760773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,1,power_law_1.01,0.038489601016044615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,1,power_law_1.01,0.03934719860553741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,1,power_law_1.01,0.04140160083770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,1,power_law_1.01,0.04437119960784912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,1,power_law_1.01,0.0532800018787384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,1,power_law_1.01,0.060524797439575194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,1,power_law_1.01,0.0791104018688202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,1,power_law_1.01,0.08759679794311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,1,power_law_1.01,0.11169279813766479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,1,power_law_1.01,0.12323839664459228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,1,power_law_1.01,0.1571071982383728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,1,power_law_1.01,0.19196159839630128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,1,power_law_1.01,0.258022403717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,1,power_law_1.01,0.32561919689178465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,2,power_law_1.2,0.054041600227355956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,2,power_law_1.2,0.07050880193710327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,2,power_law_1.2,0.08017280101776122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,2,power_law_1.2,0.09596160054206848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,2,power_law_1.2,0.1221824049949646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,2,power_law_1.2,0.13794560432434083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,2,power_law_1.2,0.18360960483551025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,2,power_law_1.2,0.1847615957260132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,2,power_law_1.2,0.19478399753570558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,2,power_law_1.2,0.2089087963104248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,2,power_law_1.2,0.20129919052124023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,2,power_law_1.2,0.2102976083755493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,2,power_law_1.2,0.21909759044647217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,2,power_law_1.2,0.22709760665893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,2,power_law_1.2,0.2410559892654419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,2,power_law_1.2,0.24967041015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,2,power_law_1.2,0.26714239120483396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,2,power_law_1.2,0.30090880393981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,2,power_law_1.2,0.3201024055480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,2,power_law_1.2,0.37029759883880614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,2,power_law_1.2,0.4331071853637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,2,power_law_1.2,0.5524352073669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,2,power_law_1.2,0.6352960109710694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,2,power_law_1.2,0.8135680198669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,2,power_law_1.2,0.9680447578430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,1,balanced,0.04987733562787374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,1,balanced,0.05006400247414907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,1,balanced,0.053743998209635414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,2,power_law_1.2,1.4250176429748536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,2,power_law_1.2,1.901420783996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,1,balanced,0.0643146683772405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,1,balanced,0.08648000160853068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,1,balanced,0.12405866384506226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,1,balanced,0.12567999958992004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,1,balanced,0.1255466639995575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,1,balanced,0.1260373294353485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,1,balanced,0.12940800189971924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,1,balanced,0.13054399689038595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,1,balanced,0.13449066877365112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,1,balanced,0.1381386617819468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,1,balanced,0.13661866386731467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,1,balanced,0.13948266704877219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,1,balanced,0.14474133650461832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,1,balanced,0.15094932913780212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,1,balanced,0.16804265975952148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,1,balanced,0.1862186590830485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,1,balanced,0.22304532925287882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,1,balanced,0.2549813389778137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,1,balanced,0.3754826784133911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,1,balanced,0.4238986571629842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,1,balanced,0.6216586828231812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,1,balanced,0.7783413728078207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,1,balanced,1.1172107060750325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,1,balanced,1.4710933367411296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,64,power_law_1.2,0.06172800064086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,64,power_law_1.2,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,64,power_law_1.2,0.04947839975357056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,64,power_law_1.2,0.048844799399375916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,64,power_law_1.2,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,64,power_law_1.2,0.05009920001029968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,64,power_law_1.2,0.05067520141601563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,64,power_law_1.2,0.0525439977645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,64,power_law_1.2,0.05287039875984192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,64,power_law_1.2,0.05297279953956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,64,power_law_1.2,0.05399680137634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,64,power_law_1.2,0.05504000186920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,64,power_law_1.2,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,64,power_law_1.2,0.0572160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,64,power_law_1.2,0.06298239827156067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,64,power_law_1.2,0.06494719982147217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,64,power_law_1.2,0.070387202501297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,64,power_law_1.2,0.0840511977672577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,64,power_law_1.2,0.09273599982261657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,64,power_law_1.2,0.11447039842605591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,64,power_law_1.2,0.13243520259857178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,64,power_law_1.2,0.16481280326843262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,64,power_law_1.2,0.1922816038131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,64,power_law_1.2,0.252729606628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,64,power_law_1.2,0.3405632019042969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,64,power_law_1.2,0.4666816234588623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,64,power_law_1.2,0.6074175834655762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,32,1,power_law_1.01,0.08554239869117737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,32,1,power_law_1.01,0.09146239757537841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,32,1,power_law_1.01,0.099481600522995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,32,1,power_law_1.01,0.12216320037841796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,32,1,power_law_1.01,0.14652800559997559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,32,1,power_law_1.01,0.16634880304336547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,32,1,power_law_1.01,0.2012928009033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,32,1,power_law_1.01,0.20731520652770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,32,1,power_law_1.01,0.21107840538024902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,32,1,power_law_1.01,0.21967360973358155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,32,1,power_law_1.01,0.22471039295196532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,32,1,power_law_1.01,0.23136639595031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,32,1,power_law_1.01,0.23678719997406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,32,1,power_law_1.01,0.24467840194702148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,32,1,power_law_1.01,0.2575551986694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,32,1,power_law_1.01,0.264300799369812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,32,1,power_law_1.01,0.27688961029052733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,32,1,power_law_1.01,0.3231231927871704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,32,1,power_law_1.01,0.36531839370727537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,32,1,power_law_1.01,0.4538432121276855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,32,1,power_law_1.01,0.5334335803985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,32,1,power_law_1.01,0.7101247787475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,32,1,power_law_1.01,0.8769344329833985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,32,1,power_law_1.01,1.2169407844543456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,32,1,power_law_1.01,1.5613183975219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,32,1,power_law_1.01,2.236319923400879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,32,1,power_law_1.01,2.9078208923339846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,4,power_law_1.01,0.03797760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,4,power_law_1.01,0.042387199401855466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,4,power_law_1.01,0.043507200479507444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,4,power_law_1.01,0.04708479940891266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,4,power_law_1.01,0.05196800231933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,4,power_law_1.01,0.05406079888343811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,4,power_law_1.01,0.054451197385787964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,4,power_law_1.01,0.05578240156173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,4,power_law_1.01,0.055308800935745236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,4,power_law_1.01,0.05557119846343994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,4,power_law_1.01,0.05777919888496399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,4,power_law_1.01,0.060249602794647215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,4,power_law_1.01,0.06215680241584778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,4,power_law_1.01,0.06454399824142457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,4,power_law_1.01,0.06465920209884643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,4,power_law_1.01,0.06784639954566955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,4,power_law_1.01,0.07274879813194275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,4,power_law_1.01,0.08076159954071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,4,power_law_1.01,0.091430401802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,4,power_law_1.01,0.10767359733581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,4,power_law_1.01,0.12218879461288452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,4,power_law_1.01,0.15452799797058106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,4,power_law_1.01,0.19612159729003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,4,power_law_1.01,0.28173439502716063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,4,power_law_1.01,0.36922240257263184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,4,power_law_1.01,0.5792191982269287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,4,power_law_1.01,0.7510335922241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,16,power_law_1.01,0.038975998759269714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,16,power_law_1.01,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,16,power_law_1.01,0.037299200892448425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,16,power_law_1.01,0.0387584000825882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,16,power_law_1.01,0.03739520013332367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,16,power_law_1.01,0.03999359905719757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,16,power_law_1.01,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,16,power_law_1.01,0.04116480052471161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,16,power_law_1.01,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,16,power_law_1.01,0.04111360013484955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,16,power_law_1.01,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,16,power_law_1.01,0.042847999930381776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,16,power_law_1.01,0.04322560131549835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,16,power_law_1.01,0.04318720102310181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,16,power_law_1.01,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,16,power_law_1.01,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,16,power_law_1.01,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,16,power_law_1.01,0.052671998739242554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,16,power_law_1.01,0.05630720257759094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,16,power_law_1.01,0.06331520080566407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,16,power_law_1.01,0.07153279781341552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,16,power_law_1.01,0.08568320274353028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,16,power_law_1.01,0.09842560291290284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,16,power_law_1.01,0.1438976049423218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,16,power_law_1.01,0.17760640382766724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,16,power_law_1.01,0.23784959316253662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,16,power_law_1.01,0.28856959342956545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,32,4,balanced,0.021402666966120403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,32,4,balanced,0.022117334107557934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,32,4,balanced,0.024090667565663654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,32,4,balanced,0.029130667448043823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,32,4,balanced,0.03141866624355316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,32,4,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,32,4,balanced,0.037647999823093414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,32,4,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,32,4,balanced,0.04147200038035711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,32,4,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,32,4,balanced,0.04368533194065094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,32,4,balanced,0.04401599864164988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,32,4,balanced,0.046021332343419395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,32,4,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,32,4,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,32,4,balanced,0.056133334835370384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,32,4,balanced,0.06196266909440359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,32,4,balanced,0.08568533261617024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,32,4,balanced,0.09755733609199524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,32,4,balanced,0.12309333682060242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,32,4,balanced,0.1585919956366221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,32,4,balanced,0.18734399477640787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,32,4,balanced,0.2177120049794515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,32,4,balanced,0.2773226698239644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,32,4,balanced,0.3223573366800944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,32,4,balanced,0.4976373513539632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,32,4,balanced,0.6106880108515421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,4,power_law_1.01,0.09182720184326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,4,power_law_1.01,0.10546560287475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,4,power_law_1.01,0.11310720443725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,4,power_law_1.01,0.1459328055381775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,4,power_law_1.01,0.15297919511795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,4,power_law_1.01,0.1652799963951111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,4,power_law_1.01,0.18943359851837158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,4,power_law_1.01,0.1918272018432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,4,power_law_1.01,0.19680000543594361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,4,power_law_1.01,0.1969472050666809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,4,power_law_1.01,0.20207359790802001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,4,power_law_1.01,0.20536959171295166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,4,power_law_1.01,0.21239039897918702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,4,power_law_1.01,0.2126528024673462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,4,power_law_1.01,0.22251520156860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,4,power_law_1.01,0.2226111888885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,4,power_law_1.01,0.2340991973876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,4,power_law_1.01,0.24897279739379882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,4,power_law_1.01,0.2622720003128052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,4,power_law_1.01,0.3236991882324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,4,power_law_1.01,0.3833600044250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,4,power_law_1.01,0.48551039695739745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,4,power_law_1.01,0.5549888134002685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,4,power_law_1.01,0.7766848087310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,4,power_law_1.01,1.015225601196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,4,power_law_1.01,1.5147775650024413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,4,power_law_1.01,1.896998405456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,8,balanced,0.062074666221936546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,8,balanced,0.07260799904664357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,8,balanced,0.09475732843081157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,8,balanced,0.1383786698182424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,8,balanced,0.20909333229064941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,8,balanced,0.33796266714731854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,8,balanced,0.3425973256429036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,8,balanced,0.3120746612548828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,8,balanced,0.31277332703272503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,8,balanced,0.31305599212646484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,8,balanced,0.2982133428255717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,8,balanced,0.293994665145874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,8,balanced,0.31386133035024005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,8,balanced,0.3088906606038411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,8,balanced,0.3107626636823018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,8,balanced,0.3299893339474996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,8,balanced,0.3001919984817505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,8,balanced,0.33906133969624835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,8,balanced,0.3104106585184733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,8,balanced,0.32977066437403363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,8,balanced,0.3336160182952881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,8,balanced,0.3571840127309163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,8,balanced,0.3943733374277751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,8,balanced,0.6459253231684366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,8,balanced,0.6581973234812418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,8,balanced,0.9178240299224854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,8,balanced,1.1814506848653157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,2,balanced,0.05387733379999796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,2,balanced,0.05826666454474131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,2,balanced,0.07665599882602692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,2,balanced,0.10351999600728352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,2,balanced,0.15795200069745383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,2,balanced,0.25355732440948486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,2,balanced,0.25483200947443646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,2,balanced,0.25692800680796307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,2,balanced,0.2609600027402242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,2,balanced,0.2641493280728658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,2,balanced,0.2662293314933777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,2,balanced,0.2708746592203776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,2,balanced,0.27430933713912964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,2,balanced,0.2749013304710388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,2,balanced,0.28620266914367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,2,balanced,0.29319467147191364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,2,balanced,0.30268800258636475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,2,balanced,0.3275039990743001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,2,balanced,0.34359999497731525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,2,balanced,0.3860693375269572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,2,balanced,0.4277760187784831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,2,balanced,0.5387359857559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,2,balanced,0.6107680002848307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,2,balanced,0.8354880015055338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,2,balanced,0.9935306708017985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,2,balanced,1.4171892801920574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,2,balanced,1.7558399836222331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,1,power_law_1.2,0.019251200556755065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,1,power_law_1.2,0.0247871994972229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,1,power_law_1.2,0.037241598963737486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,1,power_law_1.2,0.049881601333618165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,1,power_law_1.2,0.06496000289916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,1,power_law_1.2,0.08523520231246948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,1,power_law_1.2,0.10984959602355956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,1,power_law_1.2,0.1198848009109497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,1,power_law_1.2,0.12449920177459717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,1,power_law_1.2,0.1281599998474121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,1,power_law_1.2,0.1339136004447937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,1,power_law_1.2,0.13679360151290892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,1,power_law_1.2,0.14035199880599974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,1,power_law_1.2,0.1442304015159607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,1,power_law_1.2,0.1552832007408142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,1,power_law_1.2,0.15518079996109008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,1,power_law_1.2,0.1562559962272644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,1,power_law_1.2,0.17722879648208617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,1,power_law_1.2,0.1999295949935913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,1,power_law_1.2,0.25128319263458254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,1,power_law_1.2,0.26168959140777587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,1,power_law_1.2,0.34254720211029055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,1,power_law_1.2,0.327891206741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,1,power_law_1.2,0.44839038848876955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,1,power_law_1.2,0.5544064044952393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,1,power_law_1.2,0.7260543823242187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,1,power_law_1.2,0.9429311752319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,4,power_law_1.2,0.033318400382995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,4,power_law_1.2,0.03345920145511627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,4,power_law_1.2,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,4,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,4,power_law_1.2,0.042335999011993405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,4,power_law_1.2,0.051571202278137204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,4,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,4,power_law_1.2,0.05480960011482239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,4,power_law_1.2,0.0568448007106781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,4,power_law_1.2,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,4,power_law_1.2,0.058182400465011594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,4,power_law_1.2,0.060159999132156375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,4,power_law_1.2,0.06159999966621399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,4,power_law_1.2,0.06419199705123901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,4,power_law_1.2,0.07030400037765502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,4,power_law_1.2,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,4,power_law_1.2,0.07163519859313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,4,power_law_1.2,0.08104959726333619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,4,power_law_1.2,0.08846719861030579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,4,power_law_1.2,0.10531840324401856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,4,power_law_1.2,0.11834880113601684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,4,power_law_1.2,0.1440000057220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,4,power_law_1.2,0.17120640277862548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,4,power_law_1.2,0.23134078979492187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,4,power_law_1.2,0.2644479990005493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,4,power_law_1.2,0.36376960277557374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,4,power_law_1.2,0.4753471851348877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,128,power_law_1.01,0.056806397438049314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,128,power_law_1.01,0.06097919940948486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,128,power_law_1.01,0.05172479748725891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,128,power_law_1.01,0.049209600687026976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,128,power_law_1.01,0.05132799744606018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,128,power_law_1.01,0.05056639909744263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,128,power_law_1.01,0.05126399993896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,128,power_law_1.01,0.05260800123214722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,128,power_law_1.01,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,128,power_law_1.01,0.05257599949836731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,128,power_law_1.01,0.052985602617263795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,128,power_law_1.01,0.05447040200233459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,128,power_law_1.01,0.05514879822731018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,128,power_law_1.01,0.055718398094177245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,128,power_law_1.01,0.0609279990196228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,128,power_law_1.01,0.063372802734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,128,power_law_1.01,0.06854400038719177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,128,power_law_1.01,0.08147199749946595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,128,power_law_1.01,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,128,power_law_1.01,0.10451200008392333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,128,power_law_1.01,0.1217087984085083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,128,power_law_1.01,0.15716480016708373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,128,power_law_1.01,0.1790336012840271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,128,power_law_1.01,0.22928640842437745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,128,power_law_1.01,0.2948863983154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,128,power_law_1.01,0.4180607795715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,128,power_law_1.01,0.5514495849609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,4,power_law_1.01,0.05022720098495483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,4,power_law_1.01,0.05511680245399475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,4,power_law_1.01,0.06157439947128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,4,power_law_1.01,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,4,power_law_1.01,0.06846079826354981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,4,power_law_1.01,0.07909759879112244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,4,power_law_1.01,0.08796160221099854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,4,power_law_1.01,0.09311360120773315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,4,power_law_1.01,0.09195520281791687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,4,power_law_1.01,0.09263359904289245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,4,power_law_1.01,0.09486719965934753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,4,power_law_1.01,0.09562240242958069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,2,power_law_1.01,0.03540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,4,power_law_1.01,0.09752960205078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,2,power_law_1.01,0.033843201398849485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,4,power_law_1.01,0.10250879526138305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,2,power_law_1.01,0.037171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,4,power_law_1.01,0.11125760078430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,4,power_law_1.01,0.11639679670333862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,4,power_law_1.01,0.12276480197906495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,4,power_law_1.01,0.14128639698028564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,4,power_law_1.01,0.1577855944633484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,4,power_law_1.01,0.18709759712219237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,4,power_law_1.01,0.22024319171905518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,4,power_law_1.01,0.2779455900192261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,4,power_law_1.01,0.3508671998977661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,4,power_law_1.01,0.46892800331115725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,4,power_law_1.01,0.6358143806457519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,2,power_law_1.01,0.04017280042171478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,4,power_law_1.01,0.8391103744506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,2,power_law_1.01,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,2,power_law_1.01,0.05741440057754517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,2,power_law_1.01,0.05854079723358154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,4,power_law_1.01,1.074118423461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,2,power_law_1.01,0.060063999891281125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,2,power_law_1.01,0.06258559823036194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,2,power_law_1.01,0.061414402723312375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,2,power_law_1.01,0.06515200138092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,2,power_law_1.01,0.06935679912567139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,2,power_law_1.01,0.07221119999885559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,2,power_law_1.01,0.0732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,2,power_law_1.01,0.0819711983203888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,2,power_law_1.01,0.08357759714126586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,2,power_law_1.01,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,2,power_law_1.01,0.10488959550857543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,2,power_law_1.01,0.11340800523757935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,2,power_law_1.01,0.14254080057144164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,2,power_law_1.01,0.16462080478668212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,2,power_law_1.01,0.20568320751190186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,2,power_law_1.01,0.2783616065979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,2,power_law_1.01,0.37204480171203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,2,power_law_1.01,0.5159167766571044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,2,power_law_1.01,0.7138688087463378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,2,power_law_1.01,1.0181056022644044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,2,balanced,0.01969066634774208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,2,balanced,0.02067733307679494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,2,balanced,0.021087999145189922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,2,balanced,0.020981334149837494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,2,balanced,0.021840001145998638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,2,balanced,0.021354667842388153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,2,balanced,0.02178666740655899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,2,balanced,0.023050665855407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,2,balanced,0.025066666305065155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,2,balanced,0.025231999655564625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,2,balanced,0.028336000939210255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,2,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,2,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,2,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,2,balanced,0.03980266551176707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,2,balanced,0.03811733424663544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,2,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,2,balanced,0.04599999884764353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,2,balanced,0.04782933493455251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,2,balanced,0.06850666801134746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,2,balanced,0.07682666679223378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,2,balanced,0.102101335922877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,2,balanced,0.12706666191418967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,2,balanced,0.1742666761080424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,2,balanced,0.21287999550501505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,2,balanced,0.31014933188756305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,2,balanced,0.39669867356618244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,1,power_law_1.2,0.07599999904632568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,1,power_law_1.2,0.10644479990005493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,1,power_law_1.2,0.12080639600753784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,1,power_law_1.2,0.18095359802246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,1,power_law_1.2,0.18881920576095582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,1,power_law_1.2,0.1807935953140259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,1,power_law_1.2,0.17877119779586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,1,power_law_1.2,0.18101760149002075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,1,power_law_1.2,0.179257595539093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,1,power_law_1.2,0.1803712010383606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,1,power_law_1.2,0.18024319410324097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,1,power_law_1.2,0.1806015968322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,1,power_law_1.2,0.17726720571517945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,1,power_law_1.2,0.18821760416030883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,1,power_law_1.2,0.2094655990600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,1,power_law_1.2,0.21027839183807373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,1,power_law_1.2,0.23352959156036376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,1,power_law_1.2,0.263155198097229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.2,0.31441919803619384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.2,0.3970752000808716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.2,0.46575360298156737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.2,0.6086336135864258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.2,0.7437759876251221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,1,power_law_1.01,0.01643519997596741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,1,power_law_1.01,0.016531200706958772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,1,power_law_1.01,0.016896000504493712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,1,power_law_1.01,0.017324799299240114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,1,power_law_1.01,0.01817599982023239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,1,power_law_1.01,0.01847040057182312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,1,power_law_1.01,0.0191551998257637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,1,power_law_1.01,0.01966080069541931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,1,power_law_1.01,0.02038400024175644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,1,power_law_1.01,0.020531199872493744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,1,power_law_1.01,0.021580800414085388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,1,power_law_1.01,0.023846399784088135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,1,power_law_1.01,0.027807998657226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,1,power_law_1.01,0.025228801369667053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,1,power_law_1.01,0.026719999313354493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,1,power_law_1.01,0.028812798857688903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,1,power_law_1.01,0.03089280128479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,1,power_law_1.01,0.03704319894313812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.01,0.04073599874973297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.01,0.05368959903717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.01,0.06028159856796265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.01,0.07909119725227357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.01,0.09493759870529175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.2,1.056544017791748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.01,0.12462719678878784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.01,0.15858559608459472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.2,1.3874176025390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.01,0.22136321067810058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.2,2.0001407623291017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.01,0.2848128080368042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,2,balanced,0.12931733330090842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,2,balanced,0.21408534049987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,2,balanced,0.40539201100667316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,2,balanced,0.39457066853841144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,2,balanced,0.38390934467315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,2,balanced,0.39021865526835126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,2,balanced,0.38597333431243896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,2,balanced,0.39044801394144696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,2,balanced,0.38841064771016437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,2,balanced,0.3888479868570964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,2,balanced,0.39634132385253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,2,balanced,0.3946186701456706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,2,balanced,0.39582931995391846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,2,balanced,0.41811732451121014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,2,balanced,0.4594506820042928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,2,balanced,0.45523734887441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,2,balanced,0.447215994199117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,2,balanced,0.530837337176005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,2,balanced,0.515775998433431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,2,balanced,0.7061333656311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,2,balanced,0.8135786851247152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,2,balanced,1.1157546838124592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,2,balanced,1.4522666931152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,2,balanced,2.0492053031921387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,2,balanced,2.775061289469401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,2,balanced,4.151589393615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,2,balanced,5.523413340250651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.2,2.6120128631591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,16,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,16,balanced,0.0313226655125618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,16,balanced,0.03148799886306127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,16,balanced,0.032960000137488045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,16,balanced,0.033717334270477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,16,balanced,0.03572266548871994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,16,balanced,0.034261333445707955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,16,balanced,0.035429333647092186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,16,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,16,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,16,balanced,0.036415999134381614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,16,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,16,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,16,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,16,balanced,0.04181866844495138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,16,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,16,balanced,0.043663998444875084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,16,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,16,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,16,balanced,0.05190933247407278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,16,balanced,0.0583840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,16,balanced,0.07588266829649608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,16,balanced,0.08692266543706258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,16,balanced,0.11318932970364888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,16,balanced,0.14350400368372598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,16,balanced,0.19036267201105753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,16,balanced,0.2332693338394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,4,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,4,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,4,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,4,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,4,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,4,balanced,0.0373333344856898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,4,balanced,0.03772266705830892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,4,balanced,0.038773333032925926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,4,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,4,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,4,balanced,0.039061332742373146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,4,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,4,balanced,0.03977066775163015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,4,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,4,balanced,0.04565866788228353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,4,balanced,0.04594666759173075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,4,balanced,0.045935998360315956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,4,balanced,0.05273066461086273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,4,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,4,balanced,0.0609386662642161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,4,balanced,0.07222933570543925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,4,balanced,0.09058133761088054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,4,balanced,0.11476799845695496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,4,balanced,0.14705600341161093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,4,balanced,0.17172799507776895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,4,balanced,0.237936000029246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,4,balanced,0.30018667380015057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,8,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,8,balanced,0.04808533191680908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,8,balanced,0.04854399959246317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,8,balanced,0.05176533261934916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,8,balanced,0.05422399938106537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,8,balanced,0.06481066842873891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,8,balanced,0.06623466809590657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,8,balanced,0.06630933284759521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,8,balanced,0.06688533226648967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,8,balanced,0.06830400228500366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,8,balanced,0.06821866830190022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,8,balanced,0.0683840016523997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,8,balanced,0.0691786656777064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,8,balanced,0.0707946668068568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,8,balanced,0.07628799974918365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,8,balanced,0.07852800190448761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,8,balanced,0.08144000172615051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,8,balanced,0.08872532844543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,8,balanced,0.09909866253534953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,8,balanced,0.11281067132949829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,8,balanced,0.13567466537157694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,8,balanced,0.16617600123087564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,8,balanced,0.19303999344507852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,8,balanced,0.26340800523757935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,8,balanced,0.31839466094970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,8,balanced,0.43296531836191815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,8,balanced,0.5601119995117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,4,balanced,0.08947199583053589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,4,balanced,0.09300800164540608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,64,power_law_1.01,0.06540160179138184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,64,power_law_1.01,0.058143997192382814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,64,power_law_1.01,0.05132799744606018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,64,power_law_1.01,0.05335680246353149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,64,power_law_1.01,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,64,power_law_1.01,0.05377920269966126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,64,power_law_1.01,0.054764801263809205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,64,power_law_1.01,0.056428802013397214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,64,power_law_1.01,0.05674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,64,power_law_1.01,0.05720959901809693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,64,power_law_1.01,0.05674880146980286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,64,power_law_1.01,0.05975679755210876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,64,power_law_1.01,0.060140800476074216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,64,power_law_1.01,0.06158080101013184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,64,power_law_1.01,0.06773759722709656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,64,power_law_1.01,0.06817920207977295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,64,power_law_1.01,0.07385600209236146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,64,power_law_1.01,0.08617600202560424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,64,power_law_1.01,0.09518719911575317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,64,power_law_1.01,0.11571199893951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,64,power_law_1.01,0.13244160413742065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,64,power_law_1.01,0.17144960165023804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,64,power_law_1.01,0.20412158966064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,64,power_law_1.01,0.26598401069641114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,64,power_law_1.01,0.343174409866333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,64,power_law_1.01,0.46968960762023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,64,power_law_1.01,0.6396224021911621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,4,power_law_1.01,0.048870399594306946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,4,power_law_1.01,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,4,power_law_1.01,0.03391999900341034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,4,power_law_1.01,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,4,power_law_1.01,0.03482879996299744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,4,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,4,power_law_1.01,0.038873600959777835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,4,power_law_1.01,0.036575999855995175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,4,power_law_1.01,0.03774079978466034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,4,power_law_1.01,0.03834879994392395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,4,power_law_1.01,0.04376960098743439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,4,power_law_1.01,0.043808001279830935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,4,power_law_1.01,0.050758397579193114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,4,power_law_1.01,0.05340800285339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,4,power_law_1.01,0.06325119733810425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,4,power_law_1.01,0.06264320015907288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,4,power_law_1.01,0.044352000951766966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,4,power_law_1.01,0.06815999746322632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,4,power_law_1.01,0.0868287980556488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,4,power_law_1.01,0.112716805934906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,4,power_law_1.01,0.06736639738082886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,4,power_law_1.01,0.14087040424346925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,4,power_law_1.01,0.07455999851226806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,4,power_law_1.01,0.14882559776306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,4,power_law_1.01,0.1724544048309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,4,power_law_1.01,0.08320000171661376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,4,power_law_1.01,0.19566080570220948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,4,power_law_1.01,0.20023040771484374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,4,power_law_1.01,0.1875712037086487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,4,power_law_1.01,0.20246400833129882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,4,power_law_1.01,0.20986878871917725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,4,power_law_1.01,0.2123647928237915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,4,power_law_1.01,0.2153536081314087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,4,power_law_1.01,0.228985595703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,4,power_law_1.01,0.23169920444488526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,4,power_law_1.01,0.2523776054382324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,4,power_law_1.01,0.24508159160614013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,4,power_law_1.01,0.2795072078704834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,4,power_law_1.01,0.3238719940185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,4,power_law_1.01,0.39290881156921387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,4,power_law_1.01,0.44422402381896975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,4,power_law_1.01,0.11058559417724609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,4,power_law_1.01,0.5068672180175782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,4,power_law_1.01,0.6000832080841064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,4,power_law_1.01,0.7615488052368165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,4,power_law_1.01,1.0149824142456054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,4,power_law_1.01,1.2935168266296386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,4,power_law_1.01,0.12108160257339477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,4,power_law_1.01,0.14248960018157958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,4,power_law_1.01,0.19191039800643922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,4,power_law_1.01,0.2878079891204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,4,power_law_1.01,0.31240320205688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,4,power_law_1.01,0.4548351764678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,16,power_law_1.01,0.07551360130310059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,16,power_law_1.01,0.08165760040283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,16,power_law_1.01,0.09069439768791199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,16,power_law_1.01,0.09180160164833069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,16,power_law_1.01,0.09096959829330445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,16,power_law_1.01,0.09093760251998902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,16,power_law_1.01,0.09418879747390747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,16,power_law_1.01,0.09062399864196777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,16,power_law_1.01,0.08979200124740601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,16,power_law_1.01,0.09141119718551635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,16,power_law_1.01,0.09268479943275451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,16,power_law_1.01,0.09388800263404846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,16,power_law_1.01,0.09411200284957885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,16,power_law_1.01,0.0967743992805481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,4,power_law_1.01,0.575110387802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,16,power_law_1.01,0.1058303952217102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,16,power_law_1.01,0.1089792013168335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,16,power_law_1.01,0.12159359455108643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,16,power_law_1.01,0.13134080171585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,16,power_law_1.01,0.13393280506134034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,16,power_law_1.01,0.16193920373916626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,16,power_law_1.01,0.18607360124588013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,16,power_law_1.01,0.24072320461273194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,16,power_law_1.01,0.2896640062332153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,16,power_law_1.01,0.4129792213439941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,16,power_law_1.01,0.5443647861480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,16,power_law_1.01,0.773689603805542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,16,power_law_1.01,0.9982272148132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,2,balanced,0.06436266501744588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,2,balanced,0.06637333333492279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,2,balanced,0.07220800220966339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,2,balanced,0.08829333384831746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,2,balanced,0.12785067160924277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,2,balanced,0.18965866168340048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,2,balanced,0.19302932421366373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,2,balanced,0.18992000818252563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,2,balanced,0.18920000394185385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,2,balanced,0.19370132684707642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,2,balanced,0.19403733809789023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,2,balanced,0.1977013349533081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,2,balanced,0.2013066609700521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,2,balanced,0.20524267355600992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,2,balanced,0.20948266983032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,2,balanced,0.21360532442728677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,2,balanced,0.22317334016164145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,2,balanced,0.24518932898839316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,2,balanced,0.2621866663297017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,2,balanced,0.30643200874328613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,2,balanced,0.34596800804138184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,2,balanced,0.4348906675974528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,2,balanced,0.5317333141962687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,2,balanced,0.7541013558705648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,2,balanced,0.9536960124969482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,2,balanced,1.3828426996866863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,2,balanced,1.7754613558451335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,256,power_law_1.2,0.07308800220489502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,256,power_law_1.2,0.07242239713668823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,256,power_law_1.2,0.07278720140457154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,256,power_law_1.2,0.07678080201148987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,256,power_law_1.2,0.07648640275001525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,256,power_law_1.2,0.07992320060729981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,256,power_law_1.2,0.07697280049324036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,256,power_law_1.2,0.07836800217628478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,256,power_law_1.2,0.07941120266914367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,256,power_law_1.2,0.07790719866752624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,256,power_law_1.2,0.07685120105743408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,256,power_law_1.2,0.0772159993648529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,256,power_law_1.2,0.07649919986724854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,256,power_law_1.2,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,256,power_law_1.2,0.07500159740447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,256,power_law_1.2,0.07690879702568054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,256,power_law_1.2,0.09371520280838012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,256,power_law_1.2,0.09646080136299133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,256,power_law_1.2,0.10556800365447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,256,power_law_1.2,0.1374079942703247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,256,power_law_1.2,0.15964159965515137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,256,power_law_1.2,0.20782721042633057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,256,power_law_1.2,0.27061760425567627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,256,power_law_1.2,0.36727681159973147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,256,power_law_1.2,0.4691455841064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,256,power_law_1.2,0.8029376029968261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,256,power_law_1.2,1.0498432159423827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,4,power_law_1.2,0.021478399634361267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,4,power_law_1.2,0.02590720057487488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,4,power_law_1.2,0.035078400373458864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,4,power_law_1.2,0.04520959854125976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,4,power_law_1.2,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,4,power_law_1.2,0.05438719987869263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,4,power_law_1.2,0.061401599645614625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,4,power_law_1.2,0.06475520133972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,4,power_law_1.2,0.06848639845848084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,4,power_law_1.2,0.06565759778022766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,4,power_law_1.2,0.0687936007976532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,4,power_law_1.2,0.07260800004005433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,4,power_law_1.2,0.07304319739341736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,32,balanced,0.03385066737731298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,4,power_law_1.2,0.07992960214614868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,4,power_law_1.2,0.08273280262947083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,4,power_law_1.2,0.08712319731712341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,4,power_law_1.2,0.09048960208892823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,4,power_law_1.2,0.1008255958557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,4,power_law_1.2,0.1134335994720459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,4,power_law_1.2,0.13374719619750977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,32,balanced,0.03181333343187968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,32,balanced,0.03149333347876867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,4,power_law_1.2,0.17713919878005982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,4,power_law_1.2,0.20098559856414794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,4,power_law_1.2,0.23018240928649902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,4,power_law_1.2,0.29501440525054934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,4,power_law_1.2,0.3414720058441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,32,balanced,0.033887999753157295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,32,balanced,0.035936000446478523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,32,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,32,balanced,0.03704000016053518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,32,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,4,power_law_1.2,0.5050816059112548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,4,power_law_1.2,0.6130688190460205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,32,balanced,0.037818667789300285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,32,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,32,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,32,balanced,0.039664000272750854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,32,balanced,0.03961066653331121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,32,balanced,0.03993066648642222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,32,balanced,0.046015997727712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,32,balanced,0.047354668378829956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,32,balanced,0.050197333097457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,32,balanced,0.05012799799442291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,32,balanced,0.05630933245023092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,32,balanced,0.06198933223883311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,32,balanced,0.06433600187301636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,32,balanced,0.09038399656613667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,32,balanced,0.10052800178527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,32,balanced,0.12616533041000366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,32,balanced,0.15269866585731506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,32,balanced,0.20357867081960043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,32,balanced,0.2568853298823039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,16,balanced,0.03326933334271113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,16,balanced,0.030447999636332195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,16,balanced,0.031301334500312805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,16,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,16,balanced,0.031504000226656594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,16,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,128,balanced,0.023205332458019257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,128,balanced,0.023082666099071503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,128,balanced,0.02298133323589961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,128,balanced,0.023413332800070446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,128,balanced,0.02743999908367793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,128,balanced,0.05249600112438202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,128,balanced,0.04102933406829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,128,balanced,0.04164266586303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,128,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,128,balanced,0.04164800047874451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,128,balanced,0.04126933217048645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,128,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,128,balanced,0.041482667128245033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,128,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,128,balanced,0.048623998959859215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,128,balanced,0.05026666820049286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,128,balanced,0.04610133171081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,128,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,128,balanced,0.05825066566467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,128,balanced,0.08329600095748901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,128,balanced,0.09347200393676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,128,balanced,0.12794133027394614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,128,balanced,0.16125333309173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,16,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,16,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,16,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,16,balanced,0.03410666684309641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,16,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,16,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,16,balanced,0.03708266715208689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,16,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,16,balanced,0.040063999593257904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,16,balanced,0.043882668018341064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,16,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,16,balanced,0.049738665421803795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,16,balanced,0.04557866851488749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,16,balanced,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,16,balanced,0.05417599777380625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,16,balanced,0.0743093341588974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,16,balanced,0.08448533217112224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,16,balanced,0.1106773316860199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,128,balanced,0.22890132665634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,16,balanced,0.12982933719952902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,16,balanced,0.17492800951004028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,128,balanced,0.2892746726671855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,16,balanced,0.2156533400217692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,128,balanced,0.42634133497873944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,128,balanced,0.5605760018030802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,2,power_law_1.2,0.033670398592948916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,2,power_law_1.2,0.03348479866981506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,2,power_law_1.2,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,2,power_law_1.2,0.039289599657058714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,2,power_law_1.2,0.04116480052471161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,2,power_law_1.2,0.05140479803085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,2,power_law_1.2,0.051948797702789304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,2,power_law_1.2,0.05294079780578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,2,power_law_1.2,0.05483520030975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,2,power_law_1.2,0.05400320291519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,2,power_law_1.2,0.05749760270118713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,2,power_law_1.2,0.060915201902389526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,2,power_law_1.2,0.06355199813842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,2,power_law_1.2,0.07012479901313781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,2,power_law_1.2,0.07679359912872315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,2,power_law_1.2,0.08117120265960694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,2,power_law_1.2,0.08554239869117737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,2,power_law_1.2,0.099891197681427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,2,power_law_1.2,0.11974400281906128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,2,power_law_1.2,0.14219520092010499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,2,power_law_1.2,0.18892799615859984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,2,power_law_1.2,0.23971199989318848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,2,power_law_1.2,0.288319993019104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,2,power_law_1.2,0.40805759429931643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,2,power_law_1.2,0.543552017211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,2,power_law_1.2,0.7611455917358398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,8,power_law_1.01,0.02481919974088669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,8,power_law_1.01,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,8,power_law_1.01,0.04312959909439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,8,power_law_1.01,0.04359039962291718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,8,power_law_1.01,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,8,power_law_1.01,0.047603198885917665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,8,power_law_1.01,0.054553598165512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,8,power_law_1.01,0.05459840297698974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,8,power_law_1.01,0.05690240263938904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,8,power_law_1.01,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,8,power_law_1.01,0.05804160237312317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,8,power_law_1.01,0.06056320071220398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,8,power_law_1.01,0.06446719765663148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,8,power_law_1.01,0.0606656014919281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,8,power_law_1.01,0.06638720035552978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,8,power_law_1.01,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,8,power_law_1.01,0.06673280000686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,8,power_law_1.01,0.07708799839019775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,8,power_law_1.01,0.13479679822921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,8,power_law_1.01,0.16796159744262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,8,power_law_1.01,0.1224511981010437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,8,power_law_1.01,0.13408639430999755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,8,power_law_1.01,0.1614143967628479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,8,power_law_1.01,0.1978816032409668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,8,power_law_1.01,0.24389760494232177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,8,power_law_1.01,0.34085760116577146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,8,power_law_1.01,0.4195648193359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,2,power_law_1.2,0.9507200241088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,4,power_law_1.2,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,4,power_law_1.2,0.06053119897842407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,4,power_law_1.2,0.07356799840927124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,4,power_law_1.2,0.08384640216827392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,4,power_law_1.2,0.08950399756431579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,4,power_law_1.2,0.10958080291748047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,4,power_law_1.2,0.10740480422973633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,4,power_law_1.2,0.11066240072250366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,4,power_law_1.2,0.11557120084762573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,4,power_law_1.2,0.1168063998222351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,4,power_law_1.2,0.11729919910430908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,4,power_law_1.2,0.12111999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,4,power_law_1.2,0.12591999769210815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,4,power_law_1.2,0.1305408000946045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,4,power_law_1.2,0.13658879995346068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,4,power_law_1.2,0.1436735987663269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,4,power_law_1.2,0.15010559558868408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,4,power_law_1.2,0.1743872046470642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,4,power_law_1.2,0.1827072024345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,4,power_law_1.2,0.22097280025482177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,4,power_law_1.2,0.25527679920196533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,4,power_law_1.2,0.31906559467315676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,4,power_law_1.2,0.41880321502685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,4,power_law_1.2,0.5417471885681152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,4,power_law_1.2,0.6952320098876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,4,power_law_1.01,0.06330239772796631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,4,power_law_1.01,0.06302080154418946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,4,power_law_1.01,0.06781439781188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,4,power_law_1.2,0.05500800013542175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,4,power_law_1.2,0.05161600112915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,4,power_law_1.2,0.06851199865341187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,4,power_law_1.2,0.08608639836311341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,4,power_law_1.2,0.08406400084495544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,4,power_law_1.2,0.113811194896698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,4,power_law_1.2,0.11200000047683716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,4,power_law_1.2,0.11224960088729859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,4,power_law_1.01,0.08111360073089599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,4,power_law_1.2,0.11798399686813354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,4,power_law_1.01,0.07991039752960205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,4,power_law_1.2,0.11924480199813843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,4,power_law_1.01,0.08472319841384887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,4,power_law_1.2,0.12205439805984497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,4,power_law_1.01,0.08672639727592468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,4,power_law_1.2,0.12014720439910889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,4,power_law_1.01,0.08490239977836608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,4,power_law_1.2,0.125600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,4,power_law_1.01,0.08624640107154846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,4,power_law_1.2,0.12980480194091798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,4,power_law_1.01,0.0906175971031189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,4,power_law_1.01,0.09163519740104675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,4,power_law_1.01,0.09809280037879944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,4,power_law_1.01,0.10187519788742065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,4,power_law_1.01,0.10614399909973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,4,power_law_1.2,0.13685120344161988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,4,power_law_1.01,0.12376960515975952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,4,power_law_1.01,0.13363840579986572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,4,power_law_1.01,0.14357759952545165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,4,power_law_1.2,0.14298239946365357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,4,power_law_1.2,0.14785280227661132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,4,power_law_1.01,0.1671872019767761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,4,power_law_1.2,0.16094080209732056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,4,power_law_1.2,0.1758015990257263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,4,power_law_1.01,0.20409600734710692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,4,power_law_1.2,0.2035520076751709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,4,power_law_1.2,0.22369279861450195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,4,power_law_1.2,0.2917632102966309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,4,power_law_1.2,0.3563136100769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,4,power_law_1.2,0.43815040588378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,4,power_law_1.01,0.2518719911575317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,4,power_law_1.01,0.3229439973831177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,4,power_law_1.2,0.5792511940002442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,4,power_law_1.2,0.9972224235534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,4,power_law_1.2,0.028288000822067262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,4,power_law_1.2,0.029337599873542786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,4,power_law_1.2,0.029139199852943422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,4,power_law_1.2,0.03215999901294708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,4,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,4,power_law_1.2,0.03372800052165985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,4,power_law_1.2,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,4,power_law_1.2,0.03415040075778961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,4,power_law_1.2,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,4,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,4,power_law_1.2,0.036083200573921205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,4,power_law_1.01,0.43192319869995116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,4,power_law_1.2,0.0400191992521286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,4,power_law_1.2,1.344940757751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,4,power_law_1.2,0.04033919870853424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,4,power_law_1.2,0.04442239999771118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,4,power_law_1.01,0.5469888210296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,4,power_law_1.2,0.05135999917984009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,4,power_law_1.2,0.055103999376296994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,4,power_law_1.2,0.874630355834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,4,power_law_1.2,0.060192000865936277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,4,power_law_1.01,0.845638370513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,4,power_law_1.2,0.0723136007785797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,4,power_law_1.2,0.09070720076560974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,4,power_law_1.2,0.11630079746246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,4,power_law_1.2,0.14382079839706421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,4,power_law_1.01,1.0449472427368165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,4,power_law_1.2,0.2068416118621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,4,power_law_1.2,1.0072128295898437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,4,power_law_1.2,0.2670783996582031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,4,power_law_1.2,0.3964416027069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,4,power_law_1.01,1.522169589996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,4,power_law_1.2,0.5225344181060791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,4,power_law_1.01,1.9125631332397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,4,power_law_1.2,0.8614912033081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,4,power_law_1.2,1.253651237487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,1,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,1,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,1,balanced,0.046613335609436035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,1,balanced,0.07382399837176006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,1,balanced,0.10896533727645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,1,balanced,0.1339893341064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,1,balanced,0.132314662138621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,1,balanced,0.13660266995429993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,1,balanced,0.13639466961224875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,1,balanced,0.13502933581670126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,1,balanced,0.13637333114941916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,1,balanced,0.13754133383433023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,1,balanced,0.14075199762980142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,1,balanced,0.14037332932154337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,1,balanced,0.14512532949447632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,1,balanced,0.1479520003000895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,1,balanced,0.1529813309510549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,1,balanced,0.1653386652469635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,1,balanced,0.17217065890630087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,1,balanced,0.22299200296401978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,1,balanced,0.22708266973495483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,1,balanced,0.31221334139506024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,1,balanced,0.33738664786020917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,1,balanced,0.4901333252588908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,1,balanced,0.5630880196889242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,1,balanced,0.8302720387776693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,1,balanced,0.9970880349477133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,8,power_law_1.2,0.0683135986328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,8,power_law_1.2,0.08919039964675904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,8,power_law_1.2,0.09292799830436707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,8,power_law_1.2,0.10042879581451417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,8,power_law_1.2,0.11443840265274048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,8,power_law_1.2,0.11848959922790528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,8,power_law_1.2,0.11813119649887086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,8,power_law_1.2,0.12584960460662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,8,power_law_1.2,0.12182400226593018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,8,power_law_1.2,0.1278720021247864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,8,power_law_1.2,0.13059200048446656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,8,power_law_1.2,0.13656959533691407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,8,balanced,0.05527999997138977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,8,balanced,0.055999999245007835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,8,balanced,0.05726400017738342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,8,balanced,0.062090665102005005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,8,balanced,0.07693333427111308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,8,balanced,0.0974133312702179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,8,balanced,0.09898666540781657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,8,balanced,0.09861333171526591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,8,balanced,0.10129066308339436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,8,balanced,0.09893332918485005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,8,balanced,0.1015786627928416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,8,balanced,0.10317867000897725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,8,balanced,0.10266666611035664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,8,balanced,0.10469333330790202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,8,balanced,0.11104533076286316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,8,power_law_1.2,0.14142719507217408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,8,balanced,0.11105066537857056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,8,power_law_1.2,0.1510975956916809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,8,balanced,0.11357866724332173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,8,balanced,0.12184533476829529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,8,balanced,0.12680000066757202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,8,balanced,0.14403200149536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,8,balanced,0.15853333473205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,8,balanced,0.18664532899856567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,8,balanced,0.21450666586558023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,8,balanced,0.29685866832733154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,8,balanced,0.3584373394648234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,8,balanced,0.5077013174692789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,8,power_law_1.2,0.16432000398635865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,8,balanced,0.6611040035883585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,8,power_law_1.2,0.1723456025123596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,8,power_law_1.2,0.15121279954910277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,8,power_law_1.2,0.16625920534133912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,8,power_law_1.2,0.183404803276062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,8,power_law_1.2,0.21802239418029784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,8,power_law_1.2,0.23175039291381835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,8,power_law_1.2,0.30364160537719725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,8,power_law_1.2,0.33496320247650146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,8,power_law_1.2,0.46483840942382815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,8,power_law_1.2,0.5883456230163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,8,power_law_1.2,0.7721983909606933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,8,power_law_1.2,1.049766445159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,4,balanced,0.03585600107908249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,4,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,4,balanced,0.03808533400297165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,4,balanced,0.04144533226887385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,4,balanced,0.041946664452552795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,4,balanced,0.043578664461771645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,4,balanced,0.04404266675313314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,4,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,4,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,4,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,4,balanced,0.044138665000597634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,4,balanced,0.05179200073083242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,4,balanced,0.051402668158213295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,4,balanced,0.05097599824269613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,4,balanced,0.0633546660343806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,4,balanced,0.06427733103434245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,4,balanced,0.06373333434263866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,4,balanced,0.08031466603279114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,4,balanced,0.08263466755549113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,4,balanced,0.10226133465766907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,4,balanced,0.12786666552225748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,4,balanced,0.15544000267982483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,4,balanced,0.19261866807937622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,4,balanced,0.26338666677474976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,4,balanced,0.34619200229644775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,4,balanced,0.5137759844462076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,4,balanced,0.6727626323699951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,128,power_law_1.2,0.05056639909744263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,128,power_law_1.2,0.05015680193901062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,128,power_law_1.2,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,128,power_law_1.2,0.053324800729751584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,128,power_law_1.2,0.054118400812149046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,128,power_law_1.2,0.05621119737625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,128,power_law_1.2,0.06279039978981019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,128,power_law_1.2,0.06307200193405152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,128,power_law_1.2,0.06741759777069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,128,power_law_1.2,0.06763520240783691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,128,power_law_1.2,0.0644927978515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,128,power_law_1.2,0.06848639845848084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,128,power_law_1.2,0.07211520075798035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,128,power_law_1.2,0.07057920098304749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,128,power_law_1.2,0.08369280099868774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,128,power_law_1.2,0.08785279989242553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,128,power_law_1.2,0.09231359958648681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,128,power_law_1.2,0.10607999563217163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,128,power_law_1.2,0.11705600023269654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,128,power_law_1.2,0.1487488031387329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,128,power_law_1.2,0.1696128010749817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,128,power_law_1.2,0.2388607978820801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,128,power_law_1.2,0.2941440105438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,128,power_law_1.2,0.4472511768341064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,128,power_law_1.2,0.6246784210205079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,128,power_law_1.2,0.9866623878479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,128,power_law_1.2,1.293081569671631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,256,balanced,0.047770669062932335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,256,balanced,0.04779199759165446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,256,balanced,0.0480373352766037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,256,balanced,0.05216533442338308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,256,balanced,0.04991999765237173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,256,balanced,0.05413866539796194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,256,balanced,0.05213333169619242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,256,balanced,0.05230933427810669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,256,balanced,0.05407466491063436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,256,balanced,0.054373333851496376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,256,balanced,0.054234668612480164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,256,balanced,0.0547626664241155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,256,balanced,0.0572320024172465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,256,balanced,0.05629866818586985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,256,balanced,0.06171200176080068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,256,balanced,0.06048533320426941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,256,balanced,0.06247466802597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,256,balanced,0.06851199766000111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,256,balanced,0.07252799967924754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,256,balanced,0.08134399851163228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,256,balanced,0.0879306693871816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,256,balanced,0.10656000177065532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,256,balanced,0.12261866529782613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,256,balanced,0.1606613298257192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,256,balanced,0.20360000928243002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,256,balanced,0.27342933416366577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,256,balanced,0.3465546766916911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,1,balanced,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,1,balanced,0.07627200086911519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,1,balanced,0.11990400155385335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,1,balanced,0.12143466869990031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,1,balanced,0.12062933047612508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,1,balanced,0.12059733271598816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,1,balanced,0.12321066856384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,1,balanced,0.12343999743461609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,1,balanced,0.12190933028856914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,1,balanced,0.12321066856384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,1,balanced,0.12637333075205484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,1,balanced,0.1280639966328939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,1,balanced,0.12803733348846436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,1,balanced,0.13379200299580893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,1,balanced,0.14441066980361938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,1,balanced,0.1470026671886444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,1,balanced,0.1474453310171763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,1,balanced,0.17865065733591715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,1,balanced,0.18196799357732138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,1,balanced,0.24395734071731567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,1,balanced,0.2746400038401286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,1,balanced,0.36905066172281903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,1,balanced,0.43166399002075195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,1,balanced,0.6151573260625204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,1,balanced,0.7759093443552653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,1,balanced,1.12990935643514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,1,balanced,1.4885172843933105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,128,power_law_1.01,0.03442560136318207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,128,power_law_1.01,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,128,power_law_1.01,0.03425920009613037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,128,power_law_1.01,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,128,power_law_1.01,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,128,power_law_1.01,0.03845120072364807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,128,power_law_1.01,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,128,power_law_1.01,0.03813120126724243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,128,power_law_1.01,0.03800959885120392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,128,power_law_1.01,0.03777920007705689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,128,power_law_1.01,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,128,power_law_1.01,0.038624000549316403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,128,power_law_1.01,0.03849599957466125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,128,power_law_1.01,0.03930880129337311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,128,power_law_1.01,0.042316800355911253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,128,power_law_1.01,0.04286719858646393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,128,power_law_1.01,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,128,power_law_1.01,0.04681600034236908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,128,power_law_1.01,0.04921599924564361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,128,power_law_1.01,0.055251199007034305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,128,power_law_1.01,0.059487998485565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,128,power_law_1.01,0.0723904013633728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,128,power_law_1.01,0.08492799997329711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,128,power_law_1.01,0.11816960573196411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,128,power_law_1.01,0.1451583981513977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,128,power_law_1.01,0.19313280582427977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,128,power_law_1.01,0.25427200794219973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,4,power_law_1.2,0.020422400534152986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,4,power_law_1.2,0.02028159946203232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,4,power_law_1.2,0.0216511994600296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,4,power_law_1.2,0.023801599442958832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,4,power_law_1.2,0.02504960000514984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,4,power_law_1.2,0.028672000765800475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,4,power_law_1.2,0.029529601335525513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,4,power_law_1.2,0.030873599648475646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,4,power_law_1.2,0.03238399922847748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,4,power_law_1.2,0.03218559920787811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,4,power_law_1.2,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,4,power_law_1.2,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,4,power_law_1.2,0.04021120071411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,4,power_law_1.2,0.039654400944709775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,4,power_law_1.2,0.042694398760795595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,4,power_law_1.2,0.04633600115776062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,4,power_law_1.2,0.04843519926071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,4,power_law_1.2,0.055667197704315184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.2,0.09550079703330994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.2,0.10451840162277222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.2,0.09290879964828491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.2,0.11667840480804444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.2,0.13339519500732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.2,0.1814144015312195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.2,0.2124351978302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.2,0.30415360927581786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,32,power_law_1.01,0.05140479803085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,32,power_law_1.01,0.046239998936653134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,32,power_law_1.01,0.04683519899845123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,32,power_law_1.01,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,32,power_law_1.01,0.0493120014667511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,32,power_law_1.01,0.049721598625183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,32,power_law_1.01,0.05128960013389587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,32,power_law_1.01,0.05203840136528015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,32,power_law_1.01,0.05259519815444946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,32,power_law_1.01,0.05372160077095032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,8,power_law_1.2,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,32,power_law_1.01,0.05421440005302429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,8,power_law_1.2,0.05608959794044495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,32,power_law_1.01,0.05630080103874206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,32,power_law_1.01,0.05678079724311828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,32,power_law_1.01,0.059673601388931276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,32,power_law_1.01,0.06652160286903382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,32,power_law_1.01,0.06695039868354798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,32,power_law_1.01,0.07144960165023803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,32,power_law_1.01,0.09031040072441102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,8,power_law_1.2,0.055244797468185426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,8,power_law_1.2,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,8,power_law_1.2,0.055731201171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,8,power_law_1.2,0.05857279896736145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,8,power_law_1.2,0.0579584002494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,8,power_law_1.2,0.05939199924468994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,8,power_law_1.2,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,8,power_law_1.2,0.06503679752349853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,8,power_law_1.2,0.06755840182304382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,32,power_law_1.01,0.09319040179252625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,8,power_law_1.2,0.07099519968032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,32,power_law_1.01,0.12379519939422608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,8,power_law_1.2,0.07209600210189819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,32,power_law_1.01,0.14543360471725464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,8,power_law_1.2,0.07714560031890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,32,power_law_1.01,0.19094400405883788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,8,power_law_1.2,0.09608960151672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,32,power_law_1.01,0.22548480033874513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,8,power_law_1.2,0.10161279439926148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,32,power_law_1.01,0.31217279434204104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,8,power_law_1.2,0.10151679515838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,32,power_law_1.01,0.39676799774169924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,32,power_law_1.01,0.5808320045471191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,8,power_law_1.2,0.12744959592819213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,32,power_law_1.01,0.7666751861572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.2,0.1442944049835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.2,0.18592000007629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.2,0.22282240390777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,1,power_law_1.01,0.06635519862174988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,1,power_law_1.01,0.08814079761505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,1,power_law_1.01,0.10452480316162109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,1,power_law_1.01,0.13723520040512086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.2,0.3134016036987305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,1,power_law_1.01,0.13806079626083373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,1,power_law_1.01,0.14438400268554688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,1,power_law_1.01,0.14461439847946167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,1,power_law_1.01,0.14663679599761964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,1,power_law_1.01,0.14664959907531738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,1,power_law_1.01,0.14744319915771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,1,power_law_1.01,0.15124479532241822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,1,power_law_1.01,0.15527039766311646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,1,power_law_1.01,0.16027519702911378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,1,power_law_1.01,0.16563839912414552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,1,power_law_1.01,0.18310400247573852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,1,power_law_1.01,0.18424960374832153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,1,power_law_1.01,0.20533759593963624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.2,0.3956415891647339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,1,power_law_1.01,0.23724160194396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.01,0.27204480171203616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.01,0.3363071918487549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.01,0.41254401206970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.2,0.3831295967102051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.01,0.5595263957977294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.01,0.6964735984802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.01,0.9934080123901368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.2,0.6007872104644776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.01,1.26046724319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.2,0.7592832088470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.01,1.8400127410888671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.2,1.1293824195861817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.01,2.4140607833862306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.2,1.5745280265808106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,4,power_law_1.01,0.15168000459671022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,4,power_law_1.01,0.24151039123535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,4,power_law_1.01,0.3310784101486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,4,power_law_1.01,0.3945472002029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,4,power_law_1.01,0.5194047927856446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,4,power_law_1.01,0.4853375911712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,4,power_law_1.01,0.7573631763458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,4,power_law_1.01,0.684115219116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,4,power_law_1.01,0.6956607818603515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,4,power_law_1.01,0.7154560089111328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,4,power_law_1.01,0.7032000064849854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,4,power_law_1.01,0.7222335815429688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,4,power_law_1.01,0.761407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,4,power_law_1.01,0.7740863800048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,4,power_law_1.01,0.7484096050262451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,4,power_law_1.01,0.8148223876953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,4,power_law_1.01,0.815833568572998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,4,power_law_1.01,0.9058815956115722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,4,power_law_1.01,0.9121088027954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,4,power_law_1.01,1.0411264419555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,4,power_law_1.01,1.0229375839233399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,4,power_law_1.01,1.1772095680236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,4,power_law_1.01,1.385439968109131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,4,power_law_1.01,1.6360576629638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,4,power_law_1.01,1.8909055709838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,4,power_law_1.01,2.56759033203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,4,power_law_1.01,2.8595455169677733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,4,power_law_1.2,0.03369599878787995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,4,power_law_1.2,0.03209599852561951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,4,power_law_1.2,0.031564798951148984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,4,power_law_1.2,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,4,power_law_1.2,0.03244799971580505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,4,power_law_1.2,0.03293440043926239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,4,power_law_1.2,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,4,power_law_1.2,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,4,power_law_1.2,0.03400320112705231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,4,power_law_1.2,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,4,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,4,power_law_1.2,0.038464000821113585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,4,power_law_1.2,0.03790720105171204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,4,power_law_1.2,0.040320000052452086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,4,power_law_1.2,0.046105599403381346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,4,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,4,power_law_1.2,0.049958398938179015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,4,power_law_1.2,0.06402559876441956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,4,power_law_1.2,0.06681600213050842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,4,power_law_1.2,0.09055359959602356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,32,balanced,0.04380266865094503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,32,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,32,balanced,0.04191466669241587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,32,balanced,0.04659733176231384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,32,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,32,balanced,0.05128533144791921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,32,balanced,0.05178666611512502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,32,balanced,0.05202133456865946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,32,balanced,0.05349333087603251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,32,balanced,0.05397333204746246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,32,balanced,0.05415999889373779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,32,balanced,0.05613866448402405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,32,balanced,0.058005332946777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,32,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,32,balanced,0.06320000191529591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,32,balanced,0.06434666613737743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,32,balanced,0.06623999774456024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,32,balanced,0.07454399764537811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,32,balanced,0.08062933385372162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,32,balanced,0.10657067100207011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,32,balanced,0.1260640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,32,balanced,0.1651946703592936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,32,balanced,0.19753599166870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,32,balanced,0.26544533173243207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,32,balanced,0.3394293387730916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,4,power_law_1.2,0.09535359740257263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,32,balanced,0.47754132747650146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,32,balanced,0.6037013530731201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,4,power_law_1.2,0.1262336015701294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,4,power_law_1.2,0.13726719617843627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,4,power_law_1.2,0.21069440841674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,1,power_law_1.2,0.025817599892616273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,1,power_law_1.2,0.027289599180221558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,1,power_law_1.2,0.026080000400543212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,1,power_law_1.2,0.03136639893054962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,1,power_law_1.2,0.044838398694992065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,1,power_law_1.2,0.06085119843482971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,1,power_law_1.2,0.05203199982643127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,1,power_law_1.2,0.06069759726524353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,1,power_law_1.2,0.051148802042007446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,1,power_law_1.2,0.05533440113067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,1,power_law_1.2,0.06474879980087281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,4,power_law_1.2,0.24300799369812012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,1,power_law_1.2,0.05966079831123352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,1,power_law_1.2,0.06445440053939819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,1,power_law_1.2,0.07614079713821412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,1,power_law_1.2,0.16128000020980834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,1,power_law_1.2,0.1835263967514038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,1,power_law_1.2,0.2171583890914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,1,power_law_1.2,0.11816320419311524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.2,0.14343039989471434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.2,0.20195839405059815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.2,0.25659520626068116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.2,0.3588671922683716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,4,power_law_1.2,0.10585600137710571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,4,power_law_1.2,0.11985280513763427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,4,power_law_1.2,0.15139199495315553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,4,power_law_1.2,0.161407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,4,power_law_1.2,0.17980799674987794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,4,power_law_1.2,0.2738879919052124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,4,power_law_1.2,0.2786240100860596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,4,power_law_1.2,0.28047358989715576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,4,power_law_1.2,0.29151999950408936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,4,power_law_1.2,0.28399360179901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,4,power_law_1.2,0.30881919860839846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,4,power_law_1.2,0.32115840911865234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,4,power_law_1.2,0.3288703918457031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,4,power_law_1.2,0.3378432035446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,4,power_law_1.2,0.3577280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.2,0.46698880195617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,4,power_law_1.2,0.3498944044113159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,4,power_law_1.2,0.3788608074188232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,4,power_law_1.2,0.45336318016052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,4,power_law_1.2,0.44282240867614747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,4,power_law_1.2,0.5425280094146728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,4,power_law_1.2,0.5931263923645019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,4,power_law_1.2,0.7040575981140137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,4,power_law_1.2,0.8844479560852051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,4,power_law_1.2,1.0285823822021485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,4,power_law_1.2,1.457203197479248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,4,power_law_1.2,1.9022783279418944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,4,power_law_1.2,2.3737472534179687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,4,power_law_1.2,0.33535358905792234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.2,0.675923204421997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,16,power_law_1.2,0.057132798433303836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,16,power_law_1.2,0.06775680184364319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,16,power_law_1.2,0.06431360244750976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,16,power_law_1.2,0.07275519967079162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,16,power_law_1.2,0.07200639843940734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,16,power_law_1.2,0.06748160123825073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,16,power_law_1.2,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,16,power_law_1.2,0.07244799733161926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.2,0.8886783599853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,16,power_law_1.2,0.07422720193862915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,16,power_law_1.2,0.07386879920959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,16,power_law_1.2,0.07459200024604798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,16,power_law_1.2,0.07670400142669678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,16,power_law_1.2,0.07779840230941773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,16,power_law_1.2,0.07905279994010925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,16,power_law_1.2,0.08430079817771911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,16,power_law_1.2,0.08427519798278808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,16,power_law_1.2,0.08887680172920227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,16,power_law_1.2,0.09809280037879944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,16,power_law_1.2,0.11399680376052856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,16,power_law_1.2,0.13381760120391845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,16,power_law_1.2,0.15560959577560424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,16,power_law_1.2,0.19196159839630128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,16,power_law_1.2,0.2412735939025879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,16,power_law_1.2,0.3612224102020264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,16,power_law_1.2,0.4328320026397705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,16,power_law_1.2,0.6400767803192139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,16,power_law_1.2,0.868166446685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,4,power_law_1.2,0.44403839111328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.2,1.3070207595825196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.2,1.725644874572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,32,power_law_1.01,0.05719040036201477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,32,power_law_1.01,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,32,power_law_1.01,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,32,power_law_1.01,0.050367999076843264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,32,power_law_1.01,0.05097600221633911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,32,power_law_1.01,0.05236480236053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,32,power_law_1.01,0.05124480128288269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,32,power_law_1.01,0.05279359817504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,32,power_law_1.01,0.05199360251426697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,32,power_law_1.01,0.0548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,32,power_law_1.01,0.05497599840164184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,32,power_law_1.01,0.057017600536346434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,32,power_law_1.01,0.05487359762191772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,32,power_law_1.01,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,32,power_law_1.01,0.06484479904174804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,32,power_law_1.01,0.06618880033493042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,32,power_law_1.01,0.07095680236816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,32,power_law_1.01,0.07943040132522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,32,power_law_1.01,0.08574079871177673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,32,power_law_1.01,0.10104320049285889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,32,power_law_1.01,0.11721600294113159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,32,power_law_1.01,0.14115840196609497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,32,power_law_1.01,0.17265280485153198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,32,power_law_1.01,0.22696321010589598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,32,power_law_1.01,0.2920063972473145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,32,power_law_1.01,0.40232319831848146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,32,power_law_1.01,0.5037504196166992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,4,balanced,0.03161599983771642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,4,balanced,0.03223466624816259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,4,balanced,0.03162666658560435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,4,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,4,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,4,balanced,0.0580213318268458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,4,balanced,0.056976000467936196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,4,balanced,0.058880001306533813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,4,balanced,0.05795733133951823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,4,balanced,0.05993066728115082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,4,balanced,0.05975466469923655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,4,balanced,0.05946666498978933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,4,balanced,0.0590826670328776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,4,balanced,0.06048533320426941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,4,balanced,0.06522666911284129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,4,balanced,0.06690133114655812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,4,balanced,0.06630399823188782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,4,balanced,0.07815999786059062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,4,balanced,0.07364266614119212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,4,balanced,0.08463999629020691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,4,balanced,0.08930133779843648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,4,balanced,0.10981866717338562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,4,balanced,0.12300800283749898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,4,balanced,0.16671999295552573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,4,balanced,0.18489599227905273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,4,balanced,0.24927999575932822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,4,balanced,0.3003893295923869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,8,power_law_1.01,0.03258239924907684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,8,power_law_1.01,0.0336896002292633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,8,power_law_1.01,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,8,power_law_1.01,0.03283199965953827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,8,power_law_1.01,0.033228799700737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,8,power_law_1.01,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,8,power_law_1.01,0.03744640052318573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,8,power_law_1.01,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,8,power_law_1.01,0.03948160111904144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,8,power_law_1.01,0.04090240001678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,8,power_law_1.01,0.039263999462127684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,8,power_law_1.01,0.043782401084899905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,8,power_law_1.01,0.04445439875125885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,8,power_law_1.01,0.048076799511909483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,8,power_law_1.01,0.05552639961242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,8,power_law_1.01,0.05959039926528931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,8,power_law_1.01,0.062054401636123656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,8,power_law_1.01,0.0772607982158661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,4,power_law_1.01,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,8,power_law_1.01,0.1893887996673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,4,power_law_1.01,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,8,power_law_1.01,0.08813440203666686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,4,power_law_1.01,0.07884160280227662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,8,power_law_1.01,0.3195456027984619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,4,power_law_1.01,0.09473279714584351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,8,power_law_1.01,0.25214080810546874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,4,power_law_1.01,0.10172159671783447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,4,power_law_1.01,0.11773439645767211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,8,power_law_1.01,0.35294079780578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,4,power_law_1.01,0.12933759689331054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,8,power_law_1.01,0.11610239744186401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,8,power_law_1.01,0.40101118087768556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,4,power_law_1.01,0.13240959644317626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,8,power_law_1.01,0.1344383955001831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,4,power_law_1.01,0.1325824022293091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,8,power_law_1.01,0.43351039886474607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,4,power_law_1.01,0.13236479759216307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,8,power_law_1.01,0.4145472049713135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,4,power_law_1.01,0.13269120454788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,4,power_law_1.01,0.1451583981513977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,8,power_law_1.01,0.3939519882202148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,4,power_law_1.01,0.14300160408020018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,8,power_law_1.01,0.4137279987335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,4,power_law_1.01,0.14632960557937622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,4,power_law_1.01,0.15777920484542846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,4,power_law_1.01,0.16352640390396117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,4,power_law_1.01,0.17014399766921998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,4,power_law_1.01,0.19684480428695678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,4,power_law_1.01,0.21612799167633057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,4,power_law_1.01,0.2618880033493042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,4,power_law_1.01,0.29230079650878904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,4,power_law_1.01,0.3779007911682129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,4,power_law_1.01,0.45210881233215333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,4,power_law_1.01,0.629311990737915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,4,power_law_1.01,0.7877503871917725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,8,power_law_1.01,0.4054719924926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,8,power_law_1.01,0.18848639726638794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,8,power_law_1.01,0.20570240020751954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,8,power_law_1.01,0.44087681770324705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,8,power_law_1.01,0.4263167858123779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,8,power_law_1.01,0.4289599895477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,8,power_law_1.01,0.4451263904571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,8,power_law_1.01,0.4742271900177002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,4,power_law_1.01,1.0794688224792481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,8,power_law_1.01,0.47844481468200684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,8,power_law_1.01,0.5194943904876709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,4,power_law_1.01,1.3094016075134278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,8,power_law_1.01,0.60382080078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,8,power_law_1.01,0.5572991847991944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,8,power_law_1.01,0.6744063854217529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,8,power_law_1.01,0.7164735794067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,8,power_law_1.01,0.8761216163635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,8,power_law_1.01,0.94268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,8,power_law_1.01,1.1501440048217773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,8,power_law_1.01,1.4701055526733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,8,power_law_1.01,0.30035200119018557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,8,power_law_1.01,1.915545654296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,8,power_law_1.01,2.323436737060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,8,power_law_1.01,0.3728960037231445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,8,power_law_1.01,0.5411903858184814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,8,power_law_1.01,0.7184256076812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,2,power_law_1.2,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,2,power_law_1.2,0.03371520042419433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,2,power_law_1.2,0.03466239869594574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,2,power_law_1.2,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,2,power_law_1.2,0.0389631986618042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,2,power_law_1.2,0.045203199982643126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,2,power_law_1.2,0.045433598756790164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,2,power_law_1.2,0.04615040123462677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,2,power_law_1.2,0.049414399266242984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,2,power_law_1.2,0.051475197076797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,2,power_law_1.2,0.053439998626708986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,2,power_law_1.2,0.05655679702758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,2,power_law_1.2,0.060147202014923094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,2,power_law_1.2,0.06359040141105651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,2,power_law_1.2,0.07199360132217407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,2,power_law_1.2,0.07488639950752259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,2,power_law_1.2,0.07509120106697083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,2,power_law_1.2,0.09128320217132568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,2,power_law_1.2,0.10369919538497925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,2,power_law_1.2,0.1356927990913391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,2,power_law_1.2,0.15217920541763305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,2,power_law_1.2,0.19845759868621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,2,power_law_1.2,0.2788800001144409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,2,power_law_1.2,0.37223680019378663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,2,power_law_1.2,0.5283455848693848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,2,power_law_1.2,0.7724607944488525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,32,balanced,0.07454933226108551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,32,balanced,0.0499893327554067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,32,balanced,0.05825066566467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,2,power_law_1.2,0.9819071769714356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,32,balanced,0.07793599863847096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,32,balanced,0.10563199718793233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,32,balanced,0.1151039997736613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,32,balanced,0.11549866199493408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,32,balanced,0.11745066444079082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,32,balanced,0.11617599924405415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,32,balanced,0.11741333206494649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,32,balanced,0.11823466420173645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,32,balanced,0.11823466420173645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,32,balanced,0.11834133664766948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,32,balanced,0.11962133646011353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,32,balanced,0.12503467003504434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,32,balanced,0.12705066800117493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,32,balanced,0.1297706663608551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,32,balanced,0.13593066732088724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,32,balanced,0.1357493301232656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,32,balanced,0.17453867197036743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,32,balanced,0.15500799814860025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,32,balanced,0.22842133045196533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,32,balanced,0.20383999745051065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,32,balanced,0.2958986759185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,32,balanced,0.3004053235054016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,32,balanced,0.47164801756540936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,32,balanced,0.4988853136698405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,256,balanced,0.019018666197856266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,256,balanced,0.019402666638294857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,256,balanced,0.019007999449968338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,256,balanced,0.023120000958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,256,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,256,balanced,0.03133866687615713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,256,balanced,0.03163733333349228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,256,balanced,0.031557333966096245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,256,balanced,0.03179733455181122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,256,balanced,0.031557333966096245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,256,balanced,0.031311998764673867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,256,balanced,0.032501332461833954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,256,balanced,0.03263466556866964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,256,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,256,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,256,balanced,0.03494933247566223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,256,balanced,0.035946667194366455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,256,balanced,0.03811733424663544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,256,balanced,0.039173332353432976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,256,balanced,0.04418133199214935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,256,balanced,0.05294933418432871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,256,balanced,0.06658666829268138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,256,balanced,0.08104533453782399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,256,balanced,0.11142399907112122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,256,balanced,0.13918933272361755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,256,balanced,0.19448532660802206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,256,balanced,0.2514773408571879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,1,power_law_1.2,0.16106239557266236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,1,power_law_1.2,0.24678399562835693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,1,power_law_1.2,0.3670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,1,power_law_1.2,0.5266816139221191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,1,power_law_1.2,0.7253695964813233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,1,power_law_1.2,0.9487039566040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,1,power_law_1.2,1.332044792175293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,1,power_law_1.2,1.4049344062805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,1,power_law_1.2,1.4574336051940917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,1,power_law_1.2,1.5025279998779297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,1,power_law_1.2,1.5256447792053223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,1,power_law_1.2,1.5798975944519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,1,power_law_1.2,1.5880703926086426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,1,power_law_1.2,1.6461248397827148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,1,power_law_1.2,1.632192039489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,1,power_law_1.2,1.6925888061523438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,1,power_law_1.2,1.7902208328247071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,1,power_law_1.2,1.8927871704101562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.2,1.9649408340454102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.2,2.204697608947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.2,2.4101631164550783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.2,2.828268814086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.2,3.267782211303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.2,4.1111297607421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.2,4.974700927734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.2,6.627597045898438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.2,8.125772857666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,2,power_law_1.2,0.05480319857597351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,2,power_law_1.2,0.06198400259017944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,2,power_law_1.2,0.06746879816055298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,2,power_law_1.2,0.07550719976425171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,2,power_law_1.2,0.08401920199394226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,2,power_law_1.2,0.10428800582885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,2,power_law_1.2,0.10522880554199218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,2,power_law_1.2,0.10794880390167236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,2,power_law_1.2,0.10860799551010132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,2,power_law_1.2,0.10794240236282349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,2,power_law_1.2,0.11296000480651855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,2,power_law_1.2,0.11644159555435181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,2,power_law_1.2,0.11744639873504639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,2,power_law_1.2,0.12015360593795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,2,power_law_1.2,0.11962239742279053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,2,power_law_1.2,0.1215999960899353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,2,power_law_1.2,0.1272447943687439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,2,power_law_1.2,0.14142719507217408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,2,power_law_1.2,0.16499840021133422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,2,power_law_1.2,0.19474560022354126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,2,power_law_1.2,0.23161599636077881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,2,power_law_1.2,0.315833592414856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,2,power_law_1.2,0.3926912069320679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,2,power_law_1.2,0.5661952018737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,2,power_law_1.2,0.7303232192993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,2,power_law_1.2,0.018131199479103088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,2,power_law_1.2,0.017932799458503724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,2,power_law_1.2,0.018086400628089905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,2,power_law_1.2,0.017875200510025023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,2,power_law_1.2,0.01812479943037033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,2,power_law_1.2,0.019654400646686554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,2,power_law_1.2,0.020236800611019134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,2,power_law_1.2,0.02022400051355362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,2,power_law_1.2,0.02142080068588257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,2,power_law_1.2,0.02144639939069748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,2,power_law_1.2,0.0230335995554924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,2,power_law_1.2,0.023878400027751923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,2,power_law_1.2,0.024633599817752837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,2,power_law_1.2,0.0265855997800827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,2,power_law_1.2,0.02707839906215668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,2,power_law_1.2,0.029023998975753786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,2,power_law_1.2,0.03948160111904144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,2,power_law_1.2,0.04252159893512726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.2,0.04524799883365631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.2,0.05256959795951843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,2,power_law_1.2,1.057164764404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.2,0.06183040142059326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.2,0.08449919819831848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.2,0.09898880124092102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.2,0.13377280235290528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,2,power_law_1.2,1.3968000411987305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.2,0.17110400199890136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,16,balanced,0.045791998505592346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,16,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,16,balanced,0.0460746685663859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,16,balanced,0.04841066896915436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,16,balanced,0.053898667295773826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,16,balanced,0.06573866804440816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,16,balanced,0.06480533381303151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,16,balanced,0.0681279997030894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,16,balanced,0.06607466439406078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,16,balanced,0.06687999765078227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,16,balanced,0.06763199965159099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,16,balanced,0.06986666719118755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,16,balanced,0.06845333178838094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,16,balanced,0.07152533531188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,16,balanced,0.07613333563009898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,16,balanced,0.07634666562080383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,16,balanced,0.07835733393828075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,16,balanced,0.08689066767692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,16,balanced,0.08687466382980347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,16,balanced,0.1051626702149709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,16,balanced,0.11322666207949321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,16,balanced,0.1439786652723948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,16,balanced,0.15015467007954916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,16,balanced,0.20292266209920248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,16,balanced,0.2404586672782898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,16,balanced,0.3278239965438843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,16,balanced,0.40009601910909015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.2,0.24396159648895263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,4,power_law_1.2,0.04135040044784546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,4,power_law_1.2,0.04402559995651245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,4,power_law_1.2,0.04602240025997162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,4,power_law_1.2,0.05013759732246399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,4,power_law_1.2,0.05453439950942993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,4,power_law_1.2,0.05772160291671753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,4,power_law_1.2,0.0640447974205017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,4,power_law_1.2,0.06731520295143127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,4,power_law_1.2,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,4,power_law_1.2,0.06595839858055115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,4,power_law_1.2,0.07292159795761108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,4,power_law_1.2,0.07669119834899903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,4,power_law_1.2,0.07676799893379212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,4,power_law_1.2,0.0818943977355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,4,power_law_1.2,0.09100800156593322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,4,power_law_1.2,0.09178879857063293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,4,power_law_1.2,0.09721599817276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,4,power_law_1.2,0.11358079910278321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,4,power_law_1.2,0.12316160202026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,4,power_law_1.2,0.1549504041671753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,4,power_law_1.2,0.17458560466766357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,4,power_law_1.2,0.23260159492492677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,4,power_law_1.2,0.2817280054092407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,4,power_law_1.2,0.38353919982910156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,4,power_law_1.2,0.5333504199981689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,4,power_law_1.2,0.7156799793243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,4,power_law_1.2,0.9453120231628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.2,0.304422402381897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,2,balanced,0.05269866685072581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,2,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,2,balanced,0.07218133409818013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,2,balanced,0.10173867146174113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,2,balanced,0.15404799580574036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,2,balanced,0.2520266572634379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,2,balanced,0.3613119920094808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,2,balanced,0.3593120177586873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,2,balanced,0.359386682510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,2,balanced,0.36139734586079914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,2,balanced,0.36559466520945233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,2,balanced,0.3768426577250163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,2,balanced,0.37884799639383954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,2,balanced,0.3758133252461751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,2,balanced,0.38742931683858234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,2,balanced,0.3948853413263957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,2,balanced,0.4107999801635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,2,balanced,0.43458131949106854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,2,balanced,0.4543519814809163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,2,balanced,0.4976213375727336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,2,balanced,0.5446720123291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,2,balanced,0.6314560174942017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,2,balanced,0.7444746494293213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,2,balanced,0.9285439650217692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,2,balanced,1.1916426817576091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,2,balanced,1.572938601175944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,2,balanced,2.044037342071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,4,power_law_1.2,0.06604160070419311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,4,power_law_1.2,0.07175679802894593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,4,power_law_1.2,0.07879679799079894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,4,power_law_1.2,0.08418560028076172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,4,power_law_1.2,0.09475200176239014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,4,power_law_1.2,0.106387197971344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,4,power_law_1.2,0.11149439811706544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,4,power_law_1.2,0.13541120290756226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,4,power_law_1.2,0.13496960401535035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,4,power_law_1.2,0.1300160050392151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,4,power_law_1.2,0.13159040212631226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,4,power_law_1.2,0.1362944006919861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,4,power_law_1.2,0.13704960346221923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,4,power_law_1.2,0.13749120235443116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,4,power_law_1.2,0.1460096001625061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,4,power_law_1.2,0.15117440223693848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,4,power_law_1.2,0.1599552035331726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,4,power_law_1.2,0.17902719974517822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,4,power_law_1.2,0.2008960008621216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,4,power_law_1.2,0.23773438930511476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,4,power_law_1.2,0.2790080070495605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,4,power_law_1.2,0.36678400039672854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,4,power_law_1.2,0.42559361457824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,4,power_law_1.2,0.6058815956115723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,4,power_law_1.2,0.7681215763092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,4,power_law_1.2,1.109107208251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,4,power_law_1.2,1.3930432319641113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,8,power_law_1.01,0.02200320065021515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,8,power_law_1.01,0.023071999847888946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,8,power_law_1.01,0.02489600032567978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,8,power_law_1.01,0.02602880001068115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,8,power_law_1.01,0.027750399708747864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,8,power_law_1.01,0.02901119887828827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,8,power_law_1.01,0.02908799946308136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,8,power_law_1.01,0.03025279939174652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,8,power_law_1.01,0.031065601110458373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,8,power_law_1.01,0.031379199028015135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,8,power_law_1.01,0.03225600123405457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,8,power_law_1.01,0.04069760143756866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,8,power_law_1.01,0.03984639942646027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,8,power_law_1.01,0.04165120124816894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,8,power_law_1.2,0.05259519815444946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,8,power_law_1.01,0.06718080043792725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,8,power_law_1.2,0.06821119785308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,8,power_law_1.01,0.06853119730949402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,8,power_law_1.2,0.07163519859313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,8,power_law_1.01,0.07111679911613464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,8,power_law_1.2,0.08625280261039733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,8,power_law_1.2,0.0846015989780426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,8,power_law_1.2,0.09114239811897278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,8,power_law_1.2,0.09799039959907532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,8,power_law_1.2,0.09371520280838012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,8,power_law_1.01,0.07478399872779846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.01,0.08115199804306031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.01,0.0916159987449646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.01,0.09678080081939697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.01,0.11871999502182007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,8,power_law_1.2,0.09694719910621644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,8,power_law_1.2,0.0992576003074646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.01,0.13788800239562987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,8,power_law_1.2,0.10038399696350098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.01,0.19447040557861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,8,power_law_1.2,0.10030080080032348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,8,power_law_1.2,0.10366719961166382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.01,0.23811841011047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,8,power_law_1.2,0.10970239639282227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.01,0.32337279319763185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.01,0.41021437644958497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,8,power_law_1.2,0.11519999504089355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,8,power_law_1.2,0.11918720006942748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,8,power_law_1.2,0.124236798286438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,8,power_law_1.2,0.145632004737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,8,power_law_1.2,0.1567744016647339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,8,power_law_1.2,0.1958400011062622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,8,power_law_1.2,0.20850560665130616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,8,power_law_1.2,0.25386240482330324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,8,power_law_1.2,0.3212160110473633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,8,power_law_1.2,0.4275199890136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,8,power_law_1.2,0.5231743812561035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,8,power_law_1.2,0.7386496067047119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,8,power_law_1.2,1.0266752243041992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,8,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,8,balanced,0.05171200136343638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,8,balanced,0.053632001082102455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,8,balanced,0.06437333424886067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,8,balanced,0.08282133440176646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,8,balanced,0.11347732941309611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,8,balanced,0.11369066437085469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,8,balanced,0.11692800124486287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,8,balanced,0.1151039997736613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,8,balanced,0.11822400490442912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,8,balanced,0.11762666702270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,8,balanced,0.11901332934697469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,8,balanced,0.12053866187731425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,8,balanced,0.12131200234095256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,8,balanced,0.1288159986337026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,8,balanced,0.1316106617450714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,8,balanced,0.13421333829561868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,8,balanced,0.14485333363215128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,8,balanced,0.1544373333454132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,8,balanced,0.17251733938852945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,8,balanced,0.19123733043670654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,8,balanced,0.2330133318901062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,8,balanced,0.271232008934021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,8,balanced,0.35865068435668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,8,balanced,0.4281653165817261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,8,balanced,0.5954240163167318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,8,balanced,0.7682720025380453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,128,balanced,0.019088000059127808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,128,balanced,0.021301334102948506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,128,balanced,0.021040000021457672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,128,balanced,0.021114667256673176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,128,balanced,0.02348266790310542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,128,balanced,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,128,balanced,0.034341332813103996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,128,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,128,balanced,0.034261333445707955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,128,balanced,0.034645333886146545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,128,balanced,0.03475733349720637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,128,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,128,balanced,0.033359999457995095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,128,balanced,0.03409600009520849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,128,balanced,0.03920000046491623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,128,balanced,0.037317333122094475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,128,balanced,0.037392000357309975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,128,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,128,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,128,balanced,0.04994133114814758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,128,balanced,0.06224533418814341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,128,balanced,0.07553066809972127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,128,balanced,0.09409067034721375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,128,balanced,0.12888532876968384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,128,balanced,0.16218133767445883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,128,balanced,0.22708799441655478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,128,balanced,0.29267199834187824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,64,balanced,0.03325333446264267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,64,balanced,0.03631466627120972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,64,balanced,0.031184000273545582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,64,balanced,0.031162666777769726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,64,balanced,0.03251733382542928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,64,balanced,0.03431999931732813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,64,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,64,balanced,0.03388266762097677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,64,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,64,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,64,balanced,0.0352960005402565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,64,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,64,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,64,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,64,balanced,0.037920000652472176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,64,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,64,balanced,0.039690665900707245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,64,balanced,0.03933866570393244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,64,balanced,0.04106133431196213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,64,balanced,0.0454720010360082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,64,balanced,0.04420800010363261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,64,balanced,0.05057600140571594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,64,balanced,0.058602665861447654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,64,balanced,0.07080533107121785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,64,balanced,0.08085866769154866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,64,balanced,0.09357333183288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,64,balanced,0.11989866693814595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,4,power_law_1.01,0.055283200740814206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,4,power_law_1.01,0.060057598352432254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,4,power_law_1.01,0.06516479849815368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,4,power_law_1.01,0.0681984007358551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,4,power_law_1.01,0.0725823998451233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,4,power_law_1.01,0.0867904007434845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,4,power_law_1.01,0.08988159894943237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,4,power_law_1.01,0.08357120156288148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,4,power_law_1.01,0.08826239705085755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,4,power_law_1.01,0.08472959995269776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,4,power_law_1.01,0.0864960014820099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,4,power_law_1.01,0.0881663978099823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,4,power_law_1.01,0.0871999979019165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,4,power_law_1.01,0.08994560241699219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,4,power_law_1.01,0.09560959935188293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,4,power_law_1.01,0.09627519845962525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,4,power_law_1.01,0.1035264015197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,4,power_law_1.01,0.11646080017089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.01,0.1279680013656616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.01,0.15083520412445067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.01,0.18268799781799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.01,0.22654080390930176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.01,0.30264959335327146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.01,0.3714560031890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.01,0.5176640033721924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.01,0.7237120151519776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.01,0.9320320129394531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,8,balanced,0.04207466542720795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,8,balanced,0.04381333291530609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,8,balanced,0.04426133135954539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,8,balanced,0.052058666944503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,8,balanced,0.06830933193365733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,8,balanced,0.09431466460227966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,8,balanced,0.09686400492986043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,8,balanced,0.09778133034706116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,8,balanced,0.09711999694506328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,8,balanced,0.09905067086219788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,8,balanced,0.10110400120417277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,8,balanced,0.10114133358001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,8,balanced,0.10256000359853108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,8,balanced,0.10428266723950703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,8,balanced,0.1076586643854777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,8,balanced,0.11141866445541382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,8,balanced,0.11310933033625285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,8,balanced,0.12141333023707072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,8,balanced,0.12640532851219177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,8,balanced,0.1439359982808431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,8,balanced,0.16406933466593424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,8,balanced,0.1957119901974996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,8,balanced,0.22773333390553793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,8,balanced,0.2896266579627991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,8,balanced,0.356389323870341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,8,balanced,0.4928319851557414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,8,balanced,0.6201066573460897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,64,power_law_1.01,0.06773120164871216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,64,power_law_1.01,0.06472319960594178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,64,power_law_1.01,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,64,power_law_1.01,0.06415359973907471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,64,power_law_1.01,0.06615039706230164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,64,power_law_1.01,0.06723840236663818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,64,power_law_1.01,0.06694399714469909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,64,power_law_1.01,0.06935039758682252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,64,power_law_1.01,0.06785920262336731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,64,power_law_1.01,0.06819199919700622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,64,power_law_1.01,0.0680895984172821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,64,power_law_1.01,0.0700543999671936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,64,power_law_1.01,0.07169280052185059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,64,power_law_1.01,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,64,power_law_1.01,0.07908480167388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,64,power_law_1.01,0.08037760257720947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,64,power_law_1.01,0.0869376003742218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,64,power_law_1.01,0.09932799935340882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,64,power_law_1.01,0.11201920509338378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,64,power_law_1.01,0.14111360311508178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,64,power_law_1.01,0.163756799697876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,64,power_law_1.01,0.21232640743255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,64,power_law_1.01,0.2599616050720215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,64,power_law_1.01,0.36186881065368653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,64,power_law_1.01,0.4608895778656006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,64,power_law_1.01,0.6595647811889649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,64,power_law_1.01,0.8641216278076171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,8,power_law_1.2,0.04647679924964905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,8,power_law_1.2,0.045510399341583255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,8,power_law_1.2,0.045414400100708005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,8,power_law_1.2,0.0471807986497879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,8,power_law_1.2,0.04754559993743897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,8,power_law_1.2,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,8,power_law_1.2,0.04618239998817444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,8,power_law_1.2,0.04542079865932465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,8,power_law_1.2,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,8,power_law_1.2,0.04413439929485321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,8,power_law_1.2,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,8,power_law_1.2,0.048019200563430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,8,power_law_1.2,0.04831359982490539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,8,power_law_1.2,0.04800640046596527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,8,power_law_1.2,0.053523200750350955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,8,power_law_1.2,0.05496960282325745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,8,power_law_1.2,0.06272000074386597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,8,power_law_1.2,0.082150399684906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,8,power_law_1.2,0.0936959981918335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,8,power_law_1.2,0.12101119756698608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,8,power_law_1.2,0.15371520519256593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,8,power_law_1.2,0.2082047939300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,8,power_law_1.2,0.25104000568389895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,1,power_law_1.01,0.029657599329948426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,1,power_law_1.01,0.03049600124359131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,1,power_law_1.01,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,1,power_law_1.01,0.03642880022525787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,1,power_law_1.01,0.036313599348068236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,1,power_law_1.01,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,1,power_law_1.01,0.041388800740242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,1,power_law_1.01,0.041510400176048276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,1,power_law_1.01,0.046003198623657225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,1,power_law_1.01,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,1,power_law_1.01,0.04511359930038452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,1,power_law_1.01,0.04764159917831421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,1,power_law_1.01,0.050425601005554196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,1,power_law_1.01,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,1,power_law_1.01,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,1,power_law_1.01,0.06816639900207519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,1,power_law_1.01,0.0760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,1,power_law_1.01,0.09098240137100219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,1,power_law_1.01,0.10446720123291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,8,power_law_1.2,0.3691648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,1,power_law_1.01,0.13215999603271483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,1,power_law_1.01,0.1570688009262085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,1,power_law_1.01,0.21361920833587647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,1,power_law_1.01,0.2536767959594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,1,power_law_1.01,0.36332800388336184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,1,power_law_1.01,0.46820478439331054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,8,power_law_1.2,0.48265600204467773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,1,power_law_1.01,0.6759679794311524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,8,power_law_1.2,0.6944896221160889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,2,power_law_1.01,0.1522752046585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,2,power_law_1.01,0.21407999992370605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,1,power_law_1.01,0.8865663528442382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,2,power_law_1.01,0.29155840873718264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,2,power_law_1.01,0.40900478363037107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,2,power_law_1.01,0.4861440181732178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,2,power_law_1.01,0.6110591888427734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,2,power_law_1.01,0.7742720127105713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,2,power_law_1.01,0.8632575988769531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,2,power_law_1.01,0.8043583869934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,2,power_law_1.01,0.7952127933502198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,2,power_law_1.01,0.8610560417175293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,2,power_law_1.01,0.8610431671142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,2,power_law_1.01,0.8810624122619629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,2,power_law_1.01,0.8886143684387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,2,power_law_1.01,0.8629887580871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,8,power_law_1.2,0.9300992012023925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,2,power_law_1.01,0.8903871536254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,2,power_law_1.01,0.9086272239685058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,2,power_law_1.01,0.9767040252685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.01,1.0024959564208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.01,1.1344703674316405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.01,1.2573184013366698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.01,1.4923968315124512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.01,1.7375232696533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.01,2.1999744415283202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.01,2.6718719482421873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.01,3.5326271057128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.01,4.614976119995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,2,power_law_1.01,0.04694400131702423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,2,power_law_1.01,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,2,power_law_1.01,0.05745919942855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,2,power_law_1.01,0.069651198387146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,2,power_law_1.01,0.08312960267066956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,2,power_law_1.01,0.0917248010635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,2,power_law_1.01,0.10682879686355591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,2,power_law_1.01,0.11221120357513428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,2,power_law_1.01,0.11373440027236939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,2,power_law_1.01,0.1146880030632019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,2,power_law_1.01,0.12354559898376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,2,power_law_1.01,0.1244928002357483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,2,power_law_1.01,0.12458879947662353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,2,power_law_1.01,0.131494402885437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,2,power_law_1.01,0.14400639533996581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,2,power_law_1.01,0.14963200092315673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,2,power_law_1.01,0.1577407956123352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,2,power_law_1.01,0.17967360019683837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,2,power_law_1.01,0.19267200231552123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,2,power_law_1.01,0.2434623956680298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,2,power_law_1.01,0.2687616109848022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,2,power_law_1.01,0.3534656047821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,2,power_law_1.01,0.4230912208557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,2,power_law_1.01,0.5757567882537842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,2,power_law_1.01,0.7239679813385009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,2,power_law_1.01,1.0341440200805665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,2,power_law_1.01,1.3976960182189941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,2,power_law_1.01,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,2,power_law_1.01,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,2,power_law_1.01,0.041440001130104064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,2,power_law_1.01,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,2,power_law_1.01,0.0527616024017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,2,power_law_1.01,0.07289599776268005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,2,power_law_1.01,0.0743552029132843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,2,power_law_1.01,0.07485439777374267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,2,power_law_1.01,0.07674880027770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,2,power_law_1.01,0.07633919715881347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,2,power_law_1.01,0.0846015989780426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,2,power_law_1.01,0.08526719808578491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,2,power_law_1.01,0.08739839792251587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,2,power_law_1.01,0.0877951979637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,2,power_law_1.01,0.09405440092086792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,2,power_law_1.01,0.0985152006149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,2,power_law_1.01,0.10258560180664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,2,power_law_1.01,0.11214079856872558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,2,power_law_1.01,0.11605119705200195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,2,power_law_1.01,0.14106240272521972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,2,power_law_1.01,0.15904639959335326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,2,power_law_1.01,0.1944640040397644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,2,power_law_1.01,0.2388223886489868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,2,power_law_1.01,0.3375488042831421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,2,power_law_1.01,0.40402560234069823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,2,power_law_1.01,0.5911488056182861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,2,power_law_1.01,0.7283648014068603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,2,balanced,0.02162666618824005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,2,balanced,0.02123733361562093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,2,balanced,0.021066665649414062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,2,balanced,0.02214933435122172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,2,balanced,0.025114665428797405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,2,balanced,0.025120000044504803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,2,balanced,0.02518400053183238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,2,balanced,0.02531733363866806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,2,balanced,0.027776000400384266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,2,balanced,0.027376001079877216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,2,balanced,0.02935466667016347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,2,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,2,balanced,0.05197333296140035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,2,balanced,0.05421866476535797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,2,balanced,0.04484266539414724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,2,balanced,0.03959999978542328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,2,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,2,balanced,0.05681066711743673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,2,balanced,0.060821334520975746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,2,balanced,0.08343999584515889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,2,balanced,0.10359999537467957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,2,balanced,0.13803199927012125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,2,balanced,0.16710933049519858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,2,balanced,0.23806933561960855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,2,balanced,0.3036053379376729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,2,balanced,0.4410346746444702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,2,balanced,0.5754719972610474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,8,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,8,balanced,0.04604800045490265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,8,balanced,0.0476800004641215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,8,balanced,0.053946668903032936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,8,balanced,0.07827199995517731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,8,balanced,0.0851200024286906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,8,balanced,0.0865226686000824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,8,balanced,0.08729066451390584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,8,balanced,0.08612266182899475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,8,balanced,0.08860799670219421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,8,balanced,0.08854400118192036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,8,balanced,0.08999466896057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,8,balanced,0.08893866340319316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,8,balanced,0.09200533231099446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,8,balanced,0.09551999966303508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,8,balanced,0.0976106623808543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,8,balanced,0.10133866469065349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,8,balanced,0.1127839982509613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,8,balanced,0.11376532912254333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,8,balanced,0.1377173364162445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,8,balanced,0.1460693379243215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,8,balanced,0.18634132544199625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,8,balanced,0.20829333861668906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,8,balanced,0.2831520040829976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,8,balanced,0.33404799302419025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,8,balanced,0.4670933485031128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,8,balanced,0.5797173182169596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,1,balanced,0.09947199622790019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,1,balanced,0.12921599547068277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,1,balanced,0.17947200934092203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,1,balanced,0.28414400418599445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,1,balanced,0.49694931507110596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,1,balanced,0.9101280371348063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,1,balanced,0.9179733594258627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,1,balanced,0.9249333540598551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,1,balanced,0.9262773195902506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,1,balanced,0.9313279787699381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,1,balanced,0.9371626377105713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,1,balanced,0.9440000057220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,1,balanced,0.9500213464101156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,1,balanced,0.9603679974873861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,1,balanced,0.9737866719563802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,1,balanced,0.9811733563741049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,1,balanced,1.0013279914855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,1,balanced,1.040287971496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,1,balanced,1.0731786886850994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,1,balanced,1.161247968673706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,1,balanced,1.2640266418457031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,1,balanced,1.4286773999532063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,1,balanced,1.5979626973470051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,1,balanced,2.233429272969564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,1,balanced,2.5745654106140137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,1,balanced,3.6606079737345376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,1,balanced,4.653690656026204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,16,power_law_1.01,0.04341759979724884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,16,power_law_1.01,0.05144960284233093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,16,power_law_1.01,0.04567039906978607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,16,power_law_1.01,0.04442239999771118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,16,power_law_1.01,0.0453247994184494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,16,power_law_1.01,0.045311999320983884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,16,power_law_1.01,0.046214398741722104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,16,power_law_1.01,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,16,power_law_1.01,0.048019200563430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,16,power_law_1.01,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,16,power_law_1.01,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,16,power_law_1.01,0.05057280063629151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,16,power_law_1.01,0.051801598072052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,16,power_law_1.01,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,16,power_law_1.01,0.05774719715118408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,1,power_law_1.2,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,16,power_law_1.01,0.06033920049667359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,16,power_law_1.01,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,16,power_law_1.01,0.07118719816207886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,16,power_law_1.01,0.07598080039024353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,16,power_law_1.01,0.09134719967842102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,16,power_law_1.01,0.10068479776382447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,16,power_law_1.01,0.13097599744796753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,1,power_law_1.2,0.041510400176048276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,16,power_law_1.01,0.1667904019355774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,1,power_law_1.2,0.05447040200233459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,16,power_law_1.01,0.2386944055557251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,1,power_law_1.2,0.06603519916534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,1,power_law_1.2,0.07834240198135375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,1,power_law_1.2,0.1147007942199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,1,power_law_1.2,0.11918079853057861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,1,power_law_1.2,0.12339199781417846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,1,power_law_1.2,0.12435200214385986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,1,power_law_1.2,0.1301632046699524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,1,power_law_1.2,0.13610880374908446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,1,power_law_1.2,0.14054399728775024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,1,power_law_1.2,0.14179199934005737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,1,power_law_1.2,0.14953600168228148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,1,power_law_1.2,0.16099200248718262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,1,power_law_1.2,0.1689344048500061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,1,power_law_1.2,0.176256000995636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,16,power_law_1.01,0.2981631994247437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,1,power_law_1.2,0.20324480533599854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,1,power_law_1.2,0.22817280292510986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,16,power_law_1.01,0.42535037994384767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,1,power_law_1.2,0.2802623987197876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,16,power_law_1.01,0.542137622833252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,1,power_law_1.2,0.330182409286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,1,power_law_1.2,0.4357439994812012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,1,power_law_1.2,0.5497856140136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,2,power_law_1.01,0.030969598889350893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,2,power_law_1.01,0.029798400402069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,2,power_law_1.01,0.030508801341056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,2,power_law_1.01,0.031711998581886294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,2,power_law_1.01,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,2,power_law_1.01,0.032601600885391234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,2,power_law_1.01,0.03336319923400879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,2,power_law_1.01,0.03423359990119934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,2,power_law_1.01,0.03511039912700653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,2,power_law_1.01,0.03543039858341217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,1,power_law_1.2,0.756390380859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,2,power_law_1.01,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,2,power_law_1.01,0.03873920142650604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,2,power_law_1.01,0.03930880129337311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,2,power_law_1.01,0.03945600092411041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,2,power_law_1.01,0.04387840032577515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,2,power_law_1.01,0.047539201378822324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,2,power_law_1.01,0.05060480237007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,2,power_law_1.01,0.06124160289764404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,2,power_law_1.01,0.06806399822235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,2,power_law_1.01,0.08447359800338745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,1,power_law_1.2,0.9611136436462402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,2,power_law_1.01,0.10017280578613282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,2,power_law_1.01,0.12174079418182374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,2,power_law_1.01,0.1597759962081909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,1,power_law_1.2,1.381158447265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,2,power_law_1.01,0.20214400291442872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,2,power_law_1.01,0.23170559406280516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,1,power_law_1.2,1.8005056381225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,2,power_law_1.01,0.3958591938018799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,2,power_law_1.01,0.4279168128967285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,128,power_law_1.2,0.3097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,128,power_law_1.2,0.3176383972167969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,128,power_law_1.2,0.06869120001792908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,128,power_law_1.2,0.06725119948387145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,128,power_law_1.2,0.059513598680496216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,128,power_law_1.2,0.059628802537918094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,128,power_law_1.2,0.06479359865188598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,128,power_law_1.2,0.06607999801635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,128,power_law_1.2,0.0709824025630951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,128,power_law_1.2,0.06629760265350342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,128,power_law_1.2,0.07257599830627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,128,power_law_1.2,0.07528319954872131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,128,power_law_1.2,0.07882239818572997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,128,power_law_1.2,0.08145279884338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,128,power_law_1.2,0.08960639834403991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,128,power_law_1.2,0.08854399919509888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,128,power_law_1.2,0.09752320051193238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,128,power_law_1.2,0.1124608039855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,128,power_law_1.2,0.12914559841156006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,128,power_law_1.2,0.16881279945373534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,128,power_law_1.2,0.18770560026168823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,128,power_law_1.2,0.24498560428619384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,128,power_law_1.2,0.29180800914764404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,128,power_law_1.2,0.39236481189727784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,128,power_law_1.2,0.518284797668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,128,power_law_1.2,0.7515264034271241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,128,power_law_1.2,0.9440383911132812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,4,power_law_1.2,0.025894400477409363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,4,power_law_1.2,0.02383359968662262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,4,power_law_1.2,0.02503040134906769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,4,power_law_1.2,0.03054719865322113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,4,power_law_1.2,0.03187200129032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,4,power_law_1.2,0.034867200255393985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,4,power_law_1.2,0.03425920009613037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,4,power_law_1.2,0.03603839874267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,4,power_law_1.2,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,4,power_law_1.2,0.03691520094871521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,4,power_law_1.2,0.039743998646736146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,4,power_law_1.2,0.042233601212501526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,4,power_law_1.2,0.04505600035190582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,4,power_law_1.2,0.04663040041923523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,4,power_law_1.2,0.053913599252700804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,4,power_law_1.2,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,4,power_law_1.2,0.056435197591781616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,4,power_law_1.2,0.06627200245857238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,4,power_law_1.2,0.07275519967079162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,4,power_law_1.2,0.08567039966583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,4,power_law_1.2,0.1
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,4,power_law_1.2,0.12712960243225097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,4,power_law_1.2,0.15034879446029664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,4,power_law_1.2,0.2011712074279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,4,power_law_1.2,0.2519040107727051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,4,power_law_1.2,0.35802879333496096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,4,power_law_1.2,0.45131521224975585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,128,power_law_1.2,0.018886399269104005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,128,power_law_1.2,0.02041600048542023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,128,power_law_1.2,0.01923840045928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,128,power_law_1.2,0.0208064004778862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,128,power_law_1.2,0.02515200078487396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,128,power_law_1.2,0.03728640079498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,128,power_law_1.2,0.03430399894714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,128,power_law_1.2,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,128,power_law_1.2,0.03447040021419525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,128,power_law_1.2,0.03447040021419525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,128,power_law_1.2,0.03537920117378235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,128,power_law_1.2,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,128,power_law_1.2,0.03372800052165985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,128,power_law_1.2,0.033964800834655764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,128,power_law_1.2,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,128,power_law_1.2,0.03299199938774109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,128,power_law_1.2,0.037452799081802365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,128,power_law_1.2,0.0383679986000061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,128,power_law_1.2,0.039961600303649904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,128,power_law_1.2,0.050374400615692136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,128,power_law_1.2,0.0577023983001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,128,power_law_1.2,0.07582719922065735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,128,power_law_1.2,0.09054080247879029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,128,power_law_1.2,0.1294144034385681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,128,power_law_1.2,0.15292160511016845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,128,power_law_1.2,0.2327552080154419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,128,power_law_1.2,0.30789120197296144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,2,power_law_1.01,0.03497599959373474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,2,power_law_1.01,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,2,power_law_1.01,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,2,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,2,power_law_1.01,0.03726080060005188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,2,power_law_1.01,0.03772160112857818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,2,power_law_1.01,0.03844479918479919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,2,power_law_1.01,0.03835520148277283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,2,power_law_1.01,0.03982079923152924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,2,power_law_1.01,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,2,power_law_1.01,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,2,power_law_1.01,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,2,power_law_1.01,0.047353601455688475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,2,power_law_1.01,0.047091200947761536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,2,power_law_1.01,0.05472639799118042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,2,power_law_1.01,0.0573311984539032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,2,power_law_1.01,0.061247998476028444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,2,power_law_1.01,0.06960639953613282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.01,0.07919359803199769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.01,0.10142719745635986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.01,0.11850880384445191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.01,0.1581439971923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.01,0.166867196559906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.01,0.271724796295166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,32,power_law_1.2,0.06071680188179016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,32,power_law_1.2,0.06716160178184509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,32,power_law_1.2,0.0641152024269104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,32,power_law_1.2,0.06776959896087646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,32,power_law_1.2,0.0686784029006958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,32,power_law_1.2,0.06809599995613098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,32,power_law_1.2,0.06858239769935608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,32,power_law_1.2,0.06892799735069274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.01,0.3214207887649536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,32,power_law_1.2,0.07114880084991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,32,power_law_1.2,0.0710528016090393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,32,power_law_1.2,0.07120640277862549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,32,power_law_1.2,0.0738048017024994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,32,power_law_1.2,0.07397119998931885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,32,power_law_1.2,0.07738879919052125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,32,power_law_1.2,0.08786560297012329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,32,power_law_1.2,0.09032319784164429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,32,power_law_1.2,0.09622399806976319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,32,power_law_1.2,0.11521279811859131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,32,power_law_1.2,0.12006399631500245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,32,power_law_1.2,0.14405759572982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,32,power_law_1.2,0.16799999475479127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,32,power_law_1.2,0.21966080665588378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,32,power_law_1.2,0.2644160032272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,32,power_law_1.2,0.3930047988891602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,32,power_law_1.2,0.509881591796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.01,0.447603178024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,32,power_law_1.2,0.7799104213714599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,32,power_law_1.2,1.010969638824463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.01,0.5493696212768555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,2,power_law_1.2,0.05600000023841858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,2,power_law_1.2,0.06245120167732239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,2,power_law_1.2,0.08589439988136291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,2,power_law_1.2,0.11249279975891113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,2,power_law_1.2,0.11743359565734864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,2,power_law_1.2,0.18180480003356933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,2,power_law_1.2,0.1843135952949524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,2,power_law_1.2,0.19059200286865235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,2,power_law_1.2,0.19914239645004272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,2,power_law_1.2,0.2031167984008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,2,power_law_1.2,0.19997440576553344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,2,power_law_1.2,0.2095871925354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,2,power_law_1.2,0.20972800254821777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,2,power_law_1.2,0.21320319175720215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,2,power_law_1.2,0.23061120510101318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,2,power_law_1.2,0.2430720090866089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,2,power_law_1.2,0.23557119369506835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,2,power_law_1.2,0.2741695880889893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,2,power_law_1.2,0.30230400562286375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,2,power_law_1.2,0.3596287965774536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,2,power_law_1.2,0.35642879009246825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,2,power_law_1.2,0.467244815826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,2,power_law_1.2,0.5821760177612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,2,power_law_1.2,0.7467967987060546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,2,power_law_1.2,0.8718591690063476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,2,power_law_1.2,1.1897215843200684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,2,power_law_1.2,1.5864576339721679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,8,power_law_1.01,0.05438079833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,8,power_law_1.01,0.057017600536346434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,8,power_law_1.01,0.06270719766616821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,8,power_law_1.01,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,8,power_law_1.01,0.06883839964866638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,8,power_law_1.01,0.07684479951858521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,8,power_law_1.01,0.0746944010257721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,8,power_law_1.01,0.07162879705429077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,8,power_law_1.01,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,8,power_law_1.01,0.07368320226669312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,8,power_law_1.01,0.0760703980922699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,8,power_law_1.01,0.08225280046463013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,8,power_law_1.01,0.08321920037269592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,8,power_law_1.01,0.08112000226974488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,8,power_law_1.01,0.09072639942169189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,8,power_law_1.01,0.09149439930915833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,8,power_law_1.01,0.09710720181465149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,8,power_law_1.01,0.11101440191268921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,8,power_law_1.01,0.11925760507583619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,8,power_law_1.01,0.1414528012275696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,8,power_law_1.01,0.1644544005393982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,8,power_law_1.01,0.19484800100326538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,8,power_law_1.01,0.23230719566345215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,8,power_law_1.01,0.29960319995880125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,8,power_law_1.01,0.3846463918685913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,8,power_law_1.01,0.5933695793151855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,8,power_law_1.01,0.7223423957824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,64,balanced,0.046021332343419395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,64,balanced,0.04376000165939331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,64,balanced,0.045706664522488914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,64,balanced,0.047770669062932335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,64,balanced,0.046709333856900535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,64,balanced,0.04991999765237173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,64,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,64,balanced,0.050399998823801674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,64,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,64,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,64,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,64,balanced,0.052058666944503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,64,balanced,0.05384533107280731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,64,balanced,0.05315199991067251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,64,balanced,0.05839466551939646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,64,balanced,0.05835199852784475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,64,balanced,0.06001066664854685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,64,balanced,0.06754133105278015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,64,balanced,0.0688213308652242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,64,balanced,0.07692266503969829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,64,balanced,0.0848640004793803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,64,balanced,0.10198400417963664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,64,balanced,0.11750400066375732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,64,balanced,0.15877333283424377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,64,balanced,0.19533334175745645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,64,balanced,0.26125333706537884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,64,balanced,0.32917332649230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,64,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,64,balanced,0.04604800045490265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,64,balanced,0.04749333361784617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,64,balanced,0.049925332268079124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,64,balanced,0.050383999943733215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,64,balanced,0.053930665055910744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,64,balanced,0.054378668467203774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,64,balanced,0.054805333415667214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,64,balanced,0.058037335673967995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,64,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,64,balanced,0.05795733133951823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,64,balanced,0.06237866481145223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,64,balanced,0.06338133414586385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,64,balanced,0.06625066697597504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,64,balanced,0.07640000184377034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,64,balanced,0.0795360008875529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,64,balanced,0.07871999839941661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,64,balanced,0.09116266171137492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,64,balanced,0.09524266918500264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,64,balanced,0.11051733295122783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,64,balanced,0.12635200222333273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,64,balanced,0.1623360017935435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,64,balanced,0.1891146699587504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,64,balanced,0.24061334133148193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,64,balanced,0.28970666726430255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,64,balanced,0.391541322072347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,64,balanced,0.4926026662190755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,2,balanced,0.04837866624196371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,2,balanced,0.04757333298524221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,2,balanced,0.04674666623274485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,2,balanced,0.048245335618654885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,2,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,2,balanced,0.05596266686916351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,2,balanced,0.07317333420117696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,2,balanced,0.06985599795977275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,2,balanced,0.09202133615811665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,2,balanced,0.08272533118724823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,2,balanced,0.07721066474914551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,2,balanced,0.1376533309618632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,2,balanced,0.13121599952379862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,2,balanced,0.10286399722099304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,2,balanced,0.1455466647942861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,2,balanced,0.12613866726557413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,2,balanced,0.16368533174196878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,2,balanced,0.21304533878962198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,2,balanced,0.2762719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,2,balanced,0.3923146724700928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,2,balanced,0.5226240158081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,2,balanced,0.7457013130187988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,2,balanced,0.9915520350138346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,2,balanced,1.4561012585957844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,2,balanced,1.917450745900472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,2,balanced,2.8716160456339517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,2,balanced,3.8033119837443032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,128,balanced,0.03370666752258936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,128,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,128,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,128,balanced,0.03432533393303553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,128,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,128,balanced,0.03697066754102707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,128,balanced,0.03544000039498011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,128,balanced,0.037061333656311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,128,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,128,balanced,0.03734400123357773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,128,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,128,balanced,0.03696000079313914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,128,balanced,0.039018665750821434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,128,balanced,0.03989866624275843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,128,balanced,0.039701332648595176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,128,balanced,0.041696002086003624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,128,balanced,0.04191466669241587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,128,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,128,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,128,balanced,0.04786133269468943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,128,balanced,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,128,balanced,0.053861334919929504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,128,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,128,balanced,0.06838933130105336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,128,balanced,0.08124266564846039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,128,balanced,0.10532266894976298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,128,balanced,0.12583466370900473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,8,power_law_1.2,0.0395904004573822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,8,power_law_1.2,0.03813120126724243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,8,power_law_1.2,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,8,power_law_1.2,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,8,power_law_1.2,0.044531199336051944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,8,power_law_1.2,0.04802559912204742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,8,power_law_1.2,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,8,power_law_1.2,0.04862079918384552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,8,power_law_1.2,0.04922240078449249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,8,power_law_1.2,0.049983999133110045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,8,power_law_1.2,0.04990079998970032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,8,power_law_1.2,0.05041279792785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,8,power_law_1.2,0.05146880149841308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,8,power_law_1.2,0.05617920160293579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,8,power_law_1.2,0.059520000219345094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,8,power_law_1.2,0.06316800117492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,8,power_law_1.2,0.06962559819221496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,8,power_law_1.2,0.07720959782600403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,8,power_law_1.2,0.08399999737739564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,8,power_law_1.2,0.10402560234069824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,8,power_law_1.2,0.11389440298080444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,8,power_law_1.2,0.1351807951927185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,8,power_law_1.2,0.17843199968338014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,8,power_law_1.2,0.24487679004669188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,8,power_law_1.2,0.29871358871459963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,8,power_law_1.2,0.46983680725097654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,8,power_law_1.2,0.5731967926025391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,2,power_law_1.2,0.030124801397323608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,2,power_law_1.2,0.031033599376678468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,2,power_law_1.2,0.029849600791931153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,2,power_law_1.2,0.032576000690460204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,2,power_law_1.2,0.03374080061912536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,2,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,2,power_law_1.2,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,2,power_law_1.2,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,2,power_law_1.2,0.06316800117492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,2,power_law_1.2,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,2,power_law_1.2,0.057196801900863646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,2,power_law_1.2,0.07024000287055969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,2,power_law_1.2,0.07648640275001525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,2,power_law_1.2,0.07388160228729249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,2,power_law_1.2,0.10496000051498414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,2,power_law_1.2,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,2,power_law_1.2,0.09859200119972229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,2,power_law_1.2,0.13934719562530518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.2,0.16391040086746217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.2,0.23128321170806884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.2,0.2926271915435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.2,0.4178304195404053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.2,0.5298687934875488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,128,balanced,0.023056000471115112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,128,balanced,0.02346666653951009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,128,balanced,0.02388266722361247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,128,balanced,0.027141332626342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,128,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,128,balanced,0.054485330979029335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,128,balanced,0.051311999559402466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,128,balanced,0.0510506679614385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,128,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,128,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,128,balanced,0.05068266888459524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,128,balanced,0.04756799836953481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,128,balanced,0.046282668908437095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,128,balanced,0.04279466470082601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,128,balanced,0.05990933378537496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,128,balanced,0.06016000111897787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,128,balanced,0.05034666756788889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,128,balanced,0.07071466743946075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,128,balanced,0.0780320018529892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,128,balanced,0.10273067156473796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,128,balanced,0.12614400188128153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,128,balanced,0.17584532499313354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,128,balanced,0.21683200200398764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,128,balanced,0.31707199414571124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,128,balanced,0.41598931948343915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,128,balanced,0.6113440195719401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,128,balanced,0.8062986532847086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.2,0.7813759803771972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.2,1.0212287902832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.2,1.5138239860534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.2,2.0032512664794924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,32,power_law_1.2,0.05030400156974792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,32,power_law_1.2,0.05056639909744263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,32,power_law_1.2,0.05055360198020935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,32,power_law_1.2,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,32,power_law_1.2,0.050783997774124144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,32,power_law_1.2,0.05191680192947388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,32,power_law_1.2,0.05225600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,32,power_law_1.2,0.053830397129058835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,32,power_law_1.2,0.05356799960136414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,32,power_law_1.2,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,32,power_law_1.2,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,32,power_law_1.2,0.05658239722251892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,32,power_law_1.2,0.05687680244445801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,32,power_law_1.2,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,32,power_law_1.2,0.06446719765663148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,32,power_law_1.2,0.06784639954566955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,32,power_law_1.2,0.06956160068511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,32,power_law_1.2,0.08366720080375671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.2,0.09728639721870422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.2,0.12172160148620606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.2,0.13012479543685912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.2,0.1687872052192688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.2,0.20051839351654052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.2,0.2810431957244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.2,0.3847487926483154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.2,0.604966402053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.2,0.7209727764129639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,16,power_law_1.01,0.04764159917831421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,16,power_law_1.01,0.051545602083206174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,16,power_law_1.01,0.04973439872264862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,16,power_law_1.01,0.05097600221633911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,16,power_law_1.01,0.0525056004524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,16,power_law_1.01,0.05607039928436279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,16,power_law_1.01,0.05754240155220032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,16,power_law_1.01,0.05848960280418396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,16,power_law_1.01,0.05813760161399841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,16,power_law_1.01,0.05676159858703613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,16,power_law_1.01,0.05744640231132507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,16,power_law_1.01,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,16,power_law_1.01,0.06045439839363098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,16,power_law_1.01,0.061849600076675414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,16,power_law_1.01,0.07132800221443177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,16,power_law_1.01,0.07203840017318726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,16,power_law_1.01,0.07456640005111695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,16,power_law_1.01,0.08247039914131164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,16,power_law_1.01,0.09191679954528809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,16,power_law_1.01,0.11073919534683227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,16,power_law_1.01,0.12700799703598023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,16,power_law_1.01,0.16028800010681152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,16,power_law_1.01,0.1941375970840454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,16,power_law_1.01,0.25968639850616454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,16,power_law_1.01,0.34304640293121336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,16,power_law_1.01,0.46912641525268556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,16,power_law_1.01,0.6701632022857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,32,power_law_1.01,0.04441600143909454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,32,power_law_1.01,0.031788799166679385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,32,power_law_1.01,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,32,power_law_1.01,0.03081600069999695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,32,power_law_1.01,0.031481599807739256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,32,power_law_1.01,0.033395200967788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,32,power_law_1.01,0.034016001224517825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,32,power_law_1.01,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,32,power_law_1.01,0.03498879969120026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,32,power_law_1.01,0.03550080060958862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,32,power_law_1.01,0.03738240003585815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,32,power_law_1.01,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,32,power_law_1.01,0.041075199842453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,32,power_law_1.01,0.04066559970378876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,32,power_law_1.01,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,32,power_law_1.01,0.046911999583244324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,32,power_law_1.01,0.051846402883529666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,32,power_law_1.01,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,32,power_law_1.01,0.06085119843482971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,32,power_law_1.01,0.0753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,32,power_law_1.01,0.09588479995727539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,32,power_law_1.01,0.12327680587768555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,32,power_law_1.01,0.1437824010848999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,32,power_law_1.01,0.2074496030807495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,32,power_law_1.01,0.26179840564727785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,32,power_law_1.01,0.35338239669799804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,32,power_law_1.01,0.4837952136993408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,2,power_law_1.01,0.05100799798965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,2,power_law_1.01,0.054476797580718994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,2,power_law_1.01,0.06284160017967225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,2,power_law_1.01,0.07498239874839782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,2,power_law_1.01,0.08900480270385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,2,power_law_1.01,0.10094720125198364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,2,power_law_1.01,0.12302720546722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,2,power_law_1.01,0.12299519777297974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,2,power_law_1.01,0.1291584014892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,2,power_law_1.01,0.13292800188064574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,2,power_law_1.01,0.13563519716262817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,2,power_law_1.01,0.14090240001678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,2,power_law_1.01,0.1527616024017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,2,power_law_1.01,0.16068480014801026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,2,power_law_1.01,0.17215360403060914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,2,power_law_1.01,0.18254719972610473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,2,power_law_1.01,0.18292479515075682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,2,power_law_1.01,0.2078847885131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,2,power_law_1.01,0.23799679279327393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,2,power_law_1.01,0.29973759651184084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,2,power_law_1.01,0.3607424020767212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,2,power_law_1.01,0.47616000175476075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,2,power_law_1.01,0.5854400157928467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,2,power_law_1.01,0.7979199886322021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,2,power_law_1.01,1.0071104049682618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,2,power_law_1.01,1.4431103706359862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,2,power_law_1.01,1.8739328384399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,2,power_law_1.01,0.021331200003623964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,2,power_law_1.01,0.028595200181007384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,2,power_law_1.01,0.03735679984092712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,2,power_law_1.01,0.041894400119781496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,2,power_law_1.01,0.05074560046195984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,2,power_law_1.01,0.06348800063133239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,2,power_law_1.01,0.06707839965820313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,2,power_law_1.01,0.06752640008926392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,2,power_law_1.01,0.07045120000839233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,2,power_law_1.01,0.07051519751548767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,2,power_law_1.01,0.07370880246162415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,2,power_law_1.01,0.0744704008102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,2,power_law_1.01,0.07497599720954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,2,power_law_1.01,0.07728000283241272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,2,power_law_1.01,0.07910400032997131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,2,power_law_1.01,0.0875328004360199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,2,power_law_1.01,0.08770560026168824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,2,power_law_1.01,0.16393599510192872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.01,0.18220800161361694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.01,0.14396799802780152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.01,0.15299839973449708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.01,0.1827712059020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.01,0.21116158962249756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.01,0.2649983882904053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.01,0.33717119693756104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.01,0.4660799980163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.01,0.6082240104675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,4,balanced,0.06590400139490764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,4,balanced,0.0653653343518575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,4,balanced,0.09274133046468098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,4,balanced,0.09081066648165385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,4,balanced,0.09230933586756389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,4,balanced,0.09268266956011455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,4,balanced,0.09231467048327129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,4,balanced,0.09409067034721375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,4,balanced,0.09241066376368205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,4,balanced,0.0944106678167979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,4,balanced,0.09780266880989075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,4,balanced,0.0995199978351593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,4,balanced,0.09962667028109233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,4,balanced,0.10372266173362732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,4,balanced,0.11358933647473653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,4,balanced,0.11549866199493408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,4,balanced,0.11963733037312825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,4,balanced,0.13808533549308777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,4,balanced,0.14628799756368002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,4,balanced,0.18905067443847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,4,balanced,0.21299733718236288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,4,balanced,0.28620266914367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,4,balanced,0.3433973391850789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,4,balanced,0.48873599370320636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,4,balanced,0.6274133523305258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,4,balanced,0.9105919996897379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,4,balanced,1.1835306485493977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,256,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,256,balanced,0.04971733192602793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,256,balanced,0.049925332268079124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,256,balanced,0.05186133086681366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,256,balanced,0.05213866631189982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,256,balanced,0.055173332492510475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,256,balanced,0.05596266686916351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,256,balanced,0.057909334699312844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,256,balanced,0.06033066908518473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,256,balanced,0.06228266656398773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,256,balanced,0.05996266504128774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,256,balanced,0.06426666676998138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,256,balanced,0.07055999835332234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,256,balanced,0.06839466591676076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,256,balanced,0.08036266764005025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,256,balanced,0.08011733492215474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,256,balanced,0.08243200182914734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,256,balanced,0.09401599566141765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,256,balanced,0.09899200002352397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,256,balanced,0.10964799920717876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,256,balanced,0.12345600128173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,256,balanced,0.1601706643899282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,256,balanced,0.18690667549769083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,256,balanced,0.23975465695063272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,256,balanced,0.28885332743326825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,256,balanced,0.3906453450520833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,256,balanced,0.4868160088857015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,16,balanced,0.05399466554323832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,16,balanced,0.05426666637261709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,16,balanced,0.06029333174228668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,16,balanced,0.06949333349863689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,16,balanced,0.090421328941981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,16,balanced,0.12412800391515096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,16,balanced,0.11992533008257548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,16,balanced,0.12101866801579793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,16,balanced,0.12281066179275513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,16,balanced,0.12004799644152324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,16,balanced,0.12018666664759318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,16,balanced,0.11718400319417317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,16,balanced,0.11352533102035522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,16,balanced,0.12203199664751689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,16,balanced,0.11754666765530904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,16,balanced,0.12288533647855122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,16,balanced,0.13175466656684875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,16,balanced,0.13109866778055826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,16,balanced,0.14009066422780356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,16,balanced,0.15129599968592325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,16,balanced,0.16437333822250366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,16,balanced,0.19101333618164062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,16,balanced,0.21795199314753214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,16,balanced,0.3083466688791911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,16,balanced,0.3585493167241414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,16,balanced,0.49530665079752606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,16,balanced,0.6372960011164347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,1,power_law_1.01,0.03586559891700745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,1,power_law_1.01,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,1,power_law_1.01,0.03893119990825653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,1,power_law_1.01,0.045849600434303285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,1,power_law_1.01,0.04787200093269348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,1,power_law_1.01,0.04874880015850067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,1,power_law_1.01,0.04960640072822571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,1,power_law_1.01,0.050732797384262084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,1,power_law_1.01,0.05082880258560181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,1,power_law_1.01,0.0500544011592865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,1,power_law_1.01,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,1,power_law_1.01,0.056569600105285646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,1,power_law_1.01,0.06046720147132874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,1,power_law_1.01,0.06791039705276489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,1,power_law_1.01,0.07859839797019959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,1,power_law_1.01,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,1,power_law_1.01,0.08636800050735474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,1,power_law_1.01,0.09688320159912109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.01,0.12137600183486938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.01,0.16389119625091553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.01,0.18231680393218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.01,0.27755520343780515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.01,0.3644223928451538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.01,0.5299647808074951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.01,0.6777408123016357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,8,power_law_1.01,0.03841919898986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,8,power_law_1.01,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,8,power_law_1.01,0.048230400681495665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,8,power_law_1.01,0.04993920028209686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,8,power_law_1.01,0.04752640128135681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.01,1.013535976409912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,8,power_law_1.01,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,8,power_law_1.01,0.05495679974555969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,8,power_law_1.01,0.05548160076141358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,8,power_law_1.01,0.056979197263717654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,8,power_law_1.01,0.05824000239372253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,8,power_law_1.01,0.058246397972106935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,8,power_law_1.01,0.059494400024414064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,8,power_law_1.01,0.05988479852676391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,8,power_law_1.01,0.0639743983745575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,8,power_law_1.01,0.0726207971572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,8,power_law_1.01,0.07477120161056519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,8,power_law_1.01,0.07208319902420043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,8,power_law_1.01,0.08076800107955932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,8,power_law_1.01,0.09550079703330994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,8,power_law_1.01,0.10754560232162476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,8,power_law_1.01,0.11683839559555054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,8,power_law_1.01,0.15964800119400024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,8,power_law_1.01,0.16768640279769897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,8,power_law_1.01,0.22910079956054688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,8,power_law_1.01,0.26543359756469725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,8,power_law_1.01,0.3585536003112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,8,power_law_1.01,0.5291391849517822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,1,balanced,0.07657599945863088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,1,balanced,0.10312533378601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,1,balanced,0.1563093364238739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,1,balanced,0.2625386714935303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,1,balanced,0.47335465749104816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,1,balanced,0.5854399998982748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,1,balanced,0.5896159807840983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,1,balanced,0.5932000080744425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,1,balanced,0.5990080038706461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,1,balanced,0.5983253320058187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,1,balanced,0.604479988416036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,1,balanced,0.6133759816487631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,1,balanced,0.6182080109914144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,1,balanced,0.6256586710611979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,1,balanced,0.6422933340072632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,1,balanced,0.6471306482950846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,1,balanced,0.6673279603322347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,1,balanced,0.7574933369954427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,1,balanced,0.7376800378163656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,1,balanced,0.9190666675567627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,1,balanced,0.903439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,1,balanced,1.3325173060099285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,1,balanced,1.2730560302734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,1,balanced,1.90227206548055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,1,balanced,2.1698080698649087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,1,balanced,3.5265280405680337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,1,balanced,4.1135679880778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.01,1.3025983810424804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,16,power_law_1.01,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,16,power_law_1.01,0.1041983962059021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,16,power_law_1.01,0.1242624044418335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,16,power_law_1.01,0.12064000368118286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,16,power_law_1.01,0.1250496029853821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,16,power_law_1.01,0.12578560113906861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,16,power_law_1.01,0.11729279756546021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,16,power_law_1.01,0.12286080121994018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,16,power_law_1.01,0.1208448052406311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,16,power_law_1.01,0.11829760074615478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,16,power_law_1.01,0.11554559469223022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,16,power_law_1.01,0.11845120191574096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,16,power_law_1.01,0.1189568042755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,16,power_law_1.01,0.13076479434967042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,16,power_law_1.01,0.13045120239257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,16,power_law_1.01,0.14508800506591796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,16,power_law_1.01,0.1454848051071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,16,power_law_1.01,0.16806399822235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,16,power_law_1.01,0.17632639408111572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,16,power_law_1.01,0.22169599533081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,16,power_law_1.01,0.23863039016723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,16,power_law_1.01,0.3088576078414917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,16,power_law_1.01,0.3632128000259399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,16,power_law_1.01,0.4873663902282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,16,power_law_1.01,0.6143487930297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,16,power_law_1.01,0.8464447975158691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,16,power_law_1.01,1.0730175971984863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,1,balanced,0.2120800018310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,1,balanced,0.39471999804178876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,1,balanced,0.7239999771118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,1,balanced,0.739786704381307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,1,balanced,0.7329013347625732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,1,balanced,0.7344426314036051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,1,balanced,0.7360586325327555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,1,balanced,0.7345546881357828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,1,balanced,0.735040028889974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,1,balanced,0.7357493241628011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,1,balanced,0.7463946342468262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,1,balanced,0.749567985534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,1,balanced,0.7566560109456381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,1,balanced,0.8110613028208414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,1,balanced,0.7875626881917318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,1,balanced,0.7961013317108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,1,balanced,0.9053333600362142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,1,balanced,0.9889973004659017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,1,balanced,0.943008025487264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,1,balanced,1.4321494102478027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,1,balanced,1.4901173909505208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,1,balanced,2.0750719706217446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,1,balanced,2.8200108210245767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,1,balanced,4.293552080790202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,1,balanced,5.627503712972005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,1,balanced,8.005306879679361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,1,balanced,11.671663920084635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,1,power_law_1.2,0.035078400373458864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,1,power_law_1.2,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,1,power_law_1.2,0.04474239945411682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,1,power_law_1.2,0.053478401899337766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,1,power_law_1.2,0.06695039868354798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,1,power_law_1.2,0.09232640266418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,1,power_law_1.2,0.09566079974174499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,1,power_law_1.2,0.09740800261497498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,1,power_law_1.2,0.10270719528198242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,1,power_law_1.2,0.10439679622650147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,1,power_law_1.2,0.1056447982788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,1,power_law_1.2,0.10871039628982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,1,power_law_1.2,0.11358079910278321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,1,power_law_1.2,0.1386944055557251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,1,power_law_1.2,0.14733439683914185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,1,power_law_1.2,0.14909440279006958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,1,power_law_1.2,0.1646463990211487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,1,power_law_1.2,0.1879040002822876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,1,power_law_1.2,0.20016000270843506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,1,power_law_1.2,0.2541311979293823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,1,power_law_1.2,0.3033279895782471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,2,power_law_1.01,0.06177279949188232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,2,power_law_1.01,0.06720640063285828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,2,power_law_1.01,0.07182719707489013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,2,power_law_1.01,0.09740800261497498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,2,power_law_1.01,0.10695680379867553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,1,power_law_1.2,0.40128002166748045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,2,power_law_1.01,0.12153600454330445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,2,power_law_1.01,0.14584959745407106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,2,power_law_1.01,0.14595199823379518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,2,power_law_1.01,0.15067520141601562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,2,power_law_1.01,0.1524224042892456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,2,power_law_1.01,0.1533184051513672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,1,power_law_1.2,0.509600019454956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,2,power_law_1.01,0.15790719985961915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,2,power_law_1.01,0.16480000019073487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,2,power_law_1.01,0.1689344048500061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,2,power_law_1.01,0.17628159523010253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,2,power_law_1.01,0.18367359638214112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,2,power_law_1.01,0.1900928020477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,2,power_law_1.01,0.2155776023864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,2,power_law_1.01,0.2322943925857544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,2,power_law_1.01,0.2804095983505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,2,power_law_1.01,0.3131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,2,power_law_1.01,0.3974720001220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,2,power_law_1.01,0.4810495853424072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,2,power_law_1.01,0.6545599937438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,2,power_law_1.01,0.833516788482666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,2,power_law_1.01,1.211571216583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,2,power_law_1.01,1.5137408256530762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,1,power_law_1.2,0.6892735958099365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,16,power_law_1.2,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,16,power_law_1.2,0.03738240003585815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,16,power_law_1.2,0.036160001158714296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,16,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,16,power_law_1.2,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,16,power_law_1.2,0.04022400081157684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,16,power_law_1.2,0.040275201201438904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,16,power_law_1.2,0.04106239974498749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,16,power_law_1.2,0.040703999996185306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,16,power_law_1.2,0.04147199988365173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,16,power_law_1.2,0.041440001130104064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,16,power_law_1.2,0.04291200041770935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,16,power_law_1.2,0.0430976003408432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,16,power_law_1.2,0.043968001008033754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,16,power_law_1.2,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,16,power_law_1.2,0.048793599009513855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,16,power_law_1.2,0.0488319993019104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,16,power_law_1.2,0.0546239972114563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,16,power_law_1.2,0.05742719769477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,16,power_law_1.2,0.06964480280876159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,16,power_law_1.2,0.07612800002098083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,16,power_law_1.2,0.09793919920921326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,16,power_law_1.2,0.11793279647827148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,16,power_law_1.2,0.15895040035247804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,1,power_law_1.2,0.8846272468566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,1,power_law_1.2,1.249516773223877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,16,power_law_1.2,0.05736960172653198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,16,power_law_1.2,0.06269440054893494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,16,power_law_1.2,0.0580672025680542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,16,power_law_1.2,0.06093440055847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,16,power_law_1.2,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,32,balanced,0.025621332228183746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,32,balanced,0.025226667523384094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,32,balanced,0.027141332626342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,32,balanced,0.025263999899228413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,16,power_law_1.2,0.061267197132110596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,32,balanced,0.02532800038655599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,32,balanced,0.027280000348885853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,32,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,16,power_law_1.2,0.062015998363494876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,16,power_law_1.2,0.06291840076446534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,16,power_law_1.2,0.06188160181045532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,16,power_law_1.2,0.062041598558425906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,16,power_law_1.2,0.06361600160598754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,16,power_law_1.2,0.06588159799575806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,16,power_law_1.2,0.06689919829368592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,32,balanced,0.036101333796978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,32,balanced,0.04878933231035868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,16,power_law_1.2,0.06615039706230164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,32,balanced,0.04948799808820089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,32,balanced,0.057392001152038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,16,power_law_1.2,0.07208960056304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,16,power_law_1.2,0.07408639788627625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,16,power_law_1.2,0.07730559706687927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,16,power_law_1.2,0.08730880022048951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,16,power_law_1.2,0.09627519845962525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,16,power_law_1.2,0.11128319501876831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,16,power_law_1.2,0.1277119994163513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,16,power_law_1.2,0.16522239446640014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,16,power_law_1.2,0.1901695966720581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,32,balanced,0.08125333487987518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,32,balanced,0.07846400141716003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,32,balanced,0.08036266764005025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,32,balanced,0.08520533641179402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,32,balanced,0.08641067147254944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,32,balanced,0.11218133568763733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,32,balanced,0.14477866888046265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,32,balanced,0.17774933576583862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,32,balanced,0.26175999641418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,32,balanced,0.13915733496348062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,32,balanced,0.18717332681020102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,32,balanced,0.23981332778930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,32,balanced,0.33619733651479083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,32,balanced,0.43581334749857586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,32,balanced,0.6387093464533488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,32,balanced,0.8470453421274821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,16,power_law_1.2,0.26448640823364256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,16,power_law_1.2,0.3125823974609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,16,power_law_1.2,0.4782464027404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,16,power_law_1.2,0.6323264122009278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,16,power_law_1.2,0.22994558811187743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,16,power_law_1.2,0.2842112064361572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,1,power_law_1.2,1.6458751678466796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,16,power_law_1.2,0.3447103977203369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,32,power_law_1.2,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,32,power_law_1.2,0.052665597200393675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,32,power_law_1.2,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,32,power_law_1.2,0.055315202474594115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,32,power_law_1.2,0.05562880039215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,32,power_law_1.2,0.055270397663116456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,32,power_law_1.2,0.05669119954109192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,32,power_law_1.2,0.05681279897689819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,32,power_law_1.2,0.05768960118293762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,32,power_law_1.2,0.05790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,32,power_law_1.2,0.05868160128593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,32,power_law_1.2,0.05994880199432373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,32,power_law_1.2,0.061337602138519284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,32,power_law_1.2,0.06289280056953431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,32,power_law_1.2,0.0686784029006958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,32,power_law_1.2,0.0707584023475647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,32,power_law_1.2,0.0745855987071991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,32,power_law_1.2,0.08627840280532836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.2,0.09532160162925721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.2,0.12002559900283813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.2,0.1328703999519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.2,0.16619520187377929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.2,0.2111743927001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.2,0.28241279125213625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.2,0.3688640117645264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.2,0.5117568016052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.2,0.627180814743042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,2,balanced,0.050757333636283875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,2,balanced,0.06346666812896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,2,balanced,0.08663466572761536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,2,balanced,0.13565333684285483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,2,balanced,0.21844800313313803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,2,balanced,0.38677334785461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,2,balanced,0.3877493143081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,2,balanced,0.3900853395462036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,2,balanced,0.3899679978688558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,2,balanced,0.3961706558863322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,2,balanced,0.39582931995391846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,2,balanced,0.39982398351033527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,2,balanced,0.405349334081014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,2,balanced,0.41194132963816327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,2,balanced,0.41915734608968097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,2,balanced,0.4241226514180501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,2,balanced,0.4336906671524048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,2,balanced,0.4533226490020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,2,balanced,0.4737653334935506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,2,balanced,0.5128320058186849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,2,balanced,0.5603093306223551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,2,balanced,0.6335786581039429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,2,balanced,0.7144052982330322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,2,balanced,0.903119961420695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,2,balanced,1.0476373036702473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,2,balanced,1.4894827206929524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,2,balanced,1.8221920331319172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,16,power_law_1.01,0.0470335990190506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,16,power_law_1.01,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,16,power_law_1.01,0.03240320086479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,16,power_law_1.01,0.03478400111198425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,16,power_law_1.01,0.03244799971580505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,16,power_law_1.01,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,16,power_law_1.01,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,16,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,16,power_law_1.01,0.03427839875221252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,16,power_law_1.01,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,16,power_law_1.01,0.03650560081005096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,16,power_law_1.01,0.0383296012878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,16,power_law_1.01,0.03783040046691895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,16,power_law_1.01,0.040678399801254275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,16,power_law_1.01,0.045228800177574156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,16,power_law_1.01,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,16,power_law_1.01,0.04758400022983551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,16,power_law_1.01,0.051123201847076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,16,power_law_1.01,0.05177599787712097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,16,power_law_1.01,0.06061440110206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,16,power_law_1.01,0.07197440266609192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,16,power_law_1.01,0.08777599930763244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,16,power_law_1.01,0.10720000267028809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,16,power_law_1.01,0.13698559999465942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,16,power_law_1.01,0.16632959842681885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,16,power_law_1.01,0.2320319890975952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,16,power_law_1.01,0.2823744058609009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,8,power_law_1.2,0.08078719973564148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,8,power_law_1.2,0.08187519907951354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,8,power_law_1.2,0.08419200181961059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,8,power_law_1.2,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,8,power_law_1.2,0.09274880290031433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,8,power_law_1.2,0.09617279767990113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,8,power_law_1.2,0.09895679950714112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,8,power_law_1.2,0.10341119766235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,8,power_law_1.2,0.10522240400314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,8,power_law_1.2,0.10663679838180543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,8,power_law_1.2,0.1065600037574768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,8,power_law_1.2,0.1109887957572937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,8,power_law_1.2,0.10995199680328369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,8,power_law_1.2,0.11610239744186401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,8,power_law_1.2,0.12260479927062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,8,power_law_1.2,0.12688640356063843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,8,power_law_1.2,0.14202879667282103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,8,power_law_1.2,0.15224319696426392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,8,power_law_1.2,0.1676095962524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,8,power_law_1.2,0.2000511884689331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,8,power_law_1.2,0.2511552095413208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,8,power_law_1.2,0.31176319122314455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,8,power_law_1.2,0.3918463945388794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,8,power_law_1.2,0.5865856170654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,8,power_law_1.2,0.8339455604553223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,8,power_law_1.2,1.2943872451782226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,8,power_law_1.2,1.725529670715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,32,8,balanced,0.023423999547958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,32,8,balanced,0.0230880007147789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,32,8,balanced,0.025237334271272022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,32,8,balanced,0.025173333783944447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,32,8,balanced,0.027301333844661713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,32,8,balanced,0.03073599934577942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,32,8,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,32,8,balanced,0.033957332372665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,32,8,balanced,0.03525333354870478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,32,8,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,32,8,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,32,8,balanced,0.047040000557899475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,32,8,balanced,0.047610665361086525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,32,8,balanced,0.04900800188382467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,32,8,balanced,0.06885333359241486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,32,8,balanced,0.07056533296902974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,32,8,balanced,0.07468800246715546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,32,8,balanced,0.08809066812197368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,32,8,balanced,0.09597333272298177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,32,8,balanced,0.10977600018183391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,32,8,balanced,0.12353600064913432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,32,8,balanced,0.15552000204722086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,32,8,balanced,0.18797866503397623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,32,8,balanced,0.2375040054321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,32,8,balanced,0.29714133342107135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,32,8,balanced,0.438650647799174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,32,8,balanced,0.5489919980367025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,16,power_law_1.01,0.05504000186920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,16,power_law_1.01,0.06674559712409973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,16,power_law_1.01,0.06683520078659058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,16,power_law_1.01,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,16,power_law_1.01,0.06575359702110291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,16,power_law_1.01,0.06431360244750976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,16,power_law_1.01,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,16,power_law_1.01,0.06913920044898987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,16,power_law_1.01,0.0687936007976532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,16,power_law_1.01,0.06866559982299805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,16,power_law_1.01,0.07006080150604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,16,power_law_1.2,0.06901119947433472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,16,power_law_1.2,0.0976639986038208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,16,power_law_1.2,0.11753599643707276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,16,power_law_1.2,0.11863679885864258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,16,power_law_1.2,0.12054400444030762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,16,power_law_1.01,0.07208960056304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,16,power_law_1.2,0.1241919994354248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,16,power_law_1.01,0.07617920041084289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,16,power_law_1.2,0.12314239740371705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,16,power_law_1.01,0.07628160119056701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,16,power_law_1.01,0.08245760202407837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,16,power_law_1.2,0.12177920341491699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,16,power_law_1.01,0.08491520285606384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,16,power_law_1.01,0.09390720129013061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,16,power_law_1.2,0.1209663987159729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,16,power_law_1.01,0.10727039575576783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,16,power_law_1.2,0.10662399530410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.01,0.11969920396804809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,16,power_law_1.2,0.11683839559555054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,16,power_law_1.2,0.12188160419464111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,16,power_law_1.2,0.11315840482711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,16,power_law_1.2,0.12449920177459717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,16,power_law_1.2,0.12798080444335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,16,power_law_1.2,0.13098880052566528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,16,power_law_1.2,0.14554879665374756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,16,power_law_1.2,0.15290240049362183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,16,power_law_1.2,0.16856319904327394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,16,power_law_1.2,0.2056960105895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,8,power_law_1.2,0.027065598964691163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,16,power_law_1.2,0.2190335988998413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,8,power_law_1.2,0.027603200078010558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.01,0.13956480026245116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,8,power_law_1.2,0.030406400561332703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.01,0.16425600051879882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,8,power_law_1.2,0.03949440121650696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.01,0.1993280053138733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,8,power_law_1.2,0.04226559996604919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.01,0.2522495985031128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,8,power_law_1.2,0.04584319889545441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.01,0.337228798866272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,8,power_law_1.2,0.04552960097789764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.01,0.39419519901275635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,8,power_law_1.2,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,16,power_law_1.2,0.283027195930481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,8,power_law_1.2,0.04530560076236725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,16,power_law_1.2,0.3284480094909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,16,power_law_1.2,0.4277376174926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,16,power_law_1.2,0.5228415966033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,16,power_law_1.2,0.7082496166229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.01,0.5788415908813477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,8,power_law_1.2,0.04723199903964996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,16,power_law_1.2,0.9645631790161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.01,0.7332672119140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,8,power_law_1.2,0.05253120064735413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,8,power_law_1.2,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,8,power_law_1.2,0.056505602598190305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,8,power_law_1.2,0.05869439840316772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,8,power_law_1.2,0.058278399705886844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,8,power_law_1.2,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,8,power_law_1.2,0.06273279786109924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,8,power_law_1.2,0.07422080039978027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,8,power_law_1.2,0.12863999605178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,8,power_law_1.2,0.1607743978500366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,8,power_law_1.2,0.12215679883956909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,8,power_law_1.2,0.1348736047744751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,8,power_law_1.2,0.17073919773101806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,8,power_law_1.2,0.22699520587921143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,8,power_law_1.2,0.28705921173095705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,8,power_law_1.2,0.40224637985229494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,8,power_law_1.2,0.49195518493652346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,64,power_law_1.2,0.08602240085601806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,64,power_law_1.2,0.07090560197830201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,64,power_law_1.2,0.07130240201950074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,64,power_law_1.2,0.06709120273590088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,64,power_law_1.2,0.06757760047912598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,64,power_law_1.2,0.06807680130004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,64,power_law_1.2,0.07177600264549255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,64,power_law_1.2,0.0716543972492218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,64,power_law_1.2,0.07494400143623352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,64,power_law_1.2,0.07360640168190002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,64,power_law_1.2,0.07368959784507752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,64,power_law_1.2,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,64,power_law_1.2,0.07731840014457703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,64,power_law_1.2,0.0785152018070221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,64,power_law_1.2,0.08627200126647949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,64,power_law_1.2,0.0900223970413208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,64,power_law_1.2,0.09585279822349549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,64,power_law_1.2,0.11338239908218384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,64,power_law_1.2,0.1171455979347229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,64,power_law_1.2,0.14464000463485718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,64,power_law_1.2,0.1697343945503235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,64,power_law_1.2,0.21230719089508057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,64,power_law_1.2,0.2442239999771118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,64,power_law_1.2,0.33787519931793214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,64,power_law_1.2,0.4483839988708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,64,power_law_1.2,0.8743040084838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,64,power_law_1.2,0.7655807971954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,1,power_law_1.01,0.046105599403381346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,1,power_law_1.01,0.06852480173110961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,1,power_law_1.01,0.11408640146255493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,1,power_law_1.01,0.180511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,1,power_law_1.01,0.2691904067993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,1,power_law_1.01,0.36890881061553954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,1,power_law_1.01,0.4993279933929443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,1,power_law_1.01,0.5331711769104004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,1,power_law_1.01,0.5652480125427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,1,power_law_1.01,0.5843008041381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,1,power_law_1.01,0.6091328144073487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,1,power_law_1.01,0.6320767879486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,1,power_law_1.01,0.6362175941467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,1,power_law_1.01,0.7145535945892334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,1,power_law_1.01,0.6570367813110352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,1,power_law_1.01,0.6971648216247559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,1,power_law_1.01,0.7460927963256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,1,power_law_1.01,0.7402880191802979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,1,power_law_1.01,0.7956992149353027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,1,power_law_1.01,0.8402303695678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,1,power_law_1.01,0.9118656158447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,1,power_law_1.01,2.280019187927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,1,power_law_1.01,2.669900894165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,1,power_law_1.01,1.544108772277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,1,power_law_1.01,1.7618560791015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,1,power_law_1.01,2.2537792205810545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,1,power_law_1.01,2.695705604553223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,2,power_law_1.01,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,2,power_law_1.01,0.09515519738197327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,2,power_law_1.01,0.11973760128021241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,2,power_law_1.01,0.16061439514160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,2,power_law_1.01,0.20205440521240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,2,power_law_1.01,0.2402048110961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,2,power_law_1.01,0.31110401153564454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,2,power_law_1.01,0.31307520866394045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,2,power_law_1.01,0.3331264019012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,2,power_law_1.01,0.33544321060180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,2,power_law_1.01,0.35253760814666746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,2,power_law_1.01,0.34922239780426023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,2,power_law_1.01,0.35486080646514895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,2,power_law_1.01,0.37906560897827146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,2,power_law_1.01,0.4040639877319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,2,power_law_1.01,0.4080959796905518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,2,power_law_1.01,0.40960001945495605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,2,power_law_1.01,0.4790336132049561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,2,power_law_1.01,0.4810175895690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,2,power_law_1.01,0.5620351791381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,2,power_law_1.01,0.6150144100189209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,2,power_law_1.01,0.7468863964080811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,2,power_law_1.01,0.8634048461914062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,2,power_law_1.01,1.1369664192199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,2,power_law_1.01,1.396345615386963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,2,power_law_1.01,1.930771255493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,2,power_law_1.01,2.497056007385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,1,power_law_1.01,0.20947840213775634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,1,power_law_1.01,0.33267199993133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,1,power_law_1.01,0.4592127799987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,1,power_law_1.01,0.7111167907714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,1,power_law_1.01,0.7282815933227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,1,power_law_1.01,0.7270783901214599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,1,power_law_1.01,0.7361279964447022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,1,power_law_1.01,0.7557631969451905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,1,power_law_1.01,0.7681727886199952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,1,power_law_1.01,0.7701632022857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,1,power_law_1.01,0.7774720191955566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,1,power_law_1.01,0.801580810546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,1,power_law_1.01,0.817568016052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,1,power_law_1.01,0.8719103813171387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,1,power_law_1.01,0.9035776138305665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,1,power_law_1.01,0.9184831619262696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,1,power_law_1.01,1.0035840034484864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,1,power_law_1.01,1.1935296058654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,1,power_law_1.01,1.335148811340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,1,power_law_1.01,1.572383975982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,1,power_law_1.01,1.9030080795288087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,1,power_law_1.01,2.529862403869629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,1,power_law_1.01,3.4787967681884764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,1,power_law_1.01,5.11976318359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,2,power_law_1.01,0.10088959932327271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,2,power_law_1.01,0.15008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,2,power_law_1.01,0.2487936019897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,2,power_law_1.01,0.30897281169891355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,2,power_law_1.01,0.4747456073760986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,2,power_law_1.01,0.6431104183197022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,2,power_law_1.01,0.682425594329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,2,power_law_1.01,0.8894463539123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,2,power_law_1.01,0.935974407196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,2,power_law_1.01,0.960268783569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,2,power_law_1.01,0.9959232330322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,2,power_law_1.01,1.0210111618041993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,2,power_law_1.01,0.9587648391723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,2,power_law_1.01,1.041811180114746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,2,power_law_1.01,1.0439359664916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,2,power_law_1.01,1.0858559608459473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,2,power_law_1.01,1.1284159660339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,2,power_law_1.01,1.186457633972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,2,power_law_1.01,1.2495360374450684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,2,power_law_1.01,1.407481575012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,2,power_law_1.01,1.4677632331848145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,2,power_law_1.01,1.6690687179565429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,2,power_law_1.01,1.6817216873168945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,2,power_law_1.01,2.013107109069824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,2,power_law_1.01,2.40503044128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,2,power_law_1.01,3.157548713684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,2,power_law_1.01,4.064588928222657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,1,power_law_1.01,7.213126373291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,1,power_law_1.01,10.143059539794923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,1,power_law_1.01,11.573785400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,32,power_law_1.01,0.021875199675559998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,32,power_law_1.01,0.021862399578094483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,32,power_law_1.01,0.02202879935503006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,32,power_law_1.01,0.021855999529361726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,32,power_law_1.01,0.023577600717544556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,32,power_law_1.01,0.029728001356124877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,32,power_law_1.01,0.049856001138687135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,32,power_law_1.01,0.049779200553894044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,32,power_law_1.01,0.05018879771232605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,32,power_law_1.01,0.04622719883918762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,32,power_law_1.01,0.047417598962783816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,32,power_law_1.01,0.047065600752830505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,32,power_law_1.01,0.047219198942184445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,32,power_law_1.01,0.04725759923458099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,32,power_law_1.01,0.04679040014743805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,32,power_law_1.01,0.04690560102462769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,32,power_law_1.01,0.05023999810218811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,32,power_law_1.01,0.04917759895324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.01,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.01,0.06528639793395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.01,0.07927680015563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.01,0.09841279983520508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.01,0.1235200047492981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.01,0.1626431941986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.01,0.21841280460357665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.01,0.3036992073059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.01,0.3733247995376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,16,balanced,0.049679999550183616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,16,balanced,0.03516799956560135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,16,balanced,0.03327466547489166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,16,balanced,0.03526933242877325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,16,balanced,0.04985066751639048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,16,balanced,0.05813866853713989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,16,balanced,0.05804799993832906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,16,balanced,0.05815466741720835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,16,balanced,0.057392001152038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,16,balanced,0.0580320010582606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,16,balanced,0.05942399799823761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,16,balanced,0.059765333930651345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,16,balanced,0.058592001597086586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,16,balanced,0.05977599819501241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,16,balanced,0.06356800099213918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,16,balanced,0.06412800153096516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,16,balanced,0.06388266881306966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,16,balanced,0.06618133187294006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,16,balanced,0.0665226678053538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,16,balanced,0.07372800012429555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,16,balanced,0.07284266750017802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,16,balanced,0.08872532844543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,16,balanced,0.08684800068537395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,16,balanced,0.12569600343704224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,16,balanced,0.11742400129636128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,16,balanced,0.15012266238530478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,16,balanced,0.17270400126775107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,4,power_law_1.01,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,4,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,4,power_law_1.01,0.03480319976806641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,4,power_law_1.01,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,4,power_law_1.01,0.03438720107078552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,4,power_law_1.01,0.03540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,4,power_law_1.01,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,4,power_law_1.01,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,4,power_law_1.01,0.03585279881954193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,4,power_law_1.01,0.036831998825073244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,4,power_law_1.01,0.03686400055885315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,4,power_law_1.01,0.03818239867687225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,4,power_law_1.01,0.03919999897480011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,4,power_law_1.01,0.03918719887733459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,4,power_law_1.01,0.04376960098743439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,4,power_law_1.01,0.04457600116729736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,4,power_law_1.01,0.04833920001983642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,4,power_law_1.01,0.05249279737472534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,4,power_law_1.01,0.05906559824943543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,4,power_law_1.01,0.07096959948539734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,4,power_law_1.01,0.07747200131416321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,4,power_law_1.01,0.09909120202064514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,4,power_law_1.01,0.13035520315170288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,4,power_law_1.01,0.19056639671325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,1,power_law_1.2,0.10130560398101807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,1,power_law_1.2,0.12637439966201783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,1,power_law_1.2,0.19381120204925537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,1,power_law_1.2,0.27607040405273436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,1,power_law_1.2,0.3753024101257324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,1,power_law_1.2,0.648134422302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,1,power_law_1.2,0.6950079917907714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,4,power_law_1.01,0.21060481071472167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,1,power_law_1.2,0.733190393447876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,1,power_law_1.2,0.7441855907440186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,1,power_law_1.2,0.7622975826263427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,1,power_law_1.2,0.8082367897033691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,1,power_law_1.2,0.8315967559814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,1,power_law_1.2,0.8535872459411621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,1,power_law_1.2,0.8741632461547851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,1,power_law_1.2,0.9070591926574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,1,power_law_1.2,0.9292863845825196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,1,power_law_1.2,0.9701760292053223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,1,power_law_1.2,1.0750656127929688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,1,power_law_1.2,1.171225643157959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,1,power_law_1.2,1.3860480308532714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,1,power_law_1.2,1.4261440277099608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,1,power_law_1.2,1.7578367233276366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,1,power_law_1.2,2.0448768615722654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,1,power_law_1.2,2.6632831573486326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,1,power_law_1.2,3.553228759765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,4,power_law_1.01,0.3267712116241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,1,power_law_1.2,4.7919872283935545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,1,power_law_1.2,5.82874870300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,4,power_law_1.01,0.3848128080368042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,4,power_law_1.2,0.09798399806022644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,4,power_law_1.2,0.09945600032806397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,4,power_law_1.2,0.09941120147705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,4,power_law_1.2,0.11909760236740112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,4,power_law_1.2,0.1192512035369873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,4,power_law_1.2,0.1266816020011902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,4,power_law_1.2,0.13089920282363893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,4,power_law_1.2,0.12806400060653686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,4,power_law_1.2,0.13655680418014526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,4,power_law_1.2,0.1487231969833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,4,power_law_1.2,0.14179840087890624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,4,power_law_1.2,0.15156480073928832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,4,power_law_1.2,0.15811840295791627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,4,power_law_1.2,0.17181440591812133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,4,power_law_1.2,0.203110408782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,4,power_law_1.2,0.21408638954162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,4,power_law_1.2,0.21755518913269042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,4,power_law_1.2,0.2744191884994507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,4,power_law_1.2,0.3269824028015137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,4,power_law_1.2,0.5090688228607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,4,power_law_1.2,0.44888958930969236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,4,power_law_1.2,0.674393606185913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,4,power_law_1.2,0.8791808128356934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,4,power_law_1.2,1.157158374786377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,4,power_law_1.2,1.7157567977905273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,4,power_law_1.2,2.6625280380249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,2,power_law_1.01,0.03283840119838714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,2,power_law_1.01,0.031763198971748355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,2,power_law_1.01,0.034246399998664856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,2,power_law_1.01,0.03442560136318207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,2,power_law_1.01,0.0350271999835968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,2,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,2,power_law_1.01,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,2,power_law_1.01,0.04185599982738495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,2,power_law_1.01,0.04469760060310364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,2,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,2,power_law_1.01,0.04758400022983551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,2,power_law_1.01,0.0492031991481781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,2,power_law_1.01,0.050297600030899045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,2,power_law_1.01,0.05286399722099304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,2,power_law_1.01,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,64,balanced,0.035743998984495796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,64,balanced,0.03401066611210505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,64,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,64,balanced,0.03399466723203659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,64,balanced,0.03524799893299738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,64,balanced,0.037674665451049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,64,balanced,0.03587199995915095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,64,balanced,0.03631466627120972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,64,balanced,0.03631466627120972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,64,balanced,0.03622400015592575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,64,balanced,0.03796799977620443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,64,balanced,0.03786666691303253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,64,balanced,0.037871999045213066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,64,balanced,0.03786666691303253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,64,balanced,0.041738669077555336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,64,balanced,0.042026668787002563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,2,power_law_1.01,0.06640639901161194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,2,power_law_1.01,0.07077760100364686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,2,power_law_1.01,0.08286719918251037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,64,balanced,0.04031466692686081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,64,balanced,0.042223999897638954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,64,balanced,0.044122666120529175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,64,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,64,balanced,0.04633066554864248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,64,balanced,0.05035200218359629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,64,balanced,0.053082664807637535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,64,balanced,0.06305600206057231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,64,balanced,0.07500799993673961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,64,balanced,0.09341333309809367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,64,balanced,0.10970133543014526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,2,power_law_1.01,0.09872639775276185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,2,power_law_1.01,0.12711039781570435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,2,power_law_1.01,0.15207680463790893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,2,power_law_1.01,0.20213119983673095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,2,power_law_1.01,0.23267199993133544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,4,power_law_1.2,3.4056766510009764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,2,power_law_1.01,0.34163200855255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,2,power_law_1.01,0.40584959983825686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,2,power_law_1.01,0.641209602355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,2,power_law_1.01,0.7919360160827636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,2,power_law_1.2,0.10606080293655396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,2,power_law_1.2,0.13193600177764891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,2,power_law_1.2,0.15779839754104613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,2,power_law_1.2,0.21091840267181397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,2,power_law_1.2,0.21123840808868408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,2,power_law_1.2,0.21010560989379884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,2,power_law_1.2,0.2118527889251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,2,power_law_1.2,0.2271872043609619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,2,power_law_1.2,0.23648641109466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,2,power_law_1.2,0.2322240114212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,2,power_law_1.2,0.2392064094543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,2,power_law_1.2,0.2437376022338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,2,power_law_1.2,0.24199039936065675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,2,power_law_1.2,0.2566783905029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,2,power_law_1.2,0.27248640060424806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,2,power_law_1.2,0.2880255937576294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,2,power_law_1.2,0.3046335935592651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,2,power_law_1.2,0.35798399448394774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,2,power_law_1.2,0.4393792152404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,2,power_law_1.2,0.5569664001464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,2,power_law_1.2,0.6517695903778076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,2,power_law_1.2,0.8941568374633789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,2,power_law_1.2,1.1871999740600585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,2,power_law_1.2,1.665510368347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,2,power_law_1.2,1.963596725463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,256,balanced,0.07860800127188365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,256,balanced,0.07869333525498708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,256,balanced,0.07860266665617625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,256,balanced,0.08181333541870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,256,balanced,0.08109866579373677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,256,balanced,0.0828000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,256,balanced,0.08447999755541484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,256,balanced,0.08428266644477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,256,balanced,0.0836906631787618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,256,balanced,0.08474133412043254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,256,balanced,0.08616532882054646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,256,balanced,0.08684266606966655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,256,balanced,0.08830400307973225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,256,balanced,0.08713600039482117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,256,balanced,0.0928053359190623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,4,power_law_1.01,0.06352000236511231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,256,balanced,0.09272000193595886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,4,power_law_1.01,0.06934400200843811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,256,balanced,0.09569066762924194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,256,balanced,0.10258133212725322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,256,balanced,0.10423466563224792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,4,power_law_1.01,0.07973120212554932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,256,balanced,0.1148426632086436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,256,balanced,0.12504532933235168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,4,power_law_1.01,0.07916160225868225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,4,power_law_1.01,0.09353600144386291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,4,power_law_1.01,0.10278400182723998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,4,power_law_1.01,0.10494719743728638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,4,power_law_1.01,0.11850880384445191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,256,balanced,0.14517866571744284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,4,power_law_1.01,0.11882239580154419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,256,balanced,0.16644799709320068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,256,balanced,0.20674665768941244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,4,power_law_1.01,0.11682560443878173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,256,balanced,0.24661332368850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,256,balanced,0.3300960063934326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,256,balanced,0.4082026481628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,4,power_law_1.01,0.12447999715805054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,4,power_law_1.01,0.12357120513916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,4,power_law_1.01,0.12640639543533325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,4,power_law_1.01,0.125600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,4,power_law_1.01,0.13855359554290772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,4,power_law_1.01,0.13722879886627198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,4,power_law_1.01,0.14933120012283324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,4,power_law_1.01,0.16829440593719483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,4,power_law_1.01,0.18651520013809203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,4,power_law_1.01,0.22209279537200927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,4,power_law_1.01,0.2585983991622925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,4,power_law_1.01,0.3344127893447876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,4,power_law_1.01,0.4073472023010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,4,power_law_1.01,0.5730368137359619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,4,power_law_1.01,0.7436992168426514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,4,power_law_1.01,1.0101183891296386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,4,power_law_1.01,1.2944000244140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,2,power_law_1.2,3.448953628540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,1,balanced,0.03779733429352442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,1,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,1,balanced,0.04118400067090988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,1,balanced,0.04699199895064036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,1,balanced,0.06861866513888042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,1,balanced,0.07222933570543925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,1,balanced,0.07372266550858815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,1,balanced,0.07249600191911061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,1,balanced,0.07468800246715546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,1,balanced,0.07458133498827617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,1,balanced,0.07858666777610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,1,balanced,0.07980266710122426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,1,balanced,0.08284266789754231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,1,balanced,0.08931199709574382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,1,balanced,0.09516266981760661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,1,balanced,0.09955733021100362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,1,balanced,0.1043893297513326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,1,balanced,0.12585066755612692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,1,balanced,0.1460693379243215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,1,balanced,0.18742932875951132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,1,balanced,0.20983999967575073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,1,balanced,0.3206239938735962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,1,balanced,0.3857119878133138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,1,balanced,0.5817813475926717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,1,balanced,0.7569866975148519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,1,balanced,1.1182453632354736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,1,balanced,1.4669706026713054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,2,power_law_1.2,3.4293376922607424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,2,balanced,0.07230400045712788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,2,balanced,0.07451733450094859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,2,balanced,0.08049599826335907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,2,balanced,0.0971999963124593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,2,balanced,0.13467199603716531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,2,balanced,0.19944000244140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,2,balanced,0.2653599977493286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,2,balanced,0.2531573375066121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,2,balanced,0.25282132625579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,2,balanced,0.2578666607538859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,2,balanced,0.2594933311144511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,2,balanced,0.2618933320045471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,2,balanced,0.26396799087524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,2,balanced,0.26899733146031696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,2,balanced,0.2777600089708964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,2,balanced,0.2774133284886678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,2,balanced,0.2910240093866984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,2,balanced,0.31666133801142377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,2,balanced,0.33429865042368573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,2,balanced,0.38195733229319256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,2,balanced,0.4230399926503499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,2,balanced,0.5213546752929688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,2,balanced,0.6268159945805868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,2,balanced,0.8283572991689047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,2,balanced,1.1073493162790935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,2,balanced,1.5210720698038738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,2,balanced,2.0033226013183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,2,power_law_1.2,0.06431360244750976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,2,power_law_1.2,0.0852735996246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,2,power_law_1.2,0.09154559969902039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,2,power_law_1.2,0.11990400552749633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,2,power_law_1.2,0.11909120082855225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,2,power_law_1.2,0.12088320255279542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,2,power_law_1.2,0.1399168014526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,2,power_law_1.2,0.12515840530395508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,2,power_law_1.2,0.12999680042266845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,2,power_law_1.2,0.13203840255737304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,2,power_law_1.2,0.13207679986953735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,2,power_law_1.2,0.1354367971420288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,2,power_law_1.2,0.14282239675521852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,2,power_law_1.2,0.15130239725112915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,2,power_law_1.2,0.1570304036140442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,2,power_law_1.2,0.16403199434280397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,2,power_law_1.2,0.1833024024963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,2,power_law_1.2,0.21136000156402587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,2,power_law_1.2,0.23569920063018798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,2,power_law_1.2,0.2878079891204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,2,power_law_1.2,0.34866559505462646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,2,power_law_1.2,0.4666624069213867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,2,power_law_1.2,0.5719808101654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,2,power_law_1.2,0.8872256278991699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,2,power_law_1.2,1.0691200256347657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,64,power_law_1.2,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,64,power_law_1.2,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,64,power_law_1.2,0.06110720038414001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,64,power_law_1.2,0.0632960021495819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,64,power_law_1.2,0.06497920155525208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,64,power_law_1.2,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,64,power_law_1.2,0.06919040083885193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,64,power_law_1.2,0.066457599401474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,64,power_law_1.2,0.0676800012588501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,64,power_law_1.2,0.0666815996170044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,64,power_law_1.2,0.06657919883728028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,64,power_law_1.2,0.06938239932060242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,64,power_law_1.2,0.07028480172157288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,64,power_law_1.2,0.07196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,64,power_law_1.2,0.07723519802093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,64,power_law_1.2,0.07964800000190735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,64,power_law_1.2,0.08599039912223816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,64,power_law_1.2,0.10380799770355224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,64,power_law_1.2,0.11637760400772094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,64,power_law_1.2,0.1450495958328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,64,power_law_1.2,0.16858880519866942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,64,power_law_1.2,0.22387199401855468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,64,power_law_1.2,0.27099521160125734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,64,power_law_1.2,0.37784318923950194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,64,power_law_1.2,0.4737152099609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,64,power_law_1.2,0.7351935863494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,64,power_law_1.2,0.9291839599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,2,power_law_1.2,1.554905605316162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,2,power_law_1.2,1.719513511657715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,4,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,4,balanced,0.037802666425704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,4,balanced,0.03939199944337209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,4,balanced,0.043621331453323364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,4,balanced,0.051856001218159996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,4,balanced,0.0537066658337911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,4,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,4,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,4,balanced,0.05407999952634176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,4,balanced,0.05365333457787832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,4,balanced,0.05612266560395559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,4,balanced,0.056143999099731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,4,balanced,0.05851200222969055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,4,balanced,0.05856533348560333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,4,balanced,0.06216000020503998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,4,balanced,0.06705600023269653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,4,balanced,0.06885333359241486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,4,balanced,0.0758186678091685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,4,balanced,0.08481066425641377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,4,balanced,0.09776533643404643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,4,balanced,0.11051199833552043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,4,balanced,0.141157329082489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,4,balanced,0.1625173290570577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,4,balanced,0.22285334269205728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,4,balanced,0.2834239999453227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,4,balanced,0.39700265725453693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,4,balanced,0.5138826767603556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,128,power_law_1.2,0.029728001356124877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,4,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,4,balanced,0.03955200066169103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,128,power_law_1.2,0.028044798970222475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,4,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,4,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,4,balanced,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,4,balanced,0.07459199925263722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,4,balanced,0.07630933324495952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,4,balanced,0.07727999985218048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,4,balanced,0.07666666805744171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,4,balanced,0.07892799874146779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,4,balanced,0.07780799766381581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,4,balanced,0.08036266764005025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,4,balanced,0.0827466646830241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,4,balanced,0.0845973292986552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,128,power_law_1.2,0.027449598908424376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,4,balanced,0.08905067046483357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,128,power_law_1.2,0.027820798754692077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,4,balanced,0.09061866998672485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,4,balanced,0.0953493316968282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,4,balanced,0.10487999518712361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,4,balanced,0.10985599954922994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,4,balanced,0.1288533310095469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,4,balanced,0.14126933614412943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,128,power_law_1.2,0.028492799401283263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,128,power_law_1.2,0.02975359857082367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,4,balanced,0.1843093236287435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,128,power_law_1.2,0.030137598514556885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,128,power_law_1.2,0.02983039915561676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,128,power_law_1.2,0.030092799663543703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,128,power_law_1.2,0.030291199684143066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,128,power_law_1.2,0.0315200001001358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,128,power_law_1.2,0.031916800141334536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,128,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,128,power_law_1.2,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,128,power_law_1.2,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,128,power_law_1.2,0.037459200620651244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,4,balanced,0.21074666579564413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,128,power_law_1.2,0.03940480053424835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,4,balanced,0.29686933755874634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,128,power_law_1.2,0.04439679980278015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,4,balanced,0.37300801277160645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,128,power_law_1.2,0.04984959959983826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,4,balanced,0.5346506834030151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,4,balanced,0.6983359654744467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,128,power_law_1.2,0.05596799850463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,128,power_law_1.2,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,128,power_law_1.2,0.07842559814453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,128,power_law_1.2,0.08598399758338929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,128,power_law_1.2,0.12987519502639772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,128,power_law_1.2,0.165664005279541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,128,power_law_1.2,0.24183039665222167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,2,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,2,balanced,0.04121066629886627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,2,balanced,0.046256000796953835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,2,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,2,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,2,balanced,0.046442667643229164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,2,balanced,0.04560533165931702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,2,balanced,0.04795733094215393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,2,balanced,0.04834666848182678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,2,balanced,0.0459146648645401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,2,balanced,0.04586133360862732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,2,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,2,balanced,0.048351998130480446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,2,balanced,0.04752533137798309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,2,balanced,0.051813334226608276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,2,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,2,balanced,0.05590933561325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,2,balanced,0.06837333242098491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,2,balanced,0.06834666430950165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,2,balanced,0.08481066425641377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,2,balanced,0.09762666622797649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,2,balanced,0.1253919998804728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,2,balanced,0.15212266643842062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,2,balanced,0.20968000094095865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,2,balanced,0.26263999938964844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,2,balanced,0.370410680770874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,2,balanced,0.4766293366750081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,16,balanced,0.04892266790072123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,16,balanced,0.04913066824277242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,16,balanced,0.04983466863632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,16,balanced,0.05282133320967356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,16,balanced,0.05568000177542368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,16,balanced,0.07277333239714305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,16,balanced,0.07306133210659027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,16,balanced,0.07329066594441731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,16,balanced,0.07452266911665599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,16,balanced,0.07363200187683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,16,balanced,0.07456533114115398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,16,balanced,0.07682666679223378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,16,balanced,0.07642666498819987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,16,balanced,0.07860266665617625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,16,balanced,0.09099200367927551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,16,balanced,0.09110400080680847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,16,balanced,0.09493866562843323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,16,balanced,0.10896533727645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,16,balanced,0.11127466956774394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,16,balanced,0.12784000237782797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,16,balanced,0.14215999841690063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,16,balanced,0.17165333032608032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,16,balanced,0.20080000162124634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,16,balanced,0.26871999104817706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,16,balanced,0.32973333199818927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,16,balanced,0.4691893259684245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,16,balanced,0.6219146649042765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,2,balanced,0.021168000996112823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,2,balanced,0.020687999824682873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,2,balanced,0.01934933289885521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,2,balanced,0.01926933353145917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,2,balanced,0.021082667013009388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,2,balanced,0.021146667500336964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,2,balanced,0.02712533374627431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,2,balanced,0.025253333151340485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,2,balanced,0.029167999823888142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,2,balanced,0.02787200113137563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,2,balanced,0.029301332930723827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,2,balanced,0.031130666534105938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,2,balanced,0.03486400097608566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,2,balanced,0.035605333745479584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,2,balanced,0.04560533165931702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,2,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,2,balanced,0.045797333121299744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,2,balanced,0.05909333129723867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,2,balanced,0.06715199848016103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,2,balanced,0.08805867036183675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,2,balanced,0.10838400324185689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,2,balanced,0.14788800477981567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,2,balanced,0.18117332458496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,2,balanced,0.25651200612386066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,2,balanced,0.3290826678276062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,2,balanced,0.4756693442662557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,2,balanced,0.6226880153020223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,128,power_law_1.2,0.29611520767211913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,64,power_law_1.01,0.0838591992855072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,64,power_law_1.01,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,64,power_law_1.01,0.04819200038909912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,64,power_law_1.01,0.05028480291366577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,64,power_law_1.01,0.04800640046596527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,64,power_law_1.01,0.051097601652145386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,64,power_law_1.01,0.05247359871864319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,64,power_law_1.01,0.051571202278137204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,64,power_law_1.01,0.05310080051422119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,64,power_law_1.01,0.05459200143814087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,64,power_law_1.01,0.05677440166473389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,64,power_law_1.01,0.05812479853630066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,64,power_law_1.01,0.060102397203445436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,64,power_law_1.01,0.06956160068511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,64,power_law_1.01,0.0765824019908905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,64,power_law_1.01,0.07789440155029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,64,power_law_1.01,0.08409600257873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,64,power_law_1.01,0.10365439653396606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,64,power_law_1.01,0.10497280359268188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,64,power_law_1.01,0.13275519609451295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,64,power_law_1.01,0.14568320512771607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,64,power_law_1.01,0.196563196182251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,64,power_law_1.01,0.23685119152069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,64,power_law_1.01,0.3309695959091187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,64,power_law_1.01,0.4079423904418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,64,power_law_1.01,0.5945343971252441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,64,power_law_1.01,0.7491775989532471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,64,power_law_1.01,0.021503999829292297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,64,power_law_1.01,0.01865600049495697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,64,power_law_1.01,0.01849599927663803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,64,power_law_1.01,0.019308799505233766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,64,power_law_1.01,0.02131199985742569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,64,power_law_1.01,0.025203201174736022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,64,power_law_1.01,0.025248000025749208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,64,power_law_1.01,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,64,power_law_1.01,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,64,power_law_1.01,0.038553598523139956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,64,power_law_1.01,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,64,power_law_1.01,0.03837440013885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,64,power_law_1.01,0.03839359879493713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,64,power_law_1.01,0.03384959995746613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,64,power_law_1.01,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,64,power_law_1.01,0.040992000699043275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,64,power_law_1.01,0.04261760115623474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,64,power_law_1.01,0.04131839871406555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.01,0.048582398891448976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.01,0.057715201377868654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.01,0.06652160286903382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.01,0.08672639727592468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.01,0.10869760513305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.01,0.14431359767913818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.01,0.17765120267868043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.01,0.27587840557098386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.01,0.34910080432891843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,8,power_law_1.01,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,8,power_law_1.01,0.03774079978466034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,8,power_law_1.01,0.03790079951286316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,8,power_law_1.01,0.040300801396369934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,8,power_law_1.01,0.0398719996213913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,8,power_law_1.01,0.040038400888442995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,8,power_law_1.01,0.039955198764801025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,8,power_law_1.01,0.03918719887733459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,8,power_law_1.01,0.038790398836135866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,8,power_law_1.01,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,8,power_law_1.01,0.03781760036945343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,8,power_law_1.01,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,8,power_law_1.01,0.042316800355911253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,8,power_law_1.01,0.04128639996051788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,8,power_law_1.01,0.047040000557899475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,8,power_law_1.01,0.04907520115375519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,8,power_law_1.01,0.05351679921150208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,8,power_law_1.01,0.05975040197372437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.01,0.06821119785308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.01,0.0798143982887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.01,0.08904320001602173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.01,0.12806400060653686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.01,0.15575679540634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.01,0.2149183988571167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.01,0.27733120918273924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.01,0.3766016006469727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.01,0.471014404296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,4,power_law_1.2,0.048019200563430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,4,power_law_1.2,0.05118719935417175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,4,power_law_1.2,0.05552639961242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,4,power_law_1.2,0.0711296021938324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,4,power_law_1.2,0.08015360236167908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,4,power_law_1.2,0.0866752028465271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,4,power_law_1.2,0.09289600253105164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,4,power_law_1.2,0.09735040068626404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,4,power_law_1.2,0.09866880178451538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,4,power_law_1.2,0.09960319995880126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,4,power_law_1.2,0.10670720338821411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,4,power_law_1.2,0.1118783950805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,4,power_law_1.2,0.11553280353546143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,4,power_law_1.2,0.11881599426269532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,4,power_law_1.2,0.13153280019760133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,4,power_law_1.2,0.13319679498672485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,4,power_law_1.2,0.14342399835586547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,4,power_law_1.2,0.1629439949989319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,4,power_law_1.2,0.18061439990997313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,4,power_law_1.2,0.22624640464782714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,4,power_law_1.2,0.2657279968261719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,4,power_law_1.2,0.3343679904937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,4,power_law_1.2,0.4269567966461182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,4,power_law_1.2,0.5590911865234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,4,power_law_1.2,0.7067967891693115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,4,power_law_1.2,1.0380672454833983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,4,power_law_1.2,1.3559679985046387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,8,power_law_1.2,0.04631040096282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,8,power_law_1.2,0.050432002544403075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,8,power_law_1.2,0.054502397775650024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,8,power_law_1.2,0.056992000341415404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,8,power_law_1.2,0.061900800466537474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,8,power_law_1.2,0.06869120001792908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,8,power_law_1.2,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,8,power_law_1.2,0.07060480117797852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,8,power_law_1.2,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,8,power_law_1.2,0.07154560089111328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,8,power_law_1.2,0.0716863989830017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,8,power_law_1.2,0.07720320224761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,8,power_law_1.2,0.07864959836006165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,8,power_law_1.2,0.0846783995628357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,8,power_law_1.2,0.09644799828529357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,8,power_law_1.2,0.09547520279884339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,8,power_law_1.2,0.10913920402526855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,8,power_law_1.2,0.1300160050392151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,8,power_law_1.2,0.1381119966506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,8,power_law_1.2,0.17658239603042603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,8,power_law_1.2,0.21868159770965576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,8,power_law_1.2,0.2706048011779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,8,power_law_1.2,0.35820159912109373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,8,power_law_1.2,0.4807551860809326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,8,power_law_1.2,0.6547391891479493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,8,power_law_1.2,0.8747903823852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,8,power_law_1.2,1.1363264083862306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,1,balanced,0.03181333343187968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,2,balanced,0.03368533402681351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,2,balanced,0.035962666074434914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,2,balanced,0.05057600140571594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,2,balanced,0.07271466652552287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,2,balanced,0.10355200370152791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,2,balanced,0.12946133812268576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,2,balanced,0.13129066427548727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,2,balanced,0.13265066345532736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,2,balanced,0.1325813333193461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,2,balanced,0.1344266633192698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,2,balanced,0.13205333550771078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,2,balanced,0.13612799843152365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,2,balanced,0.13750400145848593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,2,balanced,0.1360106666882833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,2,balanced,0.14180800318717957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,2,balanced,0.1483840048313141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,2,balanced,0.14788267016410828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,1,balanced,0.03182400017976761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,2,balanced,0.1604426701863607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,2,balanced,0.16060800353686014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,2,balanced,0.21887999773025513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,2,balanced,0.18728532393773398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,2,balanced,0.2524639964103699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,2,balanced,0.2561653256416321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,2,balanced,0.3707733154296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,2,balanced,0.40689067045847577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,2,balanced,0.5814400116602579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,2,balanced,0.6774506568908691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,1,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,1,balanced,0.03738666574160258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,1,balanced,0.035445332527160645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,1,balanced,0.03572800010442734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,1,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,1,balanced,0.037589333951473236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,1,balanced,0.03849600007136663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,1,balanced,0.03752533346414566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,1,balanced,0.039818666875362396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,1,balanced,0.04182933270931244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,1,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,1,balanced,0.044981335600217186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,1,balanced,0.05242133140563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,1,balanced,0.054192001620928444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,1,balanced,0.055829331278800964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,1,balanced,0.07260799904664357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,1,balanced,0.0780213326215744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,1,balanced,0.10875733693440755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,1,balanced,0.1234826644261678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,1,balanced,0.16646933555603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,1,balanced,0.19875733057657877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,1,balanced,0.2813919981320699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,1,balanced,0.35947732130686444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,1,balanced,0.5215200185775757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,1,balanced,0.6766186555226644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,32,power_law_1.01,0.052876800298690796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,32,power_law_1.01,0.053958398103713986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,32,power_law_1.01,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,32,power_law_1.01,0.049344000220298764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,32,power_law_1.01,0.0494592010974884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,32,power_law_1.01,0.05281919836997986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,32,power_law_1.01,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,32,power_law_1.01,0.05330560207366943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,32,power_law_1.01,0.05403519868850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,32,power_law_1.01,0.05742719769477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,32,power_law_1.01,0.059910398721694944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,32,power_law_1.01,0.06198400259017944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,32,power_law_1.01,0.0639743983745575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,32,power_law_1.01,0.06561279892921448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,32,power_law_1.01,0.07422080039978027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,32,power_law_1.01,0.07871360182762147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,32,power_law_1.01,0.08214399814605713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,32,power_law_1.01,0.10126080513000488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,32,power_law_1.01,0.10464639663696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,32,power_law_1.01,0.12794239521026612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,32,power_law_1.01,0.15354880094528198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,32,power_law_1.01,0.19708800315856934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,32,power_law_1.01,0.24223999977111815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,32,power_law_1.01,0.3297919988632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,32,power_law_1.01,0.4126783847808838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,32,power_law_1.01,0.599616003036499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,32,power_law_1.01,0.757260799407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,4,balanced,0.044031997521718345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,4,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,4,balanced,0.05013333261013031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,4,balanced,0.05563200016816457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,4,balanced,0.07841599980990092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,4,balanced,0.11131733655929565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,4,balanced,0.11473600069681804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,4,balanced,0.11319999893506368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,4,balanced,0.11615467071533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,4,balanced,0.11758933464686076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,4,balanced,0.11738666892051697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,4,balanced,0.11917866269747417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,4,balanced,0.1230560044447581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,4,balanced,0.12332266569137573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,4,balanced,0.132341335217158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,4,balanced,0.13587733109792074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,4,balanced,0.13988266388575235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,4,balanced,0.15468266606330872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,4,balanced,0.16383999586105347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,4,balanced,0.1876586675643921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,4,balanced,0.21162132422129312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,4,balanced,0.2763146758079529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,4,balanced,0.3253866632779439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,4,balanced,0.43962132930755615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,4,balanced,0.5464426676432291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,4,balanced,0.7745920022328695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,4,balanced,0.9797226587931315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,16,power_law_1.2,0.05676159858703613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,16,power_law_1.2,0.07199360132217407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,16,power_law_1.2,0.05568000078201294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,16,power_law_1.2,0.06059520244598389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,16,power_law_1.2,0.05982080101966858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,16,power_law_1.2,0.058195197582244874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,16,power_law_1.2,0.06085119843482971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,16,power_law_1.2,0.060889601707458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,16,power_law_1.2,0.06212480068206787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,16,power_law_1.2,0.06152960062026978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,16,power_law_1.2,0.06305919885635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,16,power_law_1.2,0.0650816023349762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,16,power_law_1.2,0.06548479795455933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,16,power_law_1.2,0.06690559983253479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,16,power_law_1.2,0.07464320063591004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,16,power_law_1.2,0.07700480222702026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,16,power_law_1.2,0.08414720296859741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,16,power_law_1.2,0.09583359956741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,16,power_law_1.2,0.1085312008857727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,16,power_law_1.2,0.1355455994606018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,16,power_law_1.2,0.1523200035095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,16,power_law_1.2,0.19944319725036622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,16,power_law_1.2,0.2342463970184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,16,power_law_1.2,0.3250751972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,16,power_law_1.2,0.3986560106277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,16,power_law_1.2,0.5437632083892823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,16,power_law_1.2,0.8222720146179199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,128,power_law_1.01,0.030700799822807313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,128,power_law_1.01,0.030508801341056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,128,power_law_1.01,0.030623999238014222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,128,power_law_1.01,0.03281919956207276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,128,power_law_1.01,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,128,power_law_1.01,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,128,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,128,power_law_1.01,0.03447040021419525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,128,power_law_1.01,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,128,power_law_1.01,0.03455359935760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,128,power_law_1.01,0.03573119938373566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,128,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,128,power_law_1.01,0.03656319975852966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,128,power_law_1.01,0.03729279935359955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,128,power_law_1.01,0.03973760008811951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,128,power_law_1.01,0.0406143993139267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,128,power_law_1.01,0.04031359851360321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,128,power_law_1.01,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,128,power_law_1.01,0.045542401075363156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,128,power_law_1.01,0.05139840245246887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,128,power_law_1.01,0.05680000185966492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,128,power_law_1.01,0.06936320066452026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,128,power_law_1.01,0.07917439937591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,128,power_law_1.01,0.11639679670333862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,128,power_law_1.01,0.1410431981086731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,128,power_law_1.01,0.20356481075286864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,128,power_law_1.01,0.2633984088897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,2,balanced,0.085999995470047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,2,balanced,0.12774399916330972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,2,balanced,0.19276267290115356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,2,balanced,0.33909865220387775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,2,balanced,0.6075626611709595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,2,balanced,0.850821336110433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,2,balanced,0.8542239665985107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,2,balanced,0.8631093502044678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,2,balanced,0.8679413000742594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,2,balanced,0.8681653340657552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,2,balanced,0.8702613512674967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,2,balanced,0.8777013619740804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,2,balanced,0.8793813387552897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,2,balanced,0.8882880210876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,2,balanced,0.9016319910685221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,2,balanced,0.9088426431020101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,2,balanced,0.9276373386383057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,2,balanced,1.1390559673309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,2,balanced,0.9944852987925211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,2,balanced,1.2722453276316326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,2,balanced,1.1476799647013347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,2,balanced,1.6625919342041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,2,balanced,1.4154772758483887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,2,balanced,2.213541348775228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,2,balanced,2.2205920219421387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,2,balanced,3.3649654388427734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,2,balanced,3.6512266794840493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,16,balanced,0.06217599908510844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,16,balanced,0.06229866544405619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,16,balanced,0.07008533179759979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,16,balanced,0.09059733152389526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,16,balanced,0.12168000141779582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,16,balanced,0.18033599853515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,16,balanced,0.18132267395655313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,16,balanced,0.18099733193715414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,16,balanced,0.18339733282725015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,16,balanced,0.18184000253677368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,16,balanced,0.18306666612625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,16,balanced,0.18503999710083008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,16,balanced,0.1861120065053304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,16,balanced,0.18817599614461264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,16,balanced,0.20662933588027954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,16,balanced,0.2076479991277059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,16,balanced,0.21201066176096597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,16,balanced,0.23068799575169882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,16,balanced,0.2342346707979838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,16,balanced,0.2488053242365519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,16,balanced,0.2643466591835022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,16,balanced,0.29953600962956745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,16,balanced,0.3322506745656331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,16,balanced,0.4380640188852946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,16,balanced,0.5069119930267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,16,balanced,0.7052533626556396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,16,balanced,0.9136373202006022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,8,balanced,0.08666132887204488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,8,balanced,0.08958400289217631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,8,balanced,0.08776000142097473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,8,balanced,0.08724266290664673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,8,balanced,0.08639466762542725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,8,balanced,0.08710933725039165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,8,balanced,0.08853866656621297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,8,balanced,0.0885813335577647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,8,balanced,0.0904906690120697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,8,balanced,0.09077866872151692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,8,balanced,0.0943946639696757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,8,balanced,0.09904533624649048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,8,balanced,0.10111467043558757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,8,balanced,0.10444800059000652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,8,balanced,0.1053493320941925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,8,balanced,0.10711466272672017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,8,balanced,0.1090773344039917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,8,balanced,0.13539200027783713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,8,balanced,0.1302773356437683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,8,balanced,0.16672533750534058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,8,balanced,0.17364799976348877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,8,balanced,0.2513759930928548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,8,balanced,0.28833599885304767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,8,balanced,0.3938933213551839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,8,balanced,0.47545599937438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,8,balanced,0.6768693129221598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,8,balanced,0.8788159688313802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,2,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,2,power_law_1.2,0.031225600838661195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,2,power_law_1.2,0.03407360017299652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,2,power_law_1.2,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,2,power_law_1.2,0.0406143993139267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,2,power_law_1.2,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,2,power_law_1.2,0.05882880091667175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,2,power_law_1.2,0.06030719876289368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,2,power_law_1.2,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,2,power_law_1.2,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,2,power_law_1.2,0.06643199920654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,2,power_law_1.2,0.07057279944419861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,2,power_law_1.2,0.07221119999885559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,2,power_law_1.2,0.07360000014305115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,2,power_law_1.2,0.08161919713020324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,2,power_law_1.2,0.08577920198440551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,2,power_law_1.2,0.09059839844703674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,2,power_law_1.2,0.10500479936599731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,2,power_law_1.2,0.12086399793624877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,2,power_law_1.2,0.14755840301513673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,2,power_law_1.2,0.1693120002746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,2,power_law_1.2,0.21349120140075684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,2,power_law_1.2,0.2827712059020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,2,power_law_1.2,0.4028800010681152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,2,power_law_1.2,0.5169600009918213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,16,power_law_1.2,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,16,power_law_1.2,0.03893119990825653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,16,power_law_1.2,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,16,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,16,power_law_1.2,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,16,power_law_1.2,0.03570559918880463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,16,power_law_1.2,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,16,power_law_1.2,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,16,power_law_1.2,0.03653759956359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,2,power_law_1.2,0.7536767959594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,16,power_law_1.2,0.038822400569915774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,16,power_law_1.2,0.03982079923152924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,16,power_law_1.2,0.04142079949378967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,16,power_law_1.2,0.042828801274299624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,16,power_law_1.2,0.04894720017910004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,16,power_law_1.2,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,16,power_law_1.2,0.05919359922409058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,16,power_law_1.2,0.060134398937225345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,16,power_law_1.2,0.06328960061073304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,16,power_law_1.2,0.0737280011177063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,16,power_law_1.2,0.08284159898757934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,16,power_law_1.2,0.09813119769096375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,16,power_law_1.2,0.14683519601821898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,16,power_law_1.2,0.14300160408020018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,2,power_law_1.2,0.9954048156738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,16,power_law_1.2,0.19658880233764647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,16,power_law_1.2,0.26065280437469485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,16,power_law_1.2,0.3416383981704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,2,power_law_1.2,0.07638400197029113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,2,power_law_1.2,0.08294399976730346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,2,power_law_1.2,0.0923200011253357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,2,power_law_1.2,0.1053887963294983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,2,power_law_1.2,0.12496000528335571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,2,power_law_1.2,0.14379520416259767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,2,power_law_1.2,0.16350719928741456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,2,power_law_1.2,0.1850111961364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,2,power_law_1.2,0.18799359798431398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,2,power_law_1.2,0.19448959827423096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,2,power_law_1.2,0.20213119983673095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,2,power_law_1.2,0.20650238990783693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,2,power_law_1.2,0.20712320804595946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,2,power_law_1.2,0.22136321067810058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,2,power_law_1.2,0.22435200214385986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,2,power_law_1.2,0.22766718864440919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,2,power_law_1.2,0.2535871982574463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,2,power_law_1.2,0.2777343988418579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,2,power_law_1.2,0.3075263977050781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,2,power_law_1.2,0.3711168050765991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,2,power_law_1.2,0.4164480209350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,2,power_law_1.2,0.5404096126556397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,2,power_law_1.2,0.65382399559021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,2,power_law_1.2,0.9555647850036622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,2,power_law_1.2,1.1617152214050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,2,power_law_1.2,1.6533184051513672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,2,power_law_1.2,2.183340835571289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,16,power_law_1.2,0.43191680908203123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,16,balanced,0.034586665530999504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,16,balanced,0.03374933451414108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,16,balanced,0.03376533339420954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,16,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,16,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,16,balanced,0.04353600243727366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,16,balanced,0.04393066465854645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,16,balanced,0.045738667249679565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,16,balanced,0.044906665881474815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,16,balanced,0.04502399762471517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,16,balanced,0.045642669002215065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,16,balanced,0.05410666763782501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,16,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,16,balanced,0.05596266686916351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,16,balanced,0.05831466615200043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,16,balanced,0.05947199960549673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,16,balanced,0.06227200229962667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,16,balanced,0.0689333329598109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,16,balanced,0.07249066730340321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,16,balanced,0.08323733508586884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,16,balanced,0.08826667070388794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,16,balanced,0.11050132910410564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,16,balanced,0.12411733468373616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,16,balanced,0.16815465688705444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,16,balanced,0.19973333676656088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,16,balanced,0.2832213242848714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,16,balanced,0.37033601601918537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,2,power_law_1.2,0.10849920511245728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,2,power_law_1.2,0.1307968020439148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,2,power_law_1.2,0.17910399436950683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,2,power_law_1.2,0.24173440933227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,2,power_law_1.2,0.2953792095184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,2,power_law_1.2,0.4856383800506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,2,power_law_1.2,0.482374382019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,2,power_law_1.2,0.511571216583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,2,power_law_1.2,0.5357696056365967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,2,power_law_1.2,0.5196544170379639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,2,power_law_1.2,0.5486464023590087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,2,power_law_1.2,0.5759359836578369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,2,power_law_1.2,0.5753600120544433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,2,power_law_1.2,0.569920015335083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,2,power_law_1.2,0.5952320098876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,2,power_law_1.2,0.6368959903717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,2,power_law_1.2,0.6420159816741944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,2,power_law_1.2,0.689683198928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,2,power_law_1.2,0.7668735980987549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,2,power_law_1.2,0.9085311889648438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,2,power_law_1.2,1.070035171508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,2,power_law_1.2,1.268172836303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,2,power_law_1.2,1.2434368133544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,2,power_law_1.2,1.5188544273376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,2,power_law_1.2,2.2826303482055663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,2,power_law_1.2,2.628985595703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,2,power_law_1.2,3.3884609222412108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,8,balanced,0.07442666590213776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,8,balanced,0.078575998544693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,128,power_law_1.2,0.017452800273895265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,128,power_law_1.2,0.01713919937610626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,128,power_law_1.2,0.017862400412559508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,64,power_law_1.01,0.06181120276451111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,64,power_law_1.01,0.05800960063934326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,64,power_law_1.01,0.057599997520446776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,64,power_law_1.01,0.060710400342941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,64,power_law_1.01,0.05870720148086548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,64,power_law_1.01,0.058950400352478026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,128,power_law_1.2,0.01820800006389618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,64,power_law_1.01,0.060089600086212155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,128,power_law_1.2,0.021062399446964263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,64,power_law_1.01,0.06242560148239136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,128,power_law_1.2,0.03540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,64,power_law_1.01,0.06285439729690552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,128,power_law_1.2,0.030483201146125793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,64,power_law_1.01,0.06227200031280518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,128,power_law_1.2,0.030483201146125793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,64,power_law_1.01,0.06183680295944214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,64,power_law_1.01,0.06335359811782837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,64,power_law_1.01,0.0659775972366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,64,power_law_1.01,0.06641280055046081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,64,power_law_1.01,0.07127040028572082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,64,power_law_1.01,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,128,power_law_1.2,0.030527999997138976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,64,power_law_1.01,0.07401599884033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,128,power_law_1.2,0.030553600192070006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,128,power_law_1.2,0.030803200602531434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,128,power_law_1.2,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,128,power_law_1.2,0.030502399802207945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,128,power_law_1.2,0.030822399258613586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,128,power_law_1.2,0.03130879998207092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,128,power_law_1.2,0.03177599906921387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,128,power_law_1.2,0.03324800133705139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,128,power_law_1.2,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,128,power_law_1.2,0.037452799081802365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,128,power_law_1.2,0.0435263991355896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,128,power_law_1.2,0.04737919867038727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,128,power_law_1.2,0.058739197254180905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,128,power_law_1.2,0.0746944010257721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,64,power_law_1.01,0.08207359910011292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,128,power_law_1.2,0.1008255958557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,64,power_law_1.01,0.08919680118560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,64,power_law_1.01,0.10677119493484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,64,power_law_1.01,0.12272000312805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,64,power_law_1.01,0.1552832007408142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,64,power_law_1.01,0.18826240301132202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,64,power_law_1.01,0.25029759407043456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,64,power_law_1.01,0.3192255973815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,128,power_law_1.2,0.12410240173339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,64,power_law_1.01,0.4488959789276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,128,power_law_1.2,0.17871999740600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,64,power_law_1.01,0.559123182296753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,128,power_law_1.2,0.23001599311828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,8,power_law_1.2,0.037427198886871335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,8,power_law_1.2,0.039110401272773744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,8,power_law_1.2,0.04402559995651245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,8,power_law_1.2,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,8,power_law_1.2,0.04444800019264221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,8,power_law_1.2,0.05373439788818359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,8,power_law_1.2,0.053504002094268796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,8,power_law_1.2,0.054604798555374146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,8,power_law_1.2,0.05493119955062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,8,power_law_1.2,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,8,power_law_1.2,0.05816320180892944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,8,power_law_1.2,0.057792001962661745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,8,power_law_1.2,0.061964797973632815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,8,power_law_1.2,0.06296319961547851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,8,power_law_1.2,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,8,power_law_1.2,0.0704576015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,8,power_law_1.2,0.0716159999370575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,8,power_law_1.2,0.07929599881172181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,8,power_law_1.2,0.0830847978591919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,8,power_law_1.2,0.09891200065612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,8,power_law_1.2,0.10822399854660034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,8,power_law_1.2,0.1356735944747925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,8,power_law_1.2,0.16353919506072997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,8,power_law_1.2,0.19760639667510987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,8,power_law_1.2,0.2552191972732544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,8,power_law_1.2,0.3377471923828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,8,power_law_1.2,0.4119552135467529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,8,power_law_1.01,0.024959999322891235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,8,power_law_1.01,0.022681599855422972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,8,power_law_1.01,0.023814399540424348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,8,power_law_1.01,0.025472000241279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,8,power_law_1.01,0.027775999903678895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,8,power_law_1.01,0.029395198822021483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,8,power_law_1.01,0.03158400058746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,8,power_law_1.01,0.030444800853729248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,8,power_law_1.01,0.03258880078792572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,8,power_law_1.01,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,8,power_law_1.01,0.03880319893360138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,8,power_law_1.01,0.04306559860706329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,8,power_law_1.01,0.04459519982337952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,8,power_law_1.01,0.04111360013484955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,8,power_law_1.01,0.04620159864425659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,8,power_law_1.01,0.049081599712371825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,8,power_law_1.01,0.050835198163986205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,8,power_law_1.01,0.060684800148010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,8,power_law_1.01,0.09948800206184387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,8,power_law_1.01,0.1121791958808899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,8,power_law_1.01,0.10226559638977051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,8,power_law_1.01,0.12081279754638671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,8,power_law_1.01,0.13655680418014526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,8,power_law_1.01,0.18176640272140504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,8,power_law_1.01,0.23017599582672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,8,power_law_1.01,0.30823678970336915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,8,power_law_1.01,0.40390400886535643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,2,power_law_1.2,0.020351999998092653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,2,power_law_1.2,0.02104319930076599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,2,power_law_1.2,0.021030400693416596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,2,power_law_1.2,0.021747200191020964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,2,power_law_1.2,0.023238399624824525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,2,power_law_1.2,0.02353920042514801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,2,power_law_1.2,0.023628799617290495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,2,power_law_1.2,0.0247296005487442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,2,power_law_1.2,0.02648319900035858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,2,power_law_1.2,0.025196799635887147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,2,power_law_1.2,0.02766079902648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,2,power_law_1.2,0.050995200872421265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,2,power_law_1.2,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,2,power_law_1.2,0.052671998739242554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,2,power_law_1.2,0.04318079948425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,2,power_law_1.2,0.03860479891300202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,8,power_law_1.01,0.057599997520446776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,8,power_law_1.01,0.07355520129203796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,8,power_law_1.01,0.07997440099716187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,2,power_law_1.2,0.042259201407432556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,8,power_law_1.01,0.08071680068969726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,8,power_law_1.01,0.08819199800491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,2,power_law_1.2,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,8,power_law_1.01,0.08825600147247314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,8,power_law_1.01,0.09057919979095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,8,power_law_1.01,0.0931007981300354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.2,0.06028159856796265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,8,power_law_1.01,0.0918079972267151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,8,power_law_1.01,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,8,power_law_1.01,0.09464960098266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,8,power_law_1.01,0.09642879962921143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,8,power_law_1.01,0.10234240293502808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,8,power_law_1.01,0.10163840055465698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,8,power_law_1.01,0.11170560121536255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,8,power_law_1.01,0.11156480312347412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,8,power_law_1.01,0.12554880380630493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.2,0.081740802526474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,8,power_law_1.01,0.13923200368881225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.01,0.15835520029067993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.01,0.18264319896697997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.01,0.18977279663085939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.01,0.2528831958770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.01,0.29667201042175295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.01,0.40248961448669435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.01,0.5096447944641114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.01,0.8024703979492187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.01,0.9760512351989746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.2,0.1018496036529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.2,0.13607679605484008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.2,0.1654207944869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.2,0.23632640838623048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.2,0.30074880123138426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.2,0.43736958503723145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,64,power_law_1.2,0.055238401889801024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,64,power_law_1.2,0.052857601642608644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,64,power_law_1.2,0.051635199785232545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,64,power_law_1.2,0.05745919942855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,64,power_law_1.2,0.05708799958229065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,64,power_law_1.2,0.05779839754104614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,64,power_law_1.2,0.05921919941902161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,64,power_law_1.2,0.06117119789123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,64,power_law_1.2,0.0625216007232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,64,power_law_1.2,0.06353920102119445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,64,power_law_1.2,0.06424959897994995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,64,power_law_1.2,0.0705407977104187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,64,power_law_1.2,0.07348480224609374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,64,power_law_1.2,0.06859520077705383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,64,power_law_1.2,0.08134400248527526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,64,power_law_1.2,0.08537600040435792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,64,power_law_1.2,0.08567039966583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,64,power_law_1.2,0.10341119766235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,64,power_law_1.2,0.11276799440383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,64,power_law_1.2,0.1323583960533142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,64,power_law_1.2,0.15538560152053832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,64,power_law_1.2,0.21515519618988038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,64,power_law_1.2,0.2663039922714233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,64,power_law_1.2,0.3810751914978027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,64,power_law_1.2,0.5246016025543213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,64,power_law_1.2,0.7380928039550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,64,power_law_1.2,1.1003007888793945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.2,0.5722432136535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,32,balanced,0.0476800004641215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,32,balanced,0.04451199869314829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,32,balanced,0.04737600187460581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,32,balanced,0.04663466910521189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,32,balanced,0.050144001841545105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,32,balanced,0.06486933430035909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,32,balanced,0.06614399949709575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,32,balanced,0.06649066507816315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,32,balanced,0.06835733354091644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,32,balanced,0.06778133412202199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,32,balanced,0.06818666557470958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,32,balanced,0.07112533350785573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,32,balanced,0.07035733262697856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,32,balanced,0.07426133255163829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,32,balanced,0.07905066510041554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,32,balanced,0.0801386684179306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,32,balanced,0.0839573343594869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,32,balanced,0.09687999884287517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,32,balanced,0.09689600268999736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,32,balanced,0.11795733372370402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,32,balanced,0.12691199779510498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,32,balanced,0.16267200311024985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,32,balanced,0.19135467211405435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,32,balanced,0.261626660823822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,32,balanced,0.3043786684672038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,32,balanced,0.43955198923746747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,32,balanced,0.5369120041529337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,8,power_law_1.2,0.06245120167732239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,8,power_law_1.2,0.07697280049324036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,8,power_law_1.2,0.07839360237121581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,8,power_law_1.2,0.09181439876556396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,8,power_law_1.2,0.10191999673843384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,8,power_law_1.2,0.10641280412673951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,8,power_law_1.2,0.10577919483184814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,8,power_law_1.2,0.10353280305862426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,8,power_law_1.2,0.1046720027923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,8,power_law_1.2,0.10459519624710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,8,power_law_1.2,0.10510720014572143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,8,power_law_1.2,0.1086527943611145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,8,power_law_1.2,0.11230720281600952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,8,power_law_1.2,0.11730560064315795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,8,power_law_1.2,0.12756479978561402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,8,power_law_1.2,0.13020800352096557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,8,power_law_1.2,0.13752319812774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,8,power_law_1.2,0.15374720096588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,8,power_law_1.2,0.17492480278015138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,8,power_law_1.2,0.20740480422973634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,8,power_law_1.2,0.23313279151916505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,8,power_law_1.2,0.28231039047241213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,8,power_law_1.2,0.34025599956512453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,8,power_law_1.2,0.4609344005584717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,8,power_law_1.2,0.6001344203948975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,32,power_law_1.2,0.021779200434684752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,32,power_law_1.2,0.02129279971122742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,8,power_law_1.2,0.8531328201293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,32,power_law_1.2,0.021491199731826782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,8,power_law_1.2,1.0751615524291993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,32,power_law_1.2,0.022310400009155275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,32,power_law_1.2,0.021279999613761903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,32,power_law_1.2,0.020947200059890748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,32,power_law_1.2,0.021798400580883025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,32,power_law_1.2,0.02377600073814392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,32,power_law_1.2,0.024345600605010988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,32,power_law_1.2,0.0247296005487442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,32,power_law_1.2,0.029676800966262816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,32,power_law_1.2,0.02961280047893524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,32,power_law_1.2,0.03047040104866028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,32,power_law_1.2,0.039776000380516055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,32,power_law_1.2,0.040627199411392215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,32,power_law_1.2,0.04100480079650879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,32,power_law_1.2,0.04470399916172028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,32,power_law_1.2,0.05504000186920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,32,power_law_1.2,0.0605184018611908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,32,power_law_1.2,0.07470080256462097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,32,power_law_1.2,0.056831997632980344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,32,power_law_1.2,0.07098879814147949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,32,power_law_1.2,0.08655999898910523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,32,power_law_1.2,0.11734399795532227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,32,power_law_1.2,0.136953604221344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,32,power_law_1.2,0.20022399425506593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,32,power_law_1.2,0.2669055938720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,1,power_law_1.01,0.05279359817504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,1,power_law_1.01,0.06443520188331604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,1,power_law_1.01,0.08200960159301758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,1,power_law_1.01,0.10421760082244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,1,power_law_1.01,0.12910079956054688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,1,power_law_1.01,0.17745920419692993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,1,power_law_1.01,0.19073920249938964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,1,power_law_1.01,0.19675519466400146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,1,power_law_1.01,0.1979200005531311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,1,power_law_1.01,0.21008000373840333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,1,power_law_1.01,0.21347200870513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,1,power_law_1.01,0.2243583917617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,1,power_law_1.01,0.23443200588226318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,1,power_law_1.01,0.2375040054321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,1,power_law_1.01,0.25111680030822753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,1,power_law_1.01,0.262719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,1,power_law_1.01,0.28481919765472413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,1,power_law_1.01,0.3282560110092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,1,power_law_1.01,0.37156479358673095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,1,power_law_1.01,0.46996479034423827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,1,power_law_1.01,0.5491648197174073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,1,power_law_1.01,0.733523178100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,1,power_law_1.01,0.9603391647338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,1,power_law_1.01,1.2974271774291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,1,power_law_1.01,1.6797760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,1,power_law_1.01,2.3120256423950196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,1,power_law_1.01,2.923347282409668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,128,power_law_1.01,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,128,power_law_1.01,0.06577280163764954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,128,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,128,power_law_1.01,0.046342399716377256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,128,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,128,power_law_1.01,0.04863359928131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,128,power_law_1.01,0.0496832013130188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,128,power_law_1.01,0.05011199712753296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,128,power_law_1.01,0.05073919892311096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,128,power_law_1.01,0.050860798358917235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,128,power_law_1.01,0.05507199764251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,128,power_law_1.01,0.05756800174713135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,128,power_law_1.01,0.06045439839363098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,128,power_law_1.01,0.06888960003852844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,128,power_law_1.01,0.07630079984664917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,128,power_law_1.01,0.0789247989654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,128,power_law_1.01,0.08203520178794861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,128,power_law_1.01,0.09761919975280761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,128,power_law_1.01,0.10022399425506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,128,power_law_1.01,0.13169920444488525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,128,power_law_1.01,0.1429311990737915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,128,power_law_1.01,0.19397120475769042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,128,power_law_1.01,0.2272576093673706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,128,power_law_1.01,0.30614399909973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,128,power_law_1.01,0.38714239597320554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,128,power_law_1.01,0.5390399932861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,128,power_law_1.01,0.7117887973785401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,4,balanced,0.019440000255902607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,4,balanced,0.019194666296243668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,4,balanced,0.021365332106749218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,4,balanced,0.02124800036350886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,4,balanced,0.025333332518736523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,4,balanced,0.02603200078010559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,4,balanced,0.02957333376010259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,4,balanced,0.030805334448814392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,4,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,4,balanced,0.03279466678698858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,4,balanced,0.033728001018365227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,4,balanced,0.043920000394185386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,4,balanced,0.04414933423201243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,4,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,4,balanced,0.06471466521422069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,4,balanced,0.0647573322057724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,4,balanced,0.07054399947325389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,4,balanced,0.07313066720962524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,4,balanced,0.07949333389600118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,4,balanced,0.09058133761088054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,4,balanced,0.09987200299898784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,4,balanced,0.11822932958602905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,4,balanced,0.12744533022244772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,4,balanced,0.19328532616297403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,4,balanced,0.22233066956202188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,4,balanced,0.31173866987228394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,4,balanced,0.4068426688512166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,1,power_law_1.01,0.12616319656372071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,1,power_law_1.01,0.1655295968055725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,1,power_law_1.01,0.22378239631652833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,1,power_law_1.01,0.3667968034744263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,1,power_law_1.01,0.5045375823974609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,1,power_law_1.01,0.5928832054138183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,1,power_law_1.01,0.7859392166137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,1,power_law_1.01,0.8077247619628907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,1,power_law_1.01,0.8326720237731934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,1,power_law_1.01,0.850432014465332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,1,power_law_1.01,0.8673407554626464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,1,power_law_1.01,0.8860287666320801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,1,power_law_1.01,0.8811455726623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,2,power_law_1.01,0.03386240005493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,2,power_law_1.01,0.03307519853115082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,1,power_law_1.01,0.9088512420654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,2,power_law_1.01,0.033267199993133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,2,power_law_1.01,0.033548799157142636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,1,power_law_1.01,0.9291328430175781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,1,power_law_1.01,0.9361536026000976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,1,power_law_1.01,0.952787208557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,1,power_law_1.01,1.0392576217651368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,2,power_law_1.01,0.03551360070705414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.01,1.1088895797729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,2,power_law_1.01,0.04083200097084046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,2,power_law_1.01,0.04416640102863312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.01,1.244870376586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,2,power_law_1.01,0.04222719967365265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,2,power_law_1.01,0.044249600172042845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,2,power_law_1.01,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,2,power_law_1.01,0.045484799146652224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,2,power_law_1.01,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,2,power_law_1.01,0.05331839919090271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,2,power_law_1.01,0.05482239723205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.01,1.3893440246582032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.01,1.651411247253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.01,1.9649023056030273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,2,power_law_1.01,0.06474239826202392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,2,power_law_1.01,0.0689087986946106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.01,2.534111976623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.01,3.102195167541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,2,power_law_1.01,0.07145599722862243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,2,power_law_1.01,0.08179200291633607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.01,4.255283355712891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,2,power_law_1.01,0.09185280203819275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.01,5.358662414550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,2,power_law_1.01,0.11246720552444459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,2,power_law_1.01,0.1432960033416748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,2,power_law_1.01,0.18848639726638794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,32,power_law_1.2,0.11642240285873413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,2,power_law_1.01,0.22055039405822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,32,power_law_1.2,0.11342079639434814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,32,power_law_1.2,0.11020799875259399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,32,power_law_1.2,0.11526399850845337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,32,power_law_1.2,0.10711040496826171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,32,power_law_1.2,0.11301759481430054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,32,power_law_1.2,0.1147968053817749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,32,power_law_1.2,0.11146880388259887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,32,power_law_1.2,0.11637120246887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,32,power_law_1.2,0.10965119600296021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,32,power_law_1.2,0.1143231987953186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,32,power_law_1.2,0.11825920343399048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,32,power_law_1.2,0.1189695954322815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,32,power_law_1.2,0.12021119594573974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,32,power_law_1.2,0.11772799491882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,2,power_law_1.01,0.28287999629974364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,32,power_law_1.2,0.12792960405349732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,32,power_law_1.2,0.13587839603424073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,32,power_law_1.2,0.15224319696426392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,32,power_law_1.2,0.16472959518432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,32,power_law_1.2,0.20407040119171144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,32,power_law_1.2,0.23413760662078859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,32,power_law_1.2,0.29344000816345217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,32,power_law_1.2,0.37569921016693114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,32,power_law_1.2,0.515667200088501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,32,power_law_1.2,0.6442880153656005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,32,power_law_1.2,0.9790335655212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,32,power_law_1.2,1.2185215950012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,2,power_law_1.01,0.37804160118103025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,2,power_law_1.01,0.558841609954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,4,power_law_1.2,0.07065600156784058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,4,power_law_1.2,0.08737919926643371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,4,power_law_1.2,0.09144960045814514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,4,power_law_1.2,0.10097279548645019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,4,power_law_1.2,0.11504640579223632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,4,power_law_1.2,0.11838719844818116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,4,power_law_1.2,0.13292160034179687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,4,power_law_1.2,0.14338560104370118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,4,power_law_1.2,0.1372480034828186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,4,power_law_1.2,0.13690880537033082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,4,power_law_1.2,0.13941760063171388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,4,power_law_1.2,0.14922239780426025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,4,power_law_1.2,0.143014395236969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,4,power_law_1.2,0.1527616024017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,4,power_law_1.2,0.16139520406723024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,4,power_law_1.2,0.16201599836349487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,4,power_law_1.2,0.16669440269470215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,4,power_law_1.2,0.18263039588928223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,4,power_law_1.2,0.19164799451828002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,4,power_law_1.2,0.23081600666046143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,4,power_law_1.2,0.262444806098938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,4,power_law_1.2,0.33542399406433104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,4,power_law_1.2,0.39660799503326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,4,power_law_1.2,0.5386943817138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,4,power_law_1.2,0.7219456195831299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,4,power_law_1.2,1.1039679527282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,4,power_law_1.2,1.3429887771606446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,2,power_law_1.01,0.6504000186920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,4,power_law_1.01,0.04893440008163452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,4,power_law_1.01,0.051648002862930295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,4,power_law_1.01,0.05915520191192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,4,power_law_1.01,0.07573760151863099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,4,power_law_1.01,0.08361600041389465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,4,power_law_1.01,0.09195520281791687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,4,power_law_1.01,0.10396800041198731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,4,power_law_1.01,0.10132479667663574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,4,power_law_1.01,0.1060096025466919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,4,power_law_1.01,0.10938880443572999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,4,power_law_1.01,0.1066815972328186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,4,power_law_1.01,0.11403520107269287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,4,power_law_1.01,0.1190335988998413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,4,power_law_1.01,0.12407679557800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,4,power_law_1.01,0.13226879835128785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,4,power_law_1.01,0.13995519876480103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,4,power_law_1.01,0.14759680032730102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,4,power_law_1.01,0.17115520238876342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,4,power_law_1.01,0.18258559703826904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,4,power_law_1.01,0.20877439975738527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,4,power_law_1.01,0.25797760486602783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,4,power_law_1.01,0.3126591920852661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,4,power_law_1.01,0.3909375905990601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,4,power_law_1.01,0.5154111862182618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,4,power_law_1.01,0.6371903896331788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,4,power_law_1.01,0.8872384071350098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,4,power_law_1.01,1.197209644317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,8,balanced,0.01899733394384384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,8,balanced,0.018944000204404194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,8,balanced,0.018853332847356796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,8,balanced,0.019189332922299702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,8,balanced,0.02096533278624217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,8,balanced,0.024101334313551586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,8,balanced,0.02714666724205017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,8,balanced,0.028522667785485584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,8,balanced,0.02957333376010259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,8,balanced,0.02938666691382726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,8,balanced,0.0313226655125618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,8,balanced,0.03980266551176707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,8,balanced,0.039887999494870506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,8,balanced,0.039850667119026184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,8,balanced,0.05499733487764994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,8,balanced,0.05470933516820272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,8,balanced,0.058261334896087646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,8,balanced,0.06393066545327504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,8,balanced,0.06504533191521962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,8,balanced,0.08134399851163228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,8,balanced,0.08609599868456523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,8,balanced,0.09900266925493877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,8,balanced,0.11482133467992146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,8,balanced,0.1400053302447001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,8,balanced,0.16572266817092896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,8,balanced,0.2487893303235372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,8,balanced,0.2970026731491089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,32,power_law_1.01,0.016326400637626647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,32,power_law_1.01,0.016364799439907075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,32,power_law_1.01,0.017580799758434296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,32,power_law_1.01,0.016582399606704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,32,power_law_1.01,0.016908800601959227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,32,power_law_1.01,0.017375999689102174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,32,power_law_1.01,0.016953599452972413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,32,power_law_1.01,0.020211200416088104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,32,power_law_1.01,0.020448000729084016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,32,power_law_1.01,0.021004800498485566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,32,power_law_1.01,0.027622398734092713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,32,power_law_1.01,0.027955201268196107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,32,power_law_1.01,0.028966400027275085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,32,power_law_1.01,0.028326401114463808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,32,power_law_1.01,0.028460800647735596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,32,power_law_1.01,0.028889599442481994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,32,power_law_1.01,0.029772800207138062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,32,power_law_1.01,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,32,power_law_1.01,0.03368319869041443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,32,power_law_1.01,0.03871360123157501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,32,power_law_1.01,0.04157440066337585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,32,power_law_1.01,0.04700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,32,power_law_1.01,0.057158398628234866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,32,power_law_1.01,0.07522559762001038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,32,power_law_1.01,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,32,power_law_1.01,0.12588800191879274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,32,power_law_1.01,0.15923839807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,1,power_law_1.01,0.021913599967956544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,1,power_law_1.01,0.022111999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,1,power_law_1.01,0.021433599293231964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,1,power_law_1.01,0.021356800198554994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,1,power_law_1.01,0.02237440049648285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,1,power_law_1.01,0.02479359954595566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,1,power_law_1.01,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,1,power_law_1.01,0.026790401339530943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,1,power_law_1.01,0.03390080034732819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,1,power_law_1.01,0.03159680068492889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,1,power_law_1.01,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,1,power_law_1.01,0.04014720022678375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,1,power_law_1.01,0.04119040071964264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,1,power_law_1.01,0.041459199786186215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,1,power_law_1.01,0.06170240044593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,1,power_law_1.01,0.052857601642608644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,1,power_law_1.01,0.0592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,1,power_law_1.01,0.07318400144577027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.01,0.09076480269432068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.01,0.12492799758911133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.01,0.15000959634780883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.01,0.21096320152282716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.01,0.26566400527954104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.01,0.3825599908828735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.01,0.4972032070159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.01,0.7294847965240479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.01,0.9589887619018554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,64,power_law_1.2,0.6136640071868896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,64,power_law_1.2,0.08613119721412658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,64,power_law_1.2,0.08483200073242188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,64,power_law_1.2,0.08490880131721497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,64,power_law_1.2,0.08491520285606384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,64,power_law_1.2,0.075135999917984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,64,power_law_1.2,0.0824832022190094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,64,power_law_1.2,0.08771839737892151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,64,power_law_1.2,0.08804479837417603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,64,power_law_1.2,0.08817279934883118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,64,power_law_1.2,0.09023360013961793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,64,power_law_1.2,0.08730880022048951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,64,power_law_1.2,0.09313279986381531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,64,power_law_1.2,0.09567360281944275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,64,power_law_1.2,0.10546560287475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,64,power_law_1.2,0.10956799983978271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,64,power_law_1.2,0.1168511986732483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,64,power_law_1.2,0.13140480518341063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,64,power_law_1.2,0.15212160348892212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,64,power_law_1.2,0.1969472050666809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,64,power_law_1.2,0.2178431987762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,64,power_law_1.2,0.2913088083267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,64,power_law_1.2,0.35151360034942625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,64,power_law_1.2,0.46785922050476075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,64,power_law_1.2,0.5724671840667724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,64,power_law_1.2,0.8543231964111329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,64,power_law_1.2,1.0837183952331544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,8,power_law_1.2,0.05464320182800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,8,power_law_1.2,0.05426560044288635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,8,power_law_1.2,0.061375999450683595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,8,power_law_1.2,0.062009602785110474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,8,power_law_1.2,0.05982080101966858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,8,power_law_1.2,0.07493119835853576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,8,power_law_1.2,0.07524480223655701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,8,power_law_1.2,0.07827839851379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,8,power_law_1.2,0.07687680125236511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,8,power_law_1.2,0.07712000012397766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,8,power_law_1.2,0.08195199966430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,8,power_law_1.2,0.08346880078315735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,8,power_law_1.2,0.08802559971809387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,8,power_law_1.2,0.0864512026309967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,8,power_law_1.2,0.09592959880828858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,8,power_law_1.2,0.10206719636917114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,8,power_law_1.2,0.10963200330734253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,8,power_law_1.2,0.13771519660949708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,8,power_law_1.2,0.1388416051864624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,8,power_law_1.2,0.16913919448852538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,8,power_law_1.2,0.1909183979034424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,8,power_law_1.2,0.24829440116882323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,8,power_law_1.2,0.3224128007888794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,8,power_law_1.2,0.4362368106842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,8,power_law_1.2,0.05980799794197082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,8,power_law_1.2,0.07724159955978394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,8,power_law_1.2,0.07428479790687562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,8,power_law_1.2,0.08883200287818908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,8,power_law_1.2,0.09238399863243103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,8,power_law_1.2,0.5330560207366943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,8,power_law_1.2,0.10017280578613282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,8,power_law_1.2,0.09597439765930176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,8,power_law_1.2,0.09566720128059387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,8,power_law_1.2,0.0975488007068634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,8,power_law_1.2,0.09650560021400452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,8,power_law_1.2,0.09932799935340882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,8,power_law_1.2,0.09935359954833985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,8,power_law_1.2,0.10197759866714477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,8,power_law_1.2,0.10477440357208252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,8,power_law_1.2,0.11043200492858887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,8,power_law_1.2,0.10823040008544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,8,power_law_1.2,0.1121664047241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,8,power_law_1.2,0.12534400224685668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,8,power_law_1.2,0.13957120180130006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,8,power_law_1.2,0.1600640058517456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,8,power_law_1.2,0.18817919492721558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,8,power_law_1.2,0.23580799102783204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,8,power_law_1.2,0.2924544095993042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,8,power_law_1.2,0.40631680488586425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,8,power_law_1.2,0.5727615833282471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,8,power_law_1.2,0.7845376014709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,8,power_law_1.2,1.086086368560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,8,power_law_1.2,0.7592703819274902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,4,power_law_1.01,0.05759360194206238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,4,power_law_1.01,0.06206079721450806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,4,power_law_1.01,0.06732159852981567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,4,power_law_1.01,0.07772799730300903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,4,power_law_1.01,0.08387200236320495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,4,power_law_1.01,0.09340159893035889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,4,power_law_1.01,0.10008959770202637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,4,power_law_1.01,0.10024960041046142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,4,power_law_1.01,0.09957119822502136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,4,power_law_1.01,0.10068479776382447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,4,power_law_1.01,0.10250240564346313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,4,power_law_1.01,0.10913920402526855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,4,power_law_1.01,0.10661120414733886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,4,power_law_1.01,0.11610239744186401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,4,power_law_1.01,0.11960320472717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,4,power_law_1.01,0.12452479600906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,4,power_law_1.01,0.1319424033164978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,4,power_law_1.01,0.14920320510864257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,4,power_law_1.01,0.1571903944015503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,4,power_law_1.01,0.1895359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,4,power_law_1.01,0.2052095890045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,4,power_law_1.01,0.24981119632720947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,4,power_law_1.01,0.3121984004974365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,4,power_law_1.01,0.416703987121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,4,power_law_1.01,0.538431978225708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,4,power_law_1.01,0.7742464065551757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,4,power_law_1.01,0.9851008415222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,1,balanced,0.09850133458773296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,1,balanced,0.1479039986928304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,1,balanced,0.24447999397913614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,1,balanced,0.4427573283513387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,1,balanced,0.839786688486735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,1,balanced,1.6210880279541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,1,balanced,1.6252427101135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,1,balanced,1.6316213607788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,1,balanced,1.633786678314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,1,balanced,1.639893372853597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,1,balanced,1.6449012756347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,1,balanced,1.651968002319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,1,balanced,1.6603199640909831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,1,balanced,1.6723519961039226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,1,balanced,1.689066727956136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,1,balanced,1.6954879760742188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,1,balanced,1.7163732846577961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,1,balanced,1.7637440363566081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,1,balanced,1.8272746404012044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,1,balanced,1.9292693138122559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,8,power_law_1.2,0.9918527603149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,1,balanced,1.9941439628601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,1,balanced,2.172442595163981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,1,balanced,2.3599093755086265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,1,balanced,2.759573300679525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,1,balanced,3.046581268310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,1,balanced,4.758858680725098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,1,balanced,5.4240372975667315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,64,power_law_1.01,0.06251519918441772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,64,power_law_1.01,0.06385279893875122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,64,power_law_1.01,0.055340802669525145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,64,power_law_1.01,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,64,power_law_1.01,0.05775359869003296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,64,power_law_1.01,0.059494400024414064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,64,power_law_1.01,0.05858560204505921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,64,power_law_1.01,0.05983359813690185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,64,power_law_1.01,0.058719998598098753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,64,power_law_1.01,0.060172802209854125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,64,power_law_1.01,0.0615231990814209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,64,power_law_1.01,0.06599680185317994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,64,power_law_1.01,0.06670719981193543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,64,power_law_1.01,0.06875519752502442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,64,power_law_1.01,0.07463039755821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,64,power_law_1.01,0.08056960105895997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,64,power_law_1.01,0.08587520122528076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,64,power_law_1.01,0.1007423996925354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,64,power_law_1.01,0.10641280412673951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,64,power_law_1.01,0.1339840054512024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,64,power_law_1.01,0.14755200147628783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,64,power_law_1.01,0.19365760087966918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,64,power_law_1.01,0.20354559421539306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,64,power_law_1.01,0.26058239936828614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,64,power_law_1.01,0.3091840028762817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,64,power_law_1.01,0.431935977935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,64,power_law_1.01,0.5126207828521728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,64,power_law_1.2,0.02019840031862259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,64,power_law_1.2,0.022355200350284578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,64,power_law_1.2,0.022233599424362184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,64,power_law_1.2,0.018508799374103546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,64,power_law_1.2,0.01943040043115616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,64,power_law_1.2,0.022047999501228332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,64,power_law_1.2,0.022592000663280487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,64,power_law_1.2,0.028064000606536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,64,power_law_1.2,0.028748801350593566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,64,power_law_1.2,0.028460800647735596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,64,power_law_1.2,0.03991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,64,power_law_1.2,0.039059200882911684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,64,power_law_1.2,0.039929598569869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,64,power_law_1.2,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,64,power_law_1.2,0.04323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,64,power_law_1.2,0.04692479968070984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,64,power_law_1.2,0.05219200253486633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,64,power_law_1.2,0.06034560203552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,64,power_law_1.2,0.07603840231895446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,64,power_law_1.2,0.09567360281944275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,64,power_law_1.2,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,64,power_law_1.2,0.0840831995010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,64,power_law_1.2,0.10323840379714966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,64,power_law_1.2,0.13523839712142943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,64,power_law_1.2,0.17541760206222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,64,power_law_1.2,0.2721280097961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,64,power_law_1.2,0.35622398853302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,32,power_law_1.2,0.06001920104026794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,4,power_law_1.01,0.05482239723205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,4,power_law_1.01,0.06605439782142639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,32,power_law_1.2,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,4,power_law_1.01,0.0733568012714386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,32,power_law_1.2,0.05475199818611145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,4,power_law_1.01,0.08350080251693726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,32,power_law_1.2,0.05749760270118713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,4,power_law_1.01,0.09301120042800903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,4,power_law_1.01,0.10990079641342163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,32,power_law_1.2,0.05982080101966858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,4,power_law_1.01,0.1168895959854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,32,power_law_1.2,0.06241919994354248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,4,power_law_1.01,0.11385600566864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,4,power_law_1.01,0.1156607985496521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,32,power_law_1.2,0.06314240097999572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,4,power_law_1.01,0.12336640357971192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,4,power_law_1.01,0.12112640142440796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,4,power_law_1.01,0.1252671957015991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,4,power_law_1.01,0.1295680046081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,4,power_law_1.01,0.1363263964653015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,4,power_law_1.01,0.1412287950515747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,4,power_law_1.01,0.14573440551757813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,4,power_law_1.01,0.15870720148086548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,32,power_law_1.2,0.0636672019958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,4,power_law_1.01,0.18284159898757935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,32,power_law_1.2,0.06568959951400757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,4,power_law_1.01,0.19745919704437256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,32,power_law_1.2,0.06505600214004517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,4,power_law_1.01,0.24430079460144044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,4,power_law_1.01,0.27180800437927244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,32,power_law_1.2,0.06513919830322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,4,power_law_1.01,0.34751999378204346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,4,power_law_1.01,0.42731518745422364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,4,power_law_1.01,0.6057727813720704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,4,power_law_1.01,0.7780416011810303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,32,power_law_1.2,0.0698751986026764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,4,power_law_1.01,1.0840512275695802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,32,power_law_1.2,0.07025279998779296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,32,power_law_1.2,0.07419520020484924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,4,power_law_1.01,1.4066495895385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,32,power_law_1.2,0.088128000497818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,16,4,balanced,0.06897066533565521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,32,power_law_1.2,0.08926079869270324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,16,4,balanced,0.07019733389218648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,16,4,balanced,0.07427733143170674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,16,4,balanced,0.08480532964070638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,16,4,balanced,0.1009386678536733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,16,4,balanced,0.12430933117866516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,16,4,balanced,0.12569066882133484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,16,4,balanced,0.12487999598185222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,16,4,balanced,0.12843733032544455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,16,4,balanced,0.12637866536776224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,16,4,balanced,0.12775466839472452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,32,power_law_1.2,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,16,4,balanced,0.13197867075602213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,16,4,balanced,0.1325920025507609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,32,power_law_1.2,0.10409599542617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,32,power_law_1.2,0.11688319444656373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,32,power_law_1.2,0.1377344012260437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,32,power_law_1.2,0.16083199977874757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,32,power_law_1.2,0.20977280139923096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,16,4,balanced,0.1360426644484202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,16,4,balanced,0.14285332957903543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,16,4,balanced,0.14320000012715658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,16,4,balanced,0.14780799547831217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,16,4,balanced,0.16273066401481628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,16,4,balanced,0.17190933227539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,16,4,balanced,0.19956799348195395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,16,4,balanced,0.2278346618016561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,16,4,balanced,0.2908479968706767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,16,4,balanced,0.3556319872538249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,32,power_law_1.2,0.25260798931121825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,16,4,balanced,0.5167893171310425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,32,power_law_1.2,0.3707648038864136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,32,power_law_1.2,0.5179391860961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,32,power_law_1.2,0.7449024200439454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,32,power_law_1.2,1.0333184242248534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,16,4,balanced,0.6376533508300781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,16,4,balanced,0.9186773300170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,16,4,balanced,1.1979093551635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,16,balanced,0.019130667050679524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,16,balanced,0.020026666422684986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,16,balanced,0.02093333254257838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,16,balanced,0.021087999145189922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,16,balanced,0.025114665428797405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,16,balanced,0.02932800104220708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,16,balanced,0.030917334059874218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,16,balanced,0.029370665550231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,16,balanced,0.031370667119820915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,16,balanced,0.03212266663710276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,16,balanced,0.03403199960788091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,16,balanced,0.037050666908423104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,16,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,16,balanced,0.044624000787734985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,16,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,16,balanced,0.04913066824277242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,16,balanced,0.06927466889222463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,16,balanced,0.0763733337322871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,16,balanced,0.09152000149091084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,16,balanced,0.11529599626859029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,16,balanced,0.09477333227793376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,16,balanced,0.11219200491905212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,16,balanced,0.13409066200256348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,16,balanced,0.17117865880330405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,16,balanced,0.2083466649055481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,16,balanced,0.3107146620750427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,16,balanced,0.3859626849492391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,64,power_law_1.2,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,64,power_law_1.2,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,64,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,64,power_law_1.2,0.034918400645256045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,64,power_law_1.2,0.03537920117378235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,64,power_law_1.2,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,64,power_law_1.2,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,64,power_law_1.2,0.03635199964046478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,64,power_law_1.2,0.03687039911746979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,64,power_law_1.2,0.037171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,64,power_law_1.2,0.037248000502586365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,64,power_law_1.2,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,64,power_law_1.2,0.038150399923324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,64,power_law_1.2,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,64,power_law_1.2,0.04195840060710907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,64,power_law_1.2,0.042233601212501526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,64,power_law_1.2,0.04338560104370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,64,power_law_1.2,0.04808320105075836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,64,power_law_1.2,0.05013120174407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,64,power_law_1.2,0.055871999263763426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,64,power_law_1.2,0.0606656014919281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,64,power_law_1.2,0.07255039811134338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,64,power_law_1.2,0.09344000220298768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,64,power_law_1.2,0.12222080230712891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,64,power_law_1.2,0.14693119525909423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,2,power_law_1.2,0.03797760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,2,power_law_1.2,0.037561601400375365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,2,power_law_1.2,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,2,power_law_1.2,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,2,power_law_1.2,0.04322560131549835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,2,power_law_1.2,0.0436928004026413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,2,power_law_1.2,0.04505600035190582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,2,power_law_1.2,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,2,power_law_1.2,0.04583039879798889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,2,power_law_1.2,0.048640000820159915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,2,power_law_1.2,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,2,power_law_1.2,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,2,power_law_1.2,0.06207360029220581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,2,power_law_1.2,0.06369280219078063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,64,power_law_1.2,0.21758079528808594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,2,power_law_1.2,0.07148799896240235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,2,power_law_1.2,0.07173759937286377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,2,power_law_1.2,0.07952640056610108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,2,power_law_1.2,0.09438719749450683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,2,power_law_1.2,0.11486079692840576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,2,power_law_1.2,0.15336960554122925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,2,power_law_1.2,0.17084159851074218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,2,power_law_1.2,0.24437758922576905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,64,power_law_1.2,0.29317760467529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,2,power_law_1.2,0.33530240058898925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,2,power_law_1.2,0.48954238891601565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,2,balanced,0.050213331977526345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,2,balanced,0.05190933247407278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,2,balanced,0.05500799914201101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,2,balanced,0.06262399752934773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,2,balanced,0.08030933141708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,2,balanced,0.12322133779525757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,2,balanced,0.12648000319798788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,2,balanced,0.12096533179283142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,2,balanced,0.12371200323104858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,2,balanced,0.12425600488980611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,2,balanced,0.1272586683432261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,2,balanced,0.13030399878819784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,2,balanced,0.1333440045515696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,2,balanced,0.13763200243314108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,2,balanced,0.14385599891344705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,2,balanced,0.14958399534225464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,2,balanced,0.161871999502182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,2,balanced,0.18525866667429605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,2,balanced,0.204474667708079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,2,balanced,0.24973867336908975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,2,balanced,0.2886773347854614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,2,balanced,0.42241064707438153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,2,balanced,0.4907999833424886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,2,balanced,0.7235519886016846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,2,balanced,0.9128266970316569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,2,balanced,1.3342827161153157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,2,balanced,1.7317172686258953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,2,power_law_1.2,0.6142271995544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,4,balanced,0.021301334102948506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,4,balanced,0.01907733331123988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,4,balanced,0.01933866615096728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,4,balanced,0.01937599976857503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,4,balanced,0.021157334248224895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,4,balanced,0.02117866774400075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,4,balanced,0.0234400009115537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,4,balanced,0.023413332800070446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,4,balanced,0.048437332113583885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,4,balanced,0.04663999875386556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,4,balanced,0.0476693312327067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,4,balanced,0.036746665835380554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,4,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,4,balanced,0.03648533423741659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,4,balanced,0.037647999823093414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,4,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,4,balanced,0.04125333329041799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,4,balanced,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,4,balanced,0.04685866832733154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,4,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,4,balanced,0.06343466540177663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,4,balanced,0.0879253347714742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,4,balanced,0.1016480028629303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,4,balanced,0.13808533549308777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,4,balanced,0.1749066710472107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,4,balanced,0.24700800577799478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,4,balanced,0.3080853422482808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,2,power_law_1.2,0.9697024345397949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,64,balanced,0.053861334919929504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,64,balanced,0.05717866619427999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,64,balanced,0.06005333364009857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,64,balanced,0.061936000982920326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,64,balanced,0.068122665087382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,64,balanced,0.07225066423416138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,64,balanced,0.07146666447321574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,64,balanced,0.07283733288447063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,64,balanced,0.07546666761239369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,64,balanced,0.07602133353551228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,64,balanced,0.07455466687679291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,64,balanced,0.07866133252779643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,64,balanced,0.07692266503969829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,64,balanced,0.0803413341442744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,64,balanced,0.08819733063379924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,64,balanced,0.08611200253168742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,64,balanced,0.09223467111587524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,64,balanced,0.10456533233324687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,64,balanced,0.11079466342926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,64,balanced,0.13886400063832602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,64,balanced,0.1585706671079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,64,balanced,0.20937599738438925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,64,balanced,0.19441600640614828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,64,balanced,0.25600532690684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,64,balanced,0.2776693304379781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,64,balanced,0.36983466148376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,64,balanced,0.44790931542714435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,2,power_law_1.2,1.2509632110595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,64,balanced,0.04795733094215393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,64,balanced,0.04164800047874451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,64,balanced,0.04178666571776072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,64,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,64,balanced,0.04385066529115041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,64,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,64,balanced,0.045850664377212524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,64,balanced,0.046869332591692604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,64,balanced,0.047557334105173744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,64,balanced,0.04760533571243286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,64,balanced,0.04774933556715647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,64,balanced,0.04981866478919983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,64,balanced,0.04799999793370565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,64,balanced,0.05190933247407278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,64,balanced,0.055770665407180786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,64,balanced,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,64,balanced,0.059205333391825356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,64,balanced,0.0639573335647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,64,balanced,0.06685866912206014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,64,balanced,0.0769706666469574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,64,balanced,0.08272533118724823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,64,balanced,0.1086133321126302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,64,balanced,0.13269866506258646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,64,balanced,0.1828533411026001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,64,balanced,0.22691200176874796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,64,balanced,0.31256532669067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,64,balanced,0.39160001277923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,2,power_law_1.01,0.04439040124416351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,2,power_law_1.01,0.048851200938224794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,2,power_law_1.01,0.05400320291519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,2,power_law_1.01,0.061388802528381345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,2,power_law_1.01,0.06659839749336242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,2,power_law_1.01,0.08587520122528076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,2,power_law_1.01,0.08684160113334656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,2,power_law_1.01,0.09635199904441834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,2,power_law_1.01,0.09283199906349182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,2,power_law_1.01,0.09399679899215699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,2,power_law_1.01,0.09608960151672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,2,power_law_1.01,0.09897599816322326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,2,power_law_1.01,0.10241919755935669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,2,power_law_1.01,0.10451840162277222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,2,power_law_1.01,0.10888960361480712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,2,power_law_1.01,0.11447039842605591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,2,power_law_1.01,0.12488960027694702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,2,power_law_1.01,0.14426239728927612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.01,0.16457600593566896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.01,0.19736319780349731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.01,0.23437440395355225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.01,0.32769920825958254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.01,0.40439682006835936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.01,0.5549503803253174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.01,0.7131264209747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.01,1.0173439979553223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.01,1.3655232429504394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,128,power_law_1.2,0.07900800108909607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,128,power_law_1.2,0.08158720135688782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,128,power_law_1.2,0.0770687997341156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,128,power_law_1.2,0.0812287986278534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,128,power_law_1.2,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,128,power_law_1.2,0.07588480114936828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,128,power_law_1.2,0.07808640003204345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,128,power_law_1.2,0.08048639893531799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,128,power_law_1.2,0.0855296015739441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,128,power_law_1.2,0.08490880131721497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,128,power_law_1.2,0.08336640000343323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,128,power_law_1.2,0.08773120045661927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,128,power_law_1.2,0.08572800159454345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,128,power_law_1.2,0.08127999901771546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,128,power_law_1.2,0.09114239811897278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,128,power_law_1.2,0.09208959937095643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,128,power_law_1.2,0.0998848021030426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,1,balanced,0.032730666299661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,128,power_law_1.2,0.11175040006637574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,1,balanced,0.03552533437808355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,1,balanced,0.05375466744105021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,1,balanced,0.05584533512592316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,1,balanced,0.0583840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,1,balanced,0.058277333776156105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,1,balanced,0.057946667075157166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,1,balanced,0.06033066908518473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,1,balanced,0.05991999804973602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,1,balanced,0.06099733213583628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,1,balanced,0.061792001128196716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,1,balanced,0.06338666876157124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,1,balanced,0.06637333333492279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,1,balanced,0.0702400008837382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,1,balanced,0.07869333525498708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,1,balanced,0.08246933420499165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,1,balanced,0.09044266740481059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,1,balanced,0.10596799850463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,1,balanced,0.1148373285929362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,128,power_law_1.2,0.11382399797439575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,1,balanced,0.15339199701944986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,1,balanced,0.1808799902598063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,128,power_law_1.2,0.15330560207366944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,1,balanced,0.24459733565648398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,1,balanced,0.2981226642926534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,1,balanced,0.42376001675923664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,1,balanced,0.536298672358195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,1,balanced,0.7820160388946533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,1,balanced,1.0113600095113118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,128,power_law_1.2,0.17623039484024047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,128,power_law_1.2,0.21976959705352783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,128,power_law_1.2,0.2736448049545288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,128,power_law_1.2,0.37564160823822024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,128,power_law_1.2,0.4618175983428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,128,power_law_1.2,0.7115327835083007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,128,power_law_1.2,0.8495295524597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,2,power_law_1.2,0.06913920044898987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,2,power_law_1.2,0.09193599820137024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,2,power_law_1.2,0.10195839405059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,2,power_law_1.2,0.12481919527053834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,2,power_law_1.2,0.15088000297546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,2,power_law_1.2,0.1740928053855896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,2,power_law_1.2,0.23438720703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,2,power_law_1.2,0.23001599311828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,2,power_law_1.2,0.24995839595794678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,2,power_law_1.2,0.257587194442749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,2,power_law_1.2,0.2604480028152466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,2,power_law_1.2,0.26594560146331786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,2,power_law_1.2,0.2712575912475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,2,power_law_1.2,0.28273279666900636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,2,power_law_1.2,0.2982016086578369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,2,power_law_1.2,0.3065727949142456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,2,power_law_1.2,0.33356800079345705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,2,power_law_1.2,0.3738816022872925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,2,power_law_1.2,0.41212158203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,2,power_law_1.2,0.4942592144012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,2,power_law_1.2,0.5111999988555909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,2,power_law_1.2,0.6481152057647706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,2,power_law_1.2,0.7683519840240478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,2,power_law_1.2,0.9933376312255859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,2,power_law_1.2,1.289958381652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,2,power_law_1.2,1.66376953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,2,power_law_1.2,2.29866886138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,1,power_law_1.01,0.019968000054359437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,1,power_law_1.01,0.024883200228214265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,1,power_law_1.01,0.038796800374984744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,1,power_law_1.01,0.05278080105781555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,1,power_law_1.01,0.07084159851074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,1,power_law_1.01,0.09084799885749817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,1,power_law_1.01,0.11495039463043213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,1,power_law_1.01,0.12220159769058228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,1,power_law_1.01,0.12579200267791749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,1,power_law_1.01,0.130348801612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,8,power_law_1.2,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,1,power_law_1.01,0.13289599418640136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,8,power_law_1.2,0.043724799156188966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,1,power_law_1.01,0.13898240327835082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,8,power_law_1.2,0.04341759979724884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,1,power_law_1.01,0.13927040100097657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,1,power_law_1.01,0.13819520473480223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,1,power_law_1.01,0.14876799583435057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,1,power_law_1.01,0.15221760272979737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,1,power_law_1.01,0.15326080322265626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,1,power_law_1.01,0.17214720249176024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,1,power_law_1.01,0.19438719749450684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,1,power_law_1.01,0.247161602973938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,1,power_law_1.01,0.2553024053573608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,1,power_law_1.01,0.33315200805664064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,1,power_law_1.01,0.3188159942626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,1,power_law_1.01,0.4383808135986328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,1,power_law_1.01,0.5436992168426513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,8,power_law_1.2,0.04656639993190766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,8,power_law_1.2,0.045049598813056944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,1,power_law_1.01,0.7197440147399903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,8,power_law_1.2,0.045535999536514285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,8,power_law_1.2,0.0453247994184494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,8,power_law_1.2,0.045510399341583255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,1,power_law_1.01,0.9246208190917968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,8,power_law_1.2,0.04505600035190582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,8,power_law_1.2,0.045203199982643126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,8,power_law_1.2,0.04490880072116852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,8,power_law_1.2,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,8,power_law_1.2,0.04682239890098572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,8,power_law_1.2,0.04733439981937408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,8,power_law_1.2,0.05310080051422119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,8,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,8,power_law_1.2,0.05504639744758606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,8,power_law_1.2,0.059084802865982056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,8,power_law_1.2,0.06999040246009827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,8,power_law_1.2,0.07975680232048035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,8,power_law_1.2,0.09438080191612244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,8,power_law_1.2,0.12076159715652465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,16,balanced,0.029338667790095013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,16,balanced,0.031258667508761086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,16,balanced,0.029045333464940388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,16,balanced,0.033674667278925575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,16,balanced,0.033674667278925575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,8,power_law_1.2,0.14288640022277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,16,balanced,0.03572800010442734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,16,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,16,balanced,0.0356480007370313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,16,balanced,0.03735466549793879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,16,balanced,0.0373279998699824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,16,balanced,0.037605332831541695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,16,balanced,0.0388373335202535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,16,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,16,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,16,balanced,0.042549331982930504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,16,balanced,0.04154133299986521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,16,balanced,0.04380266865094503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,16,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,16,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,16,balanced,0.05982933441797892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,16,balanced,0.06625066697597504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,16,balanced,0.07965866724650066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,16,balanced,0.08678932984670003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,16,balanced,0.10627200206120808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,16,balanced,0.13079999883969626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,16,balanced,0.1627893348534902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,16,balanced,0.20336532592773438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,8,power_law_1.2,0.20892798900604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,8,power_law_1.2,0.25539839267730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,8,power_law_1.2,0.3479871988296509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,8,power_law_1.2,0.47788162231445314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,16,balanced,0.039887999494870506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,16,balanced,0.031258667508761086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,16,balanced,0.031471999982992806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,16,balanced,0.0352960005402565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,16,balanced,0.03491200009981791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,16,balanced,0.03779733429352442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,16,balanced,0.03941866755485535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,16,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,16,balanced,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,16,balanced,0.03882666677236557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,16,balanced,0.03796799977620443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,16,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,16,balanced,0.03827733298142751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,16,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,16,balanced,0.043509334325790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,16,balanced,0.05050666630268097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,16,balanced,0.04610133171081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,16,balanced,0.056234667698542275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,16,balanced,0.054341331124305725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,16,balanced,0.06876266499360402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,16,balanced,0.07147733370463054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,16,balanced,0.08805867036183675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,16,balanced,0.09588266412417094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,16,balanced,0.1269599994023641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,16,balanced,0.1414186656475067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,16,balanced,0.20006932814915976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,16,balanced,0.22317334016164145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,64,power_law_1.2,0.059935998916625974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,64,power_law_1.2,0.05939840078353882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,64,power_law_1.2,0.05847679972648621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,64,power_law_1.2,0.05999360084533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,64,power_law_1.2,0.05622400045394897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,64,power_law_1.2,0.06053119897842407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,64,power_law_1.2,0.06037120223045349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,64,power_law_1.2,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,64,power_law_1.2,0.05925760269165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,64,power_law_1.2,0.06224640011787415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,64,power_law_1.2,0.061299198865890504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,64,power_law_1.2,0.06262400150299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,64,power_law_1.2,0.06327679753303528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,64,power_law_1.2,0.06369280219078063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,64,power_law_1.2,0.06872320175170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,64,power_law_1.2,0.06982399821281433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,64,power_law_1.2,0.07336320281028748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,64,power_law_1.2,0.08078719973564148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,64,power_law_1.2,0.08740479946136474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,64,power_law_1.2,0.10351999998092651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,64,power_law_1.2,0.12298879623413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,64,power_law_1.2,0.15372159481048583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,64,power_law_1.2,0.182150399684906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,64,power_law_1.2,0.24798080921173096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,64,power_law_1.2,0.32875521183013917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,64,power_law_1.2,0.4442304134368896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,64,power_law_1.2,0.5678656101226807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,1,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,1,balanced,0.060640002290407814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,1,balanced,0.08473599950472514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,1,balanced,0.08588799834251404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,1,balanced,0.0867146650950114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,1,balanced,0.08669333656628926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,1,balanced,0.08680533369382222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,1,balanced,0.09026666482289632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,1,balanced,0.08886933326721191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,1,balanced,0.09098666906356812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,1,balanced,0.09057066837946574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,1,balanced,0.09311999877293904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,1,balanced,0.09504000345865886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,1,balanced,0.09974400202433269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,1,balanced,0.11241599917411804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,1,balanced,0.11333333452542622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,1,balanced,0.11743999520937602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,1,balanced,0.1383840044339498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,1,balanced,0.1464959979057312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,1,balanced,0.18453866243362427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,1,balanced,0.21386667092641196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,1,balanced,0.28783466418584186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,1,balanced,0.352944016456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,1,balanced,0.5016266504923502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,1,balanced,0.645365317662557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,1,balanced,0.9255466461181641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,1,balanced,1.209216038386027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,8,power_law_1.01,0.031430399417877196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,8,power_law_1.01,0.030553600192070006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,8,power_law_1.01,0.029811200499534608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,8,power_law_1.01,0.029721599817276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,8,power_law_1.01,0.03059839904308319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,8,power_law_1.01,0.031641599535942075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,8,power_law_1.01,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,8,power_law_1.01,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,8,power_law_1.01,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,8,power_law_1.01,0.03644160032272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,8,power_law_1.2,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,8,power_law_1.01,0.035462400317192076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,8,power_law_1.2,0.035392001271247864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,8,power_law_1.01,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,8,power_law_1.01,0.03996799886226654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,8,power_law_1.01,0.040064001083374025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,8,power_law_1.01,0.04520959854125976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,8,power_law_1.01,0.04793600142002106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,8,power_law_1.01,0.05200639963150024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,8,power_law_1.2,0.03343999981880188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,8,power_law_1.2,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,8,power_law_1.2,0.03509120047092438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,8,power_law_1.01,0.0624064028263092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,8,power_law_1.2,0.038438400626182555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,8,power_law_1.2,0.03874559998512268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,8,power_law_1.2,0.03785600066184998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,8,power_law_1.01,0.06815999746322632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,8,power_law_1.2,0.04007680118083954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,8,power_law_1.2,0.041068801283836366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,8,power_law_1.2,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,8,power_law_1.2,0.04360319972038269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,8,power_law_1.2,0.045875200629234315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,8,power_law_1.01,0.09493119716644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,8,power_law_1.2,0.04832639992237091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,8,power_law_1.2,0.055257600545883176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,8,power_law_1.2,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,8,power_law_1.2,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,8,power_law_1.2,0.06660479903221131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,8,power_law_1.2,0.08062719702720642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,8,power_law_1.2,0.09882879853248597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,8,power_law_1.2,0.11991679668426514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,8,power_law_1.01,0.10906879901885987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,8,power_law_1.2,0.13687679767608643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,8,power_law_1.2,0.17500159740447999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,8,power_law_1.2,0.2223423957824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,8,power_law_1.01,0.14678399562835692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,1,power_law_1.01,0.03243519961833954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,1,power_law_1.01,0.05118079781532288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,1,power_law_1.01,0.07720959782600403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,1,power_law_1.01,0.11434240341186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,1,power_law_1.01,0.15189759731292723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,1,power_law_1.01,0.224019193649292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,1,power_law_1.01,0.24961280822753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,1,power_law_1.01,0.2574656009674072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,1,power_law_1.01,0.2567039966583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,1,power_law_1.01,0.28112640380859377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,1,power_law_1.01,0.29303679466247556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,1,power_law_1.01,0.3115839958190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,8,power_law_1.01,0.17344640493392943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,1,power_law_1.01,0.29178240299224856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,1,power_law_1.01,0.31058559417724607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,1,power_law_1.01,0.32997760772705076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,1,power_law_1.01,0.3016576051712036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,1,power_law_1.01,0.3185215950012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,1,power_law_1.01,0.34720640182495116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.01,0.37585279941558836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.01,0.8405568122863769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,8,power_law_1.2,0.30780160427093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.01,0.9248767852783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.01,0.5857408046722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.01,0.6705855846405029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.01,0.7925568103790284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.01,0.9714752197265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.01,1.3103615760803222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.01,1.6674240112304688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,32,power_law_1.01,0.08381440043449402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,32,power_law_1.01,0.08126720190048217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,32,power_law_1.01,0.07734400033950806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,32,power_law_1.01,0.0822655975818634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,32,power_law_1.01,0.07699840068817139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,32,power_law_1.01,0.0793727993965149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,32,power_law_1.01,0.08141440153121948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,32,power_law_1.01,0.08701440095901489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,8,power_law_1.01,0.24223999977111815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,32,power_law_1.01,0.08599680066108703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,32,power_law_1.01,0.08240640163421631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,32,power_law_1.01,0.08644480109214783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,32,power_law_1.01,0.08862720131874084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,32,power_law_1.01,0.08917120099067688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,32,power_law_1.01,0.08943359851837158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,32,power_law_1.01,0.09508479833602905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,32,power_law_1.01,0.09540479779243469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,32,power_law_1.01,0.1003648042678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,32,power_law_1.01,0.11232000589370728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,32,power_law_1.01,0.12618240118026733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,32,power_law_1.01,0.15007359981536866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,32,power_law_1.01,0.17210240364074708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,32,power_law_1.01,0.22174079418182374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,32,power_law_1.01,0.2637120008468628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,8,power_law_1.2,0.44994559288024905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,32,power_law_1.01,0.3568320035934448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,32,power_law_1.01,0.464031982421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,32,power_law_1.01,0.6658944129943848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,32,power_law_1.01,0.8432767868041993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,8,power_law_1.01,0.2960063934326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,8,power_law_1.2,0.5168960094451904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,8,power_law_1.01,0.40240001678466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,8,power_law_1.01,0.5515007972717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,8,power_law_1.2,0.06060799956321716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,8,power_law_1.2,0.06601600050926208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,8,power_law_1.2,0.07616639733314515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,8,power_law_1.2,0.07747200131416321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,8,power_law_1.2,0.08566399812698364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,8,power_law_1.2,0.09249280095100403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,8,power_law_1.2,0.09673600196838379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,8,power_law_1.2,0.10438400506973267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,8,power_law_1.2,0.09527680277824402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,8,power_law_1.2,0.10191999673843384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,8,power_law_1.2,0.0943615972995758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,8,power_law_1.2,0.09969279766082764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,8,power_law_1.2,0.10752639770507813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,8,power_law_1.2,0.10702719688415527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,8,power_law_1.2,0.11617920398712159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,8,power_law_1.2,0.11820160150527954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,8,power_law_1.2,0.12954879999160768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,8,power_law_1.2,0.1463871955871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,8,power_law_1.2,0.16598399877548217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,8,power_law_1.2,0.2040640115737915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,8,power_law_1.2,0.20775680541992186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,8,power_law_1.2,0.2766207933425903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,8,power_law_1.2,0.3225023984909058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,8,power_law_1.2,0.44407038688659667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,8,power_law_1.2,0.579033613204956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,8,power_law_1.2,0.8571776390075684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,8,power_law_1.2,1.145529556274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,2,balanced,0.09664533535639445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,2,balanced,0.1486026644706726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,2,balanced,0.24387200673421225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,2,balanced,0.44041065375010174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,2,balanced,0.8329599698384603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,2,balanced,1.6296745936075847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,2,balanced,1.634010632832845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,2,balanced,1.6270826657613118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,2,balanced,1.6282720565795898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,2,balanced,1.6363733609517415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,2,balanced,1.6322506268819172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,2,balanced,1.63699738184611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,2,balanced,1.6457279523213704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,2,balanced,1.661834716796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,2,balanced,1.6566559473673503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,2,balanced,1.6692266464233398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,2,balanced,1.6849279403686523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,2,balanced,1.712997277577718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,2,balanced,1.7653493881225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,2,balanced,1.81441068649292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,2,balanced,1.884442647298177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,2,balanced,1.9640960693359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,2,balanced,2.196341355641683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,2,balanced,2.3751680056254068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,2,balanced,2.539050738016764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,2,balanced,3.957104047139486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,2,balanced,4.105237325032552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,1,power_law_1.2,0.03744640052318573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,1,power_law_1.2,0.05145599842071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,1,power_law_1.2,0.06977919936180114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,1,power_law_1.2,0.09007999897003174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,1,power_law_1.2,0.11056640148162841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,1,power_law_1.2,0.1732480049133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,1,power_law_1.2,0.18684799671173097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,1,power_law_1.2,0.19078400135040283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,1,power_law_1.2,0.19876480102539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,1,power_law_1.2,0.1996415972709656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,1,power_law_1.2,0.21147520542144777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,1,power_law_1.2,0.21607038974761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,1,power_law_1.2,0.21552639007568358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,1,power_law_1.2,0.22536320686340333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,1,power_law_1.2,0.23311998844146728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,1,power_law_1.2,0.2399616003036499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,1,power_law_1.2,0.24897279739379882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,1,power_law_1.2,0.2736448049545288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,1,power_law_1.2,0.28393599987030027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,1,power_law_1.2,0.3288640022277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,1,power_law_1.2,0.36570239067077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,1,power_law_1.2,0.4402751922607422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,1,power_law_1.2,0.4857215881347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,1,power_law_1.2,0.6207168102264404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,1,power_law_1.2,0.7589824199676514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,1,power_law_1.2,1.036012840270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,1,power_law_1.2,1.279097557067871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,128,power_law_1.01,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,128,power_law_1.01,0.041017600893974306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,128,power_law_1.01,0.04067200124263763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,128,power_law_1.01,0.04297600090503693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,128,power_law_1.01,0.0444927990436554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,128,power_law_1.01,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,128,power_law_1.01,0.04702720046043396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,128,power_law_1.01,0.04803839921951294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,128,power_law_1.01,0.049542400240898135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,128,power_law_1.01,0.049983999133110045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,128,power_law_1.01,0.053497600555419925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,128,power_law_1.01,0.056415998935699464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,128,power_law_1.01,0.051609599590301515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,128,power_law_1.01,0.056505602598190305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,128,power_law_1.01,0.0616703987121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,128,power_law_1.01,0.06174719929695129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,128,power_law_1.01,0.06573439836502075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,128,power_law_1.01,0.07408639788627625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,128,power_law_1.01,0.07941120266914367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,128,power_law_1.01,0.09607040286064147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,128,power_law_1.01,0.11360640525817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,128,power_law_1.01,0.15396480560302733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,128,power_law_1.01,0.1900480031967163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,128,power_law_1.01,0.2606719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,128,power_law_1.01,0.3282304048538208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,128,power_law_1.01,0.4887231826782227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,128,power_law_1.01,0.6427711963653564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,1,power_law_1.2,0.05983359813690185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,1,power_law_1.2,0.0642624020576477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,1,power_law_1.2,0.07411199808120728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,1,power_law_1.2,0.09797120094299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,1,power_law_1.2,0.1278720021247864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,1,power_law_1.2,0.15792640447616577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,1,power_law_1.2,0.18021759986877442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,1,power_law_1.2,0.21903998851776124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,1,power_law_1.2,0.22986240386962892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,1,power_law_1.2,0.23720960617065429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,1,power_law_1.2,0.2482304096221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,1,power_law_1.2,0.2651583909988403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,1,power_law_1.2,0.2785536050796509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,1,power_law_1.2,0.2953727960586548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,1,power_law_1.2,0.3159104108810425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,1,power_law_1.2,0.3380095958709717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,1,power_law_1.2,0.34035840034484866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,1,power_law_1.2,0.39940481185913085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,1,power_law_1.2,0.44296960830688475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,1,power_law_1.2,0.5451776027679444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,1,power_law_1.2,0.6327680110931396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,1,power_law_1.2,0.8307007789611817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,1,power_law_1.2,1.0408448219299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,1,power_law_1.2,1.399948787689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,1,power_law_1.2,1.7604032516479493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,1,power_law_1.2,2.492665672302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,1,power_law_1.2,3.2156608581542967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,32,power_law_1.01,0.06806399822235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,32,power_law_1.01,0.06658560037612915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,32,power_law_1.01,0.06484479904174804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,32,power_law_1.01,0.06703360080718994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,32,power_law_1.01,0.0685696005821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,32,power_law_1.01,0.07064319849014282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,32,power_law_1.01,0.07228800058364868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,32,power_law_1.01,0.07390720248222352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,32,power_law_1.01,0.07516160011291503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,32,power_law_1.01,0.08084480166435241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,32,power_law_1.01,0.08050559759140015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,32,power_law_1.01,0.08336640000343323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,32,power_law_1.01,0.08165119886398316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,32,power_law_1.01,0.08420479893684388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,32,power_law_1.01,0.09396479725837707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,32,power_law_1.01,0.08997759819030762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,32,power_law_1.01,0.10123519897460938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,32,power_law_1.01,0.10796799659729003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,32,power_law_1.01,0.1177664041519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,32,power_law_1.01,0.144051194190979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,32,power_law_1.01,0.1712767958641052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,32,power_law_1.01,0.21918718814849852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,32,power_law_1.01,0.2800832033157349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,32,power_law_1.01,0.4090559959411621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,32,power_law_1.01,0.4656383991241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,32,power_law_1.01,0.8084735870361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,32,power_law_1.01,0.9817024230957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,8,power_law_1.01,0.036236798763275145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,8,power_law_1.01,0.04723199903964996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,8,power_law_1.01,0.042745599150657655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,8,power_law_1.01,0.04798719882965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,8,power_law_1.01,0.050432002544403075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,8,power_law_1.01,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,8,power_law_1.01,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,8,power_law_1.01,0.054476797580718994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,8,power_law_1.01,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,8,power_law_1.01,0.05663359761238098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,8,power_law_1.01,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,8,power_law_1.01,0.0586624026298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,8,power_law_1.01,0.05896959900856018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,8,power_law_1.01,0.059987199306488034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,8,power_law_1.01,0.06595839858055115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,8,power_law_1.01,0.06778240203857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,8,power_law_1.01,0.06782079935073852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,8,power_law_1.01,0.07841280102729797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,8,power_law_1.01,0.08142719864845276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,8,power_law_1.01,0.09096959829330445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,8,power_law_1.01,0.10293760299682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,8,power_law_1.01,0.1178879976272583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,8,power_law_1.01,0.14218239784240722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,8,power_law_1.01,0.18240640163421631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,8,power_law_1.01,0.2202752113342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,8,power_law_1.01,0.30607359409332274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,8,power_law_1.01,0.3717695951461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,1,balanced,0.051829333106676735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,1,balanced,0.06400000055631001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,1,balanced,0.07836266855398814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,1,balanced,0.11767466862996419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,1,balanced,0.1793866753578186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,1,balanced,0.30912532409032184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,1,balanced,0.3083893259366353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,1,balanced,0.31678932905197144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,1,balanced,0.31857067346572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,1,balanced,0.316591997941335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,1,balanced,0.3177280028661092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,1,balanced,0.3282826741536458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,1,balanced,0.32622400919596356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,1,balanced,0.3301546573638916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,1,balanced,0.33842666943868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,1,balanced,0.3418453137079875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,1,balanced,0.35174401601155597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,1,balanced,0.37485333283742267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,1,balanced,0.3970880111058553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,1,balanced,0.44281065464019775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,1,balanced,0.4810773531595866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,1,balanced,0.5932960112889608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,1,balanced,0.6865653196970621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,1,balanced,0.9108533064524332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,1,balanced,1.0964159965515137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,1,balanced,1.5449013710021973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,1,balanced,1.9485333760579426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,16,balanced,0.05401599903901418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,16,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,16,balanced,0.04842666784922282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,16,balanced,0.06821333368619283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,16,balanced,0.08880533774693807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,16,balanced,0.13475733002026877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,16,balanced,0.13766400019327799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,16,balanced,0.13571733236312866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,16,balanced,0.1359999974568685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,16,balanced,0.13705066839853922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,16,balanced,0.1383626659711202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,16,balanced,0.13739200433095297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,16,balanced,0.1397173305352529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,16,balanced,0.14201600352923074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,16,balanced,0.14668800433476767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,16,balanced,0.1458453337351481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,16,balanced,0.14969600240389505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,16,balanced,0.15692800283432007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,16,balanced,0.15973333517710367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,16,balanced,0.17485866943995157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,16,balanced,0.18899200359980264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,16,balanced,0.21310933430989584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,16,balanced,0.23465599616368613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,16,balanced,0.29679999748865765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,16,balanced,0.3297920028368632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,16,balanced,0.4594666560490926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,16,balanced,0.5523519913355509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,8,power_law_1.01,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,8,power_law_1.01,0.052697598934173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,8,power_law_1.01,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,8,power_law_1.01,0.057574397325515746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,8,power_law_1.01,0.06207360029220581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,8,power_law_1.01,0.06472319960594178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,8,power_law_1.01,0.06335359811782837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,8,power_law_1.01,0.06353279948234558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,8,power_law_1.01,0.06664959788322448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,8,power_law_1.01,0.0670527994632721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,8,power_law_1.01,0.0700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,8,power_law_1.01,0.0716159999370575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,8,power_law_1.01,0.07365760207176208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,8,power_law_1.01,0.08359040021896362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,8,power_law_1.01,0.09015039801597595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,8,power_law_1.01,0.09300479888916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,8,power_law_1.01,0.10005120038986207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,8,power_law_1.01,0.11991679668426514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,8,power_law_1.01,0.12869759798049926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,8,power_law_1.01,0.1584831953048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,8,power_law_1.01,0.19258240461349488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,8,power_law_1.01,0.24104959964752198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,8,power_law_1.01,0.297542405128479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,8,power_law_1.01,0.4295616149902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,8,power_law_1.01,0.5414271831512452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,8,power_law_1.01,0.7874303817749023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,8,power_law_1.01,0.9718079566955566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,2,balanced,0.06447466711203258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,2,balanced,0.08340266346931458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,2,balanced,0.1113759974638621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,2,balanced,0.1684373418490092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,2,balanced,0.2707306742668152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,2,balanced,0.48234132925669354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,2,balanced,0.4847253163655599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,2,balanced,0.48686933517456055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,2,balanced,0.48898132642110187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,2,balanced,0.4928586483001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,2,balanced,0.4941226641337077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,2,balanced,0.5006399949391683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,2,balanced,0.5047359863917033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,2,balanced,0.5087039868036906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,2,balanced,0.5175146659215292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,2,balanced,0.522650678952535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,2,balanced,0.530784010887146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,2,balanced,0.5550400018692017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,2,balanced,0.5734826723734537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,2,balanced,0.6207520167032877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,2,balanced,0.6797440052032471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,2,balanced,0.7776959737141927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,2,balanced,0.8670079708099365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,2,balanced,1.192309300104777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,2,balanced,1.3783040046691895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,2,balanced,1.9388853708902996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,2,balanced,2.46997865041097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,4,power_law_1.2,0.078438401222229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,4,power_law_1.2,0.08167679905891419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,4,power_law_1.2,0.09848960041999817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,4,power_law_1.2,0.11579519510269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,4,power_law_1.2,0.11173759698867798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,4,power_law_1.2,0.18149759769439697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,4,power_law_1.2,0.16956160068511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,4,power_law_1.2,0.17550719976425172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,4,power_law_1.2,0.17278079986572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,4,power_law_1.2,0.17912319898605347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,4,power_law_1.2,0.18192000389099122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,4,power_law_1.2,0.1906816005706787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,4,power_law_1.2,0.19186559915542603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,4,power_law_1.2,0.1973952054977417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,4,power_law_1.2,0.21573760509490966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,4,power_law_1.2,0.22147839069366454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,4,power_law_1.2,0.21643519401550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,4,power_law_1.2,0.250380802154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,4,power_law_1.2,0.2787391901016235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,16,power_law_1.2,0.04477440118789673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,4,power_law_1.2,0.3506688117980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,16,power_law_1.2,0.06639360189437866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,4,power_law_1.2,0.3973439931869507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,16,power_law_1.2,0.06716160178184509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,16,power_law_1.2,0.06866559982299805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,16,power_law_1.2,0.06499199867248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,16,power_law_1.2,0.06307200193405152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,16,power_law_1.2,0.07305600047111512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,16,power_law_1.2,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,16,power_law_1.2,0.10394879579544067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,4,power_law_1.2,0.4972479820251465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,16,power_law_1.2,0.10005120038986207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,16,power_law_1.2,0.09152640104293823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,16,power_law_1.2,0.13485440015792846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,16,power_law_1.2,0.13771519660949708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,16,power_law_1.2,0.13609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,16,power_law_1.2,0.1352255940437317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,16,power_law_1.2,0.15031039714813232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,4,power_law_1.2,0.5982399940490722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,4,power_law_1.2,0.7945280075073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,4,power_law_1.2,1.022118377685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,16,power_law_1.2,0.139302396774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,16,power_law_1.2,0.1365504026412964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,16,power_law_1.2,0.14160000085830687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,16,power_law_1.2,0.16359039545059204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,16,power_law_1.2,0.17345279455184937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,16,power_law_1.2,0.21416959762573243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,16,power_law_1.2,0.2529216051101685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,16,power_law_1.2,0.3508863925933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,16,power_law_1.2,0.47156481742858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,4,power_law_1.2,1.440601634979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,16,power_law_1.2,0.6668352127075196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,16,power_law_1.2,0.8859007835388184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,4,power_law_1.2,1.7746879577636718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,1,power_law_1.01,0.10535039901733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,1,power_law_1.01,0.15564800500869752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,1,power_law_1.01,0.2487488031387329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,1,power_law_1.01,0.4459968090057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,1,power_law_1.01,0.6293248176574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,1,power_law_1.01,0.8516287803649902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,1,power_law_1.01,1.1633407592773437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,1,power_law_1.01,1.2440256118774413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,1,power_law_1.01,1.29301118850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,1,power_law_1.01,1.3220224380493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,1,power_law_1.01,1.3758720397949218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,1,power_law_1.01,1.404806423187256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,1,power_law_1.01,1.419660758972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,1,power_law_1.01,1.4692480087280273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,1,power_law_1.01,1.499456024169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,1,power_law_1.01,1.5256704330444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,1,power_law_1.01,1.6224512100219726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,1,power_law_1.01,1.750592041015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,1,power_law_1.01,1.786720085144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,1,power_law_1.01,2.027136039733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,1,power_law_1.01,2.042495918273926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,1,power_law_1.01,2.36492805480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,1,power_law_1.01,2.654956817626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,1,power_law_1.01,3.410131072998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,1,power_law_1.01,3.9065601348876955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,1,power_law_1.01,5.110188674926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,1,power_law_1.01,6.416300964355469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,16,balanced,0.05023466547330221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,16,balanced,0.05092266698678335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,16,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,16,balanced,0.05780800183614095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,16,balanced,0.07006399830182393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,16,balanced,0.09288000067075093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,16,balanced,0.09481599926948547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,16,balanced,0.09401599566141765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,16,balanced,0.09697600205739339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,16,balanced,0.09389866391817729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,16,balanced,0.09648000200589497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,16,balanced,0.09774399797121684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,16,balanced,0.09709866841634114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,16,balanced,0.09918399651845296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,16,balanced,0.10453333457310994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,16,balanced,0.10454932848612468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,16,balanced,0.10728533069292705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,16,balanced,0.11705066760381062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,16,balanced,0.11717333396275838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,16,balanced,0.13251200318336487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,16,balanced,0.14587199687957764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,16,balanced,0.17037866512934366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,4,power_law_1.01,0.04528000056743622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,16,balanced,0.19543999433517456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,4,power_law_1.01,0.04867199957370758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,4,power_law_1.01,0.04888960123062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,4,power_law_1.01,0.05247359871864319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,4,power_law_1.01,0.05297279953956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,4,power_law_1.01,0.0535103976726532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,4,power_law_1.01,0.052339202165603636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,4,power_law_1.01,0.05181440114974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,4,power_law_1.01,0.052832001447677614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,4,power_law_1.01,0.053523200750350955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,4,power_law_1.01,0.0525439977645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,4,power_law_1.01,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,4,power_law_1.01,0.0548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,4,power_law_1.01,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,4,power_law_1.01,0.07390720248222352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,4,power_law_1.01,0.07344639897346497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,16,balanced,0.2677813371022542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,16,balanced,0.31597866614659625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,4,power_law_1.01,0.08209919929504395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,16,balanced,0.4310346841812134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,16,balanced,0.5604533354441324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,4,power_law_1.01,0.09477760195732117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,4,power_law_1.01,0.10578559637069702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,4,power_law_1.01,0.1377087950706482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,4,power_law_1.01,0.17496320009231567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,4,power_law_1.01,0.22466559410095216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,4,power_law_1.01,0.27754878997802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,4,power_law_1.01,0.39786880016326903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,4,power_law_1.01,0.5095680236816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,4,power_law_1.01,0.7335103988647461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,8,power_law_1.2,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,8,power_law_1.2,0.04026240110397339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,8,power_law_1.2,0.04056960046291351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,8,power_law_1.2,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,8,power_law_1.2,0.041529598832130435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,8,power_law_1.2,0.04471679925918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,8,power_law_1.2,0.04465279877185822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,8,power_law_1.2,0.043961599469184875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,8,power_law_1.2,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,8,power_law_1.2,0.046540799736976626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,8,power_law_1.2,0.04708479940891266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,8,power_law_1.2,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,8,power_law_1.2,0.05043839812278748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,8,power_law_1.2,0.05249279737472534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,8,power_law_1.2,0.058719998598098753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,8,power_law_1.2,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,8,power_law_1.2,0.06674559712409973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,8,power_law_1.2,0.07544320225715637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,8,power_law_1.2,0.08846719861030579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,8,power_law_1.2,0.1037824034690857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,8,power_law_1.2,0.11379200220108032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,8,power_law_1.2,0.16350719928741456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,8,power_law_1.2,0.2053312063217163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,8,power_law_1.2,0.25470719337463377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,4,power_law_1.01,0.9874815940856934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,8,power_law_1.2,0.3813823938369751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,8,power_law_1.2,0.5672512054443359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,8,power_law_1.01,0.09155200123786926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,8,power_law_1.01,0.13367680311203003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,8,power_law_1.01,0.16760319471359253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,8,power_law_1.01,0.16687359809875488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,8,power_law_1.01,0.1844159960746765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,8,power_law_1.01,0.1832703948020935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,8,power_law_1.01,0.19809279441833497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,8,power_law_1.01,0.19375360012054443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,8,power_law_1.01,0.19736319780349731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,8,power_law_1.01,0.20404479503631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,8,power_law_1.01,0.20994560718536376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,8,power_law_1.01,0.20547199249267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,8,power_law_1.01,0.20401279926300048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,8,power_law_1.01,0.2225856065750122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,8,power_law_1.01,0.22916479110717775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,8,power_law_1.01,0.22010879516601561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,8,power_law_1.01,0.2421567916870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,8,power_law_1.01,0.2738048076629639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,8,power_law_1.01,0.29062399864196775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,8,power_law_1.01,0.3357503890991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,8,power_law_1.01,0.3516607999801636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,8,power_law_1.01,0.4142399787902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,8,power_law_1.2,0.6805376052856446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,8,power_law_1.01,0.48366718292236327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,8,power_law_1.01,0.6909759998321533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,8,power_law_1.01,0.8017600059509278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,8,power_law_1.01,1.1056639671325683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,8,power_law_1.01,1.3797696113586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,16,balanced,0.04368533194065094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,16,balanced,0.04427200059096018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,16,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,16,balanced,0.04570133487383524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,16,balanced,0.04970133304595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,16,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,16,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,16,balanced,0.05382933219273885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,16,balanced,0.054154664278030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,16,balanced,0.054234668612480164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,16,balanced,0.05402666827042898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,16,balanced,0.056176001826922096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,16,balanced,0.055455997586250305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,16,balanced,0.05832533538341522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,16,balanced,0.06311466793219249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,16,balanced,0.06411199768384297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,16,balanced,0.06618666648864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,16,balanced,0.07428266604741414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,16,balanced,0.07666133344173431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,16,balanced,0.10358933607737224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,16,balanced,0.1199626624584198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,16,balanced,0.1683573325475057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,16,balanced,0.19337066014607748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,16,balanced,0.2641226649284363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,16,balanced,0.31569067637125653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,16,balanced,0.44681068261464435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,16,balanced,0.5650346676508585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,8,power_law_1.01,0.018854400515556334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,8,power_law_1.01,0.018412800133228303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,8,power_law_1.01,0.01822720021009445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,8,power_law_1.01,0.018303999304771425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,8,power_law_1.01,0.018624000251293182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,8,power_law_1.01,0.018931199610233308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,8,power_law_1.01,0.018995200097560883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,8,power_law_1.01,0.021996800601482392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,8,power_law_1.01,0.022835199534893037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,8,power_law_1.01,0.022572800517082214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,8,power_law_1.01,0.03380480110645294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,8,power_law_1.01,0.034995201230049136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,8,power_law_1.01,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,8,power_law_1.01,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,8,power_law_1.01,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,8,power_law_1.01,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,8,power_law_1.01,0.03973760008811951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,8,power_law_1.01,0.042233601212501526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.01,0.04561919867992401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.01,0.05278720259666443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.01,0.06236799955368042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.01,0.08492799997329711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.01,0.0992959976196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.01,0.13498879671096803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.01,0.1710528016090393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.01,0.24325759410858155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.01,0.30568320751190187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,4,balanced,0.02534399926662445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,4,balanced,0.025146665672461193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,4,balanced,0.02720000098148982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,4,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,4,balanced,0.06182399888833364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,4,balanced,0.05171733101209005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,4,balanced,0.053685332338015236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,4,balanced,0.05189866820971171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,4,balanced,0.051925331354141235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,4,balanced,0.05070933202902476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,4,balanced,0.05169066786766052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,4,balanced,0.05227733155091604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,4,balanced,0.051957334081331887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,4,balanced,0.0603359987338384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,4,balanced,0.06573866804440816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,4,balanced,0.06429333488146464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,4,balanced,0.06844800213972728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,4,balanced,0.07156800230344136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,4,balanced,0.083146666487058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,4,balanced,0.09060800075531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,4,balanced,0.09340799848238628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,4,balanced,0.11346667011578877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,4,balanced,0.13532267014185587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,4,balanced,0.17561066150665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,4,balanced,0.2100693384806315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,4,balanced,0.2895359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,4,balanced,0.36365334192911786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,128,power_law_1.01,0.06112639904022217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,128,power_law_1.01,0.06495360136032105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,128,power_law_1.01,0.06085759997367859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,128,power_law_1.01,0.06753919720649719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,128,power_law_1.01,0.062035202980041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,128,power_law_1.01,0.06296319961547851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,128,power_law_1.01,0.06604160070419311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,128,power_law_1.01,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,128,power_law_1.01,0.06272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,128,power_law_1.01,0.06478719711303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,128,power_law_1.01,0.0652288019657135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,128,power_law_1.01,0.06519039869308471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,128,power_law_1.01,0.06689280271530151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,128,power_law_1.01,0.06723200082778931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,128,power_law_1.01,0.0732096016407013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,128,power_law_1.01,0.07118080258369446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,128,power_law_1.01,0.07460479736328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,128,power_law_1.01,0.08517119884490967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.01,0.08961920142173767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.01,0.1106112003326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.01,0.12299519777297974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.01,0.16391680240631104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.01,0.19612159729003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.01,0.2602240085601807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.01,0.3286400079727173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.01,0.4542272090911865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.01,0.6041152000427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,4,power_law_1.2,0.07897599935531616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,4,power_law_1.2,0.11251200437545776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,4,power_law_1.2,0.11222399473190307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,4,power_law_1.2,0.1552191972732544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,4,power_law_1.2,0.18256640434265137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,4,power_law_1.2,0.21683199405670167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,4,power_law_1.2,0.2426624059677124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,4,power_law_1.2,0.2506943941116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,4,power_law_1.2,0.27907838821411135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,4,power_law_1.2,0.2517888069152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,4,power_law_1.2,0.2715264081954956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,4,power_law_1.2,0.2736448049545288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,4,power_law_1.2,0.2866111993789673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,4,power_law_1.2,0.28682239055633546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,4,power_law_1.2,0.2871295928955078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,4,power_law_1.2,0.2995392084121704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,4,power_law_1.2,0.32083840370178224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,4,power_law_1.2,0.35513598918914796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,4,power_law_1.2,0.36671359539031984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,4,power_law_1.2,0.4299647808074951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,4,power_law_1.2,0.42924160957336427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,4,power_law_1.2,0.520684814453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,4,power_law_1.2,0.558899211883545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,4,power_law_1.2,0.6950208187103272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,4,power_law_1.2,0.8294848442077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,4,power_law_1.2,1.16429443359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,4,power_law_1.2,1.461068820953369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,16,balanced,0.050479998191197716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,16,balanced,0.050101334849993386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,16,balanced,0.052245333790779114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,16,balanced,0.058378666639328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,16,balanced,0.0584799995024999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,16,balanced,0.0666240006685257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,16,balanced,0.07056533296902974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,16,balanced,0.07088000078996022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,16,balanced,0.07156800230344136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,16,balanced,0.0726986676454544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,16,balanced,0.07261333366235097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,16,balanced,0.07479466497898102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,16,balanced,0.0784693310658137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,16,balanced,0.07725866635640462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,16,balanced,0.08410132924715678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,16,balanced,0.08293333152929942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,16,balanced,0.08894399801890056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,16,balanced,0.09705600142478943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,16,balanced,0.10735467076301575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,16,balanced,0.12825066844622293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,16,balanced,0.14622933665911356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,16,balanced,0.18545599778493246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,16,balanced,0.2229493260383606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,16,balanced,0.28829866647720337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,16,balanced,0.3751893440882365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,16,balanced,0.5063413381576538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,16,balanced,0.6578986644744873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,1,balanced,0.056426664193471275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,1,balanced,0.07814399898052216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,1,balanced,0.10624000430107117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,1,balanced,0.1623146633307139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,1,balanced,0.2733546694119771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,1,balanced,0.27668799956639606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,1,balanced,0.28143999973932904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,1,balanced,0.28067199389139813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,1,balanced,0.28096532821655273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,1,balanced,0.28544000784556073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,1,balanced,0.2869653304417928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,1,balanced,0.2916746735572815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,1,balanced,0.2966773311297099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,1,balanced,0.299834668636322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,1,balanced,0.3111039996147156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,1,balanced,0.3200266758600871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,1,balanced,0.337557315826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,1,balanced,0.3705120086669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,1,balanced,0.41177066167195636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,1,balanced,0.47042131423950195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,1,balanced,0.5317386786142985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,1,balanced,0.7602933247884115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,1,balanced,0.878928025563558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,1,balanced,1.2563733259836833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,1,balanced,1.5535732905069988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,1,balanced,2.2599679629007974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,1,balanced,2.9440479278564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,16,power_law_1.01,0.021830399334430695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,16,power_law_1.01,0.023097600042819976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,16,power_law_1.01,0.023129600286483764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,16,power_law_1.01,0.02316800057888031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,16,power_law_1.01,0.024332800507545473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,16,power_law_1.01,0.024300800263881685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,16,power_law_1.01,0.026495999097824095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,16,power_law_1.01,0.027027198672294618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,16,power_law_1.01,0.0355648010969162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,16,power_law_1.01,0.03372159898281098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,16,power_law_1.01,0.03506560027599335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,16,power_law_1.01,0.057017600536346434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,16,power_law_1.01,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,16,power_law_1.01,0.05959039926528931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,16,power_law_1.01,0.058380800485610965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,16,power_law_1.01,0.05719040036201477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,16,power_law_1.01,0.061510401964187625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,16,power_law_1.01,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.01,0.07071359753608704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.01,0.07809280157089234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.01,0.08542720079421998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.01,0.10798079967498779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.01,0.1388864040374756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.01,0.17641600370407104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.01,0.21679999828338622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.01,0.3074944019317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.01,0.40355839729309084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,1,balanced,0.13131733735402426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,1,balanced,0.2195840080579122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,1,balanced,0.38650667667388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,1,balanced,0.7244053681691488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,1,balanced,1.394741376241048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,1,balanced,2.7302347819010415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,1,balanced,2.7318827311197915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,1,balanced,2.7460479736328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,1,balanced,2.7592639923095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,1,balanced,2.7592798868815103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,1,balanced,2.7692693074544272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,1,balanced,2.7714560826619468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,1,balanced,2.785594622294108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,1,balanced,2.8085705439249673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,1,balanced,2.8196694056193032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,1,balanced,2.8325866063435874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,1,balanced,2.8607041041056314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,1,balanced,2.9119625091552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,1,balanced,2.994293212890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,1,balanced,3.094026565551758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,1,balanced,3.188058535257975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,1,balanced,3.3427626291910806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,1,balanced,3.7288106282552085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,1,balanced,3.903674761454264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,1,balanced,4.253909428914388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,1,balanced,6.472090403238933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,1,balanced,7.295573552449544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,128,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,128,balanced,0.05806933343410492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,128,balanced,0.05176533261934916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,128,balanced,0.07583466668923695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,128,balanced,0.08191466828187306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,128,balanced,0.08060266574223836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,128,balanced,0.08412266770998637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,128,balanced,0.08269333342711131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,128,balanced,0.08272533118724823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,128,balanced,0.08504533767700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,128,balanced,0.08249600231647491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,128,balanced,0.0867146650950114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,128,balanced,0.08683733145395915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,128,balanced,0.0865226686000824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,128,balanced,0.09265066186587016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,128,balanced,0.09273599584897359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,128,balanced,0.09610133369763692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,128,balanced,0.11336533228556316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,128,balanced,0.11045333743095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,128,balanced,0.13752532998720804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,128,balanced,0.14409066239992777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,128,balanced,0.1894773244857788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,128,balanced,0.1770133376121521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,128,balanced,0.23113600413004556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,128,balanced,0.24016000827153525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,128,balanced,0.3141760031382243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,128,balanced,0.36588799953460693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,32,balanced,0.023013333479563396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,32,balanced,0.023365333676338196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,32,balanced,0.020901332298914593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,32,balanced,0.02126399924357732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,32,balanced,0.023103999594847362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,32,balanced,0.024010665714740753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,32,balanced,0.03525333354870478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,32,balanced,0.029525332152843475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,32,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,32,balanced,0.03869866579771042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,32,balanced,0.04368533194065094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,32,balanced,0.059936001896858215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,32,balanced,0.05989866455396017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,32,balanced,0.061679999033610024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,32,balanced,0.0642080008983612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,32,balanced,0.06428266565004985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,32,balanced,0.08453333377838135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,32,balanced,0.10859733819961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,32,balanced,0.1469439963499705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,32,balanced,0.19344000021616617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,32,balanced,0.10949333508809407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,32,balanced,0.1444000005722046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,32,balanced,0.18369599183400473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,32,balanced,0.256058673063914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,32,balanced,0.33292800188064575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,32,balanced,0.4838293393452962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,32,balanced,0.6358453432718912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,4,power_law_1.2,0.04004479944705963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,4,power_law_1.2,0.032076799869537355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,4,power_law_1.2,0.031200000643730165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,4,power_law_1.2,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,4,power_law_1.2,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,4,power_law_1.2,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,4,power_law_1.2,0.04062080085277557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,4,power_law_1.2,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,4,power_law_1.2,0.04505600035190582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,4,power_law_1.2,0.04789760112762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,4,power_law_1.2,0.04762240052223206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,4,power_law_1.2,0.05089920163154602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,4,power_law_1.2,0.04999040067195892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,4,power_law_1.2,0.051718401908874514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,4,power_law_1.2,0.06379520297050476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,4,power_law_1.2,0.0663424015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,4,power_law_1.2,0.07327359914779663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,4,power_law_1.2,0.08577280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,4,power_law_1.2,0.08652799725532531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,4,power_law_1.2,0.10513919591903687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,4,power_law_1.2,0.12513279914855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,4,power_law_1.2,0.15210880041122438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,4,power_law_1.2,0.18922879695892333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,4,power_law_1.2,0.2583872079849243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,4,power_law_1.2,0.30398719310760497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,4,power_law_1.2,0.43775358200073244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,2,balanced,0.0813973347345988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,2,balanced,0.10368000467618306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,2,balanced,0.15377066532770792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,2,balanced,0.15794666608174643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,2,balanced,0.16036267081896463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,2,balanced,0.15924800435702005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,2,balanced,0.1593119998772939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,2,balanced,0.16197866201400757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,2,balanced,0.15988799929618835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,2,balanced,0.16148799657821655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,2,balanced,0.1607253352801005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,2,balanced,0.16275200247764587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,2,balanced,0.1625759998957316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,2,balanced,0.16511999567349753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,2,balanced,0.1721493403116862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,2,balanced,0.17614400386810303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,2,balanced,0.17722666263580322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,2,balanced,0.2571360071500142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,2,balanced,0.24714134136835733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,2,balanced,0.32974932591120404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,2,balanced,0.42362133661905926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,2,balanced,0.5888853470484415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,2,balanced,0.7441973686218262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,2,balanced,1.060751994450887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,2,balanced,1.3717333475748699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,2,balanced,2.0500052769978843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,2,balanced,2.680762608846029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,64,balanced,0.04770666857560476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,64,balanced,0.04413333535194397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,64,balanced,0.04623466730117798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,64,balanced,0.04373333354791006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,64,balanced,0.04580266773700714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,64,balanced,0.04966933528582255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,64,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,64,balanced,0.050053333242734276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,64,balanced,0.05178666611512502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,64,balanced,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,64,balanced,0.05018133421738943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,64,balanced,0.053770666321118675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,64,balanced,0.05230933427810669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,64,balanced,0.05425066749254862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,64,balanced,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,64,balanced,0.05795200169086456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,64,balanced,0.061994666854540505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,64,balanced,0.06717866659164429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,64,balanced,0.06950933237870534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,64,balanced,0.0805866668621699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,64,balanced,0.09302933017412822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,64,balanced,0.12171733379364014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,64,balanced,0.14526399970054626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,64,balanced,0.19090133905410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,64,balanced,0.2355466683705648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,64,balanced,0.31802666187286377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,64,balanced,0.4021173318227132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,4,power_law_1.2,0.5597055912017822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,16,power_law_1.2,0.050860798358917235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,16,power_law_1.2,0.06888960003852844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,16,power_law_1.2,0.052134400606155394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,16,power_law_1.2,0.054022401571273804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,16,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,16,power_law_1.2,0.0523904025554657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,16,power_law_1.2,0.05432320237159729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,16,power_law_1.2,0.054579198360443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,16,power_law_1.2,0.05507199764251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,16,power_law_1.2,0.0546176016330719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,16,power_law_1.2,0.05605120062828064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,16,power_law_1.2,0.057171201705932616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,16,power_law_1.2,0.058924800157546996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,16,power_law_1.2,0.0596671998500824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,16,power_law_1.2,0.06533759832382202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,16,power_law_1.2,0.06731520295143127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,16,power_law_1.2,0.07271040081977845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,16,power_law_1.2,0.08392959833145142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,16,power_law_1.2,0.0884223997592926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,16,power_law_1.2,0.10823680162429809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,16,power_law_1.2,0.1215999960899353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,16,power_law_1.2,0.1579519987106323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,16,power_law_1.2,0.18770560026168823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,16,power_law_1.2,0.24643199443817138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,16,power_law_1.2,0.32108159065246583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,16,power_law_1.2,0.46698880195617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,16,power_law_1.2,0.6115007877349854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,16,balanced,0.030554667115211487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,16,balanced,0.033946665624777474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,16,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,16,balanced,0.031285333136717476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,16,balanced,0.0315733328461647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,16,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,16,balanced,0.03326933334271113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,16,balanced,0.033589333295822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,16,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,16,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,16,balanced,0.035242666800816856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,16,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,16,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,16,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,16,balanced,0.03935466706752777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,16,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,16,balanced,0.042208001017570496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,16,balanced,0.041946664452552795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,16,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,16,balanced,0.050714666644732155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,16,balanced,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,16,balanced,0.06622933348019917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,16,balanced,0.07707199951012929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,16,balanced,0.09638933340708415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,16,balanced,0.12031466762224834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,16,balanced,0.15016532937685648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,16,balanced,0.1874986688296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,32,balanced,0.046053335070610046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,32,balanced,0.04604800045490265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,32,balanced,0.046682665745417275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,32,balanced,0.046394666035970054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,32,balanced,0.05436799923578898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,32,balanced,0.07518399755160014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,32,balanced,0.07632000247637431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,32,balanced,0.07673066854476929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,32,balanced,0.07893333335717519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,32,balanced,0.07904000083605449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,32,balanced,0.07881066699822743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,32,balanced,0.08281066517035167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,32,balanced,0.08235733211040497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,32,balanced,0.08450667063395183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,32,balanced,0.0906773308912913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,32,balanced,0.09088533123334248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,32,balanced,0.09618666768074036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,32,balanced,0.10585066676139832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,32,balanced,0.112335999806722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,32,balanced,0.13242666920026144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,32,balanced,0.14985066652297974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,32,balanced,0.1845973332722982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,32,balanced,0.22351467609405518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,32,balanced,0.2903680006663005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,32,balanced,0.36699732144673664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,32,balanced,0.5104106664657593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,32,balanced,0.6431519985198975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,2,balanced,0.023557332654794056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,2,balanced,0.025199999411900837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,2,balanced,0.03458133339881897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,2,balanced,0.051034669081370033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,2,balanced,0.07670400043328603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,2,balanced,0.12267733613650005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,2,balanced,0.1255466639995575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,2,balanced,0.12871467073758444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,2,balanced,0.13190933068593344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,2,balanced,0.13531200091044107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,2,balanced,0.14218133687973022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,2,balanced,0.14008000493049622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,2,balanced,0.14358933766682944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,2,balanced,0.1511786679426829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,2,balanced,0.15757866700490317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,2,balanced,0.15847466389338175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,2,balanced,0.1646399994691213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,2,balanced,0.21243733167648315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,2,balanced,0.21002666155497232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,2,balanced,0.3630400101343791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,2,balanced,0.39928531646728516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,2,balanced,0.37763198216756183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,2,balanced,0.4044479926427205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,2,balanced,0.4488746722539266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,2,balanced,0.4929493268330892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,2,balanced,0.8242613474527994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,2,balanced,0.9222453435262045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,128,power_law_1.01,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,128,power_law_1.01,0.03430399894714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,128,power_law_1.01,0.03421440124511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,128,power_law_1.01,0.03620480000972748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,128,power_law_1.01,0.037036800384521486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,128,power_law_1.01,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,128,power_law_1.01,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,128,power_law_1.01,0.03770880103111267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,128,power_law_1.01,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,128,power_law_1.01,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,128,power_law_1.01,0.03832319974899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,128,power_law_1.01,0.03866879940032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,128,power_law_1.01,0.03951359987258911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,128,power_law_1.01,0.039852800965309146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,128,power_law_1.01,0.04292480051517487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,128,power_law_1.01,0.04281600117683411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,128,power_law_1.01,0.04368639886379242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,128,power_law_1.01,0.04509440064430237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,128,power_law_1.01,0.04572800099849701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,128,power_law_1.01,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,128,power_law_1.01,0.05060480237007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,128,power_law_1.01,0.0605567991733551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,128,power_law_1.01,0.06691840291023254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,128,power_law_1.01,0.08579840064048767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,128,power_law_1.01,0.11663360595703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,128,power_law_1.01,0.1535040020942688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,128,power_law_1.01,0.18677120208740233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,4,power_law_1.2,0.074508798122406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,4,power_law_1.2,0.0923904001712799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,4,power_law_1.2,0.09320319890975952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,4,power_law_1.2,0.10905599594116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,4,power_law_1.2,0.11906559467315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,4,power_law_1.2,0.12934399843215943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,4,power_law_1.2,0.14880640506744386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,4,power_law_1.2,0.14256639480590821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,4,power_law_1.2,0.14582400321960448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,4,power_law_1.2,0.1458624005317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,4,power_law_1.2,0.14995839595794677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,4,power_law_1.2,0.15118080377578735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,4,power_law_1.2,0.15983359813690184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,4,power_law_1.2,0.16620800495147706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,4,power_law_1.2,0.1705664038658142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,4,power_law_1.2,0.17196799516677858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,4,power_law_1.2,0.18133759498596191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,4,power_law_1.2,0.20010240077972413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,4,power_law_1.2,0.22485120296478273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,4,power_law_1.2,0.2869760036468506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,4,power_law_1.2,0.3444607973098755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,4,power_law_1.2,0.44919681549072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,4,power_law_1.2,0.5656832218170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,4,power_law_1.2,0.7558144092559814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,4,power_law_1.2,1.0288576126098632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,4,power_law_1.2,1.474726390838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,4,power_law_1.2,1.782566452026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,32,power_law_1.01,0.11598720550537109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,32,power_law_1.01,0.11274240016937256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,32,power_law_1.01,0.11034879684448243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,32,power_law_1.01,0.11312639713287354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,32,power_law_1.01,0.11253119707107544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,32,power_law_1.01,0.11319680213928222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,32,power_law_1.01,0.11527680158615113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,32,power_law_1.01,0.11504000425338745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,32,power_law_1.01,0.11468160152435303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,32,power_law_1.01,0.11582080125808716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,32,power_law_1.01,0.11746560335159302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,32,power_law_1.01,0.12455040216445923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,32,power_law_1.01,0.12406400442123414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,32,power_law_1.01,0.12743680477142333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,32,power_law_1.01,0.1453760027885437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,32,power_law_1.01,0.14103039503097534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,32,power_law_1.01,0.14890880584716798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,32,power_law_1.01,0.15491199493408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,32,power_law_1.01,0.16817920207977294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,32,power_law_1.01,0.18835840225219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,32,power_law_1.01,0.2110975980758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,32,power_law_1.01,0.2716480016708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,32,power_law_1.01,0.3410624027252197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,32,power_law_1.01,0.45652480125427247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,32,power_law_1.01,0.5802112102508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,32,power_law_1.01,0.7984511852264404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,32,power_law_1.01,1.1613056182861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,1,power_law_1.2,0.06369919776916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,1,power_law_1.2,0.08604800105094909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,1,power_law_1.2,0.10705280303955078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,1,power_law_1.2,0.15610239505767823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,1,power_law_1.2,0.223801589012146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,1,power_law_1.2,0.2996223926544189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,1,power_law_1.2,0.34780800342559814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,1,power_law_1.2,0.4536384105682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,1,power_law_1.2,0.4902400016784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,1,power_law_1.2,0.5027584075927735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,1,power_law_1.2,0.5140031814575196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,1,power_law_1.2,0.5516287803649902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,1,power_law_1.2,0.5671872138977051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,1,power_law_1.2,0.5885248184204102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,1,power_law_1.2,0.5997695922851562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,1,power_law_1.2,0.6249279975891113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,1,power_law_1.2,0.6579775810241699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,1,power_law_1.2,0.7331711769104003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,1,power_law_1.2,0.8095040321350098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,1,power_law_1.2,0.9652607917785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,1,power_law_1.2,1.0470656394958495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,1,power_law_1.2,1.3102463722229003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,1,power_law_1.2,1.5704383850097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,1,power_law_1.2,2.0194879531860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,1,power_law_1.2,2.4294336318969725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,1,power_law_1.2,3.2368831634521484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,1,power_law_1.2,4.219500732421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,2,balanced,0.021136000752449036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,2,balanced,0.023205332458019257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,2,balanced,0.025008000433444977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,2,balanced,0.03324799984693527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,2,balanced,0.054048001766204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,2,balanced,0.07683733105659485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,2,balanced,0.08002666632334392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,2,balanced,0.08161066472530365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,2,balanced,0.08470933636029561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,2,balanced,0.08686400453249614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,2,balanced,0.0888266662756602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,2,balanced,0.09198400378227234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,2,balanced,0.09489599863688152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,2,balanced,0.09629866480827332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,2,balanced,0.09886399904886882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,2,balanced,0.10146666566530864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,2,balanced,0.10971732934315999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,2,balanced,0.12578133742014566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,2,balanced,0.14260266224543253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,2,balanced,0.18069332838058472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,2,balanced,0.18319465716679892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,2,balanced,0.2536533276240031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,2,balanced,0.3033600052197774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,2,balanced,0.3439893325169881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,2,balanced,0.3843199809392293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,2,balanced,0.634224017461141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,2,balanced,0.7234453360239664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,16,power_law_1.01,0.07563520073890687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,16,power_law_1.01,0.09210240244865417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,16,power_law_1.01,0.0885312020778656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,16,power_law_1.01,0.09297919869422913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,16,power_law_1.01,0.0911296010017395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,16,power_law_1.01,0.08262400031089782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,16,power_law_1.01,0.09393280148506164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,16,power_law_1.01,0.09265279769897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,16,power_law_1.01,0.09306880235671997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,16,power_law_1.01,0.09141759872436524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,16,power_law_1.01,0.09512959718704224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,16,power_law_1.01,0.09389439821243287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,16,power_law_1.01,0.09797760248184204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,16,power_law_1.01,0.09767040014266967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,16,power_law_1.01,0.10746239423751831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,16,power_law_1.01,0.10978560447692871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,16,power_law_1.01,0.11452159881591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,16,power_law_1.01,0.1266111969947815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,16,power_law_1.01,0.13635200262069702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,16,power_law_1.01,0.16168320178985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,16,power_law_1.01,0.18201600313186644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,16,power_law_1.01,0.2566272020339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,16,power_law_1.01,0.30142080783843994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,16,power_law_1.01,0.4084671974182129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,16,power_law_1.01,0.5566912174224854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,16,power_law_1.01,0.7897535800933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,16,power_law_1.01,0.9915583610534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,64,balanced,0.050160000721613564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,64,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,64,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,64,balanced,0.05318933228651682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,64,balanced,0.05418133238951365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,64,balanced,0.06542933483918507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,64,balanced,0.06324266890684764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,64,balanced,0.06420266628265381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,64,balanced,0.06204266846179962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,64,balanced,0.062319998939832054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,64,balanced,0.06403733293215434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,64,balanced,0.06618666648864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,64,balanced,0.0681333343187968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,64,balanced,0.06358933448791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,64,balanced,0.07077866792678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,64,balanced,0.06921066840489705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,64,balanced,0.07234133283297221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,64,balanced,0.07602666815121968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,64,balanced,0.08035733302434285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,64,balanced,0.08999466896057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,64,balanced,0.09708799918492635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,64,balanced,0.11726933717727661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,64,balanced,0.13684800267219543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,64,balanced,0.1825973391532898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,64,balanced,0.2225333253542582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,64,balanced,0.2995733420054118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,64,balanced,0.3803733189900716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,32,power_law_1.01,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,32,power_law_1.01,0.06595839858055115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,32,power_law_1.01,0.06414719820022582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,32,power_law_1.01,0.0691968023777008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,32,power_law_1.01,0.06912000179290771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,32,power_law_1.01,0.06710399985313416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,32,power_law_1.01,0.07048320174217224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,32,power_law_1.01,0.07013760209083557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,32,power_law_1.01,0.07166720032691956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,32,power_law_1.01,0.0732096016407013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,32,power_law_1.01,0.07284479737281799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,32,power_law_1.01,0.0782912015914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,32,power_law_1.01,0.07876480221748353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,32,power_law_1.01,0.08158720135688782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,32,power_law_1.01,0.09703680276870727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,32,power_law_1.01,0.09651839733123779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,32,power_law_1.01,0.09586560130119323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,32,power_law_1.01,0.11299200057983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,32,power_law_1.01,0.12003200054168701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,32,power_law_1.01,0.1414720058441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,32,power_law_1.01,0.15512959957122802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,32,power_law_1.01,0.19617279767990112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,32,power_law_1.01,0.23240320682525634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,32,power_law_1.01,0.3473727941513062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,32,power_law_1.01,0.4111616134643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,32,power_law_1.01,0.6115839958190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,32,power_law_1.01,0.862003231048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,2,power_law_1.2,0.04291839897632599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,2,power_law_1.2,0.044787201285362246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,2,power_law_1.2,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,2,power_law_1.2,0.05759360194206238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,2,power_law_1.2,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,2,power_law_1.2,0.08748160004615783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,2,power_law_1.2,0.09022719860076904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,2,power_law_1.2,0.09055359959602356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,2,power_law_1.2,0.0923904001712799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,2,power_law_1.2,0.09505919814109802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,2,power_law_1.2,0.09640960097312927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,2,power_law_1.2,0.1005247950553894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,2,power_law_1.2,0.10586240291595458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,2,power_law_1.2,0.1050495982170105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,2,power_law_1.2,0.111571204662323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,2,power_law_1.2,0.11692160367965698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,2,power_law_1.2,0.12651519775390624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,2,power_law_1.2,0.14963200092315673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.2,0.1643903970718384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.2,0.19728000164031984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.2,0.2360703945159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.2,0.3247488021850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.2,0.39290239810943606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.2,0.5984255790710449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.2,0.7100031852722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.2,1.0833024024963378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,256,power_law_1.2,0.060031998157501223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,256,power_law_1.2,0.06007680296897888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,256,power_law_1.2,0.060070401430130003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,256,power_law_1.2,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,256,power_law_1.2,0.06460800170898437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,256,power_law_1.2,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,256,power_law_1.2,0.06513919830322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,256,power_law_1.2,0.06520320177078247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,256,power_law_1.2,0.06577919721603394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,256,power_law_1.2,0.06513919830322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,256,power_law_1.2,0.06617599725723267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,256,power_law_1.2,0.06651520133018493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,256,power_law_1.2,0.06610559821128845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,256,power_law_1.2,0.06611199975013733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,256,power_law_1.2,0.0698303997516632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,256,power_law_1.2,0.06936960220336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,256,power_law_1.2,0.08044800162315369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,256,power_law_1.2,0.09023360013961793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,256,power_law_1.2,0.10620160102844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,256,power_law_1.2,0.1402943968772888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,256,power_law_1.2,0.16800639629364014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,256,power_law_1.2,0.2229248046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,256,power_law_1.2,0.28139519691467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,256,power_law_1.2,0.4047679901123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,256,power_law_1.2,0.5161791801452636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,256,power_law_1.2,0.8131135940551758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,256,power_law_1.2,1.044320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,4,power_law_1.2,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.2,1.5429439544677734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,4,power_law_1.2,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,4,power_law_1.2,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,4,power_law_1.2,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,4,power_law_1.2,0.039059200882911684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,4,power_law_1.2,0.04469760060310364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,4,power_law_1.2,0.04614399969577789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,4,power_law_1.2,0.04755200147628784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,4,power_law_1.2,0.04888960123062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,4,power_law_1.2,0.05100160241127014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,4,power_law_1.2,0.05441280007362366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,4,power_law_1.2,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,4,power_law_1.2,0.06350719928741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,4,power_law_1.2,0.06321920156478882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,4,power_law_1.2,0.07011839747428894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,4,power_law_1.2,0.07698559761047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,4,power_law_1.2,0.07461119890213012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,4,power_law_1.2,0.09354879856109619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,4,power_law_1.2,0.10522240400314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,4,power_law_1.2,0.13303680419921876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,4,power_law_1.2,0.1503872036933899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,4,power_law_1.2,0.19705599546432495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,4,power_law_1.2,0.26673920154571534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,4,power_law_1.2,0.3503232002258301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,4,power_law_1.2,0.4847743988037109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,4,power_law_1.2,0.7186816215515137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,16,power_law_1.2,0.10654079914093018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,16,power_law_1.2,0.15576959848403932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,16,power_law_1.2,0.19249279499053956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,16,power_law_1.2,0.18780800104141235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,16,power_law_1.2,0.20117759704589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,16,power_law_1.2,0.19303679466247559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,16,power_law_1.2,0.18131200075149537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,16,power_law_1.2,0.19082239866256714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,16,power_law_1.2,0.17525119781494142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,16,power_law_1.2,0.19032959938049315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,16,power_law_1.2,0.19708800315856934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,16,power_law_1.2,0.18098560571670533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,16,power_law_1.2,0.19623039960861205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,16,power_law_1.2,0.18805760145187378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,16,power_law_1.2,0.2002687931060791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,16,power_law_1.2,0.20774400234222412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,16,power_law_1.2,0.21925759315490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,16,power_law_1.2,0.25136001110076905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,16,power_law_1.2,0.278272008895874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,16,power_law_1.2,0.3338495969772339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,16,power_law_1.2,0.3408512115478516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,16,power_law_1.2,0.4318079948425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,16,power_law_1.2,0.5361023902893066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,16,power_law_1.2,0.7820032119750977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,16,power_law_1.2,0.7932288169860839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,16,power_law_1.2,1.1230655670166017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,16,power_law_1.2,1.4699647903442383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,4,power_law_1.2,0.925055980682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,4,power_law_1.01,0.0906495988368988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,4,power_law_1.01,0.12671999931335448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,4,power_law_1.01,0.12324479818344117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,4,power_law_1.01,0.14552320241928102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,4,power_law_1.01,0.17741440534591674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,4,power_law_1.01,0.23004159927368165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,4,power_law_1.01,0.23400959968566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,4,power_law_1.01,0.22590720653533936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,4,power_law_1.01,0.23400321006774902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,4,power_law_1.01,0.2381056070327759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,4,power_law_1.01,0.242956805229187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,4,power_law_1.01,0.24579839706420897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,4,power_law_1.01,0.25696640014648436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,4,power_law_1.01,0.2615423917770386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,4,power_law_1.01,0.27413120269775393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,4,power_law_1.01,0.2815743923187256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,4,power_law_1.01,0.2939136028289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,4,power_law_1.01,0.33201279640197756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,4,power_law_1.01,0.35411839485168456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,4,power_law_1.01,0.4222720146179199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,4,power_law_1.01,0.48856320381164553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,4,power_law_1.01,0.6132224082946778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,4,power_law_1.01,0.6733695983886718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,4,power_law_1.01,0.8779071807861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,4,power_law_1.01,1.128211212158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,4,power_law_1.01,1.5613375663757325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,4,power_law_1.01,1.9541183471679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,1,balanced,0.05966933568318685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,1,balanced,0.07164266705513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,1,balanced,0.09635733564694722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,1,balanced,0.14231466253598532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,1,balanced,0.22640534241994223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,1,balanced,0.39669867356618244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,1,balanced,0.4012266794840495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,1,balanced,0.40645865599314374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,1,balanced,0.4132479826609294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,1,balanced,0.41869866847991943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,1,balanced,0.4192533493041992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,1,balanced,0.4294240077336629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,1,balanced,0.4294346570968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,1,balanced,0.43724266688028973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,1,balanced,0.45030399163564044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,1,balanced,0.4550559918085734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,1,balanced,0.46798400084177655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,1,balanced,0.5060586531956991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,1,balanced,0.5342880090077718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,1,balanced,0.6024800141652426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,1,balanced,0.6670933564503988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,1,balanced,0.855616013209025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,1,balanced,0.9476746718088785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,1,balanced,1.3031466801961262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,1,balanced,1.5765280723571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,1,balanced,2.2097973823547363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,1,balanced,2.7468694051106772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,32,power_law_1.01,0.022316800057888032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,32,power_law_1.01,0.021862399578094483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,32,power_law_1.01,0.01968639940023422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,32,power_law_1.01,0.020230400562286376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,32,power_law_1.01,0.02160000056028366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,32,power_law_1.01,0.023660799860954283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,32,power_law_1.01,0.02462719976902008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,32,power_law_1.01,0.030444800853729248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,32,power_law_1.01,0.030118399858474733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,32,power_law_1.01,0.03128319978713989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,32,power_law_1.01,0.042828801274299624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,32,power_law_1.01,0.04345600008964538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,32,power_law_1.01,0.04391680061817169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,32,power_law_1.01,0.04526079893112182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,32,power_law_1.01,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,32,power_law_1.01,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,32,power_law_1.01,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,32,power_law_1.01,0.08228480219841003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.01,0.09736319780349731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.01,0.1423359990119934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.01,0.08026880025863647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.01,0.10522880554199218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.01,0.11800320148468017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.01,0.19114240407943725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.01,0.22794880867004394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.01,0.3055999994277954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.01,0.39673600196838377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,8,power_law_1.2,0.02172800004482269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,8,power_law_1.2,0.021190400421619415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,8,power_law_1.2,0.020787200331687926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,8,power_law_1.2,0.020921599864959717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,8,power_law_1.2,0.021721599996089934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,8,power_law_1.2,0.025183999538421632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,8,power_law_1.2,0.032902398705482484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,8,power_law_1.2,0.028518399596214293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,8,power_law_1.2,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,8,power_law_1.2,0.031763198971748355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,8,power_law_1.2,0.03656960129737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,8,power_law_1.2,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,8,power_law_1.2,0.05944960117340088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,8,power_law_1.2,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,8,power_law_1.2,0.0645695984363556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,8,power_law_1.2,0.05509759783744812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,8,power_law_1.2,0.06282879710197449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,8,power_law_1.2,0.08271999955177307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.2,0.09836800098419189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.2,0.1360703945159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.2,0.1706496000289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.2,0.23795840740203858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.2,0.29863040447235106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.2,0.4341248035430908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.2,0.5624832153320313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.2,0.8278847694396972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.2,1.0918720245361329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,1,balanced,0.03792533278465271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,1,balanced,0.037605332831541695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,1,balanced,0.044437333941459656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,1,balanced,0.04780800143877665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,1,balanced,0.048512001832326256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,1,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,1,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,1,balanced,0.05213866631189982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,1,balanced,0.051925331354141235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,1,balanced,0.05211733281612396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,1,balanced,0.052416001756985985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,1,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,1,balanced,0.05589866638183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,1,balanced,0.05820266902446747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,1,balanced,0.06717866659164429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,1,balanced,0.06810666620731354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,1,balanced,0.06819733480612437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,1,balanced,0.08658132950464885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,1,balanced,0.0935040016969045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,1,balanced,0.11599999666213989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,1,balanced,0.13990933696428934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,1,balanced,0.19271999597549438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,1,balanced,0.2542933424313863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,1,balanced,0.3647306760152181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,1,balanced,0.4654346704483032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,1,balanced,0.6749546527862549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,1,balanced,0.8778773148854574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,32,power_law_1.01,0.06508799791336059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,32,power_law_1.01,0.06447359919548035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,32,power_law_1.01,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,32,power_law_1.01,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,32,power_law_1.01,0.06312320232391358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,32,power_law_1.01,0.06305279731750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,32,power_law_1.01,0.06543999910354614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,32,power_law_1.01,0.06535680294036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,32,power_law_1.01,0.06817920207977295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,32,power_law_1.01,0.06636160016059875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,32,power_law_1.01,0.06810879707336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,32,power_law_1.01,0.06753919720649719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,32,power_law_1.01,0.06914560198783874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,32,power_law_1.01,0.07361279726028443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,32,power_law_1.01,0.07634559869766236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,32,power_law_1.01,0.07697280049324036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,32,power_law_1.01,0.08167679905891419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,32,power_law_1.01,0.097952002286911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,32,power_law_1.01,0.10752639770507813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,32,power_law_1.01,0.13125760555267335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,32,power_law_1.01,0.15553280115127563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,32,power_law_1.01,0.19428479671478271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,32,power_law_1.01,0.24376959800720216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,32,power_law_1.01,0.32695679664611815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,32,power_law_1.01,0.4234816074371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,32,power_law_1.01,0.5849855899810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,32,power_law_1.01,0.8203968048095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,16,balanced,0.054229333996772766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,16,balanced,0.04437866806983948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,16,balanced,0.044010668992996216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,16,balanced,0.04827199876308441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,16,balanced,0.05231466889381409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,16,balanced,0.07621866464614868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,16,balanced,0.07519466678301494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,16,balanced,0.07781333227952321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,16,balanced,0.07729599873224895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,16,balanced,0.07739200194676717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,16,balanced,0.07673599819342296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,16,balanced,0.08005333443482716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,16,balanced,0.07806933422883351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,16,balanced,0.080485333998998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,16,balanced,0.08659733335177104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,16,balanced,0.08674133817354839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,16,balanced,0.09060800075531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,16,balanced,0.09778666496276855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,16,balanced,0.10155199964841206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,16,balanced,0.11329600214958191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,16,balanced,0.12710932890574136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,16,balanced,0.15401066342989603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,16,balanced,0.1810879906018575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,16,balanced,0.22759999831517538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,16,balanced,0.2866026759147644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,16,balanced,0.3783200184504191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,16,balanced,0.478277325630188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,1,balanced,0.15014933546384177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,1,balanced,0.24913066625595093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,1,balanced,0.44410133361816406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,1,balanced,0.8411839803059896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,1,balanced,1.613375981648763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,1,balanced,3.1740798950195312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,1,balanced,3.1806348164876304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,1,balanced,3.183354695638021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,1,balanced,3.1981067657470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,1,balanced,3.206970532735189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,1,balanced,3.1960318883260093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,1,balanced,3.212538719177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,1,balanced,3.235775947570801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,1,balanced,3.258805274963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,1,balanced,3.275247891743978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,1,balanced,3.2859039306640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,1,balanced,3.3325494130452475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,1,balanced,3.380336125691732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,1,balanced,3.4702186584472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,1,balanced,3.595264116923014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,1,balanced,3.687493324279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,1,balanced,3.8352905909220376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,1,balanced,4.209744135538737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,1,balanced,4.8048906326293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,1,balanced,4.933338801066081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,1,balanced,7.28441047668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,1,balanced,8.017632166544596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,4,power_law_1.01,0.018188799917697906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,4,power_law_1.01,0.01993599981069565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,4,power_law_1.01,0.022438399493694305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,4,power_law_1.01,0.02377600073814392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,4,power_law_1.01,0.024825599789619446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,4,power_law_1.01,0.0262719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,4,power_law_1.01,0.02869119942188263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,4,power_law_1.01,0.02858240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,4,power_law_1.01,0.03225600123405457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,4,power_law_1.01,0.031097599864006044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,4,power_law_1.01,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,4,power_law_1.01,0.041817599534988405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,4,power_law_1.01,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,4,power_law_1.01,0.03883520066738129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,4,power_law_1.01,0.058387202024459836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,4,power_law_1.01,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,4,power_law_1.01,0.06042879819869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,4,power_law_1.01,0.06632959842681885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,4,power_law_1.01,0.06931840181350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,4,power_law_1.01,0.06925439834594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,4,power_law_1.01,0.07803519964218139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,4,power_law_1.01,0.09181439876556396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,4,power_law_1.01,0.10914560556411743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,4,power_law_1.01,0.14212479591369628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,4,power_law_1.01,0.18135679960250856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,4,power_law_1.01,0.24627199172973632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,4,power_law_1.01,0.30535039901733396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,16,power_law_1.01,0.06572800278663635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,16,power_law_1.01,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,16,power_law_1.01,0.07596160173416137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,16,power_law_1.01,0.0762943983078003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,16,power_law_1.01,0.07592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,16,power_law_1.01,0.06605439782142639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,16,power_law_1.01,0.07088639736175537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,16,power_law_1.01,0.07024000287055969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,16,power_law_1.01,0.07189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,16,power_law_1.01,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,16,power_law_1.01,0.07592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,16,power_law_1.01,0.07746559977531434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,16,power_law_1.01,0.08124160170555114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,16,power_law_1.01,0.08294399976730346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,16,power_law_1.01,0.08841599822044373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,16,power_law_1.01,0.09103360176086425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,16,power_law_1.01,0.09279360175132752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,16,power_law_1.01,0.10339200496673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,16,power_law_1.01,0.10544639825820923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,16,power_law_1.01,0.12392959594726563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,16,power_law_1.01,0.1323904037475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,16,power_law_1.01,0.16891520023345946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,16,power_law_1.01,0.1836351990699768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,16,power_law_1.01,0.23878400325775145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,16,power_law_1.01,0.3108288049697876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,16,power_law_1.01,0.4305600166320801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,16,power_law_1.01,0.5697728157043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,1,power_law_1.01,0.08167679905891419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,1,power_law_1.01,0.11886719465255738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,1,power_law_1.01,0.1752575993537903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,1,power_law_1.01,0.30035200119018557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,1,power_law_1.01,0.4294271945953369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,1,power_law_1.01,0.5699520111083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,1,power_law_1.01,0.7780288219451904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,1,power_law_1.01,0.8254783630371094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,1,power_law_1.01,0.8469759941101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,1,power_law_1.01,0.8623744010925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,1,power_law_1.01,0.8893312454223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,1,power_law_1.01,0.8841279983520508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,1,power_law_1.01,0.9294143676757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,1,power_law_1.01,0.9487615585327148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,1,power_law_1.01,0.9736831665039063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,1,power_law_1.01,1.0162879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,1,power_law_1.01,1.0651968002319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,1,power_law_1.01,1.1527999877929687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,1,power_law_1.01,1.1605567932128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,1,power_law_1.01,1.3052543640136718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,1,power_law_1.01,1.2786751747131349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,1,power_law_1.01,1.4459391593933106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,1,power_law_1.01,1.6123392105102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,1,power_law_1.01,1.9999296188354492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,1,power_law_1.01,2.398854446411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,1,power_law_1.01,3.044927978515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,1,power_law_1.01,3.6635841369628905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,16,balanced,0.04971200227737427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,16,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,16,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,16,balanced,0.0813920001188914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,16,balanced,0.10400000214576721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,16,balanced,0.10874666770299275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,16,balanced,0.10941333572069804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,16,balanced,0.10930132865905762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,16,balanced,0.11035199960072835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,16,balanced,0.11215466260910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,16,balanced,0.1109333336353302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,16,balanced,0.11321600278218587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,16,balanced,0.11387200156847636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,16,balanced,0.11382399996121724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,16,balanced,0.11934933066368103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,16,balanced,0.12098133563995361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,16,balanced,0.12680533528327942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,16,balanced,0.14150399963061014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,16,balanced,0.14645866552988687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,16,balanced,0.18862400452295938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,16,balanced,0.19313599665959677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,16,balanced,0.272597332795461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,16,balanced,0.2551199992497762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,16,balanced,0.33214400211970013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,16,balanced,0.3797866503397624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,16,balanced,0.5301973422368368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,16,balanced,0.6144213279088339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,8,power_law_1.01,0.01852799952030182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,8,power_law_1.01,0.018745599687099455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,8,power_law_1.01,0.018963199853897095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,8,power_law_1.01,0.019116799533367156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,8,power_law_1.01,0.01935359984636307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,8,power_law_1.01,0.020735999941825865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,8,power_law_1.01,0.022169600427150726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,8,power_law_1.01,0.022598400712013245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,8,power_law_1.01,0.02910720109939575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,8,power_law_1.01,0.027327999472618103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,8,power_law_1.01,0.02803199887275696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,8,power_law_1.01,0.04485760033130646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,8,power_law_1.01,0.0440447986125946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,8,power_law_1.01,0.04467200040817261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,8,power_law_1.01,0.04430719912052154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,8,power_law_1.01,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,8,power_law_1.01,0.044947201013565065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,8,power_law_1.01,0.04672000110149384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,8,power_law_1.01,0.05086719989776611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,8,power_law_1.01,0.055327999591827395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,8,power_law_1.01,0.06327679753303528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,8,power_law_1.01,0.08503040075302123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,8,power_law_1.01,0.09812480211257935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,8,power_law_1.01,0.12974079847335815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,8,power_law_1.01,0.15601279735565185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,8,power_law_1.01,0.21549439430236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,8,power_law_1.01,0.294867205619812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,16,power_law_1.01,0.02111999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,16,power_law_1.01,0.02222079932689667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,16,power_law_1.01,0.022278399765491487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,16,power_law_1.01,0.02261119931936264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,16,power_law_1.01,0.02300799936056137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,16,power_law_1.01,0.022655999660491942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,16,power_law_1.01,0.024780799448490144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,16,power_law_1.01,0.02632319927215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,16,power_law_1.01,0.032153600454330446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,16,power_law_1.01,0.03253760039806366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,16,power_law_1.01,0.03219839930534363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,16,power_law_1.01,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,16,power_law_1.01,0.054028797149658206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,16,power_law_1.01,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,16,power_law_1.01,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,16,power_law_1.01,0.05534719824790955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,16,power_law_1.01,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,16,power_law_1.01,0.057811200618743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,16,power_law_1.01,0.06272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,16,power_law_1.01,0.07011839747428894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,16,power_law_1.01,0.07923200130462646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,16,power_law_1.01,0.10166399478912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,16,power_law_1.01,0.12371200323104858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,16,power_law_1.01,0.15656319856643677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,16,power_law_1.01,0.20500481128692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,16,power_law_1.01,0.27393920421600343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,16,power_law_1.01,0.3548736095428467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,128,balanced,0.08256533245245616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,128,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,128,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,128,balanced,0.043605332573254905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,128,balanced,0.04562666515509287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,128,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,128,balanced,0.04790399968624115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,128,balanced,0.04834666848182678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,128,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,128,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,128,balanced,0.04970666766166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,128,balanced,0.05171733101209005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,128,balanced,0.05165866514046987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,128,balanced,0.052426666021347046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,128,balanced,0.05760000149408976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,128,balanced,0.05588266750176748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,128,balanced,0.05991466840108236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,128,balanced,0.0643146683772405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,128,balanced,0.06619733572006226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,128,balanced,0.07635733485221863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,128,balanced,0.08471999565760295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,128,balanced,0.10543466607729594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,128,balanced,0.12781866391499838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,128,balanced,0.18083733320236206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,128,balanced,0.22323733568191528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,128,balanced,0.3025439977645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,128,balanced,0.38331735134124756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,4,power_law_1.01,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,4,power_law_1.01,0.04360319972038269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,4,power_law_1.01,0.046419200301170346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,4,power_law_1.01,0.055769598484039305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,4,power_law_1.01,0.058310401439666745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,4,power_law_1.01,0.07041919827461243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,4,power_law_1.01,0.07297279834747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,4,power_law_1.01,0.07212160229682922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,4,power_law_1.01,0.07483519911766053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,4,power_law_1.01,0.07583360075950622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,4,power_law_1.01,0.07639039754867553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,4,power_law_1.01,0.08109440207481385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,4,power_law_1.01,0.0825984001159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,4,power_law_1.01,0.08421760201454162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,4,power_law_1.01,0.09354879856109619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,4,power_law_1.01,0.09570559859275818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,4,power_law_1.01,0.10494719743728638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,4,power_law_1.01,0.11848959922790528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.01,0.12538880109786987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.01,0.14602240324020385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.01,0.16087039709091186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.01,0.2065664052963257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.01,0.25693440437316895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.01,0.3437824010848999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.01,0.4976895809173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.01,0.6743552207946777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.01,0.8949503898620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,32,balanced,0.021327999730904896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,32,balanced,0.02000533292690913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,32,balanced,0.019258666783571243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,32,balanced,0.019333332777023315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,32,balanced,0.020303999384244282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,32,balanced,0.023013333479563396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,32,balanced,0.025114665428797405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,32,balanced,0.023760000864664715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,32,balanced,0.029525332152843475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,32,balanced,0.027434666951497395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,32,balanced,0.038389332592487335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,32,balanced,0.037045332292715706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,32,balanced,0.03638399889071783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,32,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,32,balanced,0.045653333266576133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,32,balanced,0.045312002301216125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,32,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,32,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,32,balanced,0.056927998860677086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,32,balanced,0.06805866460005443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,32,balanced,0.08265066643555959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,32,balanced,0.10640533765157063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,32,balanced,0.1360160013039907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,32,balanced,0.1823093295097351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,32,balanced,0.23493866125742593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,32,balanced,0.3336106538772583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,32,balanced,0.4350186586380005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,64,power_law_1.01,0.03441919982433319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,64,power_law_1.01,0.030156800150871278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,64,power_law_1.01,0.029126399755477907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,64,power_law_1.01,0.028896000981330872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,64,power_law_1.01,0.030003198981285097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,64,power_law_1.01,0.03130879998207092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,64,power_law_1.01,0.030988800525665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,64,power_law_1.01,0.031379199028015135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,64,power_law_1.01,0.03159680068492889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,64,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,64,power_law_1.01,0.03283840119838714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,64,power_law_1.01,0.03301759958267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,64,power_law_1.01,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,64,power_law_1.01,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,64,power_law_1.01,0.03860479891300202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,64,power_law_1.01,0.0394239991903305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,64,power_law_1.01,0.03945600092411041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,64,power_law_1.01,0.041875201463699344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,64,power_law_1.01,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,64,power_law_1.01,0.05015040040016174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,64,power_law_1.01,0.05384960174560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,64,power_law_1.01,0.06287360191345215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,64,power_law_1.01,0.0710528016090393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,64,power_law_1.01,0.10450559854507446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,64,power_law_1.01,0.1344256043434143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,64,power_law_1.01,0.18904320001602173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,64,power_law_1.01,0.23462400436401368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,32,1,power_law_1.01,0.021158400177955627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,32,1,power_law_1.01,0.022361600399017335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,32,1,power_law_1.01,0.02529920041561127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,32,1,power_law_1.01,0.03206399977207184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,32,1,power_law_1.01,0.041555199027061465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,32,1,power_law_1.01,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,32,1,power_law_1.01,0.07512959837913513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,32,1,power_law_1.01,0.07809280157089234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,32,1,power_law_1.01,0.08084480166435241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,32,1,power_law_1.01,0.08458240032196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,32,1,power_law_1.01,0.08417279720306396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,32,1,power_law_1.01,0.08755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,32,1,power_law_1.01,0.09109119772911071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,32,1,power_law_1.01,0.0913856029510498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,32,1,power_law_1.01,0.09568639993667602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,32,1,power_law_1.01,0.0997759997844696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,32,1,power_law_1.01,0.10766079425811767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,32,1,power_law_1.01,0.12868479490280152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.01,0.14202239513397216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.01,0.17816959619522094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.01,0.21685760021209716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.01,0.2854399919509888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.01,0.3522239923477173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.01,0.4362495899200439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.01,0.5114880084991456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,32,1,power_law_1.01,0.6807744026184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,32,1,power_law_1.01,0.857267189025879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,32,2,power_law_1.01,0.019545599818229675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,32,2,power_law_1.01,0.021087999641895293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,32,2,power_law_1.01,0.02271360009908676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,32,2,power_law_1.01,0.027263998985290527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,32,2,power_law_1.01,0.03460479974746704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,32,2,power_law_1.01,0.03709439933300018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,32,2,power_law_1.01,0.0411327987909317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,32,2,power_law_1.01,0.04311679899692535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,32,2,power_law_1.01,0.04584319889545441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,32,2,power_law_1.01,0.048102399706840514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,32,2,power_law_1.01,0.04798719882965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,32,2,power_law_1.01,0.050758397579193114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,32,2,power_law_1.01,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,32,2,power_law_1.01,0.05267840027809143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,32,2,power_law_1.01,0.05741440057754517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,32,2,power_law_1.01,0.0602944016456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,32,2,power_law_1.01,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,32,2,power_law_1.01,0.08461440205574036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,32,2,power_law_1.01,0.09608960151672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,32,2,power_law_1.01,0.12441600561141967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,32,2,power_law_1.01,0.14402559995651246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,32,2,power_law_1.01,0.1931391954421997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,32,2,power_law_1.01,0.22045440673828126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,32,2,power_law_1.01,0.27381761074066163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,32,2,power_law_1.01,0.33626880645751955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,32,2,power_law_1.01,0.4436480045318604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,32,2,power_law_1.01,0.5797952175140381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,4,balanced,0.021370666722456615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,4,balanced,0.022101332743962605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,4,balanced,0.029167999823888142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,4,balanced,0.03879466652870178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,4,balanced,0.05444799860318502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,4,balanced,0.06330666442712148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,4,balanced,0.06529066463311513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,4,balanced,0.0671253353357315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,4,balanced,0.06836799780527751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2688,1856,6,128,1,4,balanced,0.07148266832033794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2688,1856,6,128,1,4,balanced,0.07218133409818013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2688,1856,6,128,1,4,balanced,0.07429333527882893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2688,1856,6,128,1,4,balanced,0.07659199833869934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2688,1856,6,128,1,4,balanced,0.08752000331878662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2688,1856,6,128,1,4,balanced,0.08758399883906047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,16,balanced,0.02327466756105423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,16,balanced,0.02516799916823705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,16,balanced,0.02345066765944163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,16,balanced,0.023130667706330616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,16,balanced,0.02716800073782603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,16,balanced,0.029167999823888142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,16,balanced,0.03445333242416382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,16,balanced,0.031248000760873158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,16,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,16,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,16,balanced,0.043151999513308205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,16,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,16,balanced,0.05171200136343638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,16,balanced,0.06402133405208588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,16,balanced,0.08277866741021474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,16,balanced,0.08533866206804912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,16,balanced,0.08876799543698628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,16,balanced,0.12052266796429952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,16,balanced,0.08928533395131429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,16,balanced,0.1135040024916331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,16,balanced,0.12392533818880717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,16,balanced,0.15801599621772766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,16,balanced,0.1893226703008016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,16,balanced,0.2762719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,16,balanced,0.34537065029144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,16,balanced,0.49461865425109863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,16,balanced,0.654698650042216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,32,8,power_law_1.2,0.06780800223350525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,32,8,power_law_1.2,0.07461119890213012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,32,8,power_law_1.2,0.07201279997825623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,32,8,power_law_1.2,0.07969279885292054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,32,8,power_law_1.2,0.08633599877357483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,32,8,power_law_1.2,0.08065279722213745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,32,8,power_law_1.2,0.08655359745025634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,32,8,power_law_1.2,0.08855040073394775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,32,8,power_law_1.2,0.08968960046768189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,32,8,power_law_1.2,0.08870400190353393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,32,8,power_law_1.2,0.09288319945335388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,32,8,power_law_1.2,0.09285119771957398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,32,8,power_law_1.2,0.09286400079727172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,32,8,power_law_1.2,0.0949184000492096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,32,8,power_law_1.2,0.1002303957939148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,32,8,power_law_1.2,0.10060800313949585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,32,8,power_law_1.2,0.10837119817733765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,32,8,power_law_1.2,0.1217919945716858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,32,8,power_law_1.2,0.1366976022720337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,32,8,power_law_1.2,0.15838719606399537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,32,8,power_law_1.2,0.1788607954978943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,32,8,power_law_1.2,0.23130240440368652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,32,8,power_law_1.2,0.290009593963623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,32,8,power_law_1.2,0.3924288034439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,32,8,power_law_1.2,0.48757119178771974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,32,8,power_law_1.2,0.695251178741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,32,8,power_law_1.2,0.9510975837707519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,4,power_law_1.01,0.021484799683094025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,4,power_law_1.01,0.02454400062561035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,4,power_law_1.01,0.029260799288749695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,4,power_law_1.01,0.03080959916114807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,4,power_law_1.01,0.041145598888397215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,4,power_law_1.01,0.04301440119743347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,4,power_law_1.01,0.04679679870605469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,4,power_law_1.01,0.04734080135822296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,4,power_law_1.01,0.05223039984703064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,4,power_law_1.01,0.05007359981536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,4,power_law_1.01,0.04999040067195892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,4,power_law_1.01,0.06232960224151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,4,power_law_1.01,0.05999360084533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,4,power_law_1.01,0.05608959794044495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,4,power_law_1.01,0.08613119721412658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,4,power_law_1.01,0.08928639888763427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,4,power_law_1.01,0.10165120363235473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,4,power_law_1.01,0.09252480268478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.01,0.09497600197792053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.01,0.10562560558319092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.01,0.11308159828186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.01,0.13326719999313355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.01,0.16064000129699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.01,0.2048896074295044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,4,power_law_1.01,0.07067520022392274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.01,0.24942080974578856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,4,power_law_1.01,0.09102079868316651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.01,0.3609407901763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,4,power_law_1.01,0.08849920034408569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,4,power_law_1.01,0.10477440357208252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.01,0.45186557769775393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,4,power_law_1.01,0.12239999771118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,4,power_law_1.01,0.14608000516891478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,4,power_law_1.01,0.14707839488983154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,4,power_law_1.01,0.14540799856185913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,4,power_law_1.01,0.14867199659347535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,4,power_law_1.01,0.14819200038909913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,4,power_law_1.01,0.15642240047454833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,4,power_law_1.01,0.1637887954711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,4,power_law_1.01,0.16689280271530152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,4,power_law_1.01,0.168230402469635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,4,power_law_1.01,0.18223999738693236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,4,power_law_1.01,0.18973439931869507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,4,power_law_1.01,0.19928959608078003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,4,power_law_1.01,0.23240320682525634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,4,power_law_1.01,0.2540607929229736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,4,power_law_1.01,0.3208575963973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,4,power_law_1.01,0.3835968017578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,4,power_law_1.01,0.4960319995880127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,4,power_law_1.01,0.5870463848114014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,4,power_law_1.01,0.7734464168548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,4,power_law_1.01,0.9446975708007812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,4,power_law_1.01,1.297612762451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,4,power_law_1.01,1.7027711868286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,2,power_law_1.01,0.04460160136222839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,2,power_law_1.01,0.047635200619697574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,2,power_law_1.01,0.05063679814338684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,2,power_law_1.01,0.05704960227012634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,2,power_law_1.01,0.062457597255706786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,2,power_law_1.01,0.07605119943618774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,2,power_law_1.01,0.07698559761047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,2,power_law_1.01,0.07927680015563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,2,power_law_1.01,0.08085119724273682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,2,power_law_1.01,0.08195840120315552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,2,power_law_1.01,0.08360959887504578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,2,power_law_1.01,0.08854399919509888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,2,power_law_1.01,0.08832640051841736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,2,power_law_1.01,0.0932416021823883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,2,power_law_1.01,0.09876480102539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,2,power_law_1.01,0.10426239967346192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,2,power_law_1.01,0.11510399580001832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,2,power_law_1.01,0.13286399841308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,2,power_law_1.01,0.14871679544448851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,2,power_law_1.01,0.18329600095748902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,2,power_law_1.01,0.21809918880462648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,2,power_law_1.01,0.290009593963623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,2,power_law_1.01,0.38280959129333497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,2,power_law_1.01,0.5398655891418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,2,power_law_1.01,0.7250624179840088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,2,power_law_1.01,0.9955391883850098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,2,power_law_1.01,1.257376003265381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,1,power_law_1.01,0.09241600036621093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,1,power_law_1.01,0.10687359571456909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,1,power_law_1.01,0.13282560110092162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,1,power_law_1.01,0.18684159517288207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,1,power_law_1.01,0.2270143985748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,1,power_law_1.01,0.26878719329833983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,1,power_law_1.01,0.3423424005508423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,1,power_law_1.01,0.34167680740356443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,1,power_law_1.01,0.3608128070831299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,1,power_law_1.01,0.36360321044921873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,1,power_law_1.01,0.3687551975250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,1,power_law_1.01,0.38327679634094236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,1,power_law_1.01,0.38209919929504393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,1,power_law_1.01,0.39091200828552247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,1,power_law_1.01,0.39434239864349363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,1,power_law_1.01,0.41023998260498046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,1,power_law_1.01,0.4252352237701416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,1,power_law_1.01,0.46604161262512206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,2,power_law_1.2,0.08714879751205444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,1,power_law_1.01,0.5101439952850342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,2,power_law_1.2,0.09719039797782898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,2,power_law_1.2,0.11553280353546143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,2,power_law_1.2,0.13249919414520264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,2,power_law_1.2,0.14790400266647338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,2,power_law_1.2,0.16381440162658692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,1,power_law_1.01,0.5934271812438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,2,power_law_1.2,0.1865280032157898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,1,power_law_1.01,0.6843776226043701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,2,power_law_1.2,0.18831360340118408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,1,power_law_1.01,0.8456064224243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,2,power_law_1.2,0.1903872013092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,2,power_law_1.2,0.19552639722824097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,2,power_law_1.2,0.2016511917114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,2,power_law_1.2,0.2023103952407837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,2,power_law_1.2,0.20989439487457276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,2,power_law_1.2,0.22057600021362306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,1,power_law_1.01,1.0044416427612304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,2,power_law_1.2,0.22398080825805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,2,power_law_1.2,0.22933759689331054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,1,power_law_1.01,1.3371647834777831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,2,power_law_1.2,0.2451456069946289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,2,power_law_1.2,0.28264319896698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,1,power_law_1.01,1.661996841430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,2,power_law_1.2,0.3184639930725098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,2,power_law_1.2,0.39492480754852294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,1,power_law_1.01,2.303104019165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,2,power_law_1.2,0.47718400955200196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,2,power_law_1.2,0.6320703983306885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,1,power_law_1.01,2.9375295639038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,2,power_law_1.2,0.7650239944458008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,2,power_law_1.2,1.1093695640563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,2,power_law_1.2,1.371820831298828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,2,power_law_1.2,1.9919872283935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,2,power_law_1.2,2.687545585632324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,2,power_law_1.2,0.07708799839019775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,2,power_law_1.2,0.08860160112380981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,2,power_law_1.2,0.10126719474792481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,2,power_law_1.2,0.1327679991722107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,2,power_law_1.2,0.15563520193099975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,2,power_law_1.2,0.21979520320892335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,2,power_law_1.2,0.2331007957458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,2,power_law_1.2,0.23751039505004884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,2,power_law_1.2,0.24122240543365478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,2,power_law_1.2,0.2458048105239868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,2,power_law_1.2,0.2517184019088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,2,power_law_1.2,0.2635200023651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,2,power_law_1.2,0.2732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,2,power_law_1.2,0.2846719980239868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,2,power_law_1.2,0.2964864015579224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,2,power_law_1.2,0.3062335968017578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,2,power_law_1.2,0.3223167896270752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,2,power_law_1.2,0.39105920791625975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,2,power_law_1.2,0.4156991958618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,2,power_law_1.2,0.5167424201965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,2,power_law_1.2,0.6263423919677734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,2,power_law_1.2,0.7747327804565429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,2,power_law_1.2,0.9144255638122558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,2,power_law_1.2,1.2477952003479005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,2,power_law_1.2,1.5490559577941894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,2,power_law_1.2,2.094976043701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,2,power_law_1.2,2.7367551803588865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,32,power_law_1.01,0.0604095995426178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,32,power_law_1.01,0.04635519981384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,32,power_law_1.01,0.046393600106239316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,32,power_law_1.01,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,32,power_law_1.01,0.04678399860858917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,32,power_law_1.01,0.04693120121955872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,32,power_law_1.01,0.04755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,32,power_law_1.01,0.04848000109195709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,32,power_law_1.01,0.04893440008163452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,32,power_law_1.01,0.04907520115375519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,32,power_law_1.01,0.05022720098495483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,32,power_law_1.01,0.05242879986763001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,32,power_law_1.01,0.05240319967269898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,32,power_law_1.01,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,32,power_law_1.01,0.05990399718284607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,32,power_law_1.01,0.0643392026424408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,32,power_law_1.01,0.06815360188484192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,32,power_law_1.01,0.07681279778480529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,32,power_law_1.01,0.08262400031089782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,32,power_law_1.01,0.09948800206184387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,32,power_law_1.01,0.11255680322647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,32,power_law_1.01,0.14139519929885863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,32,power_law_1.01,0.17025279998779297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,32,power_law_1.01,0.22218880653381348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,32,power_law_1.01,0.2886336088180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,32,power_law_1.01,0.4123199939727783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,32,power_law_1.01,0.5131840229034423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,128,power_law_1.01,0.0322816014289856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,128,power_law_1.01,0.032204800844192506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,128,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,128,power_law_1.01,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,128,power_law_1.01,0.0348800003528595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,128,power_law_1.01,0.034860798716545106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,128,power_law_1.01,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,128,power_law_1.01,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,128,power_law_1.01,0.03506560027599335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,128,power_law_1.01,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,128,power_law_1.01,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,128,power_law_1.01,0.0367680013179779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,128,power_law_1.01,0.03761920034885406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,128,power_law_1.01,0.03729279935359955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,128,power_law_1.01,0.03993600010871887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,128,power_law_1.01,0.04059520065784454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,128,power_law_1.01,0.04022400081157684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,128,power_law_1.01,0.04172160029411316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,128,power_law_1.01,0.04293760061264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,128,power_law_1.01,0.046393600106239316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,128,power_law_1.01,0.049158400297164916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,128,power_law_1.01,0.05813760161399841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,128,power_law_1.01,0.06925439834594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,128,power_law_1.01,0.08580480217933655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,128,power_law_1.01,0.10860799551010132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,128,power_law_1.01,0.15092480182647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,128,power_law_1.01,0.1854591965675354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,8,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,8,balanced,0.05286400020122528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,8,balanced,0.07262933254241943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,8,balanced,0.09782933195432027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,8,balanced,0.15025066335995993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,8,balanced,0.24665600061416626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,8,balanced,0.24589866399765015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,8,balanced,0.24698134263356528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,8,balanced,0.24765866994857788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,8,balanced,0.25044800837834674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,8,balanced,0.25011199712753296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,8,balanced,0.2548373341560364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,8,balanced,0.2550293405850728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,8,balanced,0.25738133986790973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,8,balanced,0.25982399781545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,8,balanced,0.26316799720128375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,8,balanced,0.26689600944519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,8,balanced,0.28039999802907306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,8,balanced,0.2869493365287781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,8,balanced,0.3078773419062297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,8,balanced,0.32952000697453815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,8,balanced,0.35917866230010986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,8,balanced,0.3944106499354045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,8,balanced,0.4926240046819051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,8,balanced,0.5483359893163046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,8,balanced,0.7672106424967448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,8,balanced,0.8648053010304769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,16,balanced,0.04258666435877482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,16,balanced,0.04359466830889384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,16,balanced,0.044309332966804504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,16,balanced,0.04797333478927612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,16,balanced,0.04960533479849497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,16,balanced,0.05855466425418854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,16,balanced,0.05850133299827576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,16,balanced,0.05925333499908447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,16,balanced,0.06048533320426941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,16,balanced,0.06217066446940104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,16,balanced,0.06261866788069408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,16,balanced,0.06417599817117055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,16,balanced,0.06659199794133504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,16,balanced,0.07071466743946075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,16,balanced,0.07707199951012929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,16,balanced,0.07913599908351898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,16,balanced,0.08487466971079509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,16,balanced,0.09884799520174663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,16,balanced,0.10780800382296245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,16,balanced,0.127509335676829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,16,balanced,0.14428800344467163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,16,balanced,0.1869279940923055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,16,balanced,0.22568533817927042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,16,balanced,0.30369067192077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,16,balanced,0.3874400059382121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,16,balanced,0.5527199904123942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,16,balanced,0.6939146518707275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,1,balanced,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,1,balanced,0.03955733279387156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,1,balanced,0.055488000313440956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,1,balanced,0.08451199531555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,1,balanced,0.12941333651542664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,1,balanced,0.13061867157618204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,1,balanced,0.13338133692741394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,1,balanced,0.1341653366883596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,1,balanced,0.13614400227864584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,1,balanced,0.13780799508094788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,1,balanced,0.140474667151769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,1,balanced,0.14215466380119324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,1,balanced,0.14667733510335287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,1,balanced,0.15033066272735596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,1,balanced,0.15756266315778097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,1,balanced,0.16409066319465637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,1,balanced,0.1722773313522339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,1,balanced,0.19504533211390176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,1,balanced,0.21827733516693115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,1,balanced,0.2688159942626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,1,balanced,0.31249066193898517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,1,balanced,0.4155413309733073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,1,balanced,0.5083946784337362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,1,balanced,0.7284320195515951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,1,balanced,0.9215093453725179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,1,balanced,1.3401439984639485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,1,balanced,1.7813013394673665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,1,power_law_1.2,0.15397759675979614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,1,power_law_1.2,0.24789121150970458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,1,power_law_1.2,0.4374080181121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,1,power_law_1.2,0.7652736186981202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,1,power_law_1.2,1.101638412475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,1,power_law_1.2,1.471174430847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,1,power_law_1.2,2.2521024703979493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,1,power_law_1.2,2.350092887878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,1,power_law_1.2,2.5411712646484377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,1,power_law_1.2,2.5146047592163088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,1,power_law_1.2,2.625823974609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,1,power_law_1.2,2.7416000366210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,1,power_law_1.2,2.7291519165039064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,1,power_law_1.2,2.8297216415405275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,1,power_law_1.2,2.9191680908203126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,1,power_law_1.2,3.0527807235717774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,1,power_law_1.2,3.281804656982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,1,power_law_1.2,3.4532928466796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,1,power_law_1.2,3.6257598876953123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,1,power_law_1.2,4.112627029418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,1,power_law_1.2,3.9171070098876952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,1,power_law_1.2,4.3733055114746096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,1,power_law_1.2,4.828646469116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,1,power_law_1.2,5.818137741088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,1,power_law_1.2,6.847673797607422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,1,power_law_1.2,9.079078674316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,32,power_law_1.01,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,32,power_law_1.01,0.034380799531936644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,32,power_law_1.01,0.03160319924354553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,32,power_law_1.01,0.03132160007953644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,32,power_law_1.01,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,32,power_law_1.01,0.03229439854621887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,32,power_law_1.01,0.03243519961833954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,32,power_law_1.01,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,1,power_law_1.2,10.785906982421874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,32,power_law_1.01,0.0331712007522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,32,power_law_1.01,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,32,power_law_1.01,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,32,power_law_1.01,0.036051198840141296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,32,power_law_1.01,0.03660799860954285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,32,power_law_1.01,0.03780480027198792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,32,power_law_1.01,0.040627199411392215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,32,power_law_1.01,0.04196479916572571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,32,power_law_1.01,0.044819200038909913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,32,power_law_1.01,0.050969600677490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,32,power_law_1.01,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,32,power_law_1.01,0.05980799794197082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,32,power_law_1.01,0.06680319905281067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,32,power_law_1.01,0.07859839797019959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,32,power_law_1.01,0.10173439979553223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,32,power_law_1.01,0.13976320028305053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,32,power_law_1.01,0.17960959672927856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,32,power_law_1.01,0.25886719226837157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,32,power_law_1.01,0.3053375959396362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,1,power_law_1.01,0.08155519962310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,1,power_law_1.01,0.0856000006198883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,1,power_law_1.01,0.09463679790496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,1,power_law_1.01,0.11223679780960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,1,power_law_1.01,0.1347584009170532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,1,power_law_1.01,0.15119999647140503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,1,power_law_1.01,0.18276480436325074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,1,power_law_1.01,0.19245439767837524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,1,power_law_1.01,0.19104000329971313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,1,power_law_1.01,0.19717760086059571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,1,power_law_1.01,0.20360960960388183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,1,power_law_1.01,0.2107327938079834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,1,power_law_1.01,0.2152575969696045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,1,power_law_1.01,0.2268224000930786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,1,power_law_1.01,0.23134078979492187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,1,power_law_1.01,0.244268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,1,power_law_1.01,0.2601664066314697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,1,power_law_1.01,0.29345920085906985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,1,power_law_1.01,0.3323647975921631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,1,power_law_1.01,0.4095104217529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,1,power_law_1.01,0.47640318870544435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,1,power_law_1.01,0.6213183879852295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,1,power_law_1.01,0.7692671775817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,1,power_law_1.01,1.0695167541503907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,1,power_law_1.01,1.3676992416381837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,1,power_law_1.01,1.9642816543579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,1,power_law_1.01,2.5435455322265623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,128,power_law_1.01,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,128,power_law_1.01,0.052985602617263795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,128,power_law_1.01,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,128,power_law_1.01,0.05583999752998352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,128,power_law_1.01,0.05745919942855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,128,power_law_1.01,0.059084802865982056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,128,power_law_1.01,0.05959039926528931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,128,power_law_1.01,0.058880001306533813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,128,power_law_1.01,0.060313600301742556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,128,power_law_1.01,0.06247040033340454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,128,power_law_1.01,0.06392959952354431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,128,power_law_1.01,0.074726402759552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,128,power_law_1.01,0.07630720138549804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,128,power_law_1.01,0.07085440158843995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,128,power_law_1.01,0.08325759768486023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,128,power_law_1.01,0.08807680010795593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,128,power_law_1.01,0.0889855980873108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,128,power_law_1.01,0.10469759702682495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,128,power_law_1.01,0.11386879682540893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,128,power_law_1.01,0.13849600553512573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,128,power_law_1.01,0.15971839427947998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,128,power_law_1.01,0.2059391975402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,128,power_law_1.01,0.24198400974273682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,128,power_law_1.01,0.3310528039932251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,128,power_law_1.01,0.4259647846221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,128,power_law_1.01,0.6055871963500976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,128,power_law_1.01,0.8333632469177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,1,power_law_1.2,0.03324800133705139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,1,power_law_1.2,0.03227519989013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,1,power_law_1.2,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,1,power_law_1.2,0.039468801021575926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,1,power_law_1.2,0.047916799783706665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,1,power_law_1.2,0.071424001455307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,1,power_law_1.2,0.06314240097999572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,1,power_law_1.2,0.07138559818267823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,1,power_law_1.2,0.06142079830169678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,1,power_law_1.2,0.06667519807815551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,1,power_law_1.2,0.07574399709701538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,1,power_law_1.2,0.07144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,1,power_law_1.2,0.07615360021591186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,1,power_law_1.2,0.08836479783058167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,1,power_law_1.2,0.17543679475784302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,1,power_law_1.2,0.20690560340881348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,1,power_law_1.2,0.25680639743804934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,1,power_law_1.2,0.13774720430374146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.2,0.16308480501174927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,1,power_law_1.2,0.017868800461292265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.2,0.2307391881942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,1,power_law_1.2,0.018572799861431122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,1,power_law_1.2,0.021536000072956085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,1,power_law_1.2,0.029164800047874452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,1,power_law_1.2,0.034016001224517825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,1,power_law_1.2,0.04353919923305512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,1,power_law_1.2,0.053273600339889524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,1,power_law_1.2,0.055270397663116456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,1,power_law_1.2,0.058316802978515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,1,power_law_1.2,0.059513598680496216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,1,power_law_1.2,0.06199679970741272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,1,power_law_1.2,0.06332799792289734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,1,power_law_1.2,0.06553599834442139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,1,power_law_1.2,0.06787199974060058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,1,power_law_1.2,0.06977279782295227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,1,power_law_1.2,0.07162879705429077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,1,power_law_1.2,0.0765824019908905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,1,power_law_1.2,0.08557440042495727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,1,power_law_1.2,0.09578239917755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,1,power_law_1.2,0.11973119974136352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,1,power_law_1.2,0.1410688042640686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,1,power_law_1.2,0.1847488045692444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,1,power_law_1.2,0.20174078941345214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,1,power_law_1.2,0.2663615942001343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.2,0.2942784070968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,1,power_law_1.2,0.34776959419250486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,1,power_law_1.2,0.47511677742004393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,1,power_law_1.2,0.611513614654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.2,0.41889281272888185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.2,0.5296703815460205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,128,power_law_1.2,0.034246399998664856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,128,power_law_1.2,0.03462400138378143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,128,power_law_1.2,0.03379839956760407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,128,power_law_1.2,0.03644160032272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,128,power_law_1.2,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.2,0.7813183784484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,128,power_law_1.2,0.03814400136470795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,128,power_law_1.2,0.03777920007705689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,128,power_law_1.2,0.03765760064125061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,128,power_law_1.2,0.03827199935913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,128,power_law_1.2,0.0380160003900528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,128,power_law_1.2,0.038726401329040525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,128,power_law_1.2,0.03929600119590759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,128,power_law_1.2,0.03948160111904144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,32,1,balanced,0.05614933371543884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,128,power_law_1.2,0.04056960046291351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,32,1,balanced,0.060080001751581825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,32,1,balanced,0.06429333488146464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,128,power_law_1.2,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,32,1,balanced,0.08880533774693807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,32,1,balanced,0.1281546652317047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,128,power_law_1.2,0.04474239945411682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,32,1,balanced,0.1943733294804891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,32,1,balanced,0.19656533002853394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,32,1,balanced,0.19995200634002686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,32,1,balanced,0.20129066705703735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,32,1,balanced,0.2035413384437561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,32,1,balanced,0.20472000042597452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,32,1,balanced,0.2098026672999064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,32,1,balanced,0.20975999037424722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,32,1,balanced,0.2153653303782145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,32,1,balanced,0.2217493255933126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,32,1,balanced,0.22616000970204672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,32,1,balanced,0.236026664574941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,32,1,balanced,0.27583465973536175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,32,1,balanced,0.2961600025494893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,32,1,balanced,0.36368000507354736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,128,power_law_1.2,0.04688639938831329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,32,1,balanced,0.42763733863830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,128,power_law_1.2,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,32,1,balanced,0.6001173257827759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.2,0.056518399715423585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,32,1,balanced,0.6721546649932861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.2,0.06498559713363647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.2,0.07417600154876709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.2,0.09960960149765015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,32,1,balanced,1.000111977259318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,32,1,balanced,1.2492960294087727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,32,1,balanced,1.8007946014404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.2,0.12803839445114135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,32,1,balanced,2.338805357615153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.2,0.15982719659805297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.2,0.21281919479370118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.2,1.020595169067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.2,0.30617599487304686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,2,power_law_1.01,0.051718401908874514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,2,power_law_1.01,0.05480319857597351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,2,power_law_1.01,0.07011839747428894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,2,power_law_1.01,0.0898751974105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,2,power_law_1.01,0.10240000486373901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.2,1.5142656326293946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,2,power_law_1.01,0.11652480363845825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,2,power_law_1.01,0.14877439737319947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,2,power_law_1.01,0.1487231969833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,2,power_law_1.01,0.15324800014495848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,2,power_law_1.01,0.16024320125579833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,2,power_law_1.01,0.1645248055458069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,2,power_law_1.01,0.17756799459457398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,2,power_law_1.01,0.1744704008102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,2,power_law_1.01,0.1856384038925171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,2,power_law_1.01,0.19726719856262206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,2,power_law_1.01,0.2054208040237427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,2,power_law_1.01,0.2275007963180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,2,power_law_1.01,0.2630975961685181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,2,power_law_1.01,0.2692032098770142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,2,power_law_1.01,0.33280000686645506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,2,power_law_1.01,0.3788928031921387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,2,power_law_1.01,0.46124801635742185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,2,power_law_1.01,0.5944384098052978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,2,power_law_1.01,0.7716544151306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,2,power_law_1.01,0.9848768234252929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,2,power_law_1.01,1.3312512397766114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,2,power_law_1.01,1.726041603088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.2,0.4041920185089111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.2,2.003980827331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,128,power_law_1.01,0.08283519744873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,128,power_law_1.01,0.08143360018730164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,128,power_law_1.01,0.08229759931564332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,128,power_law_1.01,0.08524159789085388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,128,power_law_1.01,0.08780800104141236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,128,power_law_1.01,0.09431040287017822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,128,power_law_1.01,0.09287679791450501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,128,power_law_1.01,0.08858240246772767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,128,power_law_1.01,0.09037439823150635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,128,power_law_1.01,0.09260159730911255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,128,power_law_1.01,0.09105280041694641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,128,power_law_1.01,0.09006720185279846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,128,power_law_1.01,0.09429759979248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,128,power_law_1.01,0.09107199907302857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,128,power_law_1.01,0.0942911982536316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,128,power_law_1.01,0.09525120258331299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,128,power_law_1.01,0.09639040231704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,128,power_law_1.01,0.10897279977798462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,128,power_law_1.01,0.1153216004371643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,128,power_law_1.01,0.13957760334014893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,128,power_law_1.01,0.15921920537948608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,128,power_law_1.01,0.1958016037940979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,128,power_law_1.01,0.2540607929229736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,128,power_law_1.01,0.337939190864563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,128,power_law_1.01,0.40573439598083494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,128,power_law_1.01,0.5628736019134521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,128,power_law_1.01,0.7544320106506348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,4,balanced,0.03790933390458425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,4,balanced,0.03884266565243403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,4,balanced,0.04173333446184794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,4,balanced,0.047744000951449074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,4,balanced,0.05916800101598104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,4,balanced,0.05826666454474131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,4,balanced,0.058245331048965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,4,balanced,0.05867200096448263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,4,balanced,0.05829866727193197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,4,balanced,0.05807466804981232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,4,balanced,0.059578667084376015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,4,balanced,0.060309335589408875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,4,balanced,0.062319998939832054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,4,balanced,0.06222933530807495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,4,balanced,0.06618133187294006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,4,balanced,0.06842666864395142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,4,balanced,0.07040533423423767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,4,balanced,0.07508266468842824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,4,balanced,0.0786186655362447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,4,balanced,0.09196799993515015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,4,balanced,0.10116799672444661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,4,balanced,0.1366986632347107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,4,balanced,0.1611733337243398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,4,balanced,0.2205866575241089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,4,balanced,0.28005866209665936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,4,balanced,0.39179734388987225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,4,balanced,0.505402684211731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,2,power_law_1.01,0.1010815978050232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,2,power_law_1.01,0.1327296018600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,2,power_law_1.01,0.13560960292816163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,2,power_law_1.01,0.18191360235214232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,2,power_law_1.01,0.1878592014312744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,2,power_law_1.01,0.18776320219039916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,2,power_law_1.01,0.1918720006942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,2,power_law_1.01,0.18379520177841185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,2,power_law_1.01,0.18683520555496216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,2,power_law_1.01,0.18197760581970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,2,power_law_1.01,0.1778496026992798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,2,power_law_1.01,0.18406399488449096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,2,power_law_1.01,0.177510404586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,2,power_law_1.01,0.20232319831848145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,2,power_law_1.01,0.19981440305709838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,2,power_law_1.01,0.22293760776519775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,2,power_law_1.01,0.25421440601348877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,2,power_law_1.2,0.07379199862480164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,2,power_law_1.2,0.09708799719810486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,2,power_law_1.01,0.2851968050003052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,2,power_law_1.2,0.10595840215682983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.01,0.35269761085510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,2,power_law_1.2,0.135206401348114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,2,power_law_1.2,0.1628224015235901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,2,power_law_1.2,0.19852800369262696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.01,0.46640639305114745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,2,power_law_1.2,0.26445438861846926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,2,power_law_1.2,0.27012479305267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,2,power_law_1.2,0.2752511978149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,2,power_law_1.2,0.2685120105743408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,2,power_law_1.2,0.2925184011459351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,2,power_law_1.2,0.29831039905548096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,2,power_law_1.2,0.3014911890029907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,2,power_law_1.2,0.3036799907684326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,2,power_law_1.2,0.3304703950881958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,2,power_law_1.2,0.343507194519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,2,power_law_1.2,0.37306880950927734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.01,0.5011648178100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,2,power_law_1.2,0.4175424098968506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.2,0.45725440979003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.2,0.5532991886138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.2,0.5454527854919433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.2,0.7140607833862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.2,0.8273407936096191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.2,1.1249792098999023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.2,1.3965439796447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.2,2.047750473022461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.01,0.7217343807220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.2,2.494246482849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.01,0.9971967697143554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.01,1.367302417755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,1,power_law_1.01,0.054201602935791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,1,power_law_1.01,0.057708799839019775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,1,power_law_1.01,0.06535680294036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,1,power_law_1.01,0.07736319899559022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,1,power_law_1.01,0.08771839737892151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,1,power_law_1.01,0.10922880172729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.01,1.936582374572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,1,power_law_1.01,0.11180800199508667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,1,power_law_1.01,0.11422079801559448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,1,power_law_1.01,0.11585279703140258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,1,power_law_1.01,0.11713279485702514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,1,power_law_1.01,0.12302080392837525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,1,power_law_1.01,0.12311040163040161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,1,power_law_1.01,0.12578560113906861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,1,power_law_1.01,0.13292800188064574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,1,power_law_1.01,0.1376960039138794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,1,power_law_1.01,0.1441472053527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,1,power_law_1.01,0.15756800174713134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,1,power_law_1.01,0.1833024024963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,1,power_law_1.01,0.21351680755615235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,1,power_law_1.01,0.26766080856323243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,1,power_law_1.01,0.31731839179992677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,1,power_law_1.01,0.432038402557373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,1,power_law_1.01,0.5402688026428223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,1,power_law_1.01,0.7639423847198487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,1,power_law_1.01,0.9858752250671386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,1,power_law_1.01,1.4287296295166017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,1,power_law_1.01,1.860793685913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,8,balanced,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,8,balanced,0.044735997915267944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,8,balanced,0.05235733091831207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,8,balanced,0.06398400167624156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,8,balanced,0.08786666393280029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,8,balanced,0.1442506710688273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,8,balanced,0.15689599514007568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,8,balanced,0.15106133619944254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,8,balanced,0.15872533122698465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,8,balanced,0.16934933265050253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,8,balanced,0.17918932437896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,8,balanced,0.1918933391571045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,8,balanced,0.20561599731445312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,8,balanced,0.20220265785853067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,8,balanced,0.22107199827829996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,8,balanced,0.23833600680033365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,8,balanced,0.2493333419164022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,8,balanced,0.2645919919013977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,8,balanced,0.33189332485198975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,8,balanced,0.3649653196334839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,8,balanced,0.41650664806365967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,8,balanced,0.4840799967447917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,8,balanced,0.5598453283309937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,8,balanced,0.7065066496531168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,8,balanced,0.8578293323516846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,8,balanced,1.35588804880778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,8,balanced,1.6441067059834797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.01,2.3300352096557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,1,power_law_1.2,0.09242879748344421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,1,power_law_1.2,0.10764800310134888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,1,power_law_1.2,0.13426560163497925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,1,power_law_1.2,0.17910399436950683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,1,power_law_1.2,0.2182080030441284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,1,power_law_1.2,0.25445120334625243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,1,power_law_1.2,0.33943679332733157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,1,power_law_1.2,0.3523904085159302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,1,power_law_1.2,0.35571200847625734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,1,power_law_1.2,0.35754880905151365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,1,power_law_1.2,0.3645695924758911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,1,power_law_1.2,0.3831935882568359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,1,power_law_1.2,0.3868288040161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,1,power_law_1.2,0.40556797981262205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,1,power_law_1.2,0.4088768005371094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,1,power_law_1.2,0.4133632183074951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,1,power_law_1.2,0.4299776077270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,1,power_law_1.2,0.4908927917480469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,1,power_law_1.2,0.5287615776062011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,1,power_law_1.2,0.6184576034545899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,1,power_law_1.2,0.6983232021331787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,1,power_law_1.2,0.857868766784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,1,power_law_1.2,1.0134079933166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,1,power_law_1.2,1.3386431694030763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,1,power_law_1.2,1.6588415145874023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,1,power_law_1.2,2.2855552673339843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,1,power_law_1.2,2.917094421386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.01,3.4457729339599608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,4,power_law_1.2,0.05567359924316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,4,power_law_1.2,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,4,power_law_1.2,0.058323198556900026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,4,power_law_1.2,0.06901119947433472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,4,power_law_1.2,0.07191039919853211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,4,power_law_1.2,0.08256000280380249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,4,power_law_1.2,0.08558719754219055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,4,power_law_1.2,0.0890175998210907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,4,power_law_1.2,0.08384640216827392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,4,power_law_1.2,0.08743680119514466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,4,power_law_1.2,0.08611199855804444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,4,power_law_1.2,0.08702719807624817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,4,power_law_1.2,0.08826239705085755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,4,power_law_1.2,0.09078400135040283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,4,power_law_1.2,0.09767040014266967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,4,power_law_1.2,0.09873920083045959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,4,power_law_1.2,0.10524159669876099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,4,power_law_1.2,0.11661440134048462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.2,0.13434879779815673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.2,0.15720319747924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.2,0.18472959995269775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.2,0.2537600040435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.2,0.3276160001754761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.2,0.43557119369506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.2,0.48974080085754396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.2,0.8075072288513183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,8,power_law_1.2,0.03475199937820435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,8,power_law_1.2,0.035641598701477054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,8,power_law_1.2,0.04261119961738587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,8,power_law_1.2,0.04537599980831146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,8,power_law_1.2,0.054073601961135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,8,power_law_1.2,0.05134720206260681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,8,power_law_1.2,0.05170559883117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,8,power_law_1.2,0.05247359871864319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,8,power_law_1.2,0.05358719825744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,8,power_law_1.2,0.054502397775650024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,8,power_law_1.2,0.05447040200233459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,8,power_law_1.2,0.05603839755058289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,8,power_law_1.2,0.058233600854873654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,8,power_law_1.2,0.05982080101966858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,8,power_law_1.2,0.06583679914474487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,8,power_law_1.2,0.06710399985313416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,8,power_law_1.2,0.06849920153617858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,8,power_law_1.2,0.07271680235862732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,8,power_law_1.2,0.07568640112876893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,8,power_law_1.2,0.08817920088768005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,8,power_law_1.2,0.09548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,8,power_law_1.2,0.10959359407424926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,8,power_law_1.2,0.1344704031944275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,8,power_law_1.2,0.16352640390396117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,8,power_law_1.2,0.21921279430389404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,8,power_law_1.2,0.2704960107803345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.2,1.1292927742004395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,8,power_law_1.2,0.34420480728149416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,8,power_law_1.01,0.06857600212097167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,8,power_law_1.01,0.0773311972618103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,8,power_law_1.01,0.08331519961357117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,8,power_law_1.01,0.08577280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,8,power_law_1.01,0.0892416000366211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,8,power_law_1.01,0.08948479890823365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,8,power_law_1.01,0.09349759817123413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,8,power_law_1.01,0.08650239706039428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,8,power_law_1.01,0.09110400080680847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,8,power_law_1.01,0.09300479888916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,8,power_law_1.01,0.0937279999256134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,8,power_law_1.01,0.09777920246124268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,8,power_law_1.01,0.09448320269584656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,8,power_law_1.01,0.09879040122032165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,8,power_law_1.01,0.10325119495391846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,8,power_law_1.01,0.10556800365447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,8,power_law_1.01,0.11143679618835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,8,power_law_1.01,0.12331520318984986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,8,power_law_1.01,0.13738880157470704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,8,power_law_1.01,0.16462719440460205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,8,power_law_1.01,0.18343679904937743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,8,power_law_1.01,0.23496320247650146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,8,power_law_1.01,0.28012158870697024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,8,power_law_1.01,0.3781440019607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,8,power_law_1.01,0.4751296043395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,8,power_law_1.01,0.6860159873962403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,8,power_law_1.01,0.9131967544555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,1,power_law_1.2,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,1,power_law_1.2,0.047167998552322385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,1,power_law_1.2,0.04755200147628784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,1,power_law_1.2,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,1,power_law_1.2,0.0656000018119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,1,power_law_1.2,0.06628479957580566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,1,power_law_1.2,0.06689919829368592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,1,power_law_1.2,0.06762880086898804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,1,power_law_1.2,0.06807680130004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,1,power_law_1.2,0.06748160123825073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,1,power_law_1.2,0.06873599886894226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,1,power_law_1.2,0.07239680290222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,1,power_law_1.2,0.07374079823493958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,1,power_law_1.2,0.08081279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,1,power_law_1.2,0.0965503990650177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,1,power_law_1.2,0.0975488007068634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,1,power_law_1.2,0.10078719854354859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,1,power_law_1.2,0.11550079584121704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.2,0.13880959749221802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.2,0.17929600477218627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.2,0.21235840320587157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.2,0.2915136098861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.2,0.3915839910507202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.2,0.5596479892730712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.2,0.7137983798980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.2,1.0381248474121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,16,balanced,0.025370667378107708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,16,balanced,0.026741333305835724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,16,balanced,0.026389333109060924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,16,balanced,0.027552001178264618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,16,balanced,0.03160000095764796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,16,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,16,balanced,0.037733333806196846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,16,balanced,0.036474667489528656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,16,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,16,balanced,0.03730133424202601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,16,balanced,0.03860799968242645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,16,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,16,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,16,balanced,0.03791466603676478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,16,balanced,0.039674667020638786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,16,balanced,0.04177066683769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,16,balanced,0.043866669138272606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,16,balanced,0.04363200068473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,16,balanced,0.04822400212287903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,16,balanced,0.05143466591835022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,16,balanced,0.05614933371543884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,16,balanced,0.06963733335336049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,16,balanced,0.07502933343251546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,16,balanced,0.08890133102734883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,16,balanced,0.10728533069292705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,16,balanced,0.13321066896120706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,16,balanced,0.15682133038838705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.2,1.363417625427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,32,power_law_1.2,0.1208575963973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,32,power_law_1.2,0.16278400421142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,32,power_law_1.2,0.1561792016029358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,32,power_law_1.2,0.15845119953155518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,32,power_law_1.2,0.1582208037376404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,32,power_law_1.2,0.13237119913101197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,32,power_law_1.2,0.14161280393600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,32,power_law_1.2,0.15629440546035767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,32,power_law_1.2,0.1592128038406372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,32,power_law_1.2,0.14759039878845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,32,power_law_1.2,0.14425599575042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,32,power_law_1.2,0.1515391945838928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,32,power_law_1.2,0.1595199942588806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,32,power_law_1.2,0.15406080484390258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,32,power_law_1.2,0.16064000129699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,32,power_law_1.2,0.15745279788970948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,32,power_law_1.2,0.17445119619369506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,32,power_law_1.2,0.178713595867157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,32,power_law_1.2,0.19100159406661987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,32,power_law_1.2,0.21370880603790282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,32,power_law_1.2,0.25718400478363035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,32,power_law_1.2,0.33356800079345705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,32,power_law_1.2,0.3769216060638428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,32,power_law_1.2,0.4808767795562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,32,power_law_1.2,0.5951871871948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,32,power_law_1.2,0.8203200340270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,32,power_law_1.2,1.0737279891967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,256,power_law_1.2,0.018169599771499633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,256,power_law_1.2,0.018348799645900728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,256,power_law_1.2,0.01926400065422058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,256,power_law_1.2,0.021081599593162536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,256,power_law_1.2,0.035046398639678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,256,power_law_1.2,0.02983039915561676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,256,power_law_1.2,0.0302592009305954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,256,power_law_1.2,0.03022719919681549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,256,power_law_1.2,0.030451199412345885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,256,power_law_1.2,0.030527999997138976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,256,power_law_1.2,0.03048959970474243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,256,power_law_1.2,0.030937600135803222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,256,power_law_1.2,0.030508801341056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,256,power_law_1.2,0.031276801228523256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,256,power_law_1.2,0.0324288010597229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,256,power_law_1.2,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,256,power_law_1.2,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,256,power_law_1.2,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,256,power_law_1.2,0.03847039937973022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,256,power_law_1.2,0.04154880046844482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,256,power_law_1.2,0.05061119794845581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,256,power_law_1.2,0.05942400097846985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,256,power_law_1.2,0.07519360184669495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,256,power_law_1.2,0.10382720232009887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,256,power_law_1.2,0.12788480520248413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,256,power_law_1.2,0.1828160047531128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,256,power_law_1.2,0.23245439529418946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,32,power_law_1.2,0.043059200048446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,32,power_law_1.2,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,32,power_law_1.2,0.04229759871959686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,32,power_law_1.2,0.04286080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,32,power_law_1.2,0.042822399735450746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,32,power_law_1.2,0.04407039880752563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,32,power_law_1.2,0.044633600115776065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,32,power_law_1.2,0.045414400100708005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,32,power_law_1.2,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,32,power_law_1.2,0.04577920138835907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,32,power_law_1.2,0.048076799511909483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,32,power_law_1.2,0.04892799854278564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,32,power_law_1.2,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,32,power_law_1.2,0.05149440169334411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,32,power_law_1.2,0.05609599947929382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,32,power_law_1.2,0.0572160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,32,power_law_1.2,0.061343997716903687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,32,power_law_1.2,0.06964480280876159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,32,power_law_1.2,0.07305600047111512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,32,power_law_1.2,0.08759040236473084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,32,power_law_1.2,0.1011072039604187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,32,power_law_1.2,0.13345279693603515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,32,power_law_1.2,0.1714367985725403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,32,power_law_1.2,0.23605759143829347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,32,power_law_1.2,0.3054784059524536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,32,power_law_1.2,0.4478720188140869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,32,power_law_1.2,0.5571263790130615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,1,4,power_law_1.2,0.09843199849128723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,1,4,power_law_1.2,0.11414400339126587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,1,4,power_law_1.2,0.11505279541015626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,1,4,power_law_1.2,0.13565440177917482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,1,4,power_law_1.2,0.13515520095825195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,1,4,power_law_1.2,0.1371008038520813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,1,4,power_law_1.2,0.1364159941673279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,1,4,power_law_1.2,0.13872640132904052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,1,4,power_law_1.2,0.14024319648742675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,1,4,power_law_1.2,0.14135040044784547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,1,4,power_law_1.2,0.14300160408020018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,1,4,power_law_1.2,0.17025920152664184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,1,4,power_law_1.2,0.1689344048500061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,1,4,power_law_1.2,0.17935999631881713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,1,4,power_law_1.2,0.2220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,1,4,power_law_1.2,0.22004480361938478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,1,4,power_law_1.2,0.280076789855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,1,4,power_law_1.2,0.3331712007522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,1,4,power_law_1.2,0.4018688201904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,1,4,power_law_1.2,0.5376192092895508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,1,4,power_law_1.2,0.7029632091522217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,1,4,power_law_1.2,0.9986687660217285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,1,4,power_law_1.2,1.2552191734313964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,16,balanced,0.02535466601451238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,16,balanced,0.021029333273569744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,16,balanced,0.021242665747801464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,16,balanced,0.022266666094462078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,16,balanced,0.022895999252796173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,16,balanced,0.02535466601451238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,16,balanced,0.02701333413521449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,16,balanced,0.027248000105222065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,16,balanced,0.02942933390537898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,16,balanced,0.02932800104220708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,16,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,16,balanced,0.03618133316437403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,16,balanced,0.033226666351159416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,16,balanced,0.05604266623655955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,16,balanced,0.058037335673967995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,16,balanced,0.060191998879114784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,16,balanced,0.05705599983533224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,16,balanced,0.0625546673933665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,16,balanced,0.07179733117421468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,16,balanced,0.09603200356165568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,16,balanced,0.10570666193962097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,16,balanced,0.13964800039927164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,16,balanced,0.15993600090344748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,16,balanced,0.23550933599472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,16,balanced,0.2901386618614197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,16,balanced,0.41670934359232586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,16,balanced,0.5450133482615153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,1,4,power_law_1.2,1.7878976821899415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,2,power_law_1.2,0.04909439980983734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,2,power_law_1.2,0.06268159747123718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,2,power_law_1.2,0.06667519807815551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,2,power_law_1.2,0.07256960272789001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,2,power_law_1.2,0.07141119837760926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,2,power_law_1.2,0.07283200025558471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,2,power_law_1.2,0.07255679965019227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,2,power_law_1.2,0.0729088008403778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,2,power_law_1.2,0.07402880191802978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,2,power_law_1.2,0.07856640219688416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,2,power_law_1.2,0.07822719812393189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,2,power_law_1.2,0.08279039859771728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,2,power_law_1.2,0.08350719809532166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,2,power_law_1.2,0.0868287980556488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,2,power_law_1.2,0.09633920192718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,2,power_law_1.2,0.10124160051345825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,2,power_law_1.2,0.1157696008682251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,2,power_law_1.2,0.12856320142745972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.2,0.13848960399627686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.2,0.17010560035705566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.2,0.21663360595703124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.2,0.2936320066452026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,1,4,power_law_1.2,2.4700864791870116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.2,0.3432320117950439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.2,0.49146881103515627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.2,0.5326591968536377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,1,4,power_law_1.2,4.4971263885498045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.2,0.8984319686889648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,8,power_law_1.01,0.018611200153827667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,8,power_law_1.01,0.01825920045375824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,8,power_law_1.01,0.018515199422836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,8,power_law_1.01,0.018937599658966065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,8,power_law_1.01,0.018918399512767792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,8,power_law_1.01,0.02133760005235672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,8,power_law_1.01,0.044819200038909913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,8,power_law_1.01,0.045516800880432126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,8,power_law_1.01,0.032576000690460204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,8,power_law_1.01,0.03204480111598969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,8,power_law_1.01,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,8,power_law_1.01,0.035155200958251955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,8,power_law_1.01,0.03578880131244659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,8,power_law_1.01,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,8,power_law_1.01,0.03818239867687225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,8,power_law_1.01,0.03701759874820709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,8,power_law_1.01,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,8,power_law_1.01,0.04345600008964538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.01,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.01,0.06687999963760376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.01,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.01,0.09895679950714112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.01,0.12522879838943482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,1,4,power_law_1.2,4.492236709594726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.01,0.1717120051383972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.01,0.21021440029144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.2,1.2383744239807128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.01,0.3063040018081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.01,0.39328000545501707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,1,8,power_law_1.2,0.088019198179245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,1,8,power_law_1.2,0.08830080032348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,1,8,power_law_1.2,0.0886784017086029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,1,8,power_law_1.2,0.09248639941215515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,1,8,power_law_1.2,0.0923583984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,1,8,power_law_1.2,0.09319040179252625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,1,8,power_law_1.2,0.09331200122833253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,1,8,power_law_1.2,0.09455999732017517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,1,8,power_law_1.2,0.09616640210151672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,1,8,power_law_1.2,0.09654399752616882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,1,8,power_law_1.2,0.09815040230751038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,1,8,power_law_1.2,0.11436799764633179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,1,8,power_law_1.2,0.11902079582214356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,1,8,power_law_1.2,0.13890559673309327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,1,8,power_law_1.2,0.16456320285797119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,1,8,power_law_1.2,0.15639679431915282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,1,8,power_law_1.2,0.1912511944770813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,1,8,power_law_1.2,0.29256319999694824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,1,8,power_law_1.2,0.3389631986618042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,1,8,power_law_1.2,0.4604928016662598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,1,8,power_law_1.2,0.5739200115203857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,16,balanced,0.04826133449872335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,16,balanced,0.05050133168697357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,16,balanced,0.05064533154169718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,16,balanced,0.053247998158137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,16,balanced,0.054832001527150474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,16,balanced,0.060720001657803856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,16,balanced,0.06242666641871134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,16,balanced,0.06261333326498668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,16,balanced,0.06286400059858958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,16,balanced,0.06424533327420552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,16,balanced,0.06496533254782359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,16,balanced,0.06579733391602834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,16,balanced,0.0666293352842331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,16,balanced,0.06808533271153767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,16,balanced,0.07215466598669688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,16,balanced,0.07287999987602234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,16,balanced,0.07714133461316426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,16,balanced,0.08311999837557475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,16,balanced,0.08902933200200398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,16,balanced,0.1029813289642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,16,balanced,0.11871467034022014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,16,balanced,0.1539413332939148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,16,balanced,0.17893866697947183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,16,balanced,0.24048533042271933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,16,balanced,0.29290133714675903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,16,balanced,0.39795200030008954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,16,balanced,0.5083039999008179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,1,8,power_law_1.2,0.8991935729980469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,1,8,power_law_1.2,1.11080322265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,1,8,power_law_1.2,1.7325632095336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,1,8,power_law_1.2,2.118515205383301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,16,power_law_1.2,0.06275200247764587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,16,power_law_1.2,0.05913599729537964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,16,power_law_1.2,0.057171201705932616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,16,power_law_1.2,0.058156800270080564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,16,power_law_1.2,0.055801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,16,power_law_1.2,0.06354560256004334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,16,power_law_1.2,0.0640064001083374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,16,power_law_1.2,0.06551039814949036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,16,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,16,power_law_1.2,0.06528639793395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,16,power_law_1.2,0.06676480174064636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,16,power_law_1.2,0.0689087986946106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,1,8,power_law_1.2,3.0351743698120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,16,power_law_1.2,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,16,power_law_1.2,0.0720192015171051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,16,power_law_1.2,0.07594879865646362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,16,power_law_1.2,0.0822272002696991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,16,power_law_1.2,0.09299200177192687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,16,power_law_1.2,0.10109440088272095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,16,power_law_1.2,0.11563520431518555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,16,power_law_1.2,0.14759680032730102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,16,power_law_1.2,0.19148160219192506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,16,power_law_1.2,0.24583680629730226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,16,power_law_1.2,0.3060863971710205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,16,power_law_1.2,0.4289984226226807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,16,power_law_1.2,0.6348351955413818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,16,power_law_1.2,1.023622417449951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,1,8,power_law_1.2,3.9681537628173826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,2,power_law_1.2,0.1350208044052124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,2,power_law_1.2,0.21887359619140626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,2,power_law_1.2,0.2775424003601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,2,power_law_1.2,0.4135104179382324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,2,power_law_1.2,0.6110720157623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,2,power_law_1.2,0.6762239933013916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,2,power_law_1.2,1.0988927841186524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,2,power_law_1.2,0.9948032379150391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,2,power_law_1.2,1.1393024444580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,2,power_law_1.2,1.122976016998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,2,power_law_1.2,1.209440040588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,2,power_law_1.2,1.2563072204589845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,2,power_law_1.2,1.2344767570495605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,2,power_law_1.2,1.2590656280517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,2,power_law_1.2,1.2849984169006348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,2,power_law_1.2,1.350425624847412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,2,power_law_1.2,1.4201791763305665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,2,power_law_1.2,1.5941823959350585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,2,power_law_1.2,1.5452159881591796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,2,power_law_1.2,1.8250688552856444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,2,power_law_1.2,1.805036735534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,2,power_law_1.2,1.9938240051269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,2,power_law_1.2,2.2924736022949217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,2,power_law_1.2,2.8699647903442385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,2,power_law_1.2,3.1831039428710937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,2,power_law_1.2,4.053779220581054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,2,power_law_1.2,5.0330047607421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,16,power_law_1.2,1.1755328178405762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,1,power_law_1.2,0.08235520124435425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,1,power_law_1.2,0.09322239756584168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,1,power_law_1.2,0.11336319446563721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,1,power_law_1.2,0.142739200592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,1,power_law_1.2,0.17826559543609619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,1,power_law_1.2,0.21335039138793946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,1,power_law_1.2,0.28765439987182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,1,power_law_1.2,0.300601601600647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,1,power_law_1.2,0.3099391937255859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,1,power_law_1.2,0.3203520059585571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,1,power_law_1.2,0.3275455951690674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,1,power_law_1.2,0.33800320625305175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,1,power_law_1.2,0.34418559074401855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,1,power_law_1.2,0.3618560075759888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,1,power_law_1.2,0.3752703905105591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,1,power_law_1.2,0.3875135898590088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,1,power_law_1.2,0.39842560291290285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,1,power_law_1.2,0.46904959678649905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,1,power_law_1.2,0.5169280052185059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,1,power_law_1.2,0.6226367950439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,1,power_law_1.2,0.6807807922363281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,1,power_law_1.2,0.8654784202575684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,1,power_law_1.2,1.037292766571045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,1,power_law_1.2,1.4029376029968261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,1,power_law_1.2,1.806368064880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,1,power_law_1.2,2.515238380432129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,1,power_law_1.2,3.322835159301758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,32,power_law_1.01,0.044787201285362246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,32,power_law_1.01,0.04391680061817169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,32,power_law_1.01,0.04387840032577515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,32,power_law_1.01,0.043808001279830935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,32,power_law_1.01,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,32,power_law_1.01,0.045177599787712096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,32,power_law_1.01,0.04561919867992401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,32,power_law_1.01,0.04633600115776062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,32,power_law_1.01,0.0469184011220932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,32,power_law_1.01,0.04805760085582733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,32,power_law_1.01,0.049107199907302855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,32,power_law_1.01,0.05014399886131286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,32,power_law_1.01,0.05084800124168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,32,power_law_1.01,0.05203840136528015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,32,power_law_1.01,0.058387202024459836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,32,power_law_1.01,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,32,power_law_1.01,0.06242560148239136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,32,power_law_1.01,0.07191680073738098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,32,power_law_1.01,0.07499520182609558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,32,power_law_1.01,0.08906239867210389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,32,power_law_1.01,0.09970560073852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,32,power_law_1.01,0.13173120021820067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,32,power_law_1.01,0.15779839754104613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,32,power_law_1.01,0.22092800140380858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,32,power_law_1.01,0.2741568088531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,32,power_law_1.01,0.3920000076293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,32,power_law_1.01,0.4987775802612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,32,balanced,0.023242667317390442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,32,balanced,0.02111999938885371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,32,balanced,0.021354667842388153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,32,balanced,0.020981334149837494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,32,balanced,0.02128533273935318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,32,balanced,0.023189333577950794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,32,balanced,0.025093334416548412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,32,balanced,0.02513066679239273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,32,balanced,0.025477332373460133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,32,balanced,0.025770666698614757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,32,balanced,0.029482667644818623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,32,balanced,0.032960000137488045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,32,balanced,0.03163733333349228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,32,balanced,0.04534400006135305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,32,balanced,0.0487413356701533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,32,balanced,0.048026666045188904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,32,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,32,balanced,0.055589333176612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,32,balanced,0.0641653339068095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,32,balanced,0.07582933207352956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,32,balanced,0.08875200152397156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,32,balanced,0.12113599975903828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,32,balanced,0.14380266269048056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,32,balanced,0.1978293259938558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,32,balanced,0.2539466619491577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,32,balanced,0.3670773506164551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,32,balanced,0.4775466521581014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,128,power_law_1.2,0.01923840045928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,128,power_law_1.2,0.021772800385951994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,128,power_law_1.2,0.019219200313091277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,128,power_law_1.2,0.021855999529361726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,128,power_law_1.2,0.02040960043668747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,128,power_law_1.2,0.025817599892616273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,128,power_law_1.2,0.025363200902938844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,128,power_law_1.2,0.040863999724388124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,128,power_law_1.2,0.04175359904766083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,128,power_law_1.2,0.04135040044784546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,128,power_law_1.2,0.04118399918079376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,128,power_law_1.2,0.041740798950195314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,128,power_law_1.2,0.04132480025291443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,128,power_law_1.2,0.040243199467658995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,128,power_law_1.2,0.038924801349639895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,128,power_law_1.2,0.03844479918479919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,128,power_law_1.2,0.04403199851512909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,128,power_law_1.2,0.04508160054683685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,128,power_law_1.2,0.05026559829711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,1,power_law_1.2,0.07419520020484924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,128,power_law_1.2,0.06135680079460144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,1,power_law_1.2,0.10056960582733154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,1,power_law_1.2,0.14499839544296264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,128,power_law_1.2,0.07825279831886292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,128,power_law_1.2,0.10240639448165893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,128,power_law_1.2,0.12225279808044434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,128,power_law_1.2,0.17541120052337647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,1,power_law_1.2,0.2054975986480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,1,power_law_1.2,0.2722304105758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,128,power_law_1.2,0.2242432117462158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,1,power_law_1.2,0.45296640396118165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,1,power_law_1.2,0.4898111820220947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,1,power_law_1.2,0.49884800910949706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,1,power_law_1.2,0.5258560180664062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,1,power_law_1.2,0.5237631797790527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,1,power_law_1.2,0.5524864196777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,1,power_law_1.2,0.5706943988800048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,1,power_law_1.2,0.5901247978210449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,1,power_law_1.2,0.5979712009429932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,1,power_law_1.2,0.6216320037841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,1,power_law_1.2,0.6430335998535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,128,power_law_1.2,0.33016960620880126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,1,power_law_1.2,0.6626751899719239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,128,power_law_1.2,0.4474815845489502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,1,power_law_1.2,0.7523327827453613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,1,power_law_1.2,0.7710336208343506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,1,power_law_1.2,0.8907391548156738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,1,power_law_1.2,0.9270336151123046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,1,power_law_1.2,1.1262016296386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,1,power_law_1.2,1.3314623832702637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,1,power_law_1.2,1.7157119750976562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,1,power_law_1.2,2.4203264236450197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,1,power_law_1.2,3.137343978881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,1,power_law_1.2,3.8661567687988283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,1,4,balanced,0.09186133742332458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,1,4,balanced,0.09098666906356812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,1,4,balanced,0.13335466384887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,1,4,balanced,0.13847466309865317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,1,4,balanced,0.13825066884358725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,1,4,balanced,0.13799466689427695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,1,4,balanced,0.13952533404032388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,1,4,balanced,0.14082133769989014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,1,4,balanced,0.1398293375968933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,1,4,balanced,0.14172266920407614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,1,4,balanced,0.14011733730634054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,1,4,balanced,0.15849066774050394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,1,4,balanced,0.15773333112398782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,1,4,balanced,0.15806933244069418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,1,4,balanced,0.1551520029703776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,1,4,balanced,0.15651733676592508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,1,4,balanced,0.17148266235987344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,1,4,balanced,0.20201067129770914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,1,4,balanced,0.2101759910583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,1,4,balanced,0.2998560070991516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,1,4,balanced,0.34497066338857013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,1,4,balanced,0.48722132047017414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,1,4,balanced,0.6141066551208496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,1,4,balanced,0.8570453325907389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,1,4,balanced,1.097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,1,4,balanced,1.6437546412150066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,1,4,balanced,2.12827730178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,16,power_law_1.2,0.058719998598098753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,16,power_law_1.2,0.06944640278816223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,16,power_law_1.2,0.06576640009880066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,16,power_law_1.2,0.06668800115585327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,16,power_law_1.2,0.06845440268516541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,16,power_law_1.2,0.06863999962806702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,16,power_law_1.2,0.06787840127944947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,16,power_law_1.2,0.07169280052185059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,16,power_law_1.2,0.06637439727783204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,16,power_law_1.2,0.0670144021511078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,16,power_law_1.2,0.0692031979560852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,16,power_law_1.2,0.07251840233802795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,16,power_law_1.2,0.07090560197830201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,16,power_law_1.2,0.07310720086097718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,16,power_law_1.2,0.08216320276260376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,16,power_law_1.2,0.0843392014503479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,16,power_law_1.2,0.09313920140266418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,16,power_law_1.2,0.10609279870986939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,16,power_law_1.2,0.11611520051956177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,16,power_law_1.2,0.13961600065231322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,16,power_law_1.2,0.15555200576782227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,16,power_law_1.2,0.2057215929031372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,16,power_law_1.2,0.25414400100708007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,16,power_law_1.2,0.35269119739532473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,16,power_law_1.2,0.44269437789916993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,2,power_law_1.2,0.03804160058498383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,2,power_law_1.2,0.040268799662590025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,2,power_law_1.2,0.04910080134868622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,2,power_law_1.2,0.05251839756965637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,2,power_law_1.2,0.05288959741592407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,2,power_law_1.2,0.05400320291519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,2,power_law_1.2,0.05560320019721985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,2,power_law_1.2,0.05854719877243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,2,power_law_1.2,0.058899199962615965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,16,power_law_1.2,0.5986815929412842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,2,power_law_1.2,0.05978879928588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,2,power_law_1.2,0.061612802743911746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,16,power_law_1.2,0.7765056133270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,2,power_law_1.2,0.06402559876441956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,2,power_law_1.2,0.06307839751243591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,2,power_law_1.2,0.07109760046005249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,2,power_law_1.2,0.07678080201148987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,2,power_law_1.2,0.0813759982585907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,2,power_law_1.2,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,2,power_law_1.2,0.08436480164527893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,2,power_law_1.2,0.11570559740066529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,2,power_law_1.2,0.1314303994178772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,2,power_law_1.2,0.15164799690246583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,2,power_law_1.2,0.2120896100997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,2,power_law_1.2,0.23144960403442383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,2,power_law_1.2,0.3225600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,2,power_law_1.2,0.3537535905838013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,16,power_law_1.2,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,16,power_law_1.2,0.051718401908874514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,16,power_law_1.2,0.04965760111808777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,16,power_law_1.2,0.051999998092651364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,16,power_law_1.2,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,16,power_law_1.2,0.05583360195159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,16,power_law_1.2,0.05578879714012146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,16,power_law_1.2,0.057555198669433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,16,power_law_1.2,0.05634559988975525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,16,power_law_1.2,0.05568000078201294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,16,power_law_1.2,0.0567359983921051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,16,power_law_1.2,0.06047359704971313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,16,power_law_1.2,0.06220160126686096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,16,power_law_1.2,0.06135680079460144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,16,power_law_1.2,0.07066879868507385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,16,power_law_1.2,0.07259520292282104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,16,power_law_1.2,0.07559679746627808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,16,power_law_1.2,0.08487679958343505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,16,power_law_1.2,0.09621760249137878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,16,power_law_1.2,0.11637120246887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,16,power_law_1.2,0.13420159816741944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,16,power_law_1.2,0.17448960542678832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,16,power_law_1.2,0.21557118892669677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,16,power_law_1.2,0.30383360385894775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,16,power_law_1.2,0.40497918128967286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,16,power_law_1.2,0.5891776084899902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,16,power_law_1.2,0.7720704078674316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,2,power_law_1.2,0.5790016174316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,2,power_law_1.2,0.714137601852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,4,balanced,0.04148799926042557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,4,balanced,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,4,balanced,0.05977599819501241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,4,balanced,0.08241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,4,balanced,0.1272640029589335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,4,balanced,0.21797333161036173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,4,balanced,0.2295786738395691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,4,balanced,0.22891199588775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,4,balanced,0.23637332518895468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,4,balanced,0.24361066023508707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,4,balanced,0.2605973283449809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,4,balanced,0.27711466948191327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,4,balanced,0.2917226751645406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,4,balanced,0.2736000021298726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,4,balanced,0.2895146608352661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,4,balanced,0.30524800221125287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,4,balanced,0.3972586790720622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,4,balanced,0.3942933479944865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,4,balanced,0.39371732870737713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,4,balanced,0.3986186583836873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,4,balanced,0.592906673749288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,4,balanced,0.6431519985198975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,4,balanced,0.6944159666697184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,4,balanced,0.7782933712005615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,4,balanced,0.8793813387552897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,4,balanced,1.4826985994974773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,4,balanced,1.6471680005391438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,32,balanced,0.05022400120894114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,32,balanced,0.053957333167394005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,32,balanced,0.05282133320967356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,32,balanced,0.05376533170541128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,32,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,32,balanced,0.06014933188756307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,32,balanced,0.06404266754786174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,32,balanced,0.06419200201829274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,32,balanced,0.06612800061702728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,32,balanced,0.0683786670366923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,32,balanced,0.06613866488138835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,32,balanced,0.06816000243028005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,32,balanced,0.06821866830190022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,32,balanced,0.07010133564472198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,32,balanced,0.07454933226108551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,32,balanced,0.07649066547552745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,32,balanced,0.08462933699289958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,32,balanced,0.09329600135485332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,32,balanced,0.10140800476074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,32,balanced,0.11947733163833618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,32,balanced,0.13302399714787802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,32,balanced,0.1662666698296865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,32,balanced,0.19792000452677408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,32,balanced,0.25831466913223267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,32,balanced,0.32843200365702313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,32,balanced,0.4443306525548299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,32,balanced,0.5725813309351603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,32,power_law_1.2,0.0219200000166893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,32,power_law_1.2,0.022047999501228332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,32,power_law_1.2,0.019411200284957887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,32,power_law_1.2,0.02136320024728775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,32,power_law_1.2,0.02122880071401596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,32,power_law_1.2,0.023475199937820435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,32,power_law_1.2,0.02505599856376648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,32,power_law_1.2,0.029728001356124877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,32,power_law_1.2,0.030611199140548707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,32,power_law_1.2,0.03203839957714081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,32,power_law_1.2,0.042803201079368594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,32,power_law_1.2,0.04310399889945984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,32,power_law_1.2,0.04398080110549927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,32,power_law_1.2,0.047891199588775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,32,power_law_1.2,0.05496960282325745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,32,power_law_1.2,0.06219519972801209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,32,power_law_1.2,0.06355199813842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,32,power_law_1.2,0.08252800107002259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.2,0.09582080245018006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.2,0.14650880098342894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.2,0.0880895972251892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.2,0.10977280139923096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.2,0.13367040157318116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.2,0.19181439876556397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.2,0.24197759628295898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.2,0.37528960704803466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.2,0.49521918296813966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,8,power_law_1.01,0.06838399767875672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,8,power_law_1.01,0.08601599931716919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,8,power_law_1.01,0.09360640048980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,8,power_law_1.01,0.09864959716796876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,8,power_law_1.01,0.11143679618835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,8,power_law_1.01,0.10609920024871826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,8,power_law_1.01,0.10656640529632569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,2,power_law_1.01,0.044223999977111815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,8,power_law_1.01,0.11505919694900513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,2,power_law_1.01,0.04286719858646393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,8,power_law_1.01,0.11227519512176513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,8,power_law_1.01,0.1159168004989624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,8,power_law_1.01,0.11512960195541382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,8,power_law_1.01,0.11850240230560302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,8,power_law_1.01,0.11969920396804809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,8,power_law_1.01,0.12295680046081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,2,power_law_1.01,0.045075199007987975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,8,power_law_1.01,0.13311359882354737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,2,power_law_1.01,0.04769920110702515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,8,power_law_1.01,0.13893760442733766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,2,power_law_1.01,0.05226879715919495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,8,power_law_1.01,0.14360320568084717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,8,power_law_1.01,0.16607359647750855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.01,0.17890559434890746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.01,0.210534405708313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.01,0.23148159980773925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,2,power_law_1.01,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,2,power_law_1.01,0.06035199761390686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,2,power_law_1.01,0.060313600301742556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,2,power_law_1.01,0.06178560256958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,2,power_law_1.01,0.06367999911308289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,2,power_law_1.01,0.0648576021194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,2,power_law_1.01,0.06772480010986329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,2,power_law_1.01,0.06944000124931335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,2,power_law_1.01,0.07060480117797852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.01,0.2931328058242798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,2,power_law_1.01,0.07564799785614014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,2,power_law_1.01,0.08062080144882203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.01,0.3445375919342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,2,power_law_1.01,0.08696320056915283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,2,power_law_1.01,0.10133119821548461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.01,0.4532224178314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,2,power_law_1.01,0.11350400447845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,2,power_law_1.01,0.15059839487075805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.01,0.6294335842132568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,2,power_law_1.01,0.1746559977531433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,2,power_law_1.01,0.22554879188537597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.01,0.8994112014770508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,2,power_law_1.01,0.2816960096359253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,2,power_law_1.01,0.37589120864868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.01,1.0660223960876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,2,power_law_1.01,0.49730558395385743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,2,power_law_1.01,0.80697603225708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,2,power_law_1.01,0.9983039855957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,4,power_law_1.01,0.035206401348114015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,4,power_law_1.01,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,4,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,4,power_law_1.01,0.03624959886074066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,4,power_law_1.01,0.03692800104618073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,4,power_law_1.01,0.03784320056438446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,4,power_law_1.01,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,4,power_law_1.01,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,4,power_law_1.01,0.0381632000207901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,4,power_law_1.01,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,4,power_law_1.01,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,4,power_law_1.01,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,4,power_law_1.01,0.04454399943351746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,4,power_law_1.01,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,4,power_law_1.01,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,4,power_law_1.01,0.05689600110054016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,4,power_law_1.01,0.06490880250930786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,4,power_law_1.01,0.08206719756126404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,4,power_law_1.01,0.08728320002555848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,4,power_law_1.01,0.12209919691085816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,4,power_law_1.01,0.14261120557785034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,4,power_law_1.01,0.2496256113052368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,4,power_law_1.01,0.2827136039733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,4,power_law_1.01,0.41328001022338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,4,power_law_1.01,0.5540800094604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,4,power_law_1.01,0.7850880146026611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,4,power_law_1.01,1.2076416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,2,power_law_1.2,0.04595839977264404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,2,power_law_1.2,0.04925439953804016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,2,power_law_1.2,0.05257599949836731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,2,power_law_1.2,0.057766401767730714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,2,power_law_1.2,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,2,power_law_1.2,0.07503359913825988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,2,power_law_1.2,0.08593279719352723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,2,power_law_1.2,0.0874239981174469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,2,power_law_1.2,0.09149439930915833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,2,power_law_1.2,0.09257599711418152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,2,power_law_1.2,0.09612159729003907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,2,power_law_1.2,0.10375039577484131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,2,power_law_1.2,0.10409599542617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,2,power_law_1.2,0.10964479446411132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,2,power_law_1.2,0.11893759965896607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,2,power_law_1.2,0.12599680423736573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,2,power_law_1.2,0.13162239789962768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,2,power_law_1.2,0.1490048050880432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,2,power_law_1.2,0.1697983980178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,2,power_law_1.2,0.20773758888244628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,2,power_law_1.2,0.24479999542236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,2,power_law_1.2,0.3182528018951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,2,power_law_1.2,0.40070400238037107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,2,power_law_1.2,0.5674176216125488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,2,power_law_1.2,0.7092351913452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,2,power_law_1.2,0.9951423645019531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,2,power_law_1.2,1.3706303596496583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,1,power_law_1.2,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,1,power_law_1.2,0.12509440183639525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,1,power_law_1.2,0.19367040395736695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,1,power_law_1.2,0.2311232089996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,1,power_law_1.2,0.24126079082489013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,1,power_law_1.2,0.23795199394226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,1,power_law_1.2,0.23754239082336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,1,power_law_1.2,0.24861440658569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,1,power_law_1.2,0.2508224010467529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,1,power_law_1.2,0.2592895984649658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,1,power_law_1.2,0.2486975908279419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,1,power_law_1.2,0.26910719871520994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,1,power_law_1.2,0.2798336029052734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,1,power_law_1.2,0.2723392009735107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,1,power_law_1.2,0.2999039888381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,1,power_law_1.2,0.3119040012359619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,1,power_law_1.2,0.323635196685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,1,power_law_1.2,0.3699007987976074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,1,power_law_1.2,0.42169599533081054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,1,power_law_1.2,0.5143167972564697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,1,power_law_1.2,0.6136640071868896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,1,power_law_1.2,0.829856014251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,1,power_law_1.2,1.0245951652526855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,4,balanced,0.05409599840641022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,4,balanced,0.07263466715812683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,4,balanced,0.0999679962793986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,4,balanced,0.1592693328857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,4,balanced,0.2603093385696411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,1,power_law_1.2,1.4454208374023438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,4,balanced,0.31698666016260785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,4,balanced,0.3162879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,4,balanced,0.3207040031750997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,4,balanced,0.31962666908899945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,4,balanced,0.32419733206431073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,4,balanced,0.32521599531173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,4,balanced,0.33019200960795086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,4,balanced,0.3322773377100627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,4,balanced,0.33372267087300617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,4,balanced,0.34111467997233075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,4,balanced,0.34229334195454914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,4,balanced,0.3503893216451009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,4,balanced,0.39257065455118817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,4,balanced,0.3820799986521403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,4,balanced,0.4533333381017049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,4,balanced,0.4454346497853597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,4,balanced,0.6171840031941732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,4,balanced,0.5886079867680868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,4,balanced,0.8943040370941162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,4,balanced,0.9423147042592367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,4,balanced,1.5062294006347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,4,balanced,1.7345600128173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,32,power_law_1.01,0.041407999396324155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,32,power_law_1.01,0.04103679955005646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,32,power_law_1.01,0.04079999923706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,32,power_law_1.01,0.04252800047397613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,32,power_law_1.01,0.04304639995098114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,32,power_law_1.01,0.043289598822593686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,32,power_law_1.01,0.04464640021324158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,1,power_law_1.2,1.8459455490112304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,32,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,32,power_law_1.01,0.04559360146522522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,32,power_law_1.01,0.046003198623657225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,32,power_law_1.01,0.047270399332046506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,32,power_law_1.01,0.04875519871711731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,32,power_law_1.01,0.04888960123062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,32,power_law_1.01,0.05145599842071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,32,power_law_1.01,0.05615360140800476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,32,power_law_1.01,0.05602560043334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,32,power_law_1.01,0.060819202661514284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,32,power_law_1.01,0.06783999800682068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,32,power_law_1.01,0.07275519967079162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,32,power_law_1.01,0.0847487986087799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,32,power_law_1.01,0.09437440037727356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,32,power_law_1.01,0.12406400442123414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,32,power_law_1.01,0.1508031964302063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,32,power_law_1.01,0.22042880058288575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,32,power_law_1.01,0.2701632022857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,32,power_law_1.01,0.3973759889602661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,32,power_law_1.01,0.49728641510009763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,16,power_law_1.01,0.059622400999069215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,16,power_law_1.01,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,16,power_law_1.01,0.06149759888648987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,16,power_law_1.01,0.06480000019073487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,16,power_law_1.01,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,16,power_law_1.01,0.0661952018737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,16,power_law_1.01,0.06673280000686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,16,power_law_1.01,0.06904320120811462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,16,power_law_1.01,0.06995199918746949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,16,power_law_1.01,0.06997759938240052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,16,power_law_1.01,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,16,power_law_1.01,0.07130240201950074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,16,power_law_1.01,0.07166079878807068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,16,power_law_1.01,0.07528960108757018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,16,power_law_1.01,0.08631680011749268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,16,power_law_1.01,0.08832640051841736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,1,power_law_1.2,2.6408447265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,16,power_law_1.01,0.09363200068473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,16,power_law_1.01,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,16,power_law_1.01,0.11525119543075561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,16,power_law_1.01,0.13804160356521605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,16,power_law_1.01,0.1572543978691101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,16,power_law_1.01,0.21245439052581788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,16,power_law_1.01,0.25063040256500246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,16,power_law_1.01,0.35036799907684324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,16,power_law_1.01,0.48640642166137693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,16,power_law_1.01,0.7099391937255859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,16,power_law_1.01,0.8928000450134277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,1,power_law_1.2,3.450694274902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,4,power_law_1.2,0.14217599630355834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,4,power_law_1.2,0.15334399938583373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,4,power_law_1.2,0.20261120796203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,4,power_law_1.2,0.25564799308776853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,4,power_law_1.2,0.24125440120697023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,4,power_law_1.2,0.33352320194244384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,4,power_law_1.2,0.35784320831298827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,4,power_law_1.2,0.36192638874053956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,4,power_law_1.2,0.36754560470581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,4,power_law_1.2,0.36805119514465334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,4,power_law_1.2,0.38918399810791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,4,power_law_1.2,0.38964478969573973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,4,power_law_1.2,0.3792959928512573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,4,power_law_1.2,0.40123519897460935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,4,power_law_1.2,0.39079039096832274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,4,power_law_1.2,0.40351362228393556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,4,power_law_1.2,0.42334718704223634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,4,power_law_1.2,0.46730241775512693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.2,0.49068799018859866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.2,0.6013696193695068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.2,0.6751103878021241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.2,0.8409407615661622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.2,0.9538368225097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.2,1.2651583671569824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.2,1.554361629486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.2,2.218617630004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.2,2.599404716491699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,128,power_law_1.01,0.026150399446487428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,128,power_law_1.01,0.025798401236534117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,128,power_law_1.01,0.025766399502754212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,128,power_law_1.01,0.02815360128879547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,128,power_law_1.01,0.028812798857688903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,128,power_law_1.01,0.02999039888381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,128,power_law_1.01,0.030015999078750612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,128,power_law_1.01,0.030048000812530517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,128,power_law_1.01,0.030323201417922975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,128,power_law_1.01,0.03089280128479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,128,power_law_1.01,0.03206399977207184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,128,power_law_1.01,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,32,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,32,balanced,0.05769066512584686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,32,balanced,0.05994666616121928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,32,balanced,0.08218133449554443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,128,power_law_1.01,0.03402880132198334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,32,balanced,0.09318400422732036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,32,balanced,0.0953653355439504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,32,balanced,0.09582933783531189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,32,balanced,0.0950933297475179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,32,balanced,0.09608532985051473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,32,balanced,0.09763200084368388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,32,balanced,0.09733333190282185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,32,balanced,0.10085866848627727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,32,balanced,0.1011786659558614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,32,balanced,0.10186133782068889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,32,balanced,0.10688533385594685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,32,balanced,0.10892800490061443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,32,balanced,0.11388267079989116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,32,balanced,0.12680000066757202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,32,balanced,0.13566399614016214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,32,balanced,0.16611199577649435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,32,balanced,0.18036800622940063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,32,balanced,0.2362346649169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,32,balanced,0.23185600837071738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,32,balanced,0.3118133346239726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,32,balanced,0.35073598225911456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,32,balanced,0.4817066589991252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,128,power_law_1.01,0.036934399604797365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,128,power_law_1.01,0.041017600893974306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,128,power_law_1.01,0.0409280002117157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,128,power_law_1.01,0.04389120042324066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,32,balanced,0.5821173191070557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,128,power_law_1.01,0.0494592010974884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,128,power_law_1.01,0.058246397972106935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,128,power_law_1.01,0.06977919936180114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,128,power_law_1.01,0.08438400030136109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,128,power_law_1.01,0.11123839616775513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,128,power_law_1.01,0.14828799962997435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,128,power_law_1.01,0.20764799118041993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,128,power_law_1.01,0.26766080856323243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,128,power_law_1.01,0.40050559043884276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,128,power_law_1.01,0.5860671997070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,8,power_law_1.2,0.016415999829769136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,8,power_law_1.2,0.01777919977903366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,8,power_law_1.2,0.018060800433158875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,8,power_law_1.2,0.017638400197029114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,8,power_law_1.2,0.017606399953365326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,8,power_law_1.2,0.018931199610233308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,8,power_law_1.2,0.019334399700164796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,8,power_law_1.2,0.019276799261569978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,8,power_law_1.2,0.020051200687885285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,8,power_law_1.2,0.02101760059595108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,8,power_law_1.2,0.020934399962425233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,8,power_law_1.2,0.021823999285697938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,8,power_law_1.2,0.022643199563026427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,8,power_law_1.2,0.025011199712753295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,8,power_law_1.2,0.02561280131340027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,8,power_law_1.2,0.026214399933815004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,8,power_law_1.2,0.03772799968719483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,8,power_law_1.2,0.04054400026798248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,8,power_law_1.2,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,8,power_law_1.2,0.04195840060710907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,8,power_law_1.2,0.04555520117282867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,8,balanced,0.04607999821503957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,8,balanced,0.04797866443792979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,8,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,8,balanced,0.07557866473992665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,8,balanced,0.09985066453615825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,8,balanced,0.15448533495267233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,8,balanced,0.2043786644935608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,8,balanced,0.20289599895477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,8,balanced,0.20403200387954712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,8,balanced,0.2069973349571228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,8,balanced,0.20774400234222412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,8,balanced,0.21010667085647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,8,balanced,0.21286932627360025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,8,balanced,0.21522667010625204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,8,balanced,0.22268799940745035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,8,balanced,0.2258560061454773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,8,balanced,0.2342346707979838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,8,balanced,0.24636799097061157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,8,balanced,0.26128532489140827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,8,balanced,0.28989867369333905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,8,balanced,0.31382934252421063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,8,power_law_1.2,0.05997440218925476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,8,balanced,0.37802668412526447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,8,power_law_1.2,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,8,balanced,0.43691198031107586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,8,balanced,0.5487039883931478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,8,balanced,0.6828320026397705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,8,power_law_1.2,0.09077759981155395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,8,balanced,0.9088746706644694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,8,balanced,1.1678346792856853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,8,power_law_1.2,0.11111680269241334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,8,power_law_1.2,0.1555199980735779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,64,balanced,0.05011733373006185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,64,balanced,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,64,balanced,0.04969066878159841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,64,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,64,balanced,0.05186133086681366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,64,balanced,0.055871998270352684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,64,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,64,balanced,0.058304001887639366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,64,balanced,0.05809600154558817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,64,balanced,0.06014933188756307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,64,balanced,0.058058664202690125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,64,balanced,0.06168533364931742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,64,balanced,0.06264000137646993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,64,balanced,0.06222933530807495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,64,balanced,0.06880000233650208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,64,balanced,0.06834133466084798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,64,balanced,0.07266666491826375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,64,balanced,0.08041599889596303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,64,balanced,0.08656533559163411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,64,balanced,0.1046453317006429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,64,balanced,0.1197813351949056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,64,balanced,0.1520746648311615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,64,balanced,0.17972266674041748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,64,balanced,0.23597866296768188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,64,balanced,0.2961813410123189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,64,balanced,0.40221333503723145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,64,balanced,0.51692267258962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,8,power_law_1.2,0.19198720455169677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,8,power_law_1.01,0.020147199928760528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,8,power_law_1.01,0.02106879949569702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,8,power_law_1.01,0.022969600558280946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,8,power_law_1.01,0.023686400055885314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,8,power_law_1.01,0.025158399343490602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,8,power_law_1.01,0.025190401077270507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,8,power_law_1.01,0.026636800169944762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,8,power_law_1.01,0.02791680097579956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,8,power_law_1.01,0.028089600801467895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,8,power_law_1.01,0.02919679880142212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,8,power_law_1.01,0.029504001140594482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,8,power_law_1.01,0.0310591995716095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,8,power_law_1.01,0.03190400004386902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,8,power_law_1.01,0.03386240005493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,8,power_law_1.01,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,8,power_law_1.01,0.03888640105724335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,8,power_law_1.01,0.04508160054683685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,8,power_law_1.01,0.050860798358917235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,8,power_law_1.01,0.07872639894485474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,8,power_law_1.01,0.09192320108413696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,8,power_law_1.01,0.08815360069274902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,8,power_law_1.01,0.10958720445632934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,8,power_law_1.01,0.12802560329437257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,8,power_law_1.01,0.176256000995636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,8,power_law_1.01,0.2130687952041626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,8,power_law_1.01,0.297491192817688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,8,power_law_1.01,0.3683520078659058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,1,power_law_1.2,0.05633280277252197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,1,power_law_1.2,0.07191039919853211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,1,power_law_1.2,0.09649279713630676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,1,power_law_1.2,0.12641919851303102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,1,power_law_1.2,0.15413119792938232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,1,power_law_1.2,0.23928959369659425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,1,power_law_1.2,0.2548608064651489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,1,power_law_1.2,0.25689599514007566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,1,power_law_1.2,0.2640575885772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,1,power_law_1.2,0.2696703910827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,1,power_law_1.2,0.2788095951080322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,1,power_law_1.2,0.2835520029067993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,1,power_law_1.2,0.29015040397644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,1,power_law_1.2,0.2990528106689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,1,power_law_1.2,0.31378560066223143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,1,power_law_1.2,0.3266239881515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,1,power_law_1.2,0.3478271961212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,1,power_law_1.2,0.40619521141052245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.2,0.44481282234191893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.2,0.5276224136352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.2,0.6193088054656982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.2,0.7959807872772217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.2,0.9721535682678223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.2,1.3352255821228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.2,1.6808128356933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.2,2.3624000549316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.2,3.0094720840454103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,2,balanced,0.03166933357715607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,2,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,2,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,2,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,2,balanced,0.06896533568700154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,2,balanced,0.07096000015735626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,2,balanced,0.07085333267847697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,2,balanced,0.07052800059318542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,2,balanced,0.07234133283297221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,2,balanced,0.07303999861081441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,2,balanced,0.0729066679875056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,2,balanced,0.07669333120187123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,2,balanced,0.07669866581757863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,2,balanced,0.07896000146865845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,2,balanced,0.08551466464996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,2,balanced,0.08554133772850037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,2,balanced,0.09085333347320557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,2,balanced,0.10295466581980388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,2,balanced,0.11658666531244914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,2,balanced,0.13732799887657166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,2,balanced,0.15350932876269022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,2,balanced,0.21873066822687784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,2,balanced,0.27053866783777875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,2,balanced,0.3773706754048665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,2,balanced,0.46326398849487305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,2,balanced,0.67630402247111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,2,balanced,0.8682346343994141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,8,balanced,0.06201066573460897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,8,balanced,0.0721013347307841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,8,balanced,0.09077866872151692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,8,balanced,0.12044800321261089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,8,balanced,0.16550933321317038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,8,balanced,0.19619200627009073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,8,balanced,0.19754133621851602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,8,balanced,0.19914132356643677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,8,balanced,0.20216000080108643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,8,balanced,0.20336000124613443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,8,balanced,0.2041706641515096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,8,balanced,0.2092640002568563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,8,balanced,0.20992533365885416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,8,balanced,0.21422932545344034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,8,balanced,0.22200000286102295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,8,balanced,0.22141865889231363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,8,balanced,0.23311465978622437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,8,balanced,0.25205334027608234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,8,balanced,0.2662559946378072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,8,balanced,0.3463840087254842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,8,balanced,0.34432001908620197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,8,balanced,0.49141331513722736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,8,balanced,0.4994399944941203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,8,balanced,0.7151573499043783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,8,balanced,0.7920213540395101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,8,balanced,1.1559253533681233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,8,balanced,1.3951573371887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,16,power_law_1.01,0.15336960554122925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,16,power_law_1.01,0.23749120235443116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,16,power_law_1.01,0.23971199989318848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,16,power_law_1.01,0.2414400100708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,16,power_law_1.01,0.2430783987045288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,16,power_law_1.01,0.23580799102783204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,16,power_law_1.01,0.23931519985198973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,16,power_law_1.01,0.23674240112304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,16,power_law_1.01,0.20429439544677735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,16,power_law_1.01,0.23434240818023683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,16,power_law_1.01,0.23975679874420167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,16,power_law_1.01,0.21892480850219725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,16,power_law_1.01,0.2160128116607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,16,power_law_1.01,0.24462080001831055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,16,power_law_1.01,0.25742719173431394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,16,power_law_1.01,0.25949440002441404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,16,power_law_1.01,0.2726016044616699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,16,power_law_1.01,0.3032639980316162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,16,power_law_1.01,0.3101183891296387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,16,power_law_1.01,0.3491391897201538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,16,power_law_1.01,0.3524928092956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,16,power_law_1.01,0.42648959159851074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,16,power_law_1.01,0.49696640968322753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,16,power_law_1.01,0.663647985458374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,16,power_law_1.01,0.781606388092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,16,power_law_1.01,1.0348095893859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,16,power_law_1.01,1.2364416122436523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,32,power_law_1.2,0.048825600743293764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,32,power_law_1.2,0.0474047988653183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,32,power_law_1.2,0.04601599872112274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,32,power_law_1.2,0.046105599403381346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,32,power_law_1.2,0.04576640129089356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,32,power_law_1.2,0.045337599515914914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,32,power_law_1.2,0.047091200947761536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,32,power_law_1.2,0.04733439981937408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,32,power_law_1.2,0.04814079999923706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,32,power_law_1.2,0.048416000604629514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,32,power_law_1.2,0.05052800178527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,32,power_law_1.2,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,32,power_law_1.2,0.053887999057769774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,32,power_law_1.2,0.05852159857749939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,32,power_law_1.2,0.06286720037460328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,32,power_law_1.2,0.06487680077552796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,32,power_law_1.2,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,32,power_law_1.2,0.07916799783706666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,32,power_law_1.2,0.08447999954223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,32,power_law_1.2,0.10123519897460938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,32,power_law_1.2,0.12128000259399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,32,power_law_1.2,0.15210880041122438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,32,power_law_1.2,0.18471039533615113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,32,power_law_1.2,0.2549504041671753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,32,power_law_1.2,0.33775360584259034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,32,power_law_1.2,0.4708928108215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,32,power_law_1.2,0.6288000106811523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,2,power_law_1.01,0.09411839842796325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,2,power_law_1.01,0.12800639867782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,2,power_law_1.01,0.14396799802780152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,2,power_law_1.01,0.20415360927581788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,2,power_law_1.01,0.2558144092559814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,2,power_law_1.01,0.37966721057891845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,2,power_law_1.01,0.397542405128479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,2,power_law_1.01,0.39336960315704345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,2,power_law_1.01,0.4081984043121338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,2,power_law_1.01,0.4075967788696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,2,power_law_1.01,0.42876157760620115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,2,power_law_1.01,0.43396477699279784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,2,power_law_1.01,0.43944320678710935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,2,power_law_1.01,0.45066242218017577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,2,power_law_1.01,0.47136640548706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,2,power_law_1.01,0.4712319850921631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,2,power_law_1.01,0.5043647766113282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,1,power_law_1.2,0.07960320115089417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,1,power_law_1.2,0.10703999996185302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,1,power_law_1.2,0.1513152003288269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,1,power_law_1.2,0.23985280990600585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,1,power_law_1.2,0.3185663938522339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,1,power_law_1.2,0.4286911964416504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,2,power_law_1.01,0.5834176063537597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,1,power_law_1.2,0.6176447868347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,1,power_law_1.2,0.6475391864776612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,1,power_law_1.2,0.6684671878814697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,1,power_law_1.2,0.6934336185455322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,2,power_law_1.01,0.606220817565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,2,power_law_1.01,0.7332672119140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,2,power_law_1.01,0.7601471900939941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,2,power_law_1.01,0.9565567970275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,1,power_law_1.2,0.7266176223754883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,1,power_law_1.2,0.7480063915252686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,2,power_law_1.01,1.1208064079284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,1,power_law_1.2,0.763974380493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,2,power_law_1.01,1.4548735618591309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,1,power_law_1.2,0.7964608192443847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,1,power_law_1.2,0.8515904426574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,2,power_law_1.01,1.8236543655395507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,1,power_law_1.2,0.846668815612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,1,power_law_1.2,0.9108160018920899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,2,power_law_1.01,2.6033727645874025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,1,power_law_1.2,1.017465591430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,1,power_law_1.2,1.0452799797058105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,1,power_law_1.2,1.2181952476501465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,2,power_law_1.01,3.237567901611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,1,power_law_1.2,1.2645631790161134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,1,power_law_1.2,1.519871997833252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,1,power_law_1.2,1.766080093383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,1,power_law_1.2,2.286246490478516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,1,power_law_1.2,2.7658815383911133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,1,power_law_1.2,3.7171070098876955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,1,power_law_1.2,4.688851165771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,16,power_law_1.01,0.07930240035057068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,16,power_law_1.01,0.09074559807777405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,16,power_law_1.01,0.0889855980873108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,16,power_law_1.01,0.08819839954376221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,16,power_law_1.01,0.08695679903030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,16,power_law_1.01,0.08427519798278808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,16,power_law_1.01,0.0889087975025177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,16,power_law_1.01,0.08617600202560424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,16,power_law_1.01,0.08705919981002808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,16,power_law_1.01,0.08387839794158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,16,power_law_1.01,0.08833280205726624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,16,power_law_1.01,0.09111679792404175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,16,power_law_1.01,0.0920960009098053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,16,power_law_1.01,0.09268479943275451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,16,power_law_1.01,0.10285439491271972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,16,power_law_1.01,0.10309760570526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,16,power_law_1.01,0.1096384048461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,16,power_law_1.01,0.12201600074768067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,16,power_law_1.01,0.12365440130233765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,16,power_law_1.01,0.14966399669647218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,16,power_law_1.01,0.16556799411773682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,16,power_law_1.01,0.19972480535507203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,16,power_law_1.01,0.23864960670471191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,16,power_law_1.01,0.3126271963119507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,16,power_law_1.01,0.3752768039703369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,16,power_law_1.01,0.5197951793670654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,16,power_law_1.01,0.6933504104614258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,8,power_law_1.2,0.048204800486564635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,8,power_law_1.2,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,8,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,8,power_law_1.2,0.05729920268058777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,8,power_law_1.2,0.05989120006561279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,8,power_law_1.2,0.06470400094985962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,8,power_law_1.2,0.06636800169944763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,8,power_law_1.2,0.06638079881668091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,8,power_law_1.2,0.06721280217170715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,8,power_law_1.2,0.06740480065345764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,32,4,power_law_1.01,0.048403200507164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,8,power_law_1.2,0.06964480280876159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,8,power_law_1.2,0.07210879921913146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,8,power_law_1.2,0.07365120053291321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,8,power_law_1.2,0.07512320280075073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,8,power_law_1.2,0.08350719809532166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,8,power_law_1.2,0.0864192008972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,8,power_law_1.2,0.0908352017402649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,8,power_law_1.2,0.10301439762115479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.2,0.11909760236740112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,32,4,power_law_1.01,0.05343359708786011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.2,0.1409216046333313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.2,0.15190399885177613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.2,0.2117759943008423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,32,4,power_law_1.01,0.05717759728431702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,32,4,power_law_1.01,0.059987199306488034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,32,4,power_law_1.01,0.06880639791488648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,32,4,power_law_1.01,0.07470719814300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,32,4,power_law_1.01,0.084934401512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,32,4,power_law_1.01,0.08284800052642823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,32,4,power_law_1.01,0.08387839794158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.2,0.2576832056045532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,32,4,power_law_1.01,0.08545920252799988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,32,4,power_law_1.01,0.08832640051841736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,32,4,power_law_1.01,0.08804479837417603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,32,4,power_law_1.01,0.08929920196533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,32,4,power_law_1.01,0.09564160108566284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,32,4,power_law_1.01,0.10154880285263061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,32,4,power_law_1.01,0.1043776035308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,32,4,power_law_1.01,0.11471999883651733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,32,4,power_law_1.01,0.1358464002609253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,32,4,power_law_1.01,0.14300800561904908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,32,4,power_law_1.01,0.17151360511779784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,32,4,power_law_1.01,0.19559040069580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,32,4,power_law_1.01,0.2574847936630249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,32,4,power_law_1.01,0.3077568054199219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,32,4,power_law_1.01,0.4136960029602051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,32,4,power_law_1.01,0.5488192081451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,32,4,power_law_1.01,0.7461696147918702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,32,4,power_law_1.01,0.9736063957214356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.2,0.3468224048614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.2,0.4769279956817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.2,0.6405504226684571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.2,0.7212160110473633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,16,power_law_1.01,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,16,power_law_1.01,0.03196159899234772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,16,power_law_1.01,0.031225600838661195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,16,power_law_1.01,0.031193599104881287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,16,power_law_1.01,0.031385600566864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,16,power_law_1.01,0.032595199346542356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,16,power_law_1.01,0.032441601157188416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,16,power_law_1.01,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,16,power_law_1.01,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,16,power_law_1.01,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,16,power_law_1.01,0.0336896002292633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,16,power_law_1.01,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,16,power_law_1.01,0.03506560027599335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,16,power_law_1.01,0.03792639970779419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,16,power_law_1.01,0.04162560105323791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,16,power_law_1.01,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,16,power_law_1.01,0.045177599787712096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,16,power_law_1.01,0.04906240105628967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,16,power_law_1.01,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,16,power_law_1.01,0.06286079883575439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,16,power_law_1.01,0.07219200134277344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,16,power_law_1.01,0.08480640053749085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,16,power_law_1.01,0.1098688006401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,16,power_law_1.01,0.14839040040969848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,16,power_law_1.01,0.18302719593048095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,16,power_law_1.01,0.24148480892181395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,16,power_law_1.01,0.30272641181945803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,16,power_law_1.2,0.08416640162467956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,16,power_law_1.2,0.0898751974105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,16,power_law_1.2,0.09054080247879029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,16,power_law_1.2,0.09303680062294006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,16,power_law_1.2,0.08865919709205627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,16,power_law_1.2,0.09259520173072815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,16,power_law_1.2,0.09422720074653626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,16,power_law_1.2,0.09192960262298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,16,power_law_1.2,0.09626240134239197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,16,power_law_1.2,0.09478399753570557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,16,power_law_1.2,0.10063999891281128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,16,power_law_1.2,0.10328320264816285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,16,power_law_1.2,0.10615040063858032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,16,power_law_1.2,0.10905599594116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,16,power_law_1.2,0.11721600294113159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,16,power_law_1.2,0.11712000370025635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,16,power_law_1.2,0.13077759742736816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,16,power_law_1.2,0.15381759405136108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,16,power_law_1.2,0.16637439727783204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,16,power_law_1.2,0.2103424072265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,16,power_law_1.2,0.2394047975540161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,16,power_law_1.2,0.3269695997238159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,16,power_law_1.2,0.382207989692688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,16,power_law_1.2,0.5177023887634278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,16,power_law_1.2,0.6144832134246826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,16,power_law_1.2,0.9190719604492188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,16,power_law_1.2,1.1571200370788575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,4,power_law_1.01,0.04991360008716583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,4,power_law_1.01,0.05162879824638367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,4,power_law_1.01,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,4,power_law_1.01,0.06828160285949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,4,power_law_1.01,0.07985280156135559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,4,power_law_1.01,0.08920959830284118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,4,power_law_1.01,0.0933247983455658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,4,power_law_1.01,0.09823359847068787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,4,power_law_1.01,0.09827200174331666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,4,power_law_1.01,0.1011199951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,4,power_law_1.01,0.10229120254516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,4,power_law_1.01,0.10640000104904175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,4,power_law_1.01,0.1067199945449829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,4,power_law_1.01,0.10748159885406494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,4,power_law_1.01,0.12012159824371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,4,power_law_1.01,0.11901439428329467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,4,power_law_1.01,0.13058559894561766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,4,power_law_1.01,0.14415359497070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,4,power_law_1.01,0.1643839955329895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,4,power_law_1.01,0.18894720077514648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,4,power_law_1.01,0.223091197013855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,4,power_law_1.01,0.27886080741882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,4,power_law_1.01,0.34479360580444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,4,power_law_1.01,0.4623295783996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,4,power_law_1.01,0.5851967811584473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,4,power_law_1.01,0.812992000579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,4,power_law_1.01,1.0685183525085449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,4,balanced,0.04443199932575226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,4,balanced,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,4,balanced,0.0480373352766037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,4,balanced,0.054154664278030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,4,balanced,0.07422933479150136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,4,balanced,0.10280000170071919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,4,balanced,0.10548266768455505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,4,balanced,0.10698133707046509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,4,balanced,0.1067573328812917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,4,balanced,0.1088106632232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,4,balanced,0.10915199915568034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,4,balanced,0.11036800344785054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,4,balanced,0.11346667011578877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,4,balanced,0.11591999729474385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,4,balanced,0.12065066893895467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,4,balanced,0.12345066666603088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,4,balanced,0.12948800126711527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,4,balanced,0.1421066621939341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,4,balanced,0.15105066696802774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,4,balanced,0.17453867197036743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,4,balanced,0.19209599494934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,4,balanced,0.24802666902542114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,4,balanced,0.28510934114456177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,4,balanced,0.3956426779429118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,4,balanced,0.49755199750264484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,4,balanced,0.707472006479899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,4,balanced,0.8877867062886556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,2,power_law_1.2,0.030687999725341798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,2,power_law_1.2,0.03280639946460724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,2,power_law_1.2,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,2,power_law_1.2,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,2,power_law_1.2,0.033817601203918454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,2,power_law_1.2,0.03445119857788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,2,power_law_1.2,0.0377920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,2,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,2,power_law_1.2,0.040940800309181215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,2,power_law_1.2,0.044095999002456664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,2,power_law_1.2,0.043270400166511534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,2,power_law_1.2,0.047513601183891294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,2,power_law_1.2,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,2,power_law_1.2,0.04798080027103424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,2,power_law_1.2,0.05790079832077026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,2,power_law_1.2,0.0640447974205017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,2,power_law_1.2,0.061894398927688596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,2,power_law_1.2,0.07481600046157837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,2,power_law_1.2,0.08102399706840516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,2,power_law_1.2,0.09679359793663025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,2,power_law_1.2,0.11756800413131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,2,power_law_1.2,0.1313024044036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,2,power_law_1.2,0.15724159479141236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,2,balanced,0.0853653351465861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,2,balanced,0.12121066451072693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,2,balanced,0.2122453252474467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,2,balanced,0.21209067106246948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,2,balanced,0.21114667256673178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,2,balanced,0.21310933430989584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,2,balanced,0.21322667598724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,2,balanced,0.21033066511154175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,2,balanced,0.21174399058024088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,2,balanced,0.21335999170939127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,2,balanced,0.22372267643610635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,2,balanced,0.2136533260345459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,2,balanced,0.2141653299331665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,2,balanced,0.23701866467793783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,2,balanced,0.24352532625198364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,2,balanced,0.24537599086761475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,2,balanced,0.25277866919835407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,2,balanced,0.29517332712809247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,2,balanced,0.30502400795618695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,2,balanced,0.3875360091527303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,2,balanced,0.4178026517232259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,2,balanced,0.581285317738851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,2,balanced,0.7529226938883463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,2,balanced,1.058128039042155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,2,power_law_1.2,0.22037119865417482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,2,balanced,1.4751040140787761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,2,balanced,2.228794733683268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,2,balanced,2.7610133488972983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,2,power_law_1.2,0.27757439613342283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,64,power_law_1.01,0.055244797468185426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,64,power_law_1.01,0.044896000623703004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,64,power_law_1.01,0.04469119906425476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,64,power_law_1.01,0.04535039961338043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,64,power_law_1.01,0.04478079974651337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,64,power_law_1.01,0.04524799883365631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,64,power_law_1.01,0.045484799146652224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,64,power_law_1.01,0.04611839950084686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,64,power_law_1.01,0.04664320051670075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,64,power_law_1.01,0.04688639938831329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,64,power_law_1.01,0.047712001204490664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,64,power_law_1.01,0.04850560128688812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,64,power_law_1.01,0.04970879852771759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,64,power_law_1.01,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,64,power_law_1.01,0.05597440004348755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,64,power_law_1.01,0.056492799520492555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,64,power_law_1.01,0.059436798095703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,64,power_law_1.01,0.06673280000686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,64,power_law_1.01,0.07077760100364686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,64,power_law_1.01,0.08199679851531982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,64,power_law_1.01,0.09317759871482849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,64,power_law_1.01,0.1202239990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,64,power_law_1.01,0.14712320566177367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,64,power_law_1.01,0.2133631944656372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,64,power_law_1.01,0.2554431915283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,64,power_law_1.01,0.372544002532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,64,power_law_1.01,0.4835072040557861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,2,power_law_1.2,0.36630399227142335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,64,balanced,0.04724800089995066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,64,balanced,0.046442667643229164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,64,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,64,balanced,0.04996799925963084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,64,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,64,balanced,0.057999998331069946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,64,balanced,0.06010133524735769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,64,balanced,0.06015466650327047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,64,balanced,0.06437333424886067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,64,balanced,0.060085331400235496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,64,balanced,0.062394668658574425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,64,balanced,0.07047466437021892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,64,balanced,0.0718453327814738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,64,balanced,0.06673066814740498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,64,balanced,0.07761600116888683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,64,balanced,0.08076799909273784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,64,balanced,0.08270399769147237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,64,balanced,0.09497599800427754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,64,balanced,0.10079999764760335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,64,balanced,0.11450666189193726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,64,balanced,0.12825600306193033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,64,balanced,0.15915733575820923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,64,balanced,0.18560532728830972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,64,balanced,0.23711466789245605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,64,balanced,0.2835093339284261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,64,balanced,0.382698655128479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,64,balanced,0.4957386652628581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,16,balanced,0.04470400015513102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,16,balanced,0.04557333389918009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,16,balanced,0.04570133487383524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,16,balanced,0.051311999559402466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,16,balanced,0.07008000214894612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,16,balanced,0.09733333190282185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,2,power_law_1.2,0.45060482025146487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,16,balanced,0.09910399715105693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,16,balanced,0.0999786655108134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,16,balanced,0.09941333532333374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,16,balanced,0.10132267077763875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,16,balanced,0.10070400436719258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,16,balanced,0.10317867000897725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,16,balanced,0.10196266571680705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,16,balanced,0.10458667079607646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,16,balanced,0.10707733035087585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,16,balanced,0.11038933197657268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,16,balanced,0.11413333813349406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,16,balanced,0.11959466338157654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,16,balanced,0.12501866618792215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,16,balanced,0.1383680005868276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,16,balanced,0.1530506710211436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,16,balanced,0.17840532461802164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,16,balanced,0.20777599016825357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,16,balanced,0.2539786696434021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,16,balanced,0.3152800003687541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,16,balanced,0.4246879816055298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,16,balanced,0.5273439884185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,32,power_law_1.01,0.07032960057258605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,32,power_law_1.01,0.04992640018463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,32,power_law_1.01,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,32,power_law_1.01,0.0489984005689621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,32,power_law_1.01,0.050246399641036985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,32,power_law_1.01,0.05074560046195984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,32,power_law_1.01,0.0521664023399353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,32,power_law_1.01,0.055430400371551516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,32,power_law_1.01,0.05616639852523804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,32,power_law_1.01,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,32,power_law_1.01,0.057158398628234866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,32,power_law_1.01,0.057126402854919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,32,power_law_1.01,0.058329600095748904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,32,power_law_1.01,0.06217600107192993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,32,power_law_1.01,0.06769919991493226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,32,power_law_1.01,0.07023360133171082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,32,power_law_1.01,0.07749760150909424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,32,power_law_1.01,0.08539519906044006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,32,power_law_1.01,0.08967679738998413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,32,power_law_1.01,0.10821759700775146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,32,power_law_1.01,0.12440320253372192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,32,power_law_1.01,0.16219520568847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,32,power_law_1.01,0.1824895977973938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,32,power_law_1.01,0.25708799362182616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,32,power_law_1.01,0.30008320808410643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,32,power_law_1.01,0.425600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,32,power_law_1.01,0.5239935874938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,128,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,128,balanced,0.048767998814582825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,128,balanced,0.04852266609668732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,128,balanced,0.0535093347231547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,128,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,128,balanced,0.058320000767707825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,128,balanced,0.057664001981417336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,128,balanced,0.05788266658782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,128,balanced,0.05615466833114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,128,balanced,0.058090666929880776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,128,balanced,0.05797866483529409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,128,balanced,0.060266668597857155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,128,balanced,0.06029333174228668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,128,balanced,0.05852800110975901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,128,balanced,0.06445333361625671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,128,balanced,0.065461332599322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,128,balanced,0.06601066887378693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,128,balanced,0.07216000060240428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,128,balanced,0.0763626645008723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,128,balanced,0.08458667000134786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,128,balanced,0.09130133191744487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,128,balanced,0.10917333761850993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,128,balanced,0.12593600153923035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,128,balanced,0.1729653278986613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,128,balanced,0.2119306723276774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,128,balanced,0.28356266021728516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,128,balanced,0.3588053385416667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,1,balanced,0.05301333467165629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,1,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,1,balanced,0.07323733468850453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,1,balanced,0.10045866171518962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,1,balanced,0.14814399679501852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,1,balanced,0.18260266383488974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,1,balanced,0.18420799573262533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,1,balanced,0.18418665726979574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,1,balanced,0.18642133474349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,1,balanced,0.18819200992584229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,1,balanced,0.190175990263621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,1,balanced,0.1934559941291809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,1,balanced,0.19301333030064902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,1,balanced,0.1997119983037313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,1,balanced,0.20459733406702676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,1,balanced,0.21039466063181558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,1,balanced,0.22080000241597494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,1,balanced,0.24052266279856363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,1,balanced,0.25996800263722736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,1,balanced,0.38763201236724854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,1,balanced,0.3486826817194621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,1,balanced,0.5120213429133097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,1,balanced,0.5655839840571085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,1,balanced,0.8349493344624838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,1,balanced,0.9914666811625162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,1,balanced,1.4581066767374675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,1,balanced,1.8133440017700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,4,balanced,0.027162666122118633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,4,balanced,0.025199999411900837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,4,balanced,0.029125332832336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,4,balanced,0.04197333256403605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,4,balanced,0.056517332792282104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,4,balanced,0.0662720004717509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,4,balanced,0.06900266806284587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,4,balanced,0.06805866460005443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,4,balanced,0.07010133564472198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,4,balanced,0.07623466849327087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,4,balanced,0.0739519993464152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,4,balanced,0.07612266639868419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,4,balanced,0.08100800216197968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,4,balanced,0.0867680013179779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,4,balanced,0.09506133198738098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,4,balanced,0.10100266337394714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,4,balanced,0.11274133125940959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,4,balanced,0.11723732948303223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,4,balanced,0.1957333286603292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,4,balanced,0.23450666666030884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,4,balanced,0.19975467522939047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,4,balanced,0.22802132368087769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,4,balanced,0.25651733080546063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,4,balanced,0.3928266763687134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,4,balanced,0.4562346537907918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,4,balanced,0.6545173327128092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,4,balanced,0.8619679609934489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,16,power_law_1.2,0.024915200471878052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,16,power_law_1.2,0.04023039937019348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,16,power_law_1.2,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,16,power_law_1.2,0.038022398948669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,16,power_law_1.2,0.03704319894313812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,16,power_law_1.2,0.03028480112552643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,16,power_law_1.2,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,16,power_law_1.2,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,16,power_law_1.2,0.03909760117530823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,16,power_law_1.2,0.041171199083328246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,16,power_law_1.2,0.04053759872913361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,16,power_law_1.2,0.042508798837661746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,16,power_law_1.2,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,16,power_law_1.2,0.0500544011592865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,16,power_law_1.2,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,16,power_law_1.2,0.0549888014793396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,16,power_law_1.2,0.08097280263900757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,16,power_law_1.2,0.08584319949150085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,16,power_law_1.2,0.10091520547866821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,16,power_law_1.2,0.1287935972213745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,16,power_law_1.2,0.09034240245819092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,16,power_law_1.2,0.10390399694442749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,16,power_law_1.2,0.12434560060501099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,16,power_law_1.2,0.17597440481185914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,16,power_law_1.2,0.23070719242095947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,16,power_law_1.2,0.3141247987747192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,16,power_law_1.2,0.38920960426330564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,128,power_law_1.2,0.047916799783706665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,128,power_law_1.2,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,128,power_law_1.2,0.047276800870895384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,128,power_law_1.2,0.049593600630760196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,128,power_law_1.2,0.04999679923057556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,128,power_law_1.2,0.050969600677490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,128,power_law_1.2,0.05137280225753784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,128,power_law_1.2,0.051795202493667605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,128,power_law_1.2,0.05186560153961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,128,power_law_1.2,0.05252479910850525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,128,power_law_1.2,0.05309439897537231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,128,power_law_1.2,0.054201602935791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,128,power_law_1.2,0.05537279844284058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,128,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,128,power_law_1.2,0.059539198875427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,128,power_law_1.2,0.0599295973777771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,128,power_law_1.2,0.06423680186271667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,128,power_law_1.2,0.07011839747428894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,128,power_law_1.2,0.0763584017753601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,128,power_law_1.2,0.08847360014915466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,128,power_law_1.2,0.10122239589691162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,128,power_law_1.2,0.13126399517059326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,128,power_law_1.2,0.1502079963684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,128,power_law_1.2,0.20553600788116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,128,power_law_1.2,0.25971839427947996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,128,power_law_1.2,0.3641024112701416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,128,power_law_1.2,0.5074368000030518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,16,power_law_1.01,0.13435519933700563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,16,power_law_1.01,0.22064640522003173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,16,power_law_1.01,0.21587200164794923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,16,power_law_1.01,0.2170367956161499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,16,power_law_1.01,0.19985920190811157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,16,power_law_1.01,0.18845440149307252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,16,power_law_1.01,0.18483200073242187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,16,power_law_1.01,0.21143040657043458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,16,power_law_1.01,0.20099198818206787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,16,power_law_1.01,0.19704320430755615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,16,power_law_1.01,0.18846720457077026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,16,power_law_1.01,0.20432000160217284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,16,power_law_1.01,0.20793600082397462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,16,power_law_1.01,0.21422719955444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,16,power_law_1.01,0.21833600997924804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,16,power_law_1.01,0.21676158905029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,16,power_law_1.01,0.23494400978088378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,16,power_law_1.01,0.28222079277038575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,16,power_law_1.01,0.256710410118103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,16,power_law_1.01,0.3273216009140015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,16,power_law_1.01,0.31580801010131837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,16,power_law_1.01,0.3882688045501709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,16,power_law_1.01,0.4557951927185059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,16,power_law_1.01,0.5405759811401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,16,power_law_1.01,0.669209623336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,16,power_law_1.01,0.930726432800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,16,power_law_1.01,1.2207103729248048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,8,balanced,0.03586666782697042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,8,balanced,0.035829332967599235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,8,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,8,balanced,0.03782933453718821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,8,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,8,balanced,0.039173332353432976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,8,balanced,0.03956266740957896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,8,balanced,0.03993066648642222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,8,balanced,0.03987200061480204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,8,balanced,0.040778666734695435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,8,balanced,0.039834665755430855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,8,balanced,0.04374399781227112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,8,balanced,0.04192000130812327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,8,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,8,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,8,balanced,0.056186666091283165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,8,balanced,0.056090667843818665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,8,balanced,0.06418133278687795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,8,balanced,0.06676266590754192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,8,balanced,0.0832533339659373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,8,balanced,0.09166399637858073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,8,balanced,0.11935466527938843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,8,balanced,0.1485919952392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,8,balanced,0.1975839932759603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,8,balanced,0.24406399329503378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,8,balanced,0.33904532591501874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,8,balanced,0.44998931884765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,64,balanced,0.021354667842388153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,64,balanced,0.02332799881696701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,64,balanced,0.0230880007147789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,64,balanced,0.019178666174411774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,64,balanced,0.020901332298914593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,64,balanced,0.02232533444960912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,64,balanced,0.03257066756486893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,64,balanced,0.029098667204380035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,64,balanced,0.03738666574160258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,64,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,64,balanced,0.03942933430274328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,64,balanced,0.05236266553401947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,64,balanced,0.05264000097910563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,64,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,64,balanced,0.05446400245030721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,64,balanced,0.05407466491063436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,64,balanced,0.0687306672334671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,64,balanced,0.08567466338475545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,64,balanced,0.11480533083279927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,64,balanced,0.15236266454060873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,64,balanced,0.09292266766230266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,64,balanced,0.1190826694170634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,64,balanced,0.15126933654149374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,64,balanced,0.2109866738319397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,64,balanced,0.2717546621958415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,64,balanced,0.3931573232014974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,64,balanced,0.5158400138219198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,2,balanced,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,2,balanced,0.049786667029062905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,2,balanced,0.05727999905745188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,2,balanced,0.08078933258851369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,2,balanced,0.11246933539708455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,2,balanced,0.17324266831080118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,2,balanced,0.1782346765200297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,2,balanced,0.18030399084091187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,2,balanced,0.18101332585016885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,2,balanced,0.19089066982269287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,2,balanced,0.18287465969721475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,2,balanced,0.188426673412323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,2,balanced,0.19164800643920898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,2,balanced,0.19143466154734293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,2,balanced,0.19606933991114298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,2,balanced,0.20059200127919516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,2,balanced,0.21861867109934488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,2,balanced,0.24543466170628866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,2,balanced,0.2633226712544759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,2,balanced,0.3022879958152771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,2,balanced,0.3340533177057902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,2,balanced,0.4494933287302653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,2,balanced,0.5083306630452474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,2,balanced,0.7290133635203043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,2,balanced,0.8829493522644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,2,balanced,1.2668053309122722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,2,balanced,1.6155734062194824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,1,balanced,0.03336533407370249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,1,balanced,0.029466666281223297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,1,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,1,balanced,0.054714664816856384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,1,balanced,0.0864533285299937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,1,balanced,0.08749866485595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,1,balanced,0.08684800068537395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,1,balanced,0.08894399801890056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,1,balanced,0.0906773308912913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,1,balanced,0.09130666653315227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,1,balanced,0.09165866176287334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,1,balanced,0.09402133027712505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,1,balanced,0.09442133704821269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,1,balanced,0.09897599617640178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,1,balanced,0.10799466570218404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,1,balanced,0.11140799522399902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,1,balanced,0.11441066861152649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,1,balanced,0.1318933367729187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,1,balanced,0.13537066181500754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,1,balanced,0.17494400342305502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,1,balanced,0.188810666402181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,1,balanced,0.27348266045252484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,1,balanced,0.31941332419713336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,1,balanced,0.4636160135269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,1,balanced,0.5775359869003296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,1,balanced,0.8391626675923666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,1,balanced,1.0828373432159424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,8,power_law_1.2,0.022592000663280487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,8,power_law_1.2,0.023001599311828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,8,power_law_1.2,0.0244159996509552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,8,power_law_1.2,0.02492160052061081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,8,power_law_1.2,0.02648960053920746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,8,power_law_1.2,0.02863999903202057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,8,power_law_1.2,0.028678399324417115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,8,power_law_1.2,0.030156800150871278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,8,power_law_1.2,0.030771198868751525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,8,power_law_1.2,0.030399999022483824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,8,power_law_1.2,0.031436800956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,8,power_law_1.2,0.03946239948272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,8,power_law_1.2,0.03904640078544617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,8,power_law_1.2,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,8,power_law_1.2,0.06676480174064636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,8,power_law_1.2,0.06830719709396363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,8,power_law_1.2,0.07166720032691956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,8,power_law_1.2,0.07136639952659607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.2,0.08085759878158569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.2,0.08969600200653076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.2,0.09959040284156799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.2,0.11871360540390015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.2,0.14529919624328613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.2,0.1968832015991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.2,0.24637439250946044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.2,0.33445119857788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.2,0.4350592136383057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,8,balanced,0.03994666785001755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,8,balanced,0.03959999978542328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,8,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,8,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,8,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,8,balanced,0.039701332648595176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,8,balanced,0.039621333281199135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,8,balanced,0.039674667020638786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,8,balanced,0.04167466859022776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,8,balanced,0.04101333270470301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,8,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,8,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,8,balanced,0.04557333389918009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,8,balanced,0.04785599807898203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,8,balanced,0.05611200133959452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,8,balanced,0.05773333211739858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,8,balanced,0.061946665247281395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,1,power_law_1.2,0.07573760151863099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,8,balanced,0.07310933371384938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,8,balanced,0.0766186664501826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,1,power_law_1.2,0.09316480159759521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,1,power_law_1.2,0.09920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,1,power_law_1.2,0.11370240449905396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,1,power_law_1.2,0.1248128056526184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,1,power_law_1.2,0.16346240043640137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,1,power_law_1.2,0.17242879867553712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,8,balanced,0.10080533226331075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,1,power_law_1.2,0.1751360058784485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,8,balanced,0.10654933253924052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,8,balanced,0.13380799690882364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,1,power_law_1.2,0.1776128053665161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,8,balanced,0.15821866194407144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,8,balanced,0.2095253268877665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,1,power_law_1.2,0.18035199642181396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,8,balanced,0.2597759962081909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,8,balanced,0.3604213396708171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,8,balanced,0.45979201793670654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,1,power_law_1.2,0.18499840497970582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,1,power_law_1.2,0.18831360340118408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,1,power_law_1.2,0.1916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,1,power_law_1.2,0.19777920246124267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,1,power_law_1.2,0.1955135941505432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,1,power_law_1.2,0.20071039199829102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,1,power_law_1.2,0.20629119873046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,1,power_law_1.2,0.2339711904525757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,1,power_law_1.2,0.25785601139068604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,1,power_law_1.2,0.30995841026306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,1,power_law_1.2,0.36669440269470216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,1,power_law_1.2,0.4875199794769287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,1,power_law_1.2,0.6018815994262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,1,power_law_1.2,0.8445311546325683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,1,power_law_1.2,1.0941632270812989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,1,power_law_1.2,1.5718144416809081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,1,power_law_1.2,2.0428224563598634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,4,power_law_1.01,0.019948799908161164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,4,power_law_1.01,0.01930239945650101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,4,power_law_1.01,0.019603200256824493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,4,power_law_1.01,0.019603200256824493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,4,power_law_1.01,0.019916799664497376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,4,power_law_1.01,0.020377600193023683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,4,power_law_1.01,0.023180800676345825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,4,power_law_1.01,0.022726400196552275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,4,power_law_1.01,0.04723199903964996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,4,power_law_1.01,0.04649600088596344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,4,power_law_1.01,0.04784640073776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,4,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,4,power_law_1.01,0.03663359880447388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,4,power_law_1.01,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,4,power_law_1.01,0.03741439878940582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,4,power_law_1.01,0.03711999952793121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,4,power_law_1.01,0.041247999668121337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,4,power_law_1.01,0.04368639886379242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.01,0.04657280147075653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.01,0.06735360026359558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.01,0.07601280212402343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.01,0.09931520223617554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.01,0.12422399520874024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,4,balanced,0.02311466634273529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,4,balanced,0.023183998962243397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,4,balanced,0.023029332359631855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,4,balanced,0.023082666099071503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,4,balanced,0.023434666295846302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,4,balanced,0.026170666019121807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,4,balanced,0.03908266623814901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,4,balanced,0.030234667162100475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,4,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,4,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,4,balanced,0.036714665591716766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,4,balanced,0.04163199911514918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,4,balanced,0.046015997727712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,4,balanced,0.04762666424115499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,4,balanced,0.06846400101979573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,4,balanced,0.05648000041643778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,4,balanced,0.06404800216356914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,4,balanced,0.0848640004793803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,4,balanced,0.10122133294741313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,4,balanced,0.14085333546002707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,4,balanced,0.17416000366210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,4,balanced,0.24314665794372559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,4,balanced,0.3035893241564433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,4,balanced,0.43926934401194256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,4,balanced,0.5667946736017863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,4,balanced,0.8308800061543783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,4,balanced,1.0946880181630452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.01,0.1716096043586731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.01,0.20933759212493896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,8,power_law_1.01,0.04755200147628784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,8,power_law_1.01,0.05258880257606506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,8,power_law_1.01,0.054636800289154054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,8,power_law_1.01,0.0537280023097992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,8,power_law_1.01,0.05658239722251892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,8,power_law_1.01,0.05861120223999024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,8,power_law_1.01,0.060057598352432254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.01,0.3060672044754028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,8,power_law_1.01,0.062009602785110474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,8,power_law_1.01,0.06174719929695129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,8,power_law_1.01,0.061868798732757566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,8,power_law_1.01,0.06367999911308289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,8,power_law_1.01,0.06599680185317994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,8,power_law_1.01,0.0688704013824463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,8,power_law_1.01,0.07381759881973267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,8,power_law_1.01,0.0803391993045807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,8,power_law_1.01,0.08604159951210022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,8,power_law_1.01,0.08680959939956664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,8,power_law_1.01,0.1019968032836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,8,power_law_1.01,0.1196287989616394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,8,power_law_1.01,0.1565824031829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,8,power_law_1.01,0.17950079441070557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,8,power_law_1.01,0.2389888048171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,8,power_law_1.01,0.3033152103424072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,8,power_law_1.01,0.4193984031677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,8,power_law_1.01,0.5263296127319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,8,power_law_1.01,0.7778687953948975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,8,power_law_1.01,1.0003904342651366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.01,0.39231998920440675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,1,power_law_1.01,0.07338240146636962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,1,power_law_1.01,0.0875328004360199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,1,power_law_1.01,0.11361279487609863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,1,power_law_1.01,0.15820800065994262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,1,power_law_1.01,0.16151679754257203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,1,power_law_1.01,0.16647679805755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,1,power_law_1.01,0.16597119569778443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,1,power_law_1.01,0.16788480281829835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,1,power_law_1.01,0.17034879922866822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,1,power_law_1.01,0.1709247946739197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,1,power_law_1.01,0.1772160053253174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,1,power_law_1.01,0.1800447940826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,1,power_law_1.01,0.18988159894943238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,1,power_law_1.01,0.19141119718551636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,1,power_law_1.01,0.21502079963684081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,1,power_law_1.01,0.2173504114151001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,1,power_law_1.01,0.23834240436553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,1,power_law_1.01,0.27749760150909425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.01,0.3241215944290161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.01,0.40926079750061034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.01,0.4894144058227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.01,0.6683008193969726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.01,0.84835844039917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.01,1.2130240440368651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,16,power_law_1.01,0.06717439889907836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,16,power_law_1.01,0.09340800046920776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,16,power_law_1.01,0.08474239706993103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,16,power_law_1.01,0.0837823987007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,16,power_law_1.01,0.08036479949951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,16,power_law_1.01,0.0819711983203888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,16,power_law_1.01,0.08430079817771911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,16,power_law_1.01,0.08604800105094909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,16,power_law_1.01,0.08472319841384887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,16,power_law_1.01,0.08658559918403626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,16,power_law_1.01,0.08641279935836792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,16,power_law_1.01,0.08929280042648316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,16,power_law_1.01,0.0889792025089264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,16,power_law_1.01,0.09348480105400085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,16,power_law_1.01,0.09757440090179444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,16,power_law_1.01,0.10358400344848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,16,power_law_1.01,0.10780800580978393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,16,power_law_1.01,0.1270143985748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.01,0.1312064051628113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.01,1.584012794494629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.01,0.15871360301971435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.01,0.16698240041732787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.01,0.2138432025909424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.01,0.2614464044570923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.01,0.3409600019454956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.01,0.43742079734802247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.01,0.6925631999969483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.01,0.8209407806396485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,32,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,32,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,32,balanced,0.031066666046778362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,32,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,32,balanced,0.03550933301448822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,32,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,32,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,32,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,32,balanced,0.03745600084463755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,32,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,32,balanced,0.036357333262761436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,32,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,32,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,32,balanced,0.039605334401130676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,32,balanced,0.041749333341916404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,32,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,32,balanced,0.04165333261092504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,32,balanced,0.043552001317342125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,32,balanced,0.0446720023949941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,32,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,32,balanced,0.04972266654173533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,32,balanced,0.0545066644748052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,32,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,32,balanced,0.07560533285140991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,32,balanced,0.08695466319719951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,32,balanced,0.1112000048160553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,32,balanced,0.14121599992116293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.01,2.3733823776245115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.01,3.0075328826904295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,128,power_law_1.2,0.06850559711456299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,128,power_law_1.2,0.06821759939193725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,128,power_law_1.2,0.06680319905281067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,128,power_law_1.2,0.07226880192756653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,128,power_law_1.2,0.07020800113677979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,128,power_law_1.2,0.06940159797668458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,128,power_law_1.2,0.07349759936332703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,128,power_law_1.2,0.08122239708900451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,128,power_law_1.2,0.08426240086555481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,128,power_law_1.2,0.08452479839324951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,128,power_law_1.2,0.07964159846305847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,128,power_law_1.2,0.08782079815864563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,128,power_law_1.2,0.08733440041542054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,128,power_law_1.2,0.09112319946289063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,128,power_law_1.2,0.10202879905700683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,128,power_law_1.2,0.1015936017036438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,128,power_law_1.2,0.10915199518203736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,128,power_law_1.2,0.1209663987159729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,128,power_law_1.2,0.1341375946998596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,128,power_law_1.2,0.16158080101013184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,128,power_law_1.2,0.18942079544067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,128,power_law_1.2,0.25608320236206056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,128,power_law_1.2,0.3235071897506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,128,power_law_1.2,0.45623040199279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,128,power_law_1.2,0.6209280014038085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,128,power_law_1.2,0.9311871528625488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,128,power_law_1.2,1.3194047927856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,2,balanced,0.02306666721900304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,2,balanced,0.025381334125995636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,2,balanced,0.03938666731119156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,2,balanced,0.05541866521040598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,2,balanced,0.08636800448099773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,2,balanced,0.13827733198801676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,2,balanced,0.14247999588648477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,2,balanced,0.14455999930699667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,2,balanced,0.14998400211334229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,2,balanced,0.1530026694138845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,2,balanced,0.15403733650843301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,2,balanced,0.16191466649373373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,2,balanced,0.16729599237442017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,2,balanced,0.164901336034139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,2,balanced,0.170741339524587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,2,balanced,0.17697600523630777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,2,balanced,0.19149333238601685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,2,balanced,0.21820267041524252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,2,balanced,0.2331999937693278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,2,balanced,0.29638399680455524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,2,balanced,0.2678719957669576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,2,balanced,0.36797332763671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,2,balanced,0.4471893310546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,2,balanced,0.5119733413060507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,2,balanced,0.5640106598536173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,2,balanced,0.9398399988810221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,2,balanced,1.0732266902923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,256,balanced,0.04470400015513102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,256,balanced,0.04384533564249674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,256,balanced,0.043765331308046974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,256,balanced,0.046256000796953835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,256,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,256,balanced,0.04864533245563507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,256,balanced,0.04806933303674062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,256,balanced,0.04805333415667216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,128,power_law_1.2,0.03813120126724243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,128,power_law_1.2,0.03785600066184998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,128,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,128,power_law_1.2,0.03988479971885681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,128,power_law_1.2,0.039027199149131775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,128,power_law_1.2,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,128,power_law_1.2,0.04069119989871979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,128,power_law_1.2,0.040191999077796935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,128,power_law_1.2,0.0406464010477066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,128,power_law_1.2,0.0403903990983963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,128,power_law_1.2,0.041280001401901245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,128,power_law_1.2,0.04156799912452698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,128,power_law_1.2,0.0418368011713028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,256,balanced,0.05220800141493479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,256,balanced,0.05041066805521647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,128,power_law_1.2,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,128,power_law_1.2,0.04678399860858917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,128,power_law_1.2,0.047065600752830505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,128,power_law_1.2,0.049292799830436704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,128,power_law_1.2,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,128,power_law_1.2,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,128,power_law_1.2,0.07942399978637696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,128,power_law_1.2,0.09387519955635071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,128,power_law_1.2,0.1420415997505188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,256,balanced,0.05403199791908264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,256,balanced,0.058090666929880776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,256,balanced,0.05383466680844625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,128,power_law_1.2,0.17713919878005982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,256,balanced,0.057989334066708885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,256,balanced,0.06253866851329803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,256,balanced,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,256,balanced,0.06489066779613495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,256,balanced,0.07314133147398631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,256,balanced,0.07508266468842824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,256,balanced,0.08508800466855367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,256,balanced,0.09537600477536519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,128,power_law_1.2,0.245580792427063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,256,balanced,0.11851200461387634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,256,balanced,0.13869333267211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,256,balanced,0.18214933077494302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,128,power_law_1.2,0.31128320693969724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,256,balanced,0.22028799851735434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,256,balanced,0.30956800778706867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,256,balanced,0.39233601093292236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,128,power_law_1.2,0.4647808074951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,128,power_law_1.2,0.5930816173553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,1,balanced,0.05013866722583771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,1,balanced,0.05417599777380625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,1,balanced,0.06853866577148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,1,balanced,0.07436266541481018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,1,balanced,0.07021333277225494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,1,balanced,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,1,balanced,0.07187200089295705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,1,balanced,0.07220800220966339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,1,balanced,0.07228266696135204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,1,balanced,0.07431999842325847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,1,balanced,0.07222933570543925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,1,balanced,0.07295466462771098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,1,balanced,0.07233599821726482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,1,balanced,0.07453866799672444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,1,balanced,0.08111466467380524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,1,balanced,0.08081066608428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,1,balanced,0.0848586658636729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,1,balanced,0.09880000352859497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,1,balanced,0.10697600245475769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,1,balanced,0.1402239998181661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,1,balanced,0.15650666753451029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,1,balanced,0.21119999885559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,1,balanced,0.263754665851593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,1,balanced,0.3659786780675252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,1,balanced,0.457258661588033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,1,balanced,0.6701119740804037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,1,balanced,0.8707520167032877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,4,power_law_1.2,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,4,power_law_1.2,0.04240640103816986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,4,power_law_1.2,0.04202240109443665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,4,power_law_1.2,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,4,power_law_1.2,0.04567680060863495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,4,power_law_1.2,0.04555520117282867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,4,power_law_1.2,0.04588800072669983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,4,power_law_1.2,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,4,power_law_1.2,0.045977601408958436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,4,power_law_1.2,0.045721599459648134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,4,power_law_1.2,0.04521600008010864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,4,power_law_1.2,0.048390400409698484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,4,power_law_1.2,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,4,power_law_1.2,0.04822399914264679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,4,power_law_1.2,0.05370879769325256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,4,power_law_1.2,0.05720319747924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,4,power_law_1.2,0.06208639740943909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,4,power_law_1.2,0.06763520240783691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.2,0.07722240090370178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.2,0.098854398727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.2,0.11059839725494384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.2,0.14024959802627562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,4,power_law_1.01,0.05230720043182373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,4,power_law_1.01,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,4,power_law_1.01,0.08278399705886841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,4,power_law_1.01,0.08573439717292786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,4,power_law_1.01,0.09758080244064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,4,power_law_1.01,0.11327359676361085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,4,power_law_1.01,0.11399680376052856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.2,0.19302400350570678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,4,power_law_1.01,0.13021440505981446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,4,power_law_1.01,0.13215999603271483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,4,power_law_1.01,0.1278720021247864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,4,power_law_1.01,0.13019520044326782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,4,power_law_1.01,0.14131200313568115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,4,power_law_1.01,0.13962240219116212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,4,power_law_1.01,0.15495680570602416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,4,power_law_1.01,0.1556671977043152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,4,power_law_1.01,0.17486720085144042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,4,power_law_1.01,0.1876736044883728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,4,power_law_1.01,0.21536641120910643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,4,power_law_1.01,0.21950719356536866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,4,power_law_1.01,0.2695168018341064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,4,power_law_1.01,0.31101438999176023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,4,power_law_1.01,0.40519042015075685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,4,power_law_1.01,0.5017280101776123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,4,power_law_1.01,0.6708608150482178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,4,power_law_1.01,0.8310400009155273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.2,0.2479423999786377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,4,power_law_1.01,1.204531192779541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,4,power_law_1.01,1.4885567665100097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.2,0.3991231918334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.2,0.48404479026794434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,8,power_law_1.2,0.018111999332904815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.2,0.5771135807037353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,8,power_law_1.2,0.018943999707698823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,8,power_law_1.2,0.02101760059595108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,8,power_law_1.2,0.019596800208091736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,8,power_law_1.2,0.019622400403022766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,8,power_law_1.2,0.02298240065574646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,8,power_law_1.2,0.0239424005150795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,8,power_law_1.2,0.02341119945049286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,8,power_law_1.2,0.02476799935102463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,8,power_law_1.2,0.0243136003613472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,8,power_law_1.2,0.026649600267410277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,8,power_law_1.2,0.027795198559761047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,8,power_law_1.2,0.02865920066833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,8,power_law_1.2,0.033529600501060484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,8,power_law_1.2,0.035104000568389894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,8,power_law_1.2,0.037241598963737486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,8,power_law_1.2,0.0458624005317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,8,power_law_1.2,0.05605760216712952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,8,power_law_1.2,0.062009602785110474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,8,power_law_1.2,0.07485439777374267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,8,power_law_1.2,0.06281599998474122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,8,power_law_1.2,0.07538560032844543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,8,power_law_1.2,0.0930176019668579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,8,power_law_1.2,0.11864960193634033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,8,power_law_1.2,0.15799039602279663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,1,balanced,0.07434133191903432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,1,balanced,0.10299733281135559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,1,balanced,0.15465600291887918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,1,balanced,0.2524426579475403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,1,balanced,0.4503680070241292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,1,balanced,0.8441973527272543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,1,balanced,1.2469279766082764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,8,power_law_1.2,0.21028480529785157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,128,power_law_1.2,0.06097919940948486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,1,balanced,1.2438133557637532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,1,balanced,1.2525813579559326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,1,balanced,1.2555572986602783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,1,balanced,1.2591573397318523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,1,balanced,1.2802240053812664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,128,power_law_1.2,0.06360960006713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,1,balanced,1.2808213233947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,128,power_law_1.2,0.060127997398376466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,1,balanced,1.2994346618652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,128,power_law_1.2,0.0636352002620697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,1,balanced,1.3026026884714763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,1,balanced,1.3131626447041829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,1,balanced,1.3432586987813313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,2,power_law_1.2,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,1,balanced,1.3970613479614258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,2,power_law_1.2,0.06311039924621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,1,balanced,1.4436747233072917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,128,power_law_1.2,0.06121600270271301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,1,balanced,1.5716640154520671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,128,power_law_1.2,0.06149759888648987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,1,balanced,1.61735995610555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,1,balanced,1.8194559415181477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,2,power_law_1.2,0.06806399822235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,1,balanced,1.97978671391805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,2,power_law_1.2,0.0836736023426056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,2,power_law_1.2,0.10056320428848267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,1,balanced,2.339024066925049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,128,power_law_1.2,0.06376960277557372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,2,power_law_1.2,0.11338239908218384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,128,power_law_1.2,0.06308479905128479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,2,power_law_1.2,0.15218559503555298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,1,balanced,2.858229319254557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,128,power_law_1.2,0.06317440271377564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,2,power_law_1.2,0.15367679595947265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,128,power_law_1.2,0.06462079882621766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,2,power_law_1.2,0.1577664017677307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,128,power_law_1.2,0.06402559876441956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,2,power_law_1.2,0.16309759616851807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,128,power_law_1.2,0.06481919884681701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,2,power_law_1.2,0.16447999477386474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,128,power_law_1.2,0.06556159853935242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,2,power_law_1.2,0.16871039867401122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,2,power_law_1.2,0.16911360025405883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,2,power_law_1.2,0.18021119832992555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,2,power_law_1.2,0.191430401802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,2,power_law_1.2,0.1905087947845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,1,balanced,3.5702187220255532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,2,power_law_1.2,0.19987200498580932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,128,power_law_1.2,0.06610559821128845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,128,power_law_1.2,0.07169280052185059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,128,power_law_1.2,0.07029119729995728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,1,balanced,4.824458758036296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,2,power_law_1.2,0.2240895986557007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,2,power_law_1.2,0.23324799537658691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,128,power_law_1.2,0.07541760206222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,128,power_law_1.2,0.0856383979320526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.2,0.09431040287017822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.2,0.12008320093154908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.2,0.13430399894714357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.2,0.17815680503845216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.2,0.21589760780334472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,2,power_law_1.2,0.2861056089401245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.2,0.2934015989303589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,2,power_law_1.2,0.3048831939697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.2,0.3895999908447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,2,power_law_1.2,0.3802175998687744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.2,0.5732480049133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,2,power_law_1.2,0.443225622177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.2,0.7328000068664551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,2,power_law_1.2,0.5934016227722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,2,power_law_1.2,0.7100800037384033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,2,power_law_1.2,1.0038592338562011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,2,power_law_1.2,1.3185855865478515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,8,power_law_1.2,0.27258241176605225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,2,balanced,0.01934933289885521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,2,balanced,0.019461333751678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,2,balanced,0.018960000326236088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,2,balanced,0.019194666296243668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,2,balanced,0.019637333850065868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,2,balanced,0.021002667645613354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,2,balanced,0.02103466788927714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,2,balanced,0.021770666042963665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,2,balanced,0.02310933421055476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,2,balanced,0.023034666975339253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,2,balanced,0.024890666206677754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,2,balanced,0.025216000775496166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,2,balanced,0.025349333882331848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,2,balanced,0.02720533311367035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,2,balanced,0.02914133419593175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,2,balanced,0.02977599948644638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,2,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,2,balanced,0.045882667104403176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,2,balanced,0.04721599817276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,2,balanced,0.05438933273156484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,2,balanced,0.06428266565004985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,2,balanced,0.08888000249862671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,2,balanced,0.10150933265686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,2,balanced,0.13773333032925925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,2,balanced,0.17414933443069458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,2,balanced,0.24701333045959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,2,balanced,0.3075093428293864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,64,power_law_1.2,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,64,power_law_1.2,0.061945599317550656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,64,power_law_1.2,0.06286079883575439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,64,power_law_1.2,0.05996800065040588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,64,power_law_1.2,0.058348798751831056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,64,power_law_1.2,0.06076800227165222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,64,power_law_1.2,0.06082559823989868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,64,power_law_1.2,0.06170240044593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,64,power_law_1.2,0.06262400150299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,64,power_law_1.2,0.06272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,64,power_law_1.2,0.06266239881515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,64,power_law_1.2,0.06723200082778931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,64,power_law_1.2,0.06746879816055298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,64,power_law_1.2,0.07043200135231018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,64,power_law_1.2,0.07533439993858337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,64,power_law_1.2,0.07896320223808288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,64,power_law_1.2,0.08990079760551453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,64,power_law_1.2,0.1027008056640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,64,power_law_1.2,0.11658240556716919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,64,power_law_1.2,0.1545151948928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,64,power_law_1.2,0.17446399927139283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,64,power_law_1.2,0.23612160682678224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,64,power_law_1.2,0.23991680145263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,64,power_law_1.2,0.313427209854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,64,power_law_1.2,0.4027520179748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,1,balanced,0.035775999228159584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,1,balanced,0.052149335543314614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,1,balanced,0.07455466687679291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,1,balanced,0.0783733328183492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,1,balanced,0.08002133170763652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,1,balanced,0.07999466856320699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,1,balanced,0.08073066671689351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,1,balanced,0.08175466458002727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,1,balanced,0.08238400022188823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,1,balanced,0.08306666711966197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,1,balanced,0.08418132861455281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,1,balanced,0.08717866738637288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,1,balanced,0.08930133779843648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,1,balanced,0.095360000928243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,1,balanced,0.10186666250228882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,1,balanced,0.10527466734250386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,1,balanced,0.11142933368682861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,1,balanced,0.13723199566205344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,1,balanced,0.14698132872581482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,1,balanced,0.18854933977127075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,1,balanced,0.23305066426595053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,1,balanced,0.31144533554712933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,1,balanced,0.36420265833536786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,1,balanced,0.5184426705042521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,1,balanced,0.6602026621500651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,64,power_law_1.2,0.6332543849945068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,1,balanced,0.9561653137207031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,1,balanced,1.253391981124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,64,power_law_1.2,0.7195199966430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,4,power_law_1.2,0.06106879711151123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,4,power_law_1.2,0.06698880195617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,4,power_law_1.2,0.07108479738235474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,4,power_law_1.2,0.07859200239181519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,4,power_law_1.2,0.0880832016468048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,4,power_law_1.2,0.09985920190811157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,4,power_law_1.2,0.10176639556884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,4,power_law_1.2,0.125600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,4,power_law_1.2,0.11354240179061889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,4,power_law_1.2,0.11656960248947143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,4,power_law_1.2,0.11645439863204957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,4,power_law_1.2,0.12087680101394653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,4,power_law_1.2,0.12859519720077514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,4,power_law_1.2,0.12709120512008668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,4,power_law_1.2,0.13544960021972657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,4,power_law_1.2,0.14232319593429565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,4,power_law_1.2,0.15537919998168945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,4,power_law_1.2,0.17036800384521483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,4,power_law_1.2,0.18842240571975707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,4,power_law_1.2,0.2240959882736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,4,power_law_1.2,0.26487040519714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,4,power_law_1.2,0.3384320020675659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,4,power_law_1.2,0.42046079635620115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,4,power_law_1.2,0.5693568229675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,4,power_law_1.2,0.714572811126709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,16,power_law_1.2,0.020032000541687012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,16,power_law_1.2,0.02080000042915344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,4,power_law_1.2,1.0468607902526856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,4,power_law_1.2,1.4362431526184083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,16,power_law_1.2,0.018771199882030486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,16,power_law_1.2,0.01950719952583313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,16,power_law_1.2,0.01976960003376007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,16,power_law_1.2,0.020505599677562714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,16,power_law_1.2,0.020710399746894835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,16,power_law_1.2,0.021452799439430237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,16,power_law_1.2,0.021356800198554994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,16,power_law_1.2,0.021663999557495116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,16,power_law_1.2,0.025331199169158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,16,power_law_1.2,0.025305598974227905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,16,power_law_1.2,0.026348799467086792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,16,power_law_1.2,0.041433599591255185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,16,power_law_1.2,0.04046080112457275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,16,power_law_1.2,0.0414463996887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,16,power_law_1.2,0.04234879910945892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,16,power_law_1.2,0.04429439902305603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.2,0.04907520115375519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.2,0.054655998945236206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.2,0.0648959994316101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.2,0.08225280046463013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.2,0.10321279764175414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.2,0.13815679550170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.2,0.16505600214004518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.2,0.23458559513092042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.2,0.29879040718078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,16,balanced,0.03127466638882955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,16,balanced,0.02940800040960312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,16,balanced,0.03141333411137263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,16,balanced,0.03326933334271113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,16,balanced,0.033930666744709015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,16,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,16,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,16,balanced,0.037418665985266365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,16,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,16,balanced,0.03595199932654699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,16,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,16,balanced,0.03959999978542328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,16,balanced,0.038975998759269714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,16,balanced,0.03973866750796636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,16,balanced,0.04167466859022776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,16,balanced,0.04391466577847799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,16,balanced,0.04562133550643921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,16,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,16,balanced,0.05198400219281515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,16,balanced,0.05824000140031179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,16,balanced,0.06414933502674103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,16,balanced,0.0897706647713979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,16,balanced,0.11065066854159038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,16,balanced,0.15025599797566733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,16,balanced,0.18879467248916626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,16,balanced,0.2547679940859477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,16,balanced,0.32449066638946533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,32,power_law_1.2,0.06394879817962647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,32,power_law_1.2,0.0634112000465393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,32,power_law_1.2,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,32,power_law_1.2,0.04808320105075836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,32,power_law_1.2,0.047200000286102294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,32,power_law_1.2,0.04903680086135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,32,power_law_1.2,0.049670401215553286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,32,power_law_1.2,0.05002239942550659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,32,power_law_1.2,0.05084800124168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,32,power_law_1.2,0.05237119793891907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,32,power_law_1.2,0.05175679922103882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,32,power_law_1.2,0.054028797149658206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,32,power_law_1.2,0.05610880255699158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,32,power_law_1.2,0.05864319801330566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,32,power_law_1.2,0.06498559713363647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,32,power_law_1.2,0.06785920262336731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,32,power_law_1.2,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,32,power_law_1.2,0.0804032027721405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,32,power_law_1.2,0.09038720130920411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,32,power_law_1.2,0.1117184042930603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,32,power_law_1.2,0.12046719789505005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,32,power_law_1.2,0.15894399881362914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,32,power_law_1.2,0.20188798904418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,32,power_law_1.2,0.2558784008026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,32,power_law_1.2,0.3352191925048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,32,power_law_1.2,0.4803840160369873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,32,power_law_1.2,0.5882368087768555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,64,power_law_1.01,0.09770879745483399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,64,power_law_1.01,0.09667199850082397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,64,power_law_1.01,0.0939520001411438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,64,power_law_1.01,0.09646080136299133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,256,balanced,0.06452266871929169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,256,balanced,0.05726400017738342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,256,balanced,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,256,balanced,0.05635733405749003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,256,balanced,0.05453866720199585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,256,balanced,0.05578133463859558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,256,balanced,0.05671999851862589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,256,balanced,0.05669333537419637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,256,balanced,0.05819733440876007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,256,balanced,0.05770133435726166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,64,power_law_1.01,0.09672319889068604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,64,power_law_1.01,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,64,power_law_1.01,0.09841279983520508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,64,power_law_1.01,0.09881600141525268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,256,balanced,0.05727999905745188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,256,balanced,0.06086933116118113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,64,power_law_1.01,0.0982591986656189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,256,balanced,0.060693333546320595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,256,balanced,0.06087466577688853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,256,balanced,0.06668266654014587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,64,power_law_1.01,0.09796479940414429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,256,balanced,0.06870399912198384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,256,balanced,0.07065600156784058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,256,balanced,0.08302933474381764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,256,balanced,0.08711466193199158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,256,balanced,0.11229333281517029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,256,balanced,0.12846933801968893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,256,balanced,0.18361065785090128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,256,balanced,0.17545600732167563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,256,balanced,0.2267893354098002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,256,balanced,0.26317866643269855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,64,power_law_1.01,0.10634880065917969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,256,balanced,0.354362686475118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,64,power_law_1.01,0.1021183967590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,64,power_law_1.01,0.1054527997970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,64,power_law_1.01,0.1091007947921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,64,power_law_1.01,0.11767679452896118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,64,power_law_1.01,0.12009600400924683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,256,balanced,0.4322613477706909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,64,power_law_1.01,0.12459520101547242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,64,power_law_1.01,0.1327232003211975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,64,power_law_1.01,0.1453503966331482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,64,power_law_1.01,0.1664639949798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,64,power_law_1.01,0.20124800205230714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,64,power_law_1.01,0.255513596534729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,64,power_law_1.01,0.2989248037338257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,64,power_law_1.01,0.4051392078399658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,64,power_law_1.01,0.5035391807556152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,64,power_law_1.01,0.7558911800384521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,64,power_law_1.01,0.970860767364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,4,power_law_1.01,0.031711998581886294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,4,power_law_1.01,0.033292800188064575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,4,power_law_1.01,0.03505280017852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,4,power_law_1.01,0.03988479971885681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,4,power_law_1.01,0.0449535995721817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,4,power_law_1.01,0.05619840025901794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,4,power_law_1.01,0.05564799904823303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,4,power_law_1.01,0.0564736008644104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,4,power_law_1.01,0.058246397972106935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,4,power_law_1.01,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,4,power_law_1.01,0.05910400152206421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,4,power_law_1.01,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,4,power_law_1.01,0.060755199193954466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,4,power_law_1.01,0.0642304003238678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,4,power_law_1.01,0.06954879760742187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,4,power_law_1.01,0.07123200297355652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,4,power_law_1.01,0.07103360295295716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,4,power_law_1.01,0.07870079874992371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,4,power_law_1.01,0.08570880293846131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,4,power_law_1.01,0.09637119770050048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,4,power_law_1.01,0.11624319553375244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,4,power_law_1.01,0.13824000358581542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,4,power_law_1.01,0.16272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,4,power_law_1.01,0.21829121112823485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,4,power_law_1.01,0.2521087884902954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,4,power_law_1.01,0.36229119300842283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,4,power_law_1.01,0.47073922157287595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,4,power_law_1.01,0.07274879813194275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,4,power_law_1.01,0.09598079919815064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,4,power_law_1.01,0.08751999735832214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,4,power_law_1.01,0.1172287940979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,4,power_law_1.01,0.12753280401229858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,4,power_law_1.01,0.15665279626846312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,4,power_law_1.01,0.16749440431594848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,4,power_law_1.01,0.16754560470581054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,4,power_law_1.01,0.16473599672317504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,4,power_law_1.01,0.17564159631729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,4,power_law_1.01,0.17916799783706666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,4,power_law_1.01,0.1839679956436157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,4,power_law_1.01,0.18880640268325805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,4,power_law_1.01,0.1925503969192505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,4,power_law_1.01,0.20495359897613524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,4,power_law_1.01,0.21107840538024902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,4,power_law_1.01,0.21422080993652343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,4,power_law_1.01,0.24555521011352538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,4,power_law_1.01,0.2713920116424561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,4,power_law_1.01,0.3207040071487427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,4,power_law_1.01,0.37789440155029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,4,power_law_1.01,0.47845120429992677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,4,power_law_1.01,0.5545919895172119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,4,power_law_1.01,0.7230016231536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,4,power_law_1.01,0.9863424301147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,4,power_law_1.01,1.3121088027954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,4,power_law_1.01,1.5704959869384765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,64,power_law_1.2,0.017888000607490538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,64,power_law_1.2,0.01772159934043884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,64,power_law_1.2,0.017836800217628478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,64,power_law_1.2,0.01793919950723648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,64,power_law_1.2,0.018956799805164338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,64,power_law_1.2,0.02250239998102188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,64,power_law_1.2,0.022163200378417968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,64,power_law_1.2,0.03290880024433136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,64,power_law_1.2,0.032767999172210696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,64,power_law_1.2,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,64,power_law_1.2,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,64,power_law_1.2,0.03250559866428375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,64,power_law_1.2,0.03249920010566711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,64,power_law_1.2,0.03204480111598969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,64,power_law_1.2,0.03322240114212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,64,power_law_1.2,0.033990401029586795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,64,power_law_1.2,0.03694719970226288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,64,power_law_1.2,0.0414463996887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.2,0.044684800505638125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.2,0.05084159970283508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.2,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.2,0.07625600099563598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.2,0.09448320269584656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.2,0.1256832003593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.2,0.15961600542068483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.2,0.2208575963973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.2,0.2819200038909912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,128,power_law_1.01,0.018355199694633485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,128,power_law_1.01,0.020665599405765532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,128,power_law_1.01,0.01929599940776825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,128,power_law_1.01,0.01892479956150055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,128,power_law_1.01,0.022617599368095397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,128,power_law_1.01,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,128,power_law_1.01,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,128,power_law_1.01,0.03272959887981415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,128,power_law_1.01,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,128,power_law_1.01,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,128,power_law_1.01,0.033318400382995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,128,power_law_1.01,0.033766400814056394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,128,power_law_1.01,0.03237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,128,power_law_1.01,0.03283840119838714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,128,power_law_1.01,0.03404799997806549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,128,power_law_1.01,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,128,power_law_1.01,0.03739520013332367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,128,power_law_1.01,0.03948799967765808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.01,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.01,0.04822399914264679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.01,0.05440639853477478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.01,0.06935039758682252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.01,0.08695039749145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.01,0.11921919584274292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.01,0.15155839920043945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.01,0.21644160747528077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.01,0.2693887948989868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,4,power_law_1.01,0.0591871976852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,4,power_law_1.01,0.06330239772796631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,4,power_law_1.01,0.063264000415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,4,power_law_1.01,0.08240000009536744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,4,power_law_1.01,0.07322239875793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,4,power_law_1.01,0.08935040235519409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,4,power_law_1.01,0.0893887996673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,4,power_law_1.01,0.09131519794464112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,4,power_law_1.01,0.09340800046920776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,4,power_law_1.01,0.09639679789543151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,4,power_law_1.01,0.09679999947547913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,4,power_law_1.01,0.09953280091285706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,4,power_law_1.01,0.09978880286216736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,4,power_law_1.01,0.10534399747848511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,4,power_law_1.01,0.11770880222320557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,4,power_law_1.01,0.12106879949569702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,4,power_law_1.01,0.12272640466690063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,4,power_law_1.01,0.13578879833221436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,4,power_law_1.01,0.15289599895477296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,4,power_law_1.01,0.1860352039337158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,4,power_law_1.01,0.2145024061203003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,4,power_law_1.01,0.290451192855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,4,power_law_1.01,0.3679359912872314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,4,power_law_1.01,0.5099008083343506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,4,power_law_1.01,0.6405632019042968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,4,power_law_1.01,0.9743743896484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,4,power_law_1.01,1.2032768249511718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,32,power_law_1.01,0.03371520042419433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,32,power_law_1.01,0.03426559865474701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,32,power_law_1.01,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,32,power_law_1.01,0.034796801209449765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,32,power_law_1.01,0.03509120047092438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,32,power_law_1.01,0.03663359880447388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,32,power_law_1.01,0.03692800104618073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,32,power_law_1.01,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,32,power_law_1.01,0.037625598907470706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,32,power_law_1.01,0.037887999415397645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,32,power_law_1.01,0.038515201210975646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,32,power_law_1.01,0.03916159868240356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,32,power_law_1.01,0.039699199795722964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,32,power_law_1.01,0.041971200704574586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,32,power_law_1.01,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,32,power_law_1.01,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,32,power_law_1.01,0.050425601005554196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,32,power_law_1.01,0.051846402883529666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.01,0.05589759945869446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.01,0.06741120219230652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.01,0.07790079712867737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.01,0.09507840275764465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.01,0.11834239959716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.01,0.15014400482177734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.01,0.19093120098114014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.01,0.2616127967834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.01,0.3771071910858154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,1,power_law_1.2,0.12581119537353516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,1,power_law_1.2,0.1640768051147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,128,power_law_1.01,0.03781760036945343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,1,power_law_1.2,0.22083840370178223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,128,power_law_1.01,0.03889279961585999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,1,power_law_1.2,0.33114240169525144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,128,power_law_1.01,0.03819519877433777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,128,power_law_1.01,0.039929598569869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,1,power_law_1.2,0.4702144145965576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,128,power_law_1.01,0.040249601006507874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,128,power_law_1.01,0.04191359877586365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,1,power_law_1.2,0.5549888134002685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,128,power_law_1.01,0.04078719913959503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,128,power_law_1.01,0.040659201145172116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,1,power_law_1.2,0.7547264099121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,128,power_law_1.01,0.041228801012039185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,128,power_law_1.01,0.0411327987909317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,1,power_law_1.2,0.8170559883117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,1,power_law_1.2,0.8233152389526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,1,power_law_1.2,0.850374412536621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,128,power_law_1.01,0.04162560105323791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,1,power_law_1.2,0.8599871635437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,128,power_law_1.01,0.04132480025291443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,128,power_law_1.01,0.041459199786186215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,128,power_law_1.01,0.04103679955005646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,128,power_law_1.01,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,128,power_law_1.01,0.04798080027103424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,128,power_law_1.01,0.04791040122509003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,128,power_law_1.01,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.01,0.05845119953155518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.01,0.06607360243797303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.01,0.07557119727134705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,1,power_law_1.2,0.887820816040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,1,power_law_1.2,0.893177604675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,1,power_law_1.2,0.9140864372253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,1,power_law_1.2,0.9367039680480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.01,0.09256960153579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.01,0.11603840589523315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,1,power_law_1.2,0.9600959777832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.01,0.15333759784698486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.01,0.1892799973487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,1,power_law_1.2,0.9909824371337891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.01,0.28600320816040037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,1,power_law_1.2,1.0782848358154298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.01,0.3712703943252563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.2,1.1565759658813477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.2,1.3099967956542968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.2,1.4307647705078126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.2,1.7221248626708985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.2,2.010150337219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.2,2.5781503677368165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.2,3.14587516784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.2,4.288959884643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.2,5.392652893066407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,32,power_law_1.2,0.09402239918708802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,32,power_law_1.2,0.08120319843292237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,32,power_law_1.2,0.08050559759140015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,32,power_law_1.2,0.07351679801940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,32,power_law_1.2,0.06971520185470581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,32,power_law_1.2,0.0788096010684967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,32,power_law_1.2,0.0830784022808075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,32,power_law_1.2,0.08462719917297364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,32,power_law_1.2,0.08202880024909973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,32,power_law_1.2,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,32,power_law_1.2,0.0884607970714569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,32,power_law_1.2,0.09015679955482483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,32,power_law_1.2,0.09262080192565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,32,power_law_1.2,0.09548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,32,power_law_1.2,0.10574719905853272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,32,power_law_1.2,0.1130687952041626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,32,power_law_1.2,0.12215679883956909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,32,power_law_1.2,0.1434175968170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,32,power_law_1.2,0.1483839988708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,32,power_law_1.2,0.19790079593658447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,32,power_law_1.2,0.2155519962310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,32,power_law_1.2,0.30965120792388917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,32,power_law_1.2,0.3694976091384888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,32,power_law_1.2,0.5183616161346436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,32,power_law_1.2,0.6109824180603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,32,power_law_1.2,0.9242048263549805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,32,power_law_1.2,1.1422911643981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,8,power_law_1.01,0.04584319889545441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,8,power_law_1.01,0.04527359902858734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,8,power_law_1.01,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,8,power_law_1.01,0.04782080054283142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,8,power_law_1.01,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,8,power_law_1.01,0.04791040122509003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,8,power_law_1.01,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,8,power_law_1.01,0.047295999526977536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,8,power_law_1.01,0.04424319863319397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,8,power_law_1.01,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,8,power_law_1.01,0.04385280013084412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,8,power_law_1.01,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,8,power_law_1.01,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,8,power_law_1.01,0.04878720045089722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,8,power_law_1.01,0.05488640069961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,8,power_law_1.01,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,8,power_law_1.01,0.061478400230407716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,8,power_law_1.01,0.07989760041236878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,8,power_law_1.01,0.09257599711418152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,8,power_law_1.01,0.12188160419464111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,8,power_law_1.01,0.1480512022972107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,32,balanced,0.0539680023988088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,32,balanced,0.05598400036493937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,32,balanced,0.05392533540725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,32,balanced,0.06860800087451935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,32,balanced,0.07249066730340321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,32,balanced,0.08039466540018718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,32,balanced,0.0784800002972285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,32,balanced,0.08015466729799907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,32,balanced,0.07644266883532207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,32,balanced,0.07645333309968312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,32,balanced,0.07710400223731995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,32,balanced,0.07909866670767467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,32,balanced,0.0828000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,32,balanced,0.08005866905053456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,32,balanced,0.08288000027338664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,32,balanced,0.08662933111190796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,32,balanced,0.08866666754086812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,32,balanced,0.09238400061925252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,32,balanced,0.09878933429718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,32,balanced,0.11081066727638245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,32,balanced,0.11831466356913249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,32,balanced,0.15748799840609232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,32,balanced,0.1720106601715088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,32,balanced,0.230186661084493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,32,balanced,0.2914186716079712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,32,balanced,0.36869335174560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,32,balanced,0.48285865783691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,8,power_law_1.01,0.20059518814086913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,8,power_law_1.01,0.24652159214019775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,8,power_law_1.01,0.3450239896774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,4,power_law_1.2,0.05311359763145447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,4,power_law_1.2,0.058924800157546996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,4,power_law_1.2,0.07290239930152893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,4,power_law_1.2,0.07281919717788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,4,power_law_1.2,0.07882239818572997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,4,power_law_1.2,0.10833280086517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,4,power_law_1.2,0.11585279703140258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,4,power_law_1.2,0.11671040058135987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,4,power_law_1.2,0.12097920179367065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,4,power_law_1.2,0.12128000259399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,4,power_law_1.2,0.12710399627685548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,4,power_law_1.2,0.1289471983909607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,4,power_law_1.2,0.13407360315322875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,4,power_law_1.2,0.13689600229263305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,4,power_law_1.2,0.14812159538269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,4,power_law_1.2,0.1553984045982361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,4,power_law_1.2,0.15455360412597657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,4,power_law_1.2,0.17272959947586058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,4,power_law_1.2,0.1960576057434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,4,power_law_1.2,0.23943040370941163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,4,power_law_1.2,0.2681920051574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,4,power_law_1.2,0.3218496084213257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,4,power_law_1.2,0.41194877624511717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,4,power_law_1.2,0.5544447898864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,16,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,16,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,16,balanced,0.047610665361086525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,16,balanced,0.047770669062932335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,16,balanced,0.04970666766166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,16,balanced,0.06019733349482218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,16,balanced,0.07554133236408234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,16,balanced,0.07454933226108551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,16,balanced,0.07668800155321757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,16,balanced,0.07760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,16,balanced,0.07871466875076294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,16,balanced,0.08251733581225078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,16,balanced,0.08211733400821686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,16,balanced,0.0845973292986552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,16,balanced,0.09168000022570293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,16,balanced,0.09129066268603007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,8,power_law_1.01,0.48056321144104003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,16,balanced,0.10127466917037964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,16,balanced,0.1113759974638621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,16,balanced,0.12345066666603088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,16,balanced,0.1469439963499705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,16,balanced,0.16612799962361655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,16,balanced,0.21047999461491904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,8,power_law_1.01,0.12519680261611937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,4,power_law_1.2,0.6887680053710937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,8,power_law_1.01,0.192576003074646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,8,power_law_1.01,0.17226879596710204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,8,power_law_1.01,0.20762879848480226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,16,balanced,0.26691200335820514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,8,power_law_1.01,0.23393919467926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,16,balanced,0.3548106749852498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,16,balanced,0.4471946557362874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,8,power_law_1.01,0.23808000087738038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,16,balanced,0.6288693348566691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,8,power_law_1.01,0.23478400707244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,16,balanced,0.8119893074035645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,8,power_law_1.01,0.24046719074249268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,8,power_law_1.01,0.24542720317840577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,8,power_law_1.01,0.24367361068725585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,8,power_law_1.01,0.24284799098968507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,8,power_law_1.01,0.25557119846343995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,8,power_law_1.01,0.25180160999298096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,8,power_law_1.01,0.2604736089706421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,8,power_law_1.01,0.26633601188659667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,8,power_law_1.01,0.2796927928924561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,8,power_law_1.01,0.3031552076339722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,8,power_law_1.01,0.347654390335083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,8,power_law_1.01,0.3515007972717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,8,power_law_1.01,0.4246016025543213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,8,power_law_1.01,0.47712001800537107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,8,power_law_1.01,0.6196735858917236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,8,power_law_1.01,0.5963583946228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,8,power_law_1.01,0.7605760097503662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,8,power_law_1.01,0.9468352317810058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,8,power_law_1.01,1.398796844482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,8,power_law_1.01,1.494431972503662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,4,power_law_1.2,0.9362431526184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,8,power_law_1.01,0.7110079765319824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,4,power_law_1.2,1.301683235168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,8,power_law_1.01,0.9328255653381348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,64,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,64,balanced,0.029509333272775013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,64,balanced,0.029802667597929638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,64,balanced,0.03158933420976003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,64,balanced,0.03133333226044973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,64,balanced,0.031354665756225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,64,balanced,0.03317866722742716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,64,balanced,0.033200000723203026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,64,balanced,0.033471999069054924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,64,balanced,0.03205333401759466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,64,balanced,0.0332640012105306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,64,balanced,0.0351946676770846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,64,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,64,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,64,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,64,balanced,0.0393653338154157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,64,balanced,0.03933866570393244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,64,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,64,balanced,0.04177066683769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,64,balanced,0.045791998505592346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,64,balanced,0.047541335225105286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,64,balanced,0.051872000098228455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,64,balanced,0.05834666887919108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,64,balanced,0.08509332935015361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,64,balanced,0.10359467069307964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,64,balanced,0.1362933317820231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,64,balanced,0.16835200786590576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,4,power_law_1.01,0.0759552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,4,power_law_1.01,0.09442560076713562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,4,power_law_1.01,0.0984063982963562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,4,power_law_1.01,0.11474560499191284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,4,power_law_1.01,0.12471040487289428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,4,power_law_1.01,0.1331007957458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,4,power_law_1.01,0.1474303960800171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,4,power_law_1.01,0.1470720052719116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,4,power_law_1.01,0.15193599462509155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,4,power_law_1.01,0.15205119848251342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,4,power_law_1.01,0.15266560316085814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,4,power_law_1.01,0.1583680033683777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,4,power_law_1.01,0.1575935959815979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,4,power_law_1.01,0.1625920057296753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,4,power_law_1.01,0.1685312032699585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,4,power_law_1.01,0.17053439617156982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,4,power_law_1.01,0.1772223949432373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,4,power_law_1.01,0.19296640157699585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,4,power_law_1.01,0.21490559577941895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,4,power_law_1.01,0.26797440052032473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,4,power_law_1.01,0.31452159881591796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,4,power_law_1.01,0.4166975975036621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,4,power_law_1.01,0.5372799873352051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,4,power_law_1.01,0.7273024082183838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,4,power_law_1.01,0.9186047554016114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,4,power_law_1.01,1.328384017944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,4,power_law_1.01,1.7075712203979492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,32,power_law_1.01,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,32,power_law_1.01,0.03570559918880463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,32,power_law_1.01,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,32,power_law_1.01,0.039673599600791934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,32,power_law_1.01,0.04095999896526337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,32,power_law_1.01,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,32,power_law_1.01,0.04238080084323883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,32,power_law_1.01,0.04261119961738587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,8,balanced,0.04574400186538696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,8,balanced,0.048122664292653404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,8,balanced,0.0470773329337438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,8,balanced,0.04523199796676636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,8,balanced,0.04568533102671305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,8,balanced,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,8,balanced,0.04781333108743032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,8,balanced,0.04706666866938273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,8,balanced,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,8,balanced,0.04879466692606608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,8,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,8,balanced,0.050570666790008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,8,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,8,balanced,0.05049600203831991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,8,balanced,0.06635733445485432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,8,balanced,0.06758933266003926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,8,balanced,0.06619733572006226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,8,balanced,0.0746506651242574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,8,balanced,0.08292266726493835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,32,power_law_1.01,0.04780159890651703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,32,power_law_1.01,0.04756479859352112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,8,balanced,0.09408533573150635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,8,balanced,0.1039626697699229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,8,balanced,0.12526933352152506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,8,balanced,0.1476959983507792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,8,balanced,0.19312000274658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,8,balanced,0.24015466372172037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,8,balanced,0.3279946645100911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,8,balanced,0.417797327041626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,32,power_law_1.01,0.04852479994297028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,32,power_law_1.01,0.04760960042476654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,32,power_law_1.01,0.049472001194953916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,32,power_law_1.01,0.05137280225753784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,32,power_law_1.01,0.06021760106086731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,32,power_law_1.01,0.06082559823989868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,32,power_law_1.01,0.06573439836502075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,32,power_law_1.01,0.0753920018672943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,32,power_law_1.01,0.08919039964675904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,32,power_law_1.01,0.10649600028991699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,32,power_law_1.01,0.12142720222473144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,32,power_law_1.01,0.16751359701156615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,32,power_law_1.01,0.19783040285110473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,32,power_law_1.01,0.2876607894897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,32,power_law_1.01,0.3336319923400879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,32,power_law_1.01,0.5667520046234131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,32,power_law_1.01,0.7902143955230713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,256,power_law_1.2,0.05275520086288452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,256,power_law_1.2,0.05303040146827698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,256,power_law_1.2,0.053491199016571046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,256,power_law_1.2,0.05813760161399841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,256,power_law_1.2,0.06206079721450806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,256,power_law_1.2,0.06826239824295044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,256,power_law_1.2,0.06493440270423889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,4,power_law_1.01,0.06108160018920898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,256,power_law_1.2,0.06986879706382751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,4,power_law_1.01,0.07047680020332336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,4,power_law_1.01,0.06570879817008972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,4,power_law_1.01,0.08000640273094177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,4,power_law_1.01,0.07950080037117005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,4,power_law_1.01,0.07843199968338013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,4,power_law_1.01,0.07655680179595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,4,power_law_1.01,0.07929599881172181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,4,power_law_1.01,0.08017280101776122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,4,power_law_1.01,0.07486720085144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,256,power_law_1.2,0.06989439725875854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,4,power_law_1.01,0.07285119891166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,4,power_law_1.01,0.07783679962158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,4,power_law_1.01,0.09150720238685608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,4,power_law_1.01,0.08698880076408386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,4,power_law_1.01,0.09806079864501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,256,power_law_1.2,0.07131519913673401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,4,power_law_1.01,0.10332800149917602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,256,power_law_1.2,0.07336320281028748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,256,power_law_1.2,0.07787520289421082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,256,power_law_1.2,0.0779263973236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,256,power_law_1.2,0.08198400139808655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,256,power_law_1.2,0.08958079814910888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,4,power_law_1.01,0.10922880172729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,256,power_law_1.2,0.09524480104446412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,256,power_law_1.2,0.10382720232009887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,256,power_law_1.2,0.12459520101547242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,4,power_law_1.01,0.13578239679336548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,256,power_law_1.2,0.14302079677581786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,256,power_law_1.2,0.2041408061981201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,256,power_law_1.2,0.24501121044158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.01,0.1454208016395569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,256,power_law_1.2,0.3712703943252563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,256,power_law_1.2,0.4213376045227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,256,power_law_1.2,0.6434432029724121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,256,power_law_1.2,0.8347200393676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.01,0.1910591959953308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,256,power_law_1.2,1.2373056411743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.01,0.2385024070739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,256,power_law_1.2,1.7973888397216797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.01,0.3222399950027466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.01,0.44682879447937013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.01,0.6472127914428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,1,balanced,0.1409386694431305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,1,balanced,0.24283732970555624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,1,balanced,0.4593706528345744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,1,balanced,0.4542880058288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,1,balanced,0.45188268025716144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,1,balanced,0.44757334391276044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,1,balanced,0.45296533902486164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,1,balanced,0.45373332500457764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,1,balanced,0.45739734172821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,1,balanced,0.45557332038879395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,1,balanced,0.4646666844685872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,1,balanced,0.46996267636617023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,1,balanced,0.4726773500442505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,1,balanced,0.4956746498743693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,1,balanced,0.5017919937769572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,1,balanced,0.5071466763814291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,1,balanced,0.5247199932734171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,1,balanced,0.6407093207041422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.01,0.7341440200805665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,1,balanced,0.6218560139338175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,1,balanced,0.8864746888478597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,1,balanced,1.0064373016357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,1,balanced,1.4129759470621746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,1,balanced,1.653264045715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,1,balanced,2.4521919886271157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,1,balanced,3.398373285929362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,1,balanced,4.926970799763997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,1,balanced,6.453439712524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.01,1.0464192390441895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.01,1.3729215621948243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,32,balanced,0.031701333820819855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,32,balanced,0.031583999594052635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,32,balanced,0.03133333226044973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,32,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,32,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,32,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,32,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,32,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,32,balanced,0.03817066550254822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,32,balanced,0.03809066613515218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,32,balanced,0.039477333426475525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,32,balanced,0.039333333571751915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,32,balanced,0.041189332803090416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,32,balanced,0.04144533226887385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,32,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,32,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,32,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,32,balanced,0.045893331368764244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,32,balanced,0.0459146648645401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,32,balanced,0.05143466591835022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,32,balanced,0.05198400219281515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,32,balanced,0.06121066709359487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,32,balanced,0.06433066725730896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,32,balanced,0.0817386656999588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,32,balanced,0.09846400221188863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,32,balanced,0.1267626682917277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,32,balanced,0.15609600146611533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,16,balanced,0.02124800036350886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,16,balanced,0.021749332547187805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,16,balanced,0.02314666658639908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,16,balanced,0.024405332903067272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,16,balanced,0.02914133419593175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,16,balanced,0.043290664752324425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,16,balanced,0.04594666759173075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,16,balanced,0.047744000951449074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,16,balanced,0.049365331729253135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,16,balanced,0.05233600238958994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,16,balanced,0.050757333636283875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,16,balanced,0.05268266797065735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,16,balanced,0.05585066477457682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,16,balanced,0.06018666426340739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,16,balanced,0.06776000062624614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,16,balanced,0.06981333096822102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,16,balanced,0.10155199964841206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,16,balanced,0.12005333105723064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,16,balanced,0.14878933628400168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,16,balanced,0.18768533070882162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,16,balanced,0.13555199901262918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,16,balanced,0.16237333416938782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,16,balanced,0.19339199860890707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,16,balanced,0.24475733439127603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,16,balanced,0.3006346623102824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,16,balanced,0.4585706790288289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,16,balanced,0.5693600177764893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,8,power_law_1.2,0.05120639801025391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,8,power_law_1.2,0.0547327995300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,8,power_law_1.2,0.05470719933509827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,8,power_law_1.2,0.06368640065193176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,8,power_law_1.2,0.06826239824295044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,8,power_law_1.2,0.07477759718894958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,8,power_law_1.2,0.07372159957885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,8,power_law_1.2,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,8,power_law_1.2,0.07226240038871765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,8,power_law_1.2,0.07254400253295898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,8,power_law_1.2,0.07663999795913697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,8,power_law_1.2,0.0766207993030548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,8,power_law_1.2,0.08127999901771546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,8,power_law_1.2,0.08156160116195679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,8,power_law_1.2,0.09025920033454896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,8,power_law_1.2,0.09362559914588928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,8,power_law_1.2,0.09790080189704894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,8,power_law_1.2,0.11185920238494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,8,power_law_1.2,0.11694719791412353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,8,power_law_1.2,0.14523520469665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,8,power_law_1.2,0.16547199487686157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,8,power_law_1.2,0.2054271936416626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,8,power_law_1.2,0.2484544038772583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,8,power_law_1.2,0.3367232084274292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,8,power_law_1.2,0.43076481819152834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,8,power_law_1.2,0.5653952121734619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,8,power_law_1.2,0.7482880115509033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,32,power_law_1.2,0.05367040038108826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,32,power_law_1.2,0.05101439952850342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,32,power_law_1.2,0.04917120039463043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,32,power_law_1.2,0.051820802688598636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,32,power_law_1.2,0.05232639908790589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,32,power_law_1.2,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,32,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,32,power_law_1.2,0.055846399068832396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,32,power_law_1.2,0.056409597396850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,32,power_law_1.2,0.05740799903869629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,32,power_law_1.2,0.05649920105934143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,32,power_law_1.2,0.058348798751831056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,32,power_law_1.2,0.06109439730644226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,32,power_law_1.2,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,32,power_law_1.2,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,32,power_law_1.2,0.07233920097351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,32,power_law_1.2,0.08094080090522766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,32,power_law_1.2,0.09764479994773864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,32,power_law_1.2,0.10077439546585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,32,power_law_1.2,0.12663040161132813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,32,power_law_1.2,0.15481599569320678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,32,power_law_1.2,0.19100159406661987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,32,power_law_1.2,0.24506239891052245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,32,power_law_1.2,0.3208832025527954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,32,power_law_1.2,0.44071040153503416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,32,power_law_1.2,0.745030403137207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,32,power_law_1.2,0.8236991882324218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,64,power_law_1.01,0.05767040252685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,64,power_law_1.01,0.04268800020217896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,64,power_law_1.01,0.04394879937171936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,64,power_law_1.01,0.0456959992647171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,64,power_law_1.01,0.04270080029964447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,64,power_law_1.01,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,64,power_law_1.01,0.04752640128135681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,64,power_law_1.01,0.04805760085582733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,64,power_law_1.01,0.04845440089702606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,64,power_law_1.01,0.04888319969177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,64,power_law_1.01,0.05081599950790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,64,power_law_1.01,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,64,power_law_1.01,0.05680000185966492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,64,power_law_1.01,0.05906559824943543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,64,power_law_1.01,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,64,power_law_1.01,0.06516479849815368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,64,power_law_1.01,0.06167680025100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,64,power_law_1.01,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,64,power_law_1.01,0.07596799731254578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,64,power_law_1.01,0.08737919926643371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,64,power_law_1.01,0.09848960041999817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,64,power_law_1.01,0.12136960029602051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,64,power_law_1.01,0.14431359767913818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,64,power_law_1.01,0.18972159624099733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,64,power_law_1.01,0.224947190284729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,64,power_law_1.01,0.3198080062866211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,64,power_law_1.01,0.3877055883407593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,128,power_law_1.2,0.02531839907169342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,128,power_law_1.2,0.02487040013074875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,128,power_law_1.2,0.025254398584365845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,128,power_law_1.2,0.02800000011920929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,128,power_law_1.2,0.028352001309394838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,128,power_law_1.2,0.02943359911441803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,128,power_law_1.2,0.029478400945663452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,128,power_law_1.2,0.029145601391792297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,128,power_law_1.2,0.029344001412391664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,128,power_law_1.2,0.030368000268936157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,128,power_law_1.2,0.03144319951534271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,128,power_law_1.2,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,128,power_law_1.2,0.03309440016746521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,128,power_law_1.2,0.034457600116729735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,128,power_law_1.2,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,128,power_law_1.2,0.041843199729919435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,128,power_law_1.2,0.04561919867992401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,128,power_law_1.2,0.0516543984413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,128,power_law_1.2,0.06259840130805969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,128,power_law_1.2,0.07973759770393371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,128,power_law_1.2,0.099891197681427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,128,power_law_1.2,0.14084479808807374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,128,power_law_1.2,0.16725759506225585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,128,power_law_1.2,0.27790079116821287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,128,power_law_1.2,0.34503679275512694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,128,power_law_1.2,0.5165567874908448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,128,power_law_1.2,0.7147456169128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,1,balanced,0.045781334241231285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,1,balanced,0.0580320010582606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,1,balanced,0.08258133133252461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,1,balanced,0.1272266705830892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,1,balanced,0.22155733903249106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,1,balanced,0.4148000081380208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,1,balanced,0.42504000663757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,1,balanced,0.43909335136413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,1,balanced,0.45076266924540204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,1,balanced,0.4610613187154134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,1,balanced,0.4834453264872233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,1,balanced,0.5072693427403768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,1,balanced,0.5293813149134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,1,balanced,0.5753493309020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,1,balanced,0.5153973499933878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,1,balanced,0.5365279912948608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,1,balanced,0.579093337059021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,1,balanced,0.5657013257344564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,1,balanced,0.6127200126647949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,1,balanced,0.6228266557057699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,1,balanced,0.6702346801757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,1,balanced,1.5281707445780437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,1,balanced,1.7632160186767578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,1,balanced,1.114789326985677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,1,balanced,1.1993172963460286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,1,balanced,2.12555726369222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,1,balanced,2.1738239924112954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,4,balanced,0.05198400219281515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,4,balanced,0.07149333258469899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,4,balanced,0.0899679958820343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,4,balanced,0.12892799576123556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,4,balanced,0.19818133115768433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,4,balanced,0.2561546762784322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,4,balanced,0.2581920027732849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,4,balanced,0.26294400294621784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,4,balanced,0.2616639931996663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,4,balanced,0.26493332783381146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,4,balanced,0.2661760052045186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,4,balanced,0.2704426646232605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,4,balanced,0.27084267139434814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,4,balanced,0.27552000681559247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,4,balanced,0.28324800729751587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,4,balanced,0.2863946755727132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,4,balanced,0.29702399174372357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,4,balanced,0.33530131975809735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,4,balanced,0.33878934383392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,4,balanced,0.44458667437235516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,4,balanced,0.4328426520029704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,4,balanced,0.6143999894460043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,4,balanced,0.592298666636149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,4,balanced,0.8400213718414307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,4,balanced,0.9297813574473063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,4,balanced,1.3676800727844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,4,balanced,1.6349387168884277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,64,power_law_1.2,0.030348798632621764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,64,power_law_1.2,0.029203200340270997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,64,power_law_1.2,0.02948479950428009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,64,power_law_1.2,0.03144319951534271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,64,power_law_1.2,0.032358399033546446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,64,power_law_1.2,0.034092798829078674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,64,power_law_1.2,0.03361920118331909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,64,power_law_1.2,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,64,power_law_1.2,0.0349375993013382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,64,power_law_1.2,0.03500159978866577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,64,power_law_1.2,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,64,power_law_1.2,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,64,power_law_1.2,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,2,balanced,0.10152000188827515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,2,balanced,0.13914133111635843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,64,power_law_1.2,0.04010240137577057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,2,balanced,0.21726399660110474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,64,power_law_1.2,0.046342399716377256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,2,balanced,0.35022401809692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,64,power_law_1.2,0.04713599979877472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,64,power_law_1.2,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,64,power_law_1.2,0.0621504008769989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,64,power_law_1.2,0.07825919985771179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,64,power_law_1.2,0.09296000003814697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,64,power_law_1.2,0.1211135983467102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,2,balanced,0.5925813515981039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,2,balanced,0.6834506988525391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,2,balanced,0.6722186406453451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,2,balanced,0.6520373423894247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,64,power_law_1.2,0.17293440103530883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,2,balanced,0.6706240177154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,2,balanced,0.647487998008728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,64,power_law_1.2,0.21374080181121827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,2,balanced,0.66811203956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,2,balanced,0.7045813401540121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,2,balanced,0.6656266848246256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,2,balanced,0.7014079888661703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,2,balanced,0.7084106604258219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,2,balanced,0.6711359818776449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,2,balanced,0.7099786599477133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,2,balanced,0.6912319660186768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,2,balanced,0.6827200253804525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,2,balanced,0.706058661142985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,64,power_law_1.2,0.3405440092086792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,2,balanced,0.721461296081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,2,balanced,1.2877333164215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,2,balanced,1.2741813659667969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,2,balanced,1.8617226282755535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,2,balanced,2.430880069732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,2,balanced,2.836853345235189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,2,balanced,4.028047879536946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,64,power_law_1.2,0.460646390914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,64,power_law_1.2,0.6872064113616944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,2,power_law_1.01,0.017395199835300447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,2,power_law_1.01,0.01825280040502548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,2,power_law_1.01,0.019782400131225585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,2,power_law_1.01,0.024716800451278685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,2,power_law_1.01,0.029798400402069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,2,power_law_1.01,0.031276801228523256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,2,power_law_1.01,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,2,power_law_1.01,0.038848000764846805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,2,power_law_1.01,0.04023039937019348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,2,power_law_1.01,0.04164479970932007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,2,power_law_1.01,0.04264320135116577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,2,power_law_1.01,0.04337919950485229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,2,power_law_1.01,0.046291199326515195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,2,power_law_1.01,0.046675199270248414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,2,power_law_1.01,0.049081599712371825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,2,power_law_1.01,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,2,power_law_1.01,0.05533440113067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,2,power_law_1.01,0.06173440217971802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,2,power_law_1.01,0.0705407977104187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,2,power_law_1.01,0.08951039910316468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,2,power_law_1.01,0.09999359846115112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,2,power_law_1.01,0.12360320091247559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,2,power_law_1.01,0.14298239946365357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,2,power_law_1.01,0.17384320497512817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,2,power_law_1.01,0.20889599323272706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,2,power_law_1.01,0.2828864097595215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,2,power_law_1.01,0.35776638984680176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,64,power_law_1.2,0.9457728385925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,256,balanced,0.022117334107557934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,256,balanced,0.02199999988079071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,256,balanced,0.023306667804718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,256,balanced,0.029157333076000214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,256,balanced,0.052042668064435325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,256,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,256,balanced,0.04524266719818115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,256,balanced,0.04526400069395701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,256,balanced,0.04518933097521464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,256,balanced,0.044879997769991554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,256,balanced,0.04378666480382284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,256,balanced,0.043141335248947144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,256,balanced,0.04154133299986521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,256,balanced,0.042170668641726174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,256,balanced,0.04640000065167745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,256,balanced,0.049312000473340355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,256,balanced,0.04554666578769684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,256,balanced,0.04958933095137278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,256,balanced,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,256,balanced,0.0800906668106715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,256,balanced,0.0885759989420573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,256,balanced,0.1225386659304301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,256,balanced,0.15309866269429526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,256,balanced,0.21854400634765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,256,balanced,0.28202666838963825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,256,balanced,0.40826133886973065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,256,balanced,0.5357813437779745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,16,balanced,0.03169066707293192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,16,balanced,0.031557333966096245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,16,balanced,0.03161066770553589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,16,balanced,0.035487999518712364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,16,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,16,balanced,0.037802666425704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,16,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,16,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,16,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,16,balanced,0.03956799954175949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,16,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,16,balanced,0.04390933116277059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,16,balanced,0.043621331453323364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,16,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,16,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,16,balanced,0.051125332713127136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,16,balanced,0.055957332253456116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,16,balanced,0.05619200070699056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,16,balanced,0.06234666705131531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,16,balanced,0.0705973356962204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,16,balanced,0.07979733248551686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,16,balanced,0.10089600086212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,16,balanced,0.114138662815094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,16,balanced,0.14867732922236124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,16,balanced,0.17801066239674887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,16,balanced,0.2562933365503947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,16,balanced,0.3336533308029175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,32,power_law_1.2,0.09242240190505982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,32,power_law_1.2,0.0910975992679596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,32,power_law_1.2,0.08730880022048951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,32,power_law_1.2,0.089273601770401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,4,power_law_1.2,0.07438719868659974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,32,power_law_1.2,0.07985280156135559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,4,power_law_1.2,0.10181119441986083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,32,power_law_1.2,0.08280320167541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,4,power_law_1.2,0.12510720491409302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,32,power_law_1.2,0.08333439826965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,4,power_law_1.2,0.15253119468688964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,4,power_law_1.2,0.19464319944381714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,4,power_law_1.2,0.22036480903625488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,4,power_law_1.2,0.23274240493774415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,32,power_law_1.2,0.08848000168800355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,32,power_law_1.2,0.08531200289726257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,32,power_law_1.2,0.08858240246772767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,32,power_law_1.2,0.08495360016822814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,32,power_law_1.2,0.0887935996055603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,32,power_law_1.2,0.0870527982711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,32,power_law_1.2,0.08633599877357483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,4,power_law_1.2,0.3053312063217163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,32,power_law_1.2,0.09297919869422913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,4,power_law_1.2,0.3166912078857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,4,power_law_1.2,0.28448638916015623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,4,power_law_1.2,0.2672575950622559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,4,power_law_1.2,0.2948992013931274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,32,power_law_1.2,0.09342719912528992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,32,power_law_1.2,0.10347520112991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,32,power_law_1.2,0.11224960088729859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,32,power_law_1.2,0.12431360483169555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,32,balanced,0.04215999941031138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,32,power_law_1.2,0.1418239951133728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,32,power_law_1.2,0.16783360242843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,32,power_law_1.2,0.2167680025100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,4,power_law_1.2,0.33131520748138427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,32,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,32,balanced,0.042805333932240806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,4,power_law_1.2,0.3218559980392456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,32,balanced,0.04164266586303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,32,balanced,0.0458186666170756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,32,power_law_1.2,0.2441472053527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,32,balanced,0.05563200016816457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,32,balanced,0.06152533491452535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,32,balanced,0.06453866759936015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,32,power_law_1.2,0.3560192108154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,32,balanced,0.07326399783293407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,32,balanced,0.07457066575686137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,32,balanced,0.07116266588370006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,32,power_law_1.2,0.4182720184326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,32,power_law_1.2,0.6501120090484619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,32,power_law_1.2,0.7769919872283936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,4,power_law_1.2,0.3355456113815308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,4,power_law_1.2,0.3460479974746704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,32,balanced,0.08357333143552144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,4,power_law_1.2,0.35812480449676515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,4,power_law_1.2,0.4140031814575195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,4,power_law_1.2,0.43872637748718263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,4,power_law_1.2,0.5108352184295655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,4,power_law_1.2,0.5448575973510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,32,balanced,0.08507200082143147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,32,balanced,0.12770666678746542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,32,balanced,0.1602079967657725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,32,balanced,0.16141866644223532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,32,balanced,0.1953279972076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,32,balanced,0.2609386642773946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,32,balanced,0.30858665704727173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,32,balanced,0.4384746551513672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,32,balanced,0.24064532915751138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,32,balanced,0.30727465947469074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,4,power_law_1.2,0.6827136039733886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,32,balanced,0.39553598562876385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,32,balanced,0.5502080122629801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,4,power_law_1.2,0.7963520050048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,32,balanced,0.7254613240559896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,4,power_law_1.2,1.0615103721618653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,32,balanced,1.087381362915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,32,balanced,1.4181119600931804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,4,power_law_1.2,1.193183994293213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,4,power_law_1.2,1.6232128143310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,4,power_law_1.2,2.0372735977172853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,4,balanced,0.021397332350413006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,4,balanced,0.022197333474953968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,4,balanced,0.024858665963013966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,4,balanced,0.02718399961789449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,4,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,4,balanced,0.04558399816354116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,4,balanced,0.048170665899912514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,4,balanced,0.05044800043106079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,4,balanced,0.05402666827042898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,4,balanced,0.055674667159716286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,4,balanced,0.05852800110975901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,4,balanced,0.05894933144251505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,4,balanced,0.07433066765467326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,4,balanced,0.07427200178305308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,4,balanced,0.07527466615041097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,4,balanced,0.09645332892735799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,4,balanced,0.1179039975007375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,4,balanced,0.12968533237775168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,4,balanced,0.13774399956067404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,4,balanced,0.15165332953135172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,4,balanced,0.15822399655977884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,4,balanced,0.19313599665959677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,4,balanced,0.21195199092229208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,4,balanced,0.32274667421976727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,4,balanced,0.3810133139292399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,4,balanced,0.5431573390960693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,4,balanced,0.7067786852518717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,4,power_law_1.01,0.032627201080322264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,4,power_law_1.01,0.02961919903755188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,4,power_law_1.01,0.02794879972934723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,4,power_law_1.01,0.02876800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,4,power_law_1.01,0.028755199909210206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,4,power_law_1.01,0.029651200771331786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,4,power_law_1.01,0.029606398940086365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,4,power_law_1.01,0.030355200171470642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,4,power_law_1.01,0.030687999725341798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,4,power_law_1.01,0.03214719891548157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,4,power_law_1.01,0.032364800572395325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,4,power_law_1.01,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,4,power_law_1.01,0.03409920036792755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,4,power_law_1.01,0.03540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,4,power_law_1.01,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,4,power_law_1.01,0.04010879993438721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,4,power_law_1.01,0.04806399941444397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,4,power_law_1.01,0.05457280278205871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,4,power_law_1.01,0.05711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,4,power_law_1.01,0.06836479902267456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,4,power_law_1.01,0.07616000175476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,4,power_law_1.01,0.10201599597930908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,4,power_law_1.01,0.11753599643707276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,4,power_law_1.01,0.15196160078048707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,4,power_law_1.01,0.1890496015548706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,4,power_law_1.01,0.27545599937438964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,4,power_law_1.01,0.3494719982147217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,256,power_law_1.01,0.04524799883365631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,256,power_law_1.01,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,256,power_law_1.01,0.04560000002384186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,256,power_law_1.01,0.04815999865531921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,256,power_law_1.01,0.04898560047149658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,256,power_law_1.01,0.05107839703559876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,256,power_law_1.01,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,256,power_law_1.01,0.054579198360443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,256,power_law_1.01,0.05798400044441223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,256,power_law_1.01,0.058905601501464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,256,power_law_1.01,0.05681279897689819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,256,power_law_1.01,0.05960320234298706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,256,power_law_1.01,0.06275839805603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,256,power_law_1.01,0.0611519992351532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,256,power_law_1.01,0.07159039974212647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,256,power_law_1.01,0.07365120053291321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,256,power_law_1.01,0.07811200022697448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,256,power_law_1.01,0.08863999843597412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,256,power_law_1.01,0.09836159944534302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,256,power_law_1.01,0.11484160423278808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,256,power_law_1.01,0.1390079975128174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,256,power_law_1.01,0.19304319620132446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,256,power_law_1.01,0.2329472064971924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,256,power_law_1.01,0.31957759857177737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,256,power_law_1.01,0.4042496204376221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,256,power_law_1.01,0.571340799331665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,256,power_law_1.01,0.8091648101806641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,8,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,8,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,8,balanced,0.03788266579310099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,8,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,8,balanced,0.0535093347231547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,8,balanced,0.05323733389377594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,8,balanced,0.05406400064627329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,8,balanced,0.05542933444182078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,8,balanced,0.05445333321889242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,8,balanced,0.054917335510253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,8,balanced,0.05522666871547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,8,balanced,0.056986664732297264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,8,balanced,0.057573333382606506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,8,balanced,0.058037335673967995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,8,balanced,0.0692853331565857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,8,balanced,0.07019733389218648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,8,balanced,0.07031466563542683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,8,balanced,0.0746506651242574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,8,balanced,0.08386133114496867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,8,balanced,0.09198932846387227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,8,balanced,0.10265066226323445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,8,balanced,0.1264959971110026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,8,balanced,0.14542933305104574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,8,balanced,0.20436267058054605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,8,balanced,0.25868799289067584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,8,balanced,0.3749706745147705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,8,balanced,0.4902720053990682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,1,power_law_1.2,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,1,power_law_1.2,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,1,power_law_1.2,0.049542400240898135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,1,power_law_1.2,0.06441599726676941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,1,power_law_1.2,0.06510720252990723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,1,power_law_1.2,0.06626560091972351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,1,power_law_1.2,0.0660863995552063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,1,power_law_1.2,0.06698880195617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,1,power_law_1.2,0.0670144021511078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,1,power_law_1.2,0.06704639792442321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,1,power_law_1.2,0.0683456003665924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,1,power_law_1.2,0.07210239768028259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,1,power_law_1.2,0.07641599774360656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,1,power_law_1.2,0.08457599878311158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,1,power_law_1.2,0.09553279876708984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,1,power_law_1.2,0.09959040284156799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,1,power_law_1.2,0.10341119766235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,1,power_law_1.2,0.12108800411224366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.2,0.14746880531311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.2,0.19383039474487304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.2,0.2299328088760376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.2,0.3236991882324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.2,0.4379007816314697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.2,0.6174464225769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,1,balanced,0.1281599998474121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,1,balanced,0.21654399236043295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,1,balanced,0.3870240052541097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,1,balanced,0.3912159999211629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,1,balanced,0.3895680109659831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,1,balanced,0.3919839859008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,1,balanced,0.3952053387959798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,1,balanced,0.3969279925028483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,1,balanced,0.40243200461069745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,1,balanced,0.4023199876149495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,1,balanced,0.4059893290201823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,1,balanced,0.41233599185943604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,1,balanced,0.4151466687520345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,1,balanced,0.4424533446629842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,1,balanced,0.4438133239746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,1,balanced,0.44893864790598553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,1,balanced,0.46694934368133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,1,balanced,0.579205314318339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,1,balanced,0.5903893311818441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,1,balanced,0.8098133405049642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,1,balanced,0.9104959964752197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,1,balanced,1.2607359886169434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,1,balanced,1.6769439379374187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,1,balanced,2.437839984893799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,1,balanced,3.2394078572591147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,1,balanced,4.715269406636556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,1,balanced,6.222970962524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,4,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,4,balanced,0.045935998360315956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,4,balanced,0.054469332098960876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,4,balanced,0.056133334835370384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,4,balanced,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,4,balanced,0.056277334690093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,4,balanced,0.056133334835370384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,4,balanced,0.057904000083605446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,4,balanced,0.062122667829195656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,4,balanced,0.059989333152770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,4,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,4,balanced,0.05435200035572052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,4,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,4,balanced,0.05213866631189982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,4,balanced,0.05589866638183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,4,balanced,0.05589333176612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,4,balanced,0.05553600192070007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,4,balanced,0.07330133517583211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,4,balanced,0.07019733389218648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,4,balanced,0.08759466807047527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,4,balanced,0.10125866532325745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,4,balanced,0.12772799531618753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,4,balanced,0.15108266472816467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,4,balanced,0.20573333899180093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,4,balanced,0.25594667593638104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,4,balanced,0.3626933495203654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,4,balanced,0.4668373266855876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.2,0.7991104125976562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.2,1.1674304008483887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,8,power_law_1.2,0.12513920068740844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,8,power_law_1.2,0.15518720149993898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,8,power_law_1.2,0.17297279834747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,16,power_law_1.2,0.04927999973297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,8,power_law_1.2,0.19998719692230224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,16,power_law_1.2,0.06042240262031555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,16,power_law_1.2,0.04783360064029694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,8,power_law_1.2,0.2262655973434448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,16,power_law_1.2,0.05306879878044128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,16,power_law_1.2,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,16,power_law_1.2,0.05333120226860046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,16,power_law_1.2,0.05422719717025757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,16,power_law_1.2,0.053932797908782956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,16,power_law_1.2,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,16,power_law_1.2,0.05748479962348938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,8,power_law_1.2,0.2318336009979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,16,power_law_1.2,0.06085119843482971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,8,power_law_1.2,0.23367040157318114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,8,power_law_1.2,0.22024960517883302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,8,power_law_1.2,0.2290112018585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,8,power_law_1.2,0.23765120506286622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,8,power_law_1.2,0.24062080383300782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,8,power_law_1.2,0.2484800100326538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,16,power_law_1.2,0.06362239718437195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,8,power_law_1.2,0.24883840084075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,16,power_law_1.2,0.06521599888801574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,8,power_law_1.2,0.2617023944854736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,8,power_law_1.2,0.27295360565185545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,8,power_law_1.2,0.2859711885452271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,8,power_law_1.2,0.3017215967178345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,8,power_law_1.2,0.35011839866638184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,16,power_law_1.2,0.06793599724769592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,8,power_law_1.2,0.35653119087219237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,16,power_law_1.2,0.0759935975074768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,8,power_law_1.2,0.4375936031341553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,8,power_law_1.2,0.49368958473205565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,8,power_law_1.2,0.6146751880645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,8,power_law_1.2,0.6323071956634522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,16,power_law_1.2,0.07930880188941955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,8,power_law_1.2,0.7662784099578858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,16,power_law_1.2,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,16,power_law_1.2,0.10216319561004639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,16,power_law_1.2,0.10907520055770874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,16,power_law_1.2,0.1385151982307434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,8,power_law_1.2,0.9807871818542481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,16,power_law_1.2,0.15941760540008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,16,power_law_1.2,0.215667200088501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,8,power_law_1.2,1.5119615554809571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,16,power_law_1.2,0.25868799686431887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,16,power_law_1.2,0.36912639141082765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,16,power_law_1.2,0.4848896026611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,16,power_law_1.2,0.6602687835693359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,8,power_law_1.2,1.7260095596313476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,16,power_law_1.2,0.8750656127929688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.2,1.5287936210632325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,2,power_law_1.2,0.047654399275779726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,2,power_law_1.2,0.050988799333572386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,2,power_law_1.2,0.057120001316070555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,2,power_law_1.2,0.06646400094032287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,2,power_law_1.2,0.07654399871826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,2,power_law_1.2,0.08831359744071961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,2,power_law_1.2,0.10468480587005616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,2,power_law_1.2,0.10833920240402221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,2,power_law_1.2,0.10817919969558716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,2,power_law_1.2,0.11118719577789307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,2,power_law_1.2,0.11358720064163208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,2,power_law_1.2,0.11700479984283448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,2,power_law_1.2,0.12210559844970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,2,power_law_1.2,0.12799999713897706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,2,power_law_1.2,0.13649280071258546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,2,power_law_1.2,0.1429952025413513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,2,power_law_1.2,0.14905600547790526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,2,power_law_1.2,0.16670080423355102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,2,power_law_1.2,0.18447359800338745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,2,power_law_1.2,0.21914880275726317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,2,power_law_1.2,0.24843521118164064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,2,power_law_1.2,0.32045440673828124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,2,power_law_1.2,0.37715198993682864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,2,power_law_1.2,0.4976384162902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,2,power_law_1.2,0.6713151931762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,2,power_law_1.2,0.9511808395385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,8,power_law_1.01,0.03982079923152924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,8,power_law_1.01,0.03992320001125336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,2,power_law_1.2,1.1360639572143554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,8,power_law_1.01,0.0395583987236023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,8,power_law_1.01,0.041766399145126344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,8,power_law_1.01,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,8,power_law_1.01,0.04418559968471527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,8,power_law_1.01,0.044582399725914004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,8,power_law_1.01,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,8,power_law_1.01,0.045516800880432126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,8,power_law_1.01,0.04657920002937317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,8,power_law_1.01,0.04725759923458099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,8,power_law_1.01,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,8,power_law_1.01,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,8,power_law_1.01,0.04907520115375519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,8,power_law_1.01,0.053446400165557864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,8,power_law_1.01,0.05464959740638733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,8,power_law_1.01,0.05800960063934326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,8,power_law_1.01,0.06290559768676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,2,balanced,0.04576000074545542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,2,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,2,balanced,0.05421866476535797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,2,balanced,0.07898133496443431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,2,balanced,0.1157919963200887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,2,balanced,0.17961066961288452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,2,balanced,0.18047465880711874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,2,balanced,0.18659200270970663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,2,balanced,0.1844693422317505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,2,balanced,0.18628267447153726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,2,balanced,0.18853867053985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,2,balanced,0.1900906761487325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,2,balanced,0.19195733467737833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,2,balanced,0.19556266069412231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,2,balanced,0.19912532965342203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,2,balanced,0.20148267348607382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,2,balanced,0.20698134104410806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,2,balanced,0.2233333388964335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,2,balanced,0.23459200064341226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,2,balanced,0.26021866003672284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,2,balanced,0.2940586606661479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,2,balanced,0.35091733932495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,2,balanced,0.40990400314331055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,2,balanced,0.5357546806335449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,8,power_law_1.01,0.0695039987564087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,2,balanced,0.6398346821467081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,8,power_law_1.01,0.07806079983711242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,2,balanced,0.8927626609802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,8,power_law_1.01,0.08874880075454712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,8,power_law_1.01,0.11606400012969971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,8,power_law_1.01,0.12872960567474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,8,power_law_1.01,0.16895359754562378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,8,power_law_1.01,0.2300544023513794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,8,power_law_1.01,0.3334784030914307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,8,power_law_1.01,0.40529918670654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,2,balanced,1.1192800203959148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,1,power_law_1.2,0.1041599988937378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,1,power_law_1.2,0.1552832007408142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,1,power_law_1.2,0.25050880908966067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,1,power_law_1.2,0.3682944059371948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,1,power_law_1.2,0.4868031978607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,1,power_law_1.2,0.8759167671203614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,1,power_law_1.2,0.9124544143676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,1,power_law_1.2,0.9832575798034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,1,power_law_1.2,0.9892736434936523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,1,power_law_1.2,1.0201855659484864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,1,power_law_1.2,1.0619839668273925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,1,power_law_1.2,1.093881607055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,1,power_law_1.2,1.1087103843688966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,1,power_law_1.2,1.1307904243469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,1,power_law_1.2,1.178828811645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,1,power_law_1.2,1.2104000091552733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,1,power_law_1.2,1.2653823852539063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,1,power_law_1.2,1.390732765197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,1,power_law_1.2,1.4475263595581054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,1,power_law_1.2,1.638035202026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,1,power_law_1.2,1.6539583206176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,1,power_law_1.2,1.977132797241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,1,power_law_1.2,2.314566421508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,1,power_law_1.2,2.904319953918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,1,power_law_1.2,4.140268707275391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,1,power_law_1.2,5.256083297729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,1,power_law_1.2,6.203123092651367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,16,power_law_1.01,0.0719488024711609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,16,power_law_1.01,0.09552639722824097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,16,power_law_1.01,0.10241919755935669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,16,power_law_1.01,0.10298880338668823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,16,power_law_1.01,0.10272639989852905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,16,power_law_1.01,0.09501439929008484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,16,power_law_1.01,0.09279999732971192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,16,power_law_1.01,0.10404479503631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,16,power_law_1.01,0.09616640210151672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,16,power_law_1.01,0.10072959661483764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,16,power_law_1.01,0.10410879850387574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,16,power_law_1.01,0.10254720449447632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,16,power_law_1.01,0.10595200061798096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,16,power_law_1.01,0.10622080564498901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,16,power_law_1.01,0.1165120005607605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,16,power_law_1.01,0.11728639602661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,1,power_law_1.01,0.018054400384426118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,16,power_law_1.01,0.12564480304718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,1,power_law_1.01,0.020051200687885285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,16,power_law_1.01,0.1376255989074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,1,power_law_1.01,0.025836798548698425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,16,power_law_1.01,0.15333759784698486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,1,power_law_1.01,0.035148799419403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,16,power_law_1.01,0.17909120321273803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,1,power_law_1.01,0.04524160027503967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,16,power_law_1.01,0.19996800422668456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,1,power_law_1.01,0.05942400097846985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,16,power_law_1.01,0.25535359382629397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,1,power_law_1.01,0.07300480008125305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,16,power_law_1.01,0.2970815896987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,1,power_law_1.01,0.07705600261688232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,1,power_law_1.01,0.07874559760093688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,1,power_law_1.01,0.0805184006690979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,1,power_law_1.01,0.08824959993362427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,1,power_law_1.01,0.08739200234413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,1,power_law_1.01,0.09516159892082214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,1,power_law_1.01,0.10056320428848267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,1,power_law_1.01,0.09976320266723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,16,power_law_1.01,0.39006719589233396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,1,power_law_1.01,0.10461440086364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,1,power_law_1.01,0.11279360055923462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,1,power_law_1.01,0.11326719522476196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,1,power_law_1.01,0.12133760452270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,1,power_law_1.01,0.1303231954574585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,1,power_law_1.01,0.14767999649047853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,1,power_law_1.01,0.2850239992141724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,1,power_law_1.01,0.34232320785522463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,1,power_law_1.01,0.2882431983947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,1,power_law_1.01,0.33544321060180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,1,power_law_1.01,0.4351808071136475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,1,power_law_1.01,0.545465612411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,16,power_law_1.01,0.4992256164550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,16,power_law_1.01,0.7086912155151367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,16,power_law_1.01,0.8517375946044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,1,balanced,0.07258666555086772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,1,balanced,0.09635733564694722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,1,balanced,0.13132799665133157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,1,balanced,0.2019360065460205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,1,balanced,0.33722134431203205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,1,balanced,0.4076586564381917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,1,balanced,0.4126826524734497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,1,balanced,0.413263996442159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,1,balanced,0.4156533479690552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,1,balanced,0.4185493389765422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,1,balanced,0.42320533593495685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,1,balanced,0.4267573356628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,1,balanced,0.4331200122833252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,1,balanced,0.44175998369852704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,1,balanced,0.4528053204218547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,1,balanced,0.4617919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,1,balanced,0.4808479944864909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,1,balanced,0.5259146690368652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,1,balanced,0.569973349571228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,1,balanced,0.8540799617767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,1,balanced,0.7244853178660074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,1,balanced,1.0703252951304119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,1,balanced,1.1322986284891765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,1,balanced,1.7615787188212078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,1,balanced,2.0480586687723794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,1,balanced,2.948725382486979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,1,balanced,3.815967877705892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,8,power_law_1.01,0.07012479901313781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,8,power_law_1.01,0.08663679957389832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,8,power_law_1.01,0.09637119770050048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,8,power_law_1.01,0.10026880502700805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,8,power_law_1.01,0.10339839458465576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,8,power_law_1.01,0.10405759811401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,8,power_law_1.01,0.1185920000076294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,8,power_law_1.01,0.11582080125808716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,8,power_law_1.01,0.11507840156555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,8,power_law_1.01,0.11568000316619872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,8,power_law_1.01,0.11412479877471923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,8,power_law_1.01,0.11730560064315795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,8,power_law_1.01,0.1217344045639038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,8,power_law_1.01,0.12142720222473144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,8,power_law_1.01,0.13076479434967042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,8,power_law_1.01,0.13255679607391357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,8,power_law_1.01,0.1356608033180237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,8,power_law_1.01,0.15746560096740722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,8,power_law_1.01,0.16609280109405516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,8,power_law_1.01,0.1967743992805481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,8,power_law_1.01,0.21905279159545898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,8,power_law_1.01,0.28770558834075927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,8,power_law_1.01,0.36262400150299073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,8,power_law_1.01,0.4813504219055176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,8,power_law_1.01,0.6524096012115479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,8,power_law_1.01,0.920531177520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,8,power_law_1.01,1.165766429901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,8,balanced,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,8,balanced,0.05073600014050802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,8,balanced,0.05212266743183136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,8,balanced,0.05797866483529409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,8,balanced,0.07016533116499583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,8,balanced,0.08247466882069905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,8,balanced,0.08353066444396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,8,balanced,0.084389328956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,8,balanced,0.08475200335184734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,8,balanced,0.08455466230710347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,8,balanced,0.08478400111198425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,8,balanced,0.08660266796747844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,8,balanced,0.09000533819198608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,8,balanced,0.08899733424186707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,8,balanced,0.0965173343817393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,8,balanced,0.09685867031415303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,8,balanced,0.09995200236638387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,8,balanced,0.11082133650779724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,8,balanced,0.11935466527938843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,8,balanced,0.1524799962838491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,8,balanced,0.1525759994983673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,8,balanced,0.19502933820088705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,8,balanced,0.2260319987932841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,8,balanced,0.31996800502141315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,8,balanced,0.38179198900858563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,8,balanced,0.5209973255793253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,8,balanced,0.6721493403116862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,2,power_law_1.01,0.08250880241394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,2,power_law_1.01,0.10631680488586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,2,power_law_1.01,0.13891199827194214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,2,power_law_1.01,0.1562880039215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,2,power_law_1.01,0.2187903881072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,2,power_law_1.01,0.24985599517822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,2,power_law_1.01,0.2808768033981323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,2,power_law_1.01,0.3526463985443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,2,power_law_1.01,0.3528768062591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,2,power_law_1.01,0.3706943988800049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,2,power_law_1.01,0.36759679317474364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,2,power_law_1.01,0.3911360025405884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,2,power_law_1.01,0.3883392095565796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,2,power_law_1.01,0.40238080024719236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,2,power_law_1.01,0.4177152156829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,2,power_law_1.01,0.4258431911468506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,2,power_law_1.01,0.4530367851257324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,2,power_law_1.01,0.5027647972106933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,2,power_law_1.01,0.5445248126983643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,2,power_law_1.01,0.6255743980407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,2,power_law_1.01,0.6599232196807862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,2,power_law_1.01,0.8192959785461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,2,power_law_1.01,0.9901375770568848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,2,power_law_1.01,1.3001343727111816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,2,power_law_1.01,1.5839872360229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,2,power_law_1.01,2.1168832778930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,2,power_law_1.01,2.6180992126464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,32,balanced,0.02075200031201045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,32,balanced,0.018976000448067982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,32,balanced,0.01933866615096728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,32,balanced,0.019098666807015736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,32,balanced,0.023002666731675465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,32,balanced,0.025040000677108765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,32,balanced,0.027349332968393963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,32,balanced,0.027109332382678986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,32,balanced,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,32,balanced,0.03186666717131933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,32,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,32,balanced,0.03579200059175491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,32,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,32,balanced,0.043493335445721946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,32,balanced,0.05241066714127859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,32,balanced,0.05202133456865946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,32,balanced,0.05585066477457682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,32,balanced,0.07025599976380666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,32,balanced,0.07608533402283986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,32,balanced,0.10973333319028218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,32,balanced,0.08238933483759563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,32,balanced,0.09990400075912476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,32,balanced,0.11991467078526814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,32,balanced,0.16019200285275778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,32,balanced,0.20164799690246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,32,balanced,0.28964267174402875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,32,balanced,0.3721280097961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,2,power_law_1.01,0.034176000952720643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,2,power_law_1.01,0.036083200573921205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,2,power_law_1.01,0.037964800000190736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,2,power_law_1.01,0.040940800309181215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,2,power_law_1.01,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,2,power_law_1.01,0.0554751992225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,2,power_law_1.01,0.05692160129547119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,2,power_law_1.01,0.056492799520492555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,2,power_law_1.01,0.06033920049667359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,2,power_law_1.01,0.063264000415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,2,power_law_1.01,0.06477439999580384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,2,power_law_1.01,0.06707199811935424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,2,power_law_1.01,0.07134079933166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,2,power_law_1.01,0.07189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,2,power_law_1.01,0.07994239926338195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,2,power_law_1.01,0.08529279828071594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,2,power_law_1.01,0.09461119771003723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,2,power_law_1.01,0.11144319772720337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,2,power_law_1.01,0.1360576033592224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,2,power_law_1.01,0.16563839912414552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,2,power_law_1.01,0.19756799936294556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,2,power_law_1.01,0.2637631893157959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,2,power_law_1.01,0.33692800998687744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,2,power_law_1.01,0.5034431934356689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,2,power_law_1.01,0.6147647857666015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,2,power_law_1.01,0.8772928237915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,2,power_law_1.01,1.1843968391418458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,16,power_law_1.01,0.03961600065231323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,16,power_law_1.01,0.03468160033226013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,16,power_law_1.01,0.032716798782348636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,16,power_law_1.01,0.031193599104881287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,16,power_law_1.01,0.03086720108985901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,16,power_law_1.01,0.03250559866428375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,16,power_law_1.01,0.032902398705482484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,16,power_law_1.01,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,16,power_law_1.01,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,16,power_law_1.01,0.03384959995746613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,16,power_law_1.01,0.03594880104064942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,16,power_law_1.01,0.03711999952793121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,16,power_law_1.01,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,16,power_law_1.01,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,16,power_law_1.01,0.04266240000724793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,16,power_law_1.01,0.0455808013677597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,16,power_law_1.01,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,16,power_law_1.01,0.04965760111808777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,16,power_law_1.01,0.0542464017868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,16,power_law_1.01,0.061913597583770755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,16,power_law_1.01,0.06806399822235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,16,power_law_1.01,0.08399360179901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,16,power_law_1.01,0.1081536054611206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,16,power_law_1.01,0.15168000459671022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,16,power_law_1.01,0.17635200023651124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,16,power_law_1.01,0.24844799041748047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,16,power_law_1.01,0.31379199028015137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,1,balanced,0.04260266820589701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,1,balanced,0.04906666775544485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,1,balanced,0.0614879975716273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,1,balanced,0.06565333406130473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,1,balanced,0.06807466844717662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,1,balanced,0.06843199829260509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,1,balanced,0.06926933427651723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,1,balanced,0.07048533360163371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,1,balanced,0.07029333213965099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,1,balanced,0.07042666773001353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,1,balanced,0.07021866738796234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,1,balanced,0.07426133255163829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,1,balanced,0.07424533367156982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,1,balanced,0.0746666689713796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,1,balanced,0.0823359986146291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,1,balanced,0.08486933509508769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,1,balanced,0.08454400300979614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,1,balanced,0.10842667023340861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,1,balanced,0.1169599990049998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,1,balanced,0.15180266896883646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,1,balanced,0.1845973332722982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,1,balanced,0.2550613284111023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,1,balanced,0.3189653356870015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,1,balanced,0.461952010790507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,1,balanced,0.6011360088984171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,1,balanced,0.8614400227864584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,1,balanced,1.1289119720458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,16,16,power_law_1.2,0.05318400263786316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,16,16,power_law_1.2,0.05699840188026428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,16,16,power_law_1.2,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,16,16,power_law_1.2,0.058361601829528806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,16,16,power_law_1.2,0.05754240155220032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,16,16,power_law_1.2,0.06037759780883789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,16,16,power_law_1.2,0.05978239774703979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,16,16,power_law_1.2,0.06245120167732239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,16,16,power_law_1.2,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,16,16,power_law_1.2,0.06381440162658691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,16,16,power_law_1.2,0.06322559714317322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,16,16,power_law_1.2,0.06639999747276307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,16,16,power_law_1.2,0.06636160016059875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,16,16,power_law_1.2,0.0694208025932312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,16,16,power_law_1.2,0.07918720245361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,16,16,power_law_1.2,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,16,16,power_law_1.2,0.08628479838371277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,16,16,power_law_1.2,0.09717760086059571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,16,16,power_law_1.2,0.11111680269241334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,16,16,power_law_1.2,0.13445760011672975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,16,16,power_law_1.2,0.15445760488510132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,16,16,power_law_1.2,0.20376319885253907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,16,16,power_law_1.2,0.25754880905151367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,16,16,power_law_1.2,0.38794879913330077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,16,16,power_law_1.2,0.4870912075042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,16,16,power_law_1.2,0.6898111820220947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,16,16,power_law_1.2,0.9972224235534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,1,power_law_1.2,0.05368319749832153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,1,power_law_1.2,0.05782399773597717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,1,power_law_1.2,0.06883199810981751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,1,power_law_1.2,0.07975040078163147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,1,power_law_1.2,0.09476479887962341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,1,power_law_1.2,0.1322368025779724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,1,power_law_1.2,0.13856639862060546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,1,power_law_1.2,0.14200960397720336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,1,power_law_1.2,0.14517120122909546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,1,power_law_1.2,0.14584319591522216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,1,power_law_1.2,0.14921599626541138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,1,power_law_1.2,0.15779839754104613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,1,power_law_1.2,0.1599552035331726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,1,power_law_1.2,0.16530560255050658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,1,power_law_1.2,0.17458560466766357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,1,power_law_1.2,0.18272000551223755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,1,power_law_1.2,0.2009727954864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,1,power_law_1.2,0.23734400272369385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.2,0.2741760015487671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.2,0.35080320835113527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.2,0.4089536190032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.2,0.5570303916931152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.2,0.7013951778411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.2,0.9856127738952637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.2,1.2598719596862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,1,power_law_1.01,0.02106879949569702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,1,power_law_1.01,0.02279680073261261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,1,power_law_1.01,0.027635198831558228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,1,power_law_1.01,0.03518719971179962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,1,power_law_1.01,0.05303679704666138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,1,power_law_1.01,0.0716863989830017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,1,power_law_1.01,0.08391680121421814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,1,power_law_1.01,0.08725119829177856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,1,power_law_1.01,0.09146239757537841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,1,power_law_1.01,0.09641600251197815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,1,power_law_1.01,0.09886720180511474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,1,power_law_1.01,0.10216959714889526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,1,power_law_1.01,0.1061568021774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,1,power_law_1.01,0.10958720445632934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,1,power_law_1.01,0.11361279487609863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,1,power_law_1.01,0.11701120138168335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,1,power_law_1.01,0.12343679666519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,1,power_law_1.01,0.14298239946365357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.01,0.1617535948753357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.01,0.19736959934234619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.2,1.8148096084594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.01,0.23121280670166017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.01,0.30271999835968016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.01,0.3334208011627197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.01,0.4487167835235596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.01,0.5620160102844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.01,0.7968895912170411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.01,1.0337023735046387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,128,power_law_1.2,0.0328000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,128,power_law_1.2,0.03423359990119934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,128,power_law_1.2,0.032864001393318173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,128,power_law_1.2,0.03763839900493622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,128,power_law_1.2,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,128,power_law_1.2,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,128,power_law_1.2,0.03258239924907684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,128,power_law_1.2,0.0328000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,128,power_law_1.2,0.032864001393318173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,128,power_law_1.2,0.03404799997806549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,128,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,128,power_law_1.2,0.036575999855995175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,128,power_law_1.2,0.03904640078544617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,128,power_law_1.2,0.04031359851360321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,128,power_law_1.2,0.04521600008010864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,128,power_law_1.2,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,128,power_law_1.2,0.049907198548316954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,128,power_law_1.2,0.06159999966621399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,128,power_law_1.2,0.06691840291023254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,128,power_law_1.2,0.0939136028289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,128,power_law_1.2,0.0997759997844696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,128,power_law_1.2,0.1388800024986267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,128,power_law_1.2,0.15552639961242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,128,power_law_1.2,0.23774080276489257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,128,power_law_1.2,0.29672319889068605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.2,2.3716672897338866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,128,power_law_1.2,0.44022397994995116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,128,power_law_1.2,0.6390655994415283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,64,balanced,0.03369066615899404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,64,balanced,0.03405333310365677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,64,balanced,0.029487999776999157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,64,balanced,0.02935466667016347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,64,balanced,0.031397332747777305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,64,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,64,balanced,0.03162666658560435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,64,balanced,0.03136533250411352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,64,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,64,balanced,0.03141333411137263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,64,balanced,0.03170666595300039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,64,balanced,0.03344533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,64,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,64,balanced,0.033359999457995095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,64,balanced,0.0353973334034284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,64,balanced,0.03605333218971888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,64,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,64,balanced,0.036618667344252266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,64,balanced,0.03756266583998998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,64,balanced,0.04152533411979675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,64,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,64,balanced,0.04186133543650309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,64,balanced,0.042634665966033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,64,balanced,0.049728001157442726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,64,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,64,balanced,0.0659093310435613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,64,balanced,0.07625066737333934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,16,balanced,0.021541332205136616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,16,balanced,0.021018666525681812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,16,balanced,0.019498666127522785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,16,balanced,0.02046400060256322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,16,balanced,0.021333334346612293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,16,balanced,0.02146133283774058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,16,balanced,0.022053333620230358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,16,balanced,0.023141334454218548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,16,balanced,0.023530667026837666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,16,balanced,0.02316266546646754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,16,balanced,0.02587733417749405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,16,balanced,0.02922666569550832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,16,balanced,0.02938133229811986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,16,balanced,0.036357333262761436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,16,balanced,0.04260799785455068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,16,balanced,0.04055466751257578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,16,balanced,0.04079466561476389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,16,balanced,0.04517866671085358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,16,balanced,0.050383999943733215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,16,balanced,0.0613013356924057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,16,balanced,0.075354665517807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,16,balanced,0.0909440020720164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,16,balanced,0.10566400488217671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,16,balanced,0.14864533146222433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,16,balanced,0.18469866116841635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,16,balanced,0.2616639931996663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,16,balanced,0.33874666690826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,8,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,8,balanced,0.039733332892258964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,8,balanced,0.039701332648595176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,8,balanced,0.04168533285458883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,8,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,8,balanced,0.041877334316571556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,8,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,8,balanced,0.04277333120505015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,8,balanced,0.04388799766699473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,8,balanced,0.042591998974482216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,8,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,8,balanced,0.04156800111134847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,8,balanced,0.04367466767628988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,8,balanced,0.04292266567548116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,8,balanced,0.043968002001444496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,8,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,8,balanced,0.048122664292653404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,8,balanced,0.05417599777380625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,8,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,8,balanced,0.06178666651248932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,8,balanced,0.06815466781457265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,8,balanced,0.08889066179593404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,8,balanced,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,8,balanced,0.1327786644299825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,8,balanced,0.1609333356221517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,8,balanced,0.21970132986704508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,8,balanced,0.27473066250483197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,2,power_law_1.01,0.07917439937591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,2,power_law_1.01,0.11817599534988403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,2,power_law_1.01,0.14296319484710693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,2,power_law_1.01,0.20563840866088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,2,power_law_1.01,0.2657279968261719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,2,power_law_1.01,0.31595520973205565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,2,power_law_1.01,0.4274303913116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,2,power_law_1.01,0.43802881240844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,2,power_law_1.01,0.4386879920959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,2,power_law_1.01,0.45978240966796874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,2,power_law_1.01,0.4693759918212891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,2,power_law_1.01,0.4910912036895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,2,power_law_1.01,0.4755136013031006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,2,power_law_1.01,0.5122560024261474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,2,power_law_1.01,0.5032639980316163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,2,power_law_1.01,0.5491903781890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,2,power_law_1.01,0.559123182296753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,2,power_law_1.01,0.6151423931121827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,2,power_law_1.01,0.6130943775177002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,2,power_law_1.01,0.7052351951599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,2,power_law_1.01,0.7013823986053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,2,power_law_1.01,0.7999680042266846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,2,power_law_1.01,0.9005248069763183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,2,power_law_1.01,1.1113727569580079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,2,power_law_1.01,1.2956671714782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,2,power_law_1.01,1.6777599334716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,2,power_law_1.01,2.1288639068603517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,32,4,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,32,4,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,32,4,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,32,4,balanced,0.053413331508636475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,32,4,balanced,0.060005332032839455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,32,4,balanced,0.08648533622423808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,32,4,balanced,0.08867733677228291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,32,4,balanced,0.08858666817347209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,32,4,balanced,0.08772266904513042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,32,4,balanced,0.08956266442934673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,32,4,balanced,0.09075733025868733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,32,4,balanced,0.09157333771387736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,32,4,balanced,0.09405333797136943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,32,4,balanced,0.0978559950987498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,32,4,balanced,0.10152000188827515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,32,4,balanced,0.10389866431554158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,32,4,balanced,0.10821333527565002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,32,4,balanced,0.11978666981061299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,32,4,balanced,0.1306880017121633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,32,4,balanced,0.15203199783960977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,32,4,balanced,0.17282666762669882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,32,4,balanced,0.21523733933766684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,32,4,balanced,0.2608799934387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,32,4,balanced,0.369429349899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,32,4,balanced,0.4511786699295044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,32,4,balanced,0.6497493187586466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,32,4,balanced,0.8446453412373861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,8,balanced,0.020026666422684986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,8,balanced,0.021136000752449036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,8,balanced,0.022997332115968067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,8,balanced,0.024874667326609295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,8,balanced,0.03316800047953924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,8,balanced,0.04725333551565806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,8,balanced,0.059818665186564125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,8,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,8,balanced,0.05668266614278158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2688,1856,6,128,1,8,balanced,0.08493866523106892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2688,1856,6,128,1,8,balanced,0.07941866914431255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2688,1856,6,128,1,8,balanced,0.07618133227030437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,2,power_law_1.2,0.07645440101623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,2,power_law_1.2,0.11247999668121338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,2,power_law_1.2,0.1350592017173767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,2,power_law_1.2,0.1875712037086487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,2,power_law_1.2,0.24532480239868165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,2,power_law_1.2,0.329964804649353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,2,power_law_1.2,0.43829760551452634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,2,power_law_1.2,0.4398335933685303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,2,power_law_1.2,0.4943552017211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,2,power_law_1.2,0.4524543762207031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,2,power_law_1.2,0.4683839797973633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,2,power_law_1.2,0.4932415962219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,2,power_law_1.2,0.49683837890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,2,power_law_1.2,0.5005631923675538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,2,power_law_1.2,0.5184832096099854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,2,power_law_1.2,0.5407296180725097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,2,power_law_1.2,0.5964479923248291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,2,power_law_1.2,0.6393472194671631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,2,power_law_1.2,0.6398719787597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,2,power_law_1.2,0.7211071968078613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,2,power_law_1.2,0.7308671951293946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,2,power_law_1.2,0.8427904129028321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,2,power_law_1.2,0.961241626739502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,2,power_law_1.2,1.160972785949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,2,power_law_1.2,1.3654335975646972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,2,power_law_1.2,1.8338367462158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,2,power_law_1.2,2.150924873352051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,8,power_law_1.01,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,8,power_law_1.01,0.05826560258865356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,8,power_law_1.01,0.0599295973777771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,8,power_law_1.01,0.06208639740943909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,8,power_law_1.01,0.06107519865036011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,8,power_law_1.01,0.06968960165977478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,8,power_law_1.01,0.07235199809074402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,8,power_law_1.01,0.07455360293388366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,8,power_law_1.01,0.07573760151863099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,8,power_law_1.01,0.0741375982761383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,8,power_law_1.01,0.07888640165328979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,8,power_law_1.01,0.07791360020637512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,8,power_law_1.01,0.07899519801139832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,8,power_law_1.01,0.08074880242347718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,8,power_law_1.01,0.08928639888763427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,8,power_law_1.01,0.09180160164833069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,8,power_law_1.01,0.09512959718704224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,8,power_law_1.01,0.10625920295715333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,8,power_law_1.01,0.11649919748306274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,8,power_law_1.01,0.13818880319595336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,8,power_law_1.01,0.1589951992034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,8,power_law_1.01,0.20273919105529786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,8,power_law_1.01,0.24220800399780273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,8,power_law_1.01,0.33035519123077395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,8,power_law_1.01,0.4061439990997314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,8,power_law_1.01,0.5656320095062256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,8,power_law_1.01,0.7167359828948975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,16,balanced,0.06426133215427399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,16,balanced,0.06044266621271769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,16,balanced,0.07851199805736542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,16,balanced,0.1046399970849355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,16,balanced,0.1518186628818512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,16,balanced,0.1563040018081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,16,balanced,0.1564853290716807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,16,balanced,0.14942933122316995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,16,balanced,0.15933332840601602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,16,balanced,0.1487679978211721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,16,balanced,0.15661866466204324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,16,balanced,0.15588266650835672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,16,balanced,0.1532533367474874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,16,balanced,0.15453867117563883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,16,balanced,0.1616159975528717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,16,balanced,0.15240533153216043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,16,balanced,0.150325338045756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,16,balanced,0.1636319955190023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,16,balanced,0.161871999502182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,16,balanced,0.16734933853149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,16,balanced,0.1722773313522339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,16,balanced,0.27244800329208374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,16,balanced,0.2752106587092082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,16,balanced,0.389413317044576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,16,balanced,0.49910934766133624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,16,balanced,0.5786826610565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,16,balanced,0.8109013239542643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,4,power_law_1.2,0.05117440223693848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,4,power_law_1.2,0.05026559829711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,4,power_law_1.2,0.054553598165512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,4,power_law_1.2,0.058931201696395874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,4,power_law_1.2,0.05660799741744995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,4,power_law_1.2,0.06848639845848084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,4,power_law_1.2,0.07436800003051758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,4,power_law_1.2,0.07382400035858154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,4,power_law_1.2,0.07736319899559022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,4,power_law_1.2,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,4,power_law_1.2,0.08512639999389648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,4,power_law_1.2,0.09047039747238159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,4,power_law_1.2,0.09479680061340331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,4,power_law_1.2,0.09264000058174134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,4,power_law_1.2,0.101254403591156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,4,power_law_1.2,0.10611200332641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,4,power_law_1.2,0.11514240503311157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,4,power_law_1.2,0.13730560541152953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,4,power_law_1.2,0.15189759731292723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,4,power_law_1.2,0.18710399866104127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,4,power_law_1.2,0.2551935911178589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,4,power_law_1.2,0.30899839401245116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,4,power_law_1.2,0.4000383853912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,4,power_law_1.2,0.5549759864807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,4,power_law_1.2,0.6721471786499024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,4,power_law_1.2,1.105395221710205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,4,power_law_1.2,1.3122559547424317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,64,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,64,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,64,balanced,0.031727999448776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,64,balanced,0.029658667743206024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,64,balanced,0.03162666658560435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,64,balanced,0.03369600077470144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,64,balanced,0.03327466547489166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,64,balanced,0.03331733246644338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,64,balanced,0.03336533407370249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,64,balanced,0.03293866664171219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,64,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,64,balanced,0.033733333150545754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,64,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,64,balanced,0.03566399961709976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,64,balanced,0.03782933453718821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,64,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,64,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,64,balanced,0.03789333254098892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,64,balanced,0.04158399999141693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,64,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,64,balanced,0.043712000052134194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,64,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,64,balanced,0.05068266888459524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,64,balanced,0.06413866579532623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,64,balanced,0.07223466535409291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,64,balanced,0.08591999610265096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,64,balanced,0.10707733035087585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,8,power_law_1.2,0.09642879962921143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,8,power_law_1.2,0.11806080341339112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,8,power_law_1.2,0.12789759635925294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,8,power_law_1.2,0.13269120454788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,8,power_law_1.2,0.17136640548706056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,8,power_law_1.2,0.16917760372161866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,8,power_law_1.2,0.1716223955154419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,8,power_law_1.2,0.1663424015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,8,power_law_1.2,0.1737280011177063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,8,power_law_1.2,0.1720896005630493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,8,power_law_1.2,0.18492799997329712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,8,power_law_1.2,0.18764159679412842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,8,power_law_1.2,0.18653440475463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,8,power_law_1.2,0.198636794090271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,8,power_law_1.2,0.2169343948364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,8,power_law_1.2,0.22621440887451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,8,power_law_1.2,0.2413952112197876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,8,power_law_1.2,0.2756671905517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,8,power_law_1.2,0.30203518867492674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,8,power_law_1.2,0.35445759296417234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,8,power_law_1.2,0.40930562019348143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,32,power_law_1.01,0.03389439880847931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,32,power_law_1.01,0.033580800890922545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,32,power_law_1.01,0.03242239952087402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,32,power_law_1.01,0.0348800003528595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,4,balanced,0.07253866891066234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,32,power_law_1.01,0.03492479920387268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,4,balanced,0.10514666636784871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,4,balanced,0.1529973347981771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,32,power_law_1.01,0.03726719915866852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,4,balanced,0.262554665406545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,32,power_law_1.01,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,4,balanced,0.46579734484354657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,32,power_law_1.01,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,4,balanced,0.5758560101191202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,32,power_law_1.01,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,4,balanced,0.579802672068278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,32,power_law_1.01,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,8,power_law_1.2,0.5107135772705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,32,power_law_1.01,0.038412800431251524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,32,power_law_1.01,0.03893760144710541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,8,power_law_1.2,0.5448192119598388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,32,power_law_1.01,0.03924480080604553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,32,power_law_1.01,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,32,power_law_1.01,0.04332799911499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,32,power_law_1.01,0.04359039962291718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,32,power_law_1.01,0.04542079865932465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,32,power_law_1.01,0.049132800102233885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,4,balanced,0.5779146750768026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,4,balanced,0.582757314046224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,8,power_law_1.2,0.79169921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,4,balanced,0.5856800079345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,4,balanced,0.5825333197911581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,4,balanced,0.5862186749776205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,4,balanced,0.5867573420206705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,4,balanced,0.5912853479385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,4,balanced,0.6020693381627401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,4,balanced,0.6014826695124308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,32,power_law_1.01,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,4,balanced,0.6143999894460043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,32,power_law_1.01,0.05761280059814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,4,balanced,0.6533653338750204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,32,power_law_1.01,0.061337602138519284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,4,balanced,0.6394879817962646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,32,power_law_1.01,0.0754047989845276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,4,balanced,0.7896533012390137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,32,power_law_1.01,0.0856768012046814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,4,balanced,0.7052106857299805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,32,power_law_1.01,0.11216000318527222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,4,balanced,1.0958027044932048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,32,power_law_1.01,0.14958720207214354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,4,balanced,0.8436319828033447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,32,power_law_1.01,0.1810431957244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,8,power_law_1.2,0.9688320159912109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,4,balanced,1.3828746477762859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,4,balanced,1.3130666414896648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,8,power_law_1.2,1.3398207664489745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,4,balanced,2.353109359741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,4,balanced,2.433114687601725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,32,power_law_1.01,0.2492095947265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,8,power_law_1.2,1.6299007415771485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,16,balanced,0.05013866722583771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,16,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,16,balanced,0.04966933528582255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,16,balanced,0.05596266686916351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,16,balanced,0.061994666854540505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,16,balanced,0.08687999844551086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,16,balanced,0.08865599830945332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,16,balanced,0.08898133039474487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,16,balanced,0.08930133779843648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,16,balanced,0.0906826655069987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,16,balanced,0.090037335952123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,16,balanced,0.09212799866994222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,16,balanced,0.09499733646710713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,16,balanced,0.09327999750773112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,16,balanced,0.10087466239929199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,16,balanced,0.10206400354703267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,16,balanced,0.10313066840171814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,16,balanced,0.11309333642323811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,16,balanced,0.12131200234095256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,16,balanced,0.1360160013039907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,16,balanced,0.1527466674645742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,16,balanced,0.18145600954691568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,16,balanced,0.21045867602030435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,16,balanced,0.2805280089378357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,16,balanced,0.33872532844543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,16,balanced,0.4618719816207886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,16,balanced,0.586570660273234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,16,power_law_1.2,0.03266560137271881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,16,power_law_1.2,0.02892799973487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,16,power_law_1.2,0.028755199909210206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,16,power_law_1.2,0.029574400186538695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,16,power_law_1.2,0.030995199084281923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,16,power_law_1.2,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,16,power_law_1.2,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,16,power_law_1.2,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,16,power_law_1.2,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,16,power_law_1.2,0.03283199965953827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,16,power_law_1.2,0.034406399726867674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,16,power_law_1.2,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,16,power_law_1.2,0.035596799850463864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,16,power_law_1.2,0.0369024008512497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,16,power_law_1.2,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,16,power_law_1.2,0.04147199988365173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,16,power_law_1.2,0.045747199654579164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,16,power_law_1.2,0.050393599271774295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,16,power_law_1.2,0.05619840025901794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,16,power_law_1.2,0.06763520240783691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,16,power_law_1.2,0.07534719705581665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,16,power_law_1.2,0.10113919973373413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,16,power_law_1.2,0.11574399471282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,16,power_law_1.2,0.16191999912261962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,16,power_law_1.2,0.2084223985671997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,16,power_law_1.2,0.3059391975402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,2,power_law_1.01,0.051507198810577394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,2,power_law_1.01,0.05655679702758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,2,power_law_1.01,0.06912639737129211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,2,power_law_1.01,0.08231679797172546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,2,power_law_1.01,0.09564800262451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,2,power_law_1.01,0.1179967999458313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,2,power_law_1.01,0.1385472059249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,16,power_law_1.2,0.39001600742340087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,2,power_law_1.01,0.14312959909439088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,2,power_law_1.01,0.14224640130996705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,2,power_law_1.01,0.14302079677581786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,2,power_law_1.01,0.157043194770813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,2,power_law_1.01,0.1565119981765747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,2,power_law_1.01,0.1652992010116577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,2,power_law_1.01,0.1729024052619934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,2,power_law_1.01,0.18083200454711915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,2,power_law_1.01,0.1805888056755066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,2,power_law_1.01,0.18859519958496093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,2,power_law_1.01,0.21806719303131103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,2,power_law_1.01,0.24624640941619874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,2,power_law_1.01,0.3049535989761353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,2,power_law_1.01,0.3344448089599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,1,power_law_1.01,0.056959998607635495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,2,power_law_1.01,0.4153728008270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,1,power_law_1.01,0.06159359812736511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,2,power_law_1.01,0.5251200199127197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,1,power_law_1.01,0.0812287986278534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,1,power_law_1.01,0.11189119815826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,2,power_law_1.01,0.6801536083221436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,1,power_law_1.01,0.14228479862213134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,2,power_law_1.01,0.8886912345886231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,1,power_law_1.01,0.18231040239334106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,1,power_law_1.01,0.23566720485687256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,2,power_law_1.01,1.1941760063171387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,1,power_law_1.01,0.24455680847167968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,1,power_law_1.01,0.2540031909942627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,2,power_law_1.01,1.6146751403808595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,1,power_law_1.01,0.2617343902587891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,1,power_law_1.01,0.27191040515899656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,1,power_law_1.01,0.27952001094818113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,1,power_law_1.01,0.2856575965881348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,1,power_law_1.01,0.29256319999694824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,1,power_law_1.01,0.30976641178131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,1,balanced,0.03577066709597906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,1,power_law_1.01,0.3175935983657837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,1,balanced,0.04558933277924856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,1,power_law_1.01,0.3439232110977173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,1,power_law_1.01,0.394598388671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,1,power_law_1.01,0.4351359844207764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,1,power_law_1.01,0.5295680046081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,1,power_law_1.01,0.6021503925323486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,1,balanced,0.06681600213050842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,1,power_law_1.01,0.7725120067596436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,1,balanced,0.1046453317006429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,1,balanced,0.17334399620691934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,1,balanced,0.22077866395314535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,1,balanced,0.22164267301559448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,1,balanced,0.22442134221394858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,1,balanced,0.225765327612559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,1,balanced,0.2246239980061849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,1,balanced,0.22610666354497275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,1,balanced,0.22789865732192993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,1,balanced,0.23097066084543863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,1,balanced,0.2339199980099996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,1,balanced,0.23731732368469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,1,balanced,0.24153600136439005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,1,balanced,0.24732265869776407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,1,power_law_1.01,0.9541312217712402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,1,balanced,0.26892266670862836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,1,balanced,0.2711679935455322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,1,balanced,0.3306559920310974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,1,balanced,0.3145013252894084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,1,balanced,0.4435253143310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,1,balanced,0.4280159870783488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,1,balanced,0.6286293268203735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,1,balanced,0.6974186897277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,1,balanced,1.016640027364095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,1,power_law_1.01,1.2826751708984374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,1,balanced,1.1800373395284016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,1,power_law_1.01,1.5992575645446778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,1,power_law_1.01,2.2430208206176756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,1,power_law_1.01,2.876038360595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,8,power_law_1.01,0.09255679845809936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,8,power_law_1.01,0.12007039785385132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,8,power_law_1.01,0.17697919607162477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,8,power_law_1.01,0.15664639472961425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,8,power_law_1.01,0.19906560182571412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,8,power_law_1.01,0.20680320262908936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,8,power_law_1.01,0.22785279750823975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,8,power_law_1.01,0.2166975975036621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,8,power_law_1.01,0.2361151933670044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,8,power_law_1.01,0.22472960948944093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,8,power_law_1.01,0.21639680862426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,8,power_law_1.01,0.20348799228668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,8,power_law_1.01,0.23352320194244386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,8,power_law_1.01,0.2237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,8,power_law_1.01,0.2266239881515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,8,power_law_1.01,0.2431936025619507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,8,power_law_1.01,0.24670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,8,power_law_1.01,0.25620479583740235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,8,power_law_1.01,0.28881919384002686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,8,power_law_1.01,0.32036480903625486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,8,power_law_1.01,0.3642879962921143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,8,power_law_1.01,0.45618557929992676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,8,power_law_1.01,0.4893951892852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,8,power_law_1.01,0.6520256042480469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,8,power_law_1.01,0.8146368026733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,8,power_law_1.01,1.0901760101318358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,8,power_law_1.01,1.3796223640441894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,4,balanced,0.021365332106749218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,4,balanced,0.023157333334287006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,4,balanced,0.025216000775496166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,4,balanced,0.029626667499542236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,4,balanced,0.04134399940570196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,4,balanced,0.058117335041364036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,4,balanced,0.05936000247796377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,4,balanced,0.06126399834950765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,4,balanced,0.06393600006898244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,4,balanced,0.06364800035953522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,4,balanced,0.06531199812889099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,4,balanced,0.0666240006685257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,4,balanced,0.06875733534495036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,4,balanced,0.07422400017579396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,4,balanced,0.07824533184369405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,4,balanced,0.08065066734949748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,4,balanced,0.0895146628220876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,4,balanced,0.10705066720644633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,4,balanced,0.11851200461387634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,4,balanced,0.1385546624660492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,4,balanced,0.19426133235295615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,4,balanced,0.21308799584706625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,4,balanced,0.25301865736643475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,4,balanced,0.3041386604309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,4,balanced,0.3631253242492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,4,balanced,0.5548319816589355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,4,balanced,0.6789546807607015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,2,balanced,0.05395199855168661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,2,balanced,0.07239999870459239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,2,balanced,0.10732266306877136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,2,balanced,0.175546665986379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,2,balanced,0.3007520039876302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,2,balanced,0.5539040168126425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,2,balanced,0.5568373203277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,2,balanced,0.556656002998352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,2,balanced,0.557808001836141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,2,balanced,0.5598826805750529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,2,balanced,0.5646560192108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,2,balanced,0.5655999978383383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,2,balanced,0.5663679838180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,2,balanced,0.5736106634140015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,2,balanced,0.5781439940134684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,2,balanced,0.5844053427378336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,2,balanced,0.5896639823913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,2,balanced,0.6064159870147705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,2,balanced,0.6242773135503134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,2,balanced,0.6555360158284506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,2,balanced,0.6944320201873779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,2,balanced,0.7502346833546957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,2,balanced,0.8041706879933676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,2,balanced,1.0121599833170574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,2,balanced,1.066480000813802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,2,balanced,1.5212799708048503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,2,balanced,1.6941973368326824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,1,power_law_1.2,0.02874239981174469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,1,power_law_1.2,0.029068800806999206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,1,power_law_1.2,0.03091840147972107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,1,power_law_1.2,0.034643200039863584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,1,power_law_1.2,0.033855998516082765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,1,power_law_1.2,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,1,power_law_1.2,0.037484800815582274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,1,power_law_1.2,0.037503999471664426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,1,power_law_1.2,0.038700801134109494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,1,power_law_1.2,0.039776000380516055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,1,power_law_1.2,0.039340800046920775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,1,power_law_1.2,0.042438399791717527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,1,power_law_1.2,0.04387840032577515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,1,power_law_1.2,0.04496000111103058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,1,power_law_1.2,0.0546176016330719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,1,power_law_1.2,0.05694079995155334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,1,power_law_1.2,0.05756160020828247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,1,power_law_1.2,0.0748095989227295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,4,power_law_1.2,0.021670399606227873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,4,power_law_1.2,0.024083200097084045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,4,power_law_1.2,0.028960001468658448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,4,power_law_1.2,0.031788799166679385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,4,power_law_1.2,0.0383103996515274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,4,power_law_1.2,0.041606399416923526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,4,power_law_1.2,0.045177599787712096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,4,power_law_1.2,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,4,power_law_1.2,0.052716797590255736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,4,power_law_1.2,0.050393599271774295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,4,power_law_1.2,0.0483711987733841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,4,power_law_1.2,0.06558079719543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,4,power_law_1.2,0.0576960027217865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,4,power_law_1.2,0.056576001644134524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,4,power_law_1.2,0.08940160274505615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,4,power_law_1.2,0.09089919924736023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,4,power_law_1.2,0.10117119550704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,4,power_law_1.2,0.09144319891929627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.2,0.09293439984321594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.2,0.10314240455627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.2,0.11562240123748779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.2,0.14042240381240845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,1,power_law_1.2,0.08355200290679932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.2,0.16431360244750975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,1,power_law_1.2,0.10592639446258545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,1,power_law_1.2,0.12414720058441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.2,0.21628160476684571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,1,power_law_1.2,0.1654144048690796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.2,0.2688256025314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,1,power_law_1.2,0.1985152006149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,16,power_law_1.2,0.023071999847888946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,16,power_law_1.2,0.027820798754692077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,16,power_law_1.2,0.026956799626350402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,16,power_law_1.2,0.028140801191329955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,16,power_law_1.2,0.030105599761009218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,16,power_law_1.2,0.03333120048046112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,16,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,16,power_law_1.2,0.03180800080299377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,16,power_law_1.2,0.04319359958171844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,16,power_law_1.2,0.046054399013519286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,16,power_law_1.2,0.04509440064430237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,16,power_law_1.2,0.0648639976978302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,16,power_law_1.2,0.06684160232543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,16,power_law_1.2,0.06728320121765137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,16,power_law_1.2,0.07578880190849305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,16,power_law_1.2,0.07754240036010743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,16,power_law_1.2,0.07467520236968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,16,power_law_1.2,0.07378559708595275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.2,0.0877951979637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.2,0.09560959935188293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.2,0.10688639879226684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.2,0.12472319602966309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.2,0.15959680080413818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.2,0.22259199619293213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.2,0.27752320766448973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.2,0.38204159736633303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.2,0.49590401649475097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.2,0.3704832077026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,1,power_law_1.2,0.28292479515075686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,1,power_law_1.2,0.35973119735717773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.2,0.5179903984069825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,1,power_law_1.2,0.5203328132629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,2,power_law_1.2,0.01664000004529953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,2,power_law_1.2,0.018163199722766876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,2,power_law_1.2,0.02115200012922287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,2,power_law_1.2,0.021529600024223328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,2,power_law_1.2,0.0239424005150795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,2,power_law_1.2,0.02622720003128052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,2,power_law_1.2,0.027263998985290527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,2,power_law_1.2,0.027641600370407103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,2,power_law_1.2,0.030342400074005127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,2,power_law_1.2,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,2,power_law_1.2,0.031667199730873105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,2,power_law_1.2,0.03325439989566803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,2,power_law_1.2,0.03500159978866577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,2,power_law_1.2,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,2,power_law_1.2,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,2,power_law_1.2,0.041228801012039185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,2,power_law_1.2,0.041382399201393125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,2,power_law_1.2,0.04761599898338318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,2,power_law_1.2,0.053311997652053834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,2,power_law_1.2,0.06631039977073669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,2,power_law_1.2,0.08373119831085205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,2,power_law_1.2,0.09846400022506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,2,power_law_1.2,0.10910719633102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,2,power_law_1.2,0.14409600496292113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,2,power_law_1.2,0.17934080362319946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,1,power_law_1.2,0.6822463989257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,2,power_law_1.2,0.24230399131774902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,2,power_law_1.2,0.305785608291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,2,power_law_1.01,0.06685439944267273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,2,power_law_1.01,0.07264000177383423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,2,power_law_1.01,0.07953919768333435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,2,power_law_1.01,0.09031040072441102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,2,power_law_1.01,0.10192639827728271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,2,power_law_1.01,0.1137279987335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,2,power_law_1.01,0.13839999437332154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,2,power_law_1.01,0.14471039772033692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,2,power_law_1.01,0.1464319944381714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,2,power_law_1.01,0.14731520414352417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,2,power_law_1.01,0.14913920164108277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,2,power_law_1.01,0.15422719717025757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,2,power_law_1.01,0.16050560474395753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,2,power_law_1.01,0.1612223982810974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,2,power_law_1.01,0.168230402469635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,2,power_law_1.01,0.17350399494171143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,2,power_law_1.01,0.18165760040283202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,2,power_law_1.01,0.20785920619964598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,2,power_law_1.01,0.23031680583953856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,2,power_law_1.01,0.2796031951904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,2,power_law_1.01,0.32467200756073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,2,power_law_1.01,0.4334144115447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,2,power_law_1.01,0.5345280170440674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,2,power_law_1.01,0.740505599975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,2,power_law_1.01,0.9463871955871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,2,power_law_1.01,1.355884838104248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,2,power_law_1.01,1.7372608184814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,4,balanced,0.0446720023949941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,4,balanced,0.04504533112049103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,4,balanced,0.05613866448402405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,4,balanced,0.05267733335494995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,4,balanced,0.056703999638557434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,4,balanced,0.054976001381874084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,4,balanced,0.05439466734727224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,4,balanced,0.05835733314355215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,4,balanced,0.056277334690093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,4,balanced,0.054416000843048096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,4,balanced,0.05246399839719137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,4,balanced,0.05423999826113383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,4,balanced,0.058634668588638306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,4,balanced,0.05258666475613912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,4,balanced,0.05475200215975443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,4,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,4,balanced,0.055733333031336464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,4,balanced,0.0727946658929189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,4,balanced,0.07091733316580455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,4,balanced,0.08714666962623596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,4,balanced,0.1018506685892741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,4,balanced,0.12874133388201395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,4,balanced,0.1516853372255961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,4,balanced,0.20614933967590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,4,balanced,0.25788267453511554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,4,balanced,0.3630133469899495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,4,balanced,0.4656533400217692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,4,power_law_1.01,0.0549888014793396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,4,power_law_1.01,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,4,power_law_1.01,0.10426239967346192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,4,power_law_1.01,0.10625280141830444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,4,power_law_1.01,0.1326848030090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,4,power_law_1.01,0.13871359825134277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,4,power_law_1.01,0.15735039710998536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,4,power_law_1.01,0.17211519479751586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,4,power_law_1.01,0.1778048038482666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,4,power_law_1.01,0.16944639682769774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,4,power_law_1.01,0.17840640544891356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,4,power_law_1.01,0.18495359420776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,4,power_law_1.01,0.18799999952316285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,4,power_law_1.01,0.1934847950935364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,4,power_law_1.01,0.19827840328216553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,4,power_law_1.01,0.21189119815826415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,4,power_law_1.01,0.22529919147491456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,4,power_law_1.01,0.2396672010421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,4,power_law_1.01,0.26192638874053953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,4,power_law_1.01,0.30237441062927245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,4,power_law_1.01,0.34219520092010497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,4,power_law_1.01,0.4182079792022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,4,power_law_1.01,0.49724159240722654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,4,power_law_1.01,0.6656640052795411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,4,power_law_1.01,0.818239974975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,4,power_law_1.01,1.1183679580688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,4,power_law_1.01,1.3890239715576171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,64,power_law_1.2,0.09797760248184204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,64,power_law_1.2,0.09468160271644592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,64,power_law_1.2,0.09449599981307984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,64,power_law_1.2,0.09767680168151856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,64,power_law_1.2,0.09880319833755494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,64,power_law_1.2,0.09025920033454896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,64,power_law_1.2,0.09772160053253173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,64,power_law_1.2,0.09795839786529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,64,power_law_1.2,0.09970560073852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,64,power_law_1.2,0.09920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,64,power_law_1.2,0.09925119876861573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,64,power_law_1.2,0.10948480367660522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,64,power_law_1.2,0.10751999616622925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,64,power_law_1.2,0.11086080074310303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,64,power_law_1.2,0.11895040273666382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,64,power_law_1.2,0.11571199893951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,64,power_law_1.2,0.12547839879989625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,64,power_law_1.2,0.1373952031135559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,64,power_law_1.2,0.15282560586929322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,64,power_law_1.2,0.1890176057815552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,64,power_law_1.2,0.21285760402679443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,64,power_law_1.2,0.2785536050796509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,64,power_law_1.2,0.3251456022262573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,64,power_law_1.2,0.5094592094421386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,64,power_law_1.2,0.6082496166229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,64,power_law_1.2,0.9376192092895508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,64,power_law_1.2,1.202355194091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,1,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,1,balanced,0.05607999861240387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,1,balanced,0.0640533318122228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,1,balanced,0.08669867118199666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,1,balanced,0.12628799676895142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,1,balanced,0.18727999925613403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,1,balanced,0.1911146640777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,1,balanced,0.1922773321469625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,1,balanced,0.19556800524393717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,1,balanced,0.19869333505630493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,1,balanced,0.20005333423614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,1,balanced,0.20893865823745728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,1,balanced,0.21331733465194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,1,balanced,0.21758399407068887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,1,balanced,0.22866666316986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,1,balanced,0.23601067066192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,1,balanced,0.25360532601674396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,1,balanced,0.2983520030975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,1,balanced,0.3347040017445882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,1,balanced,0.40990932782491046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,1,balanced,0.47337067127227783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,1,balanced,0.6985387007395426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,1,balanced,0.8048426310221354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,1,balanced,1.210922638575236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,1,balanced,1.5099786122639973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,1,balanced,2.216261386871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,1,balanced,2.891717274983724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,64,power_law_1.01,0.030899199843406677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,64,power_law_1.01,0.030214399099349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,64,power_law_1.01,0.02991360127925873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,64,power_law_1.01,0.032601600885391234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,64,power_law_1.01,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,64,power_law_1.01,0.03473280072212219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,64,power_law_1.01,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,64,power_law_1.01,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,64,power_law_1.01,0.0367935985326767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,64,power_law_1.01,0.03699199855327606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,64,power_law_1.01,0.03741439878940582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,64,power_law_1.01,0.038975998759269714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,64,power_law_1.01,0.03976959884166718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,64,power_law_1.01,0.04439040124416351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,64,power_law_1.01,0.050393599271774295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,64,power_law_1.01,0.04899199903011322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,64,power_law_1.01,0.05343359708786011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,64,power_law_1.01,0.05911039710044861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,64,power_law_1.01,0.06788480281829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,64,power_law_1.01,0.09009280204772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,64,power_law_1.01,0.10497280359268188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,64,power_law_1.01,0.14618879556655884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,64,power_law_1.01,0.1699136018753052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,64,power_law_1.01,0.25800960063934325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,64,power_law_1.01,0.31658239364624025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,64,power_law_1.01,0.5499519824981689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,64,power_law_1.01,0.8264191627502442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,16,power_law_1.01,0.04534400105476379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,16,power_law_1.01,0.04268159866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,16,power_law_1.01,0.04004479944705963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,16,power_law_1.01,0.043270400166511534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,16,power_law_1.01,0.04237439930438995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,16,power_law_1.01,0.044896000623703004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,16,power_law_1.01,0.04523519873619079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,16,power_law_1.01,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,16,power_law_1.01,0.04552960097789764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,16,power_law_1.01,0.046758401393890384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,16,power_law_1.01,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,16,power_law_1.01,0.0471807986497879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,16,power_law_1.01,0.04683519899845123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,16,power_law_1.01,0.049107199907302855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,16,power_law_1.01,0.052185600996017455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,16,power_law_1.01,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,16,power_law_1.01,0.05591679811477661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,16,power_law_1.01,0.062105602025985716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,16,power_law_1.01,0.06937599778175355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,16,power_law_1.01,0.07754240036010743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,16,power_law_1.01,0.08760960102081299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,16,power_law_1.01,0.11944960355758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,16,power_law_1.01,0.13976320028305053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,16,power_law_1.01,0.2052608013153076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,16,power_law_1.01,0.23192319869995118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,16,power_law_1.01,0.3630336046218872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,16,power_law_1.01,0.42115201950073244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,4,power_law_1.2,0.04904319941997528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,4,power_law_1.2,0.053574401140213015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,4,power_law_1.2,0.0607807993888855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,4,power_law_1.2,0.06780160069465638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,4,power_law_1.2,0.0803712010383606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,4,power_law_1.2,0.09313920140266418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,4,power_law_1.2,0.09374079704284669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,4,power_law_1.2,0.09552000164985656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,4,power_law_1.2,0.09782400131225585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,4,power_law_1.2,0.0998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,4,power_law_1.2,0.09986559748649597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,4,power_law_1.2,0.09670400023460388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,4,power_law_1.2,0.10787839889526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,4,power_law_1.2,0.1114240050315857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,4,power_law_1.2,0.12061439752578736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,4,power_law_1.2,0.12163840532302857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,4,power_law_1.2,0.1298815965652466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,4,power_law_1.2,0.14944640398025513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.2,0.15907200574874877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.2,0.20236799716949463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.2,0.21835520267486572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.2,0.28965120315551757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.2,0.363916802406311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.2,0.4916672229766846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.2,0.6363840103149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.2,1.0807168006896972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.2,1.1108351707458497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,8,balanced,0.049738665421803795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,8,balanced,0.05303466816743215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,8,balanced,0.07418133318424225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,8,balanced,0.09738666812578838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,8,balanced,0.14923733472824097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,8,balanced,0.248581329981486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,8,balanced,0.2485599915186564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,8,balanced,0.25135467449824017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,8,balanced,0.2516000072161357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,8,balanced,0.2518933415412903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,8,balanced,0.25778132677078247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,8,balanced,0.2595626711845398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,8,balanced,0.261135995388031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,8,balanced,0.2614026665687561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,8,balanced,0.26952000459035236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,8,balanced,0.27138133843739826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,8,balanced,0.2765973409016927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,8,balanced,0.2935360074043274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,8,balanced,0.3070613344510396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,8,balanced,0.3337706724802653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,8,balanced,0.36086400349934894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,8,balanced,0.4054453372955322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,8,balanced,0.45402665932973224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,8,balanced,0.5919040044148763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,8,balanced,0.6878933111826578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,8,balanced,0.9320212999979655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,8,balanced,1.133461316426595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,8,power_law_1.2,0.033062401413917544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,8,power_law_1.2,0.031711998581886294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,8,power_law_1.2,0.030931198596954347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,8,power_law_1.2,0.03240320086479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,8,power_law_1.2,0.03317759931087494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,8,power_law_1.2,0.038540801405906676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,8,power_law_1.2,0.06755840182304382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,8,power_law_1.2,0.06945279836654664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,8,power_law_1.2,0.07272959947586059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,8,power_law_1.2,0.06789119839668274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,8,power_law_1.2,0.05978879928588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,8,power_law_1.2,0.09612159729003907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,8,power_law_1.2,0.08950399756431579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,8,power_law_1.2,0.07343360185623168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,8,power_law_1.2,0.10599039793014527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,8,power_law_1.2,0.08384640216827392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,8,power_law_1.2,0.09772160053253173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,8,power_law_1.2,0.13939199447631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.2,0.1624511957168579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.2,0.22832000255584717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.2,0.2924864053726196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,64,balanced,0.05795200169086456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,64,balanced,0.05830933153629303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,64,balanced,0.07107733190059662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,64,balanced,0.09507733583450317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,64,balanced,0.10943999886512756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,64,balanced,0.10512000322341919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,64,balanced,0.10898133118947347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,64,balanced,0.10661866267522176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,64,balanced,0.10803733269373576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,64,balanced,0.1102186640103658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,64,balanced,0.10731732845306396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,64,balanced,0.11238400141398112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,64,balanced,0.1123253305753072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,64,balanced,0.1127306620279948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,64,balanced,0.11746666828791301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,64,balanced,0.11873066425323486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,64,balanced,0.12217600146929423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,64,balanced,0.1402186652024587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,64,balanced,0.13804800311724344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,64,balanced,0.1706719994544983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,64,balanced,0.17697600523630777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,64,balanced,0.23651200532913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,64,balanced,0.21485867102940878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,64,balanced,0.2903520067532857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,64,balanced,0.29772265752156574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,64,balanced,0.4041653474171956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,64,balanced,0.4615039825439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.2,0.41860480308532716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.2,0.5283199787139893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.2,0.7838592052459716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.2,1.019596767425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,32,power_law_1.01,0.05588480234146118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,32,power_law_1.01,0.0697983980178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,32,power_law_1.01,0.06977919936180114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,32,power_law_1.01,0.06759679913520814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,32,power_law_1.01,0.06828160285949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,32,power_law_1.01,0.06474879980087281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,32,power_law_1.01,0.06225280165672302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,32,power_law_1.01,0.0668287992477417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,32,power_law_1.01,0.06892799735069274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,32,power_law_1.01,0.06971520185470581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,32,power_law_1.01,0.07085440158843995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,32,power_law_1.01,0.07359359860420227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,32,power_law_1.01,0.07208319902420043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,32,power_law_1.01,0.0778432011604309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,32,power_law_1.01,0.08427519798278808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,32,power_law_1.01,0.08688639998435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,32,power_law_1.01,0.09744639992713929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,32,power_law_1.01,0.11698559522628785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,32,power_law_1.01,0.11727360486984253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,64,power_law_1.2,0.037196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,64,power_law_1.2,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,64,power_law_1.2,0.033606401085853575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,64,power_law_1.2,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,64,power_law_1.2,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,64,power_law_1.2,0.035436800122261046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,64,power_law_1.2,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,64,power_law_1.2,0.03603839874267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,32,power_law_1.01,0.14925440549850463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,64,power_law_1.2,0.03619199991226196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,32,power_law_1.01,0.17295360565185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,64,power_law_1.2,0.03643519878387451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,32,power_law_1.01,0.22403841018676757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,64,power_law_1.2,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,64,power_law_1.2,0.039084801077842714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,64,power_law_1.2,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,64,power_law_1.2,0.03991680145263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,64,power_law_1.2,0.043584001064300534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,64,power_law_1.2,0.044486400485038755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,64,power_law_1.2,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,64,power_law_1.2,0.05141119956970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,32,power_law_1.01,0.27294719219207764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.2,0.055961602926254274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,32,power_law_1.01,0.36967039108276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,32,power_law_1.01,0.46625919342041017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,32,power_law_1.01,0.6667327880859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,32,power_law_1.01,0.9022208213806152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.2,0.06424959897994995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.2,0.07660800218582153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.2,0.1
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.2,1.5110336303710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.2,0.11943680047988892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.2,0.16272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.2,0.21194241046905518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.2,0.3063040018081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.2,0.36394879817962644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.2,1.9986303329467774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,16,balanced,0.038586666186650596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,16,balanced,0.03826666623353958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,16,balanced,0.03972266614437103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,16,balanced,0.04324266811211904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,16,balanced,0.043882668018341064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,16,balanced,0.047930667797724404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,16,balanced,0.04794666667779287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,16,balanced,0.050106664498647056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,16,balanced,0.049914668003718056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,16,balanced,0.050474668542544045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,16,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,16,balanced,0.052144000927607216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,16,balanced,0.05425066749254862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,16,balanced,0.05410666763782501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,16,balanced,0.059578667084376015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,16,balanced,0.05922666688760122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,16,balanced,0.062405332922935486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,16,balanced,0.06855466465155284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,16,balanced,0.07251733541488647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,16,balanced,0.08316799998283386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,16,balanced,0.09314133723576863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,16,balanced,0.11756267150243123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,16,balanced,0.14075199762980142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,16,balanced,0.21176532904307047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,16,balanced,0.2562026580174764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,16,balanced,0.35806934038798016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,16,balanced,0.462501327196757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,4,balanced,0.05221333106358846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,4,balanced,0.05421866476535797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,4,balanced,0.06016000111897787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,4,balanced,0.07374933362007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,4,balanced,0.10060266653696696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,4,balanced,0.14378666877746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,4,balanced,0.14338666200637817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,4,balanced,0.143994669119517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,4,balanced,0.14570132891337076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,4,balanced,0.14682132999102274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,4,balanced,0.14893333117167154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,4,balanced,0.14991999665896097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,4,balanced,0.1530453364054362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,4,balanced,0.15556800365447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,4,balanced,0.1625866691271464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,4,balanced,0.16269866625467935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,4,balanced,0.16900799671808878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,4,balanced,0.17981332540512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,4,balanced,0.18948266903559366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,4,balanced,0.2132800022761027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,4,balanced,0.2355840007464091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,4,balanced,0.28189865748087567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,4,balanced,0.3330986698468526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,4,balanced,0.45160531997680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,4,balanced,0.5422293345133463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,4,balanced,0.760591983795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,4,balanced,0.9659626483917236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,64,balanced,0.023599999646345775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,64,balanced,0.023157333334287006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,64,balanced,0.02317333221435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,64,balanced,0.02312533309062322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,64,balanced,0.025072000920772552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,64,balanced,0.02956266701221466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,64,balanced,0.05009600023428599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,64,balanced,0.0517493337392807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,64,balanced,0.050767997900644936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,64,balanced,0.051183998584747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,64,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,64,balanced,0.048122664292653404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,64,balanced,0.04565866788228353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,64,balanced,0.04161066561937332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,64,balanced,0.05820266902446747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,64,balanced,0.05770133435726166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,64,balanced,0.04748799900213877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,64,balanced,0.06676800052324931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,64,balanced,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,64,balanced,0.09870400031407674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,64,balanced,0.12016000350316365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,64,balanced,0.1669600009918213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,64,balanced,0.20501333475112915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,64,balanced,0.29626667499542236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,64,balanced,0.3808906475702922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,64,balanced,0.5576479832331339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,64,balanced,0.7474666436513265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,4,power_law_1.01,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,4,power_law_1.01,0.030956798791885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,4,power_law_1.01,0.03410559892654419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,4,power_law_1.01,0.034067198634147644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,4,power_law_1.01,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,4,power_law_1.01,0.038278400897979736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,4,power_law_1.01,0.03945600092411041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,4,power_law_1.01,0.04014720022678375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,4,power_law_1.01,0.04074879884719849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,4,power_law_1.01,0.042387199401855466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,4,power_law_1.01,0.04362240135669708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,4,power_law_1.01,0.047200000286102294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,4,power_law_1.01,0.04907520115375519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,4,power_law_1.01,0.05071359872817993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,4,power_law_1.01,0.05625600218772888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,4,power_law_1.01,0.05925120115280151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,4,power_law_1.01,0.06344959735870362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,4,power_law_1.01,0.07405440211296081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,4,power_law_1.01,0.0898751974105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,4,power_law_1.01,0.10408960580825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,4,power_law_1.01,0.11423360109329224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,4,power_law_1.01,0.14590079784393312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,4,power_law_1.01,0.20054399967193604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,4,power_law_1.01,0.2368959903717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,4,power_law_1.01,0.32069759368896483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,4,power_law_1.01,0.4845119953155518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,4,power_law_1.01,0.640940809249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,16,power_law_1.2,0.020608000457286835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,16,power_law_1.2,0.020006400346755982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,16,power_law_1.2,0.01850239932537079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,16,power_law_1.2,0.018764799833297728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,16,power_law_1.2,0.019014400243759156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,16,power_law_1.2,0.019731199741363524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,16,power_law_1.2,0.020185600221157073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,16,power_law_1.2,0.02051199972629547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,16,power_law_1.2,0.020608000457286835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,16,power_law_1.2,0.021414400637149812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,16,power_law_1.2,0.023801599442958832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,16,power_law_1.2,0.02459519952535629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,16,power_law_1.2,0.026137599349021913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,16,power_law_1.2,0.0359360009431839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,16,power_law_1.2,0.03687039911746979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,16,power_law_1.2,0.037376001477241516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,16,power_law_1.2,0.0392767995595932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,16,power_law_1.2,0.042099198698997496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,16,power_law_1.2,0.0445248007774353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,16,power_law_1.2,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,16,power_law_1.2,0.05901439785957337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,16,power_law_1.2,0.07856000065803528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,16,power_law_1.2,0.09418879747390747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,2,power_law_1.01,0.05013120174407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,2,power_law_1.01,0.05580800175666809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,2,power_law_1.01,0.06312959790229797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,2,power_law_1.01,0.07608320116996765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,2,power_law_1.01,0.07986559867858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,2,power_law_1.01,0.1027519941329956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,2,power_law_1.01,0.10295039415359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,2,power_law_1.01,0.10375679731369018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,2,power_law_1.01,0.10575360059738159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,2,power_law_1.01,0.10646400451660157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,2,power_law_1.01,0.10897279977798462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,2,power_law_1.01,0.11057920455932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,2,power_law_1.01,0.1126207947731018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,2,power_law_1.01,0.1146239995956421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,2,power_law_1.01,0.12295680046081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,16,power_law_1.2,0.12370560169219971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,2,power_law_1.01,0.12275840044021606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,2,power_law_1.01,0.1316480040550232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,2,power_law_1.01,0.14780800342559813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,2,power_law_1.01,0.1605183959007263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,2,power_law_1.01,0.19408639669418334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,2,power_law_1.01,0.22823679447174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,2,power_law_1.01,0.28935039043426514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,2,power_law_1.01,0.37888638973236083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,2,power_law_1.01,0.4867839813232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,2,power_law_1.01,0.6008384227752686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,2,power_law_1.01,0.8853631973266601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,2,power_law_1.01,1.1580927848815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,16,power_law_1.2,0.15456639528274535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,16,power_law_1.2,0.22838399410247803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,32,power_law_1.01,0.0335999995470047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,32,power_law_1.01,0.029177600145339967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,32,power_law_1.01,0.03185279965400696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,32,power_law_1.01,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,32,power_law_1.01,0.029574400186538695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,32,power_law_1.01,0.02996479868888855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,32,power_law_1.01,0.030777600407600404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,32,power_law_1.01,0.03020159900188446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,32,power_law_1.01,0.030828800797462464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,32,power_law_1.01,0.030924800038337707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,32,power_law_1.01,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,32,power_law_1.01,0.03146879971027374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,32,power_law_1.01,0.03158400058746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,32,power_law_1.01,0.03392640054225922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,32,power_law_1.01,0.037036800384521486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,32,power_law_1.01,0.037241598963737486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,32,power_law_1.01,0.0382207989692688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,32,power_law_1.01,0.043884798884391785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,32,power_law_1.01,0.04302079975605011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,32,power_law_1.01,0.05121920108795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,32,power_law_1.01,0.05089920163154602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,32,power_law_1.01,0.0621504008769989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,32,power_law_1.01,0.07678719758987426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,32,power_law_1.01,0.08777599930763244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,32,power_law_1.01,0.10206079483032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,32,power_law_1.01,0.13186559677124024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,32,power_law_1.01,0.17025920152664184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,16,power_law_1.2,0.2780159950256348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,64,power_law_1.2,0.02040960043668747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,64,power_law_1.2,0.019833600521087645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,64,power_law_1.2,0.019948799908161164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,64,power_law_1.2,0.01991039961576462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,64,power_law_1.2,0.021804800629615782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,64,power_law_1.2,0.026092800498008727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,64,power_law_1.2,0.04772480130195618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,64,power_law_1.2,0.04928640127182007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,64,power_law_1.2,0.040454399585723874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,64,power_law_1.2,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,64,power_law_1.2,0.04217599928379059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,64,power_law_1.2,0.04286719858646393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,64,power_law_1.2,0.0425024002790451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,64,power_law_1.2,0.042412799596786496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,64,power_law_1.2,0.041875201463699344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,64,power_law_1.2,0.04142720103263855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,64,power_law_1.2,0.044512000679969785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,64,power_law_1.2,0.046777600049972536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,64,power_law_1.2,0.05130239725112915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,64,power_law_1.2,0.05911039710044861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,64,power_law_1.2,0.07337599992752075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,64,power_law_1.2,0.10400639772415161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,64,power_law_1.2,0.12154239416122437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,64,power_law_1.2,0.1637887954711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,64,power_law_1.2,0.2119488000869751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,64,power_law_1.2,0.31330559253692625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,64,power_law_1.2,0.41281919479370116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,2,power_law_1.2,0.04792959988117218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,2,power_law_1.2,0.056678402423858645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,2,power_law_1.2,0.07074559926986694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,2,power_law_1.2,0.08954240083694458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,2,power_law_1.2,0.09820160269737244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,2,power_law_1.2,0.14020479917526246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,2,power_law_1.2,0.14609919786453246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,2,power_law_1.2,0.14837759733200073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,2,power_law_1.2,0.1566655993461609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,2,power_law_1.2,0.15543680191040038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,2,power_law_1.2,0.1602687954902649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,2,power_law_1.2,0.16636799573898314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,2,power_law_1.2,0.16881920099258424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,2,power_law_1.2,0.1715775966644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,2,power_law_1.2,0.18217600584030152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,2,power_law_1.2,0.19210879802703856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,2,power_law_1.2,0.20444159507751464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,2,power_law_1.2,0.23676800727844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.2,0.2604991912841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.2,0.3083904027938843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.2,0.3882047891616821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.2,0.49680638313293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.2,0.5542272090911865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.2,0.7604095935821533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,32,2,power_law_1.01,0.021529600024223328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,32,2,power_law_1.01,0.022015999257564544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,32,2,power_law_1.01,0.024371199309825897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,32,2,power_law_1.01,0.029318401217460634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,32,2,power_law_1.01,0.03782399892807007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,32,2,power_law_1.01,0.04067200124263763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,32,2,power_law_1.01,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,32,2,power_law_1.01,0.048467200994491574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,32,2,power_law_1.01,0.05103359818458557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,32,2,power_law_1.01,0.05455999970436096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,32,2,power_law_1.01,0.06198400259017944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,32,2,power_law_1.01,0.06481919884681701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,32,2,power_law_1.01,0.06766080260276794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,32,2,power_law_1.01,0.07593600153923034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,32,2,power_law_1.01,0.07157760262489318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,32,2,power_law_1.01,0.07593600153923034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,32,2,power_law_1.01,0.08168960213661194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,32,2,power_law_1.01,0.10373760461807251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,32,2,power_law_1.01,0.11449600458145141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,32,2,power_law_1.01,0.15676800012588502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,32,2,power_law_1.01,0.19402240514755248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,32,2,power_law_1.01,0.21774721145629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,32,2,power_law_1.01,0.24726400375366211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,32,2,power_law_1.01,0.32096641063690184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,32,2,power_law_1.01,0.3778559923171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,32,2,power_law_1.01,0.5236927986145019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,32,2,power_law_1.01,0.6502016067504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.2,1.0291071891784669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,2,power_law_1.01,0.08279039859771728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,2,power_law_1.01,0.12933759689331054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,2,power_law_1.01,0.11982079744338989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,2,power_law_1.01,0.17262719869613646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,2,power_law_1.01,0.17377279996871947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,2,power_law_1.01,0.17839360237121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,2,power_law_1.01,0.17223039865493775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,2,power_law_1.01,0.17469439506530762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,2,power_law_1.01,0.16856319904327394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,2,power_law_1.01,0.165830397605896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,2,power_law_1.01,0.16010240316390992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,2,power_law_1.01,0.18356479406356813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,2,power_law_1.01,0.19379839897155762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,2,power_law_1.01,0.1753983974456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,2,power_law_1.01,0.1852671980857849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,2,power_law_1.01,0.20456960201263427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,2,power_law_1.01,0.2125312089920044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,2,power_law_1.01,0.26053121089935305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.01,0.2923648118972778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.01,0.36426239013671874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.01,0.44510722160339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.01,0.6381887912750244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.2,1.3229056358337403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.01,0.8297727584838868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.01,1.2277824401855468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,64,balanced,0.05195199946562449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,64,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,64,balanced,0.052239999175071716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,64,balanced,0.05506666501363119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,64,balanced,0.05625600119431814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,64,balanced,0.0705973356962204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,64,balanced,0.07817066709200542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,64,balanced,0.0783786674340566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,64,balanced,0.07977066437403361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,64,balanced,0.07869333525498708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,64,balanced,0.090037335952123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,64,balanced,0.09318932890892029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,64,balanced,0.09292266766230266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,64,balanced,0.08899733424186707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,64,balanced,0.10497599840164185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,64,balanced,0.10505066315333049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,64,balanced,0.10947733124097188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,64,balanced,0.11829866965611775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,64,balanced,0.1269439955552419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,64,balanced,0.14217600226402283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,64,balanced,0.15847466389338175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,64,balanced,0.198527991771698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,64,balanced,0.2299519975980123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,64,balanced,0.29369600613911945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,64,balanced,0.3592533270517985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.01,1.5374015808105468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,64,balanced,0.48446400960286456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,64,balanced,0.6312906742095947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.01,2.023520088195801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.2,1.8860671997070313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.01,2.9844736099243163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,64,power_law_1.2,0.026150399446487428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,64,power_law_1.2,0.025721600651741026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,64,power_law_1.2,0.024799999594688416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,64,power_law_1.2,0.02683520019054413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,64,power_law_1.2,0.030707201361656188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,64,power_law_1.2,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,64,power_law_1.2,0.057068800926208495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,64,power_law_1.2,0.05914880037307739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,64,power_law_1.2,0.07090560197830201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,64,power_law_1.2,0.07146880030632019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,64,power_law_1.2,0.06888319849967957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,64,power_law_1.2,0.06829439997673034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,64,power_law_1.2,0.06593279838562012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,64,power_law_1.2,0.06558079719543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,64,power_law_1.2,0.06044160127639771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,64,power_law_1.2,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,64,power_law_1.2,0.06570879817008972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,64,power_law_1.2,0.07019519805908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,64,power_law_1.2,0.07541760206222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,64,power_law_1.2,0.09530240297317505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,64,power_law_1.2,0.11357439756393432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,64,power_law_1.2,0.15895040035247804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,64,power_law_1.2,0.1811840057373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,64,power_law_1.2,0.27076480388641355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,64,power_law_1.2,0.34513919353485106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,64,power_law_1.2,0.5461120128631591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,64,power_law_1.2,0.7273983955383301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,16,power_law_1.01,0.0985535979270935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,16,power_law_1.01,0.14641280174255372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,16,power_law_1.01,0.14698879718780516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,16,power_law_1.01,0.1437824010848999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,16,power_law_1.01,0.14307839870452882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,16,power_law_1.01,0.12995200157165526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,16,power_law_1.01,0.14349440336227418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,16,power_law_1.01,0.14244480133056642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,16,power_law_1.01,0.1432703971862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,16,power_law_1.01,0.14302719831466676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,16,power_law_1.01,0.14309120178222656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,16,power_law_1.01,0.14312319755554198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,16,power_law_1.01,0.14929920434951782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,16,power_law_1.01,0.14897279739379882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,16,power_law_1.01,0.15642240047454833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,16,power_law_1.01,0.15882879495620728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,16,power_law_1.01,0.16609280109405516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,16,power_law_1.01,0.18240640163421631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,16,power_law_1.01,0.18405120372772216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,16,power_law_1.01,0.22557439804077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,16,power_law_1.01,0.23968000411987306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,16,power_law_1.01,0.2947968006134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,16,power_law_1.01,0.3444159984588623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,16,power_law_1.01,0.43558402061462403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,16,power_law_1.01,0.5376832008361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,16,power_law_1.01,0.731712007522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,16,power_law_1.01,0.9103679656982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,64,power_law_1.2,0.06193280220031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,64,power_law_1.2,0.059961599111557004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,64,power_law_1.2,0.05886719822883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,64,power_law_1.2,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,64,power_law_1.2,0.059487998485565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,64,power_law_1.2,0.0640447974205017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,64,power_law_1.2,0.06440320014953613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,64,power_law_1.2,0.06472960114479065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,64,power_law_1.2,0.06424319744110107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,64,power_law_1.2,0.06538879871368408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,64,power_law_1.2,0.06731520295143127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,64,power_law_1.2,0.06764159798622131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,64,power_law_1.2,0.06856319904327393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,64,power_law_1.2,0.07140480279922486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,64,power_law_1.2,0.07441279888153077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,64,power_law_1.2,0.07530239820480347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,64,power_law_1.2,0.0845632016658783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,4,power_law_1.2,0.14918400049209596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,64,power_law_1.2,0.09710080027580262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,64,power_law_1.2,0.1087488055229187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,4,power_law_1.2,0.23749759197235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,64,power_law_1.2,0.1368127942085266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,4,power_law_1.2,0.2788671970367432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,64,power_law_1.2,0.1591231942176819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,4,power_law_1.2,0.373523211479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,64,power_law_1.2,0.20140159130096436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,64,power_law_1.2,0.2577791929244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,64,power_law_1.2,0.3680959939956665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,64,power_law_1.2,0.4488704204559326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,64,power_law_1.2,0.7005760192871093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,4,power_law_1.2,0.4643199920654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,64,power_law_1.2,0.8911423683166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,4,power_law_1.2,0.539411211013794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,4,power_law_1.2,0.6871808052062989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,4,power_law_1.2,0.6574079990386963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,4,power_law_1.2,0.7084159851074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,4,power_law_1.2,0.7274176120758057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,4,power_law_1.2,0.7178688049316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,4,power_law_1.2,0.7281087875366211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,4,power_law_1.2,0.7216639995574952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,4,power_law_1.2,0.7617919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,4,power_law_1.2,0.7858431816101075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,4,power_law_1.2,0.8195455551147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,4,power_law_1.2,0.867244815826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,4,power_law_1.2,0.9957695960998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,4,power_law_1.2,1.0032320022583008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,4,power_law_1.2,1.1821696281433105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,4,power_law_1.2,1.115065574645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,4,power_law_1.2,1.3270976066589355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,4,power_law_1.2,1.4057663917541503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,4,power_law_1.2,1.7328447341918944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,4,power_law_1.2,2.109049606323242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,4,power_law_1.2,2.7510847091674804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,4,power_law_1.2,3.4604351043701174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,2,balanced,0.07957866787910461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,2,balanced,0.0937439997990926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,2,balanced,0.13649066289265951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,2,balanced,0.14010133345921835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,2,balanced,0.14266133308410645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,2,balanced,0.1418453355630239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,2,balanced,0.1422826647758484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,2,balanced,0.14428800344467163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,2,balanced,0.14387200276056925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,2,balanced,0.1439893345038096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,2,balanced,0.14386133352915445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,2,balanced,0.14666133125623068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,2,balanced,0.14632532993952432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,2,balanced,0.14839999874432883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,2,balanced,0.15593066811561584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,2,balanced,0.15967466433842978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,2,balanced,0.1643786629041036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,2,balanced,0.21051200230916342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,2,balanced,0.21827733516693115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,2,balanced,0.288975993792216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,2,balanced,0.3615413506825765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,2,balanced,0.49054932594299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,2,balanced,0.622437318166097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,2,balanced,0.8974453608194987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,2,balanced,1.1645867029825847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,2,balanced,1.7448800404866536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,2,balanced,2.277781327565511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,1,power_law_1.01,0.06094080209732056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,1,power_law_1.01,0.08079360127449035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,1,power_law_1.01,0.10907520055770874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,1,power_law_1.01,0.15663360357284545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,1,power_law_1.01,0.24131200313568116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,1,power_law_1.01,0.32946560382843015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,1,power_law_1.01,0.3806720018386841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,1,power_law_1.01,0.47637119293212893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,1,power_law_1.01,0.48447360992431643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,1,power_law_1.01,0.5078656196594238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,1,power_law_1.01,0.5393087863922119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,1,power_law_1.01,0.5500671863555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,1,power_law_1.01,0.5701824188232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,1,power_law_1.01,0.5869696140289307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,1,power_law_1.01,0.6012991905212403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,1,power_law_1.01,0.6273600101470947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,1,power_law_1.01,0.6541312217712403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,1,power_law_1.01,0.7234879970550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,1,power_law_1.01,0.8000831604003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,1,power_law_1.01,0.9458304405212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,1,power_law_1.01,1.0233535766601562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,1,power_law_1.01,1.259059238433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,1,power_law_1.01,1.5116928100585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,1,power_law_1.01,1.9680959701538085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,1,power_law_1.01,2.3887487411499024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,1,power_law_1.01,3.1974720001220702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,1,power_law_1.01,4.142668914794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,16,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,16,balanced,0.05425066749254862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,16,balanced,0.05784533421198527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,16,balanced,0.06261333326498668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,16,balanced,0.06471466521422069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,16,balanced,0.07642666498819987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,16,balanced,0.07867733140786488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,16,balanced,0.08041599889596303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,16,balanced,0.07904533545176189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,16,balanced,0.08140799899895985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,16,balanced,0.07849066456158955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,16,balanced,0.08055466910203297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,16,balanced,0.08278400202592213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,16,balanced,0.08313600222269694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,16,balanced,0.08782399694124858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,16,balanced,0.08854933579762776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,16,balanced,0.09096533060073853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,16,balanced,0.09918933113416036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,16,balanced,0.10704533259073894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,16,balanced,0.12379733721415202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,16,balanced,0.14085333546002707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,16,balanced,0.1723733345667521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,16,balanced,0.20075732469558716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,16,balanced,0.2715199987093608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,16,balanced,0.332586665948232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,16,balanced,0.4624853531519572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,16,balanced,0.5920586585998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,8,power_law_1.01,0.07939199805259704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,8,power_law_1.01,0.08131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,8,power_law_1.01,0.08031359910964966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,8,power_law_1.01,0.07939199805259704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,8,power_law_1.01,0.08256000280380249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,8,power_law_1.01,0.0800704002380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,8,power_law_1.01,0.08429440259933471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,8,power_law_1.01,0.09182080030441284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,8,power_law_1.01,0.09316480159759521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,8,power_law_1.01,0.09813759922981262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,8,power_law_1.01,0.09470720291137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,8,power_law_1.01,0.10136959552764893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,8,power_law_1.01,0.10042239427566528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,8,power_law_1.01,0.10980479717254639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,8,power_law_1.01,0.1311360001564026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,8,power_law_1.01,0.1375808000564575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,8,power_law_1.01,0.16405119895935058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,8,power_law_1.01,0.2094719886779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,8,power_law_1.01,0.2414400100708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,8,power_law_1.01,0.3431616067886353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,8,power_law_1.01,0.391430401802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,8,power_law_1.01,0.5355711936950683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,8,power_law_1.01,0.790169620513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,16,power_law_1.2,0.042335999011993405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,16,power_law_1.2,0.038675200939178464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,16,power_law_1.2,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,16,power_law_1.2,0.036857599020004274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,16,power_law_1.2,0.037990400195121767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,16,power_law_1.2,0.03983359932899475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,16,power_law_1.2,0.040601599216461184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,16,power_law_1.2,0.04059520065784454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,16,power_law_1.2,0.04076800048351288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,16,power_law_1.2,0.04108799993991852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,16,power_law_1.2,0.04286080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,16,power_law_1.2,0.04416640102863312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,16,power_law_1.2,0.044889599084854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,16,power_law_1.2,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,16,power_law_1.2,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,16,power_law_1.2,0.057920002937316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,16,power_law_1.2,0.058316802978515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,16,power_law_1.2,0.06646400094032287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.2,0.07445120215415954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.2,0.08467199802398681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.2,0.10067199468612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.2,0.14184319972991943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,8,power_law_1.01,1.2209983825683595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.2,0.18469120264053346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.2,0.22618238925933837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.2,0.2849152088165283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,8,power_law_1.01,1.2923775672912599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,8,power_law_1.01,2.266092872619629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,8,power_law_1.01,2.6206335067749023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,256,power_law_1.01,0.053913599252700804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,256,power_law_1.01,0.06812160015106201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,256,power_law_1.01,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,256,power_law_1.01,0.042368000745773314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,256,power_law_1.01,0.042534399032592776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,256,power_law_1.01,0.04288640022277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,256,power_law_1.01,0.04348160028457641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,256,power_law_1.01,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,256,power_law_1.01,0.04524160027503967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,256,power_law_1.01,0.044761601090431216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,256,power_law_1.01,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,256,power_law_1.01,0.04640640020370483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,256,power_law_1.01,0.04831359982490539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,256,power_law_1.01,0.04906240105628967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,256,power_law_1.01,0.05306239724159241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,256,power_law_1.01,0.05491200089454651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,256,power_law_1.01,0.058873599767684935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,256,power_law_1.01,0.0652288019657135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,256,power_law_1.01,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,256,power_law_1.01,0.08249599933624267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,256,power_law_1.01,0.0921280026435852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,256,power_law_1.01,0.12090879678726196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,256,power_law_1.01,0.14740480184555055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,256,power_law_1.01,0.20341119766235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,256,power_law_1.01,0.2527616024017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,256,power_law_1.01,0.36031360626220704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,256,power_law_1.01,0.4434624195098877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.2,0.34625279903411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,1,power_law_1.2,0.09758080244064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,1,power_law_1.2,0.14146560430526733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,1,power_law_1.2,0.2185663938522339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,1,power_law_1.2,0.3652415990829468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,1,power_law_1.2,0.5092095851898193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,1,power_law_1.2,0.6656640052795411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,1,power_law_1.2,1.0160511970520019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,1,power_law_1.2,1.0513407707214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,1,power_law_1.2,1.0926655769348144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,1,power_law_1.2,1.1185152053833007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,1,power_law_1.2,1.187276840209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,1,power_law_1.2,1.2191103935241698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,1,power_law_1.2,1.205241584777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,1,power_law_1.2,1.2536191940307617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,1,power_law_1.2,1.2948415756225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,1,power_law_1.2,1.3179776191711425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,1,power_law_1.2,1.4384703636169434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,1,power_law_1.2,1.590771198272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,1,power_law_1.2,1.599135971069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,1,power_law_1.2,1.8246912002563476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,1,power_law_1.2,1.7704063415527345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,1,power_law_1.2,2.0244863510131834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,1,power_law_1.2,2.3441535949707033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,1,power_law_1.2,2.81343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,1,power_law_1.2,3.308287811279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,1,power_law_1.2,4.20503044128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,1,power_law_1.2,5.142841720581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.2,0.576857614517212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,8,power_law_1.01,0.09800959825515747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,8,power_law_1.01,0.14687999486923217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,8,power_law_1.01,0.24082560539245607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,8,power_law_1.01,0.2250688076019287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,8,power_law_1.01,0.31392641067504884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,8,power_law_1.01,0.29985918998718264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,8,power_law_1.01,0.3085376024246216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,8,power_law_1.01,0.3121279954910278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,8,power_law_1.01,0.30152320861816406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,8,power_law_1.01,0.28408958911895754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,8,power_law_1.01,0.27971200942993163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,8,power_law_1.01,0.2920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,8,power_law_1.01,0.32775681018829345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,8,power_law_1.01,0.29709439277648925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,8,power_law_1.01,0.30183680057525636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,8,power_law_1.01,0.3371328115463257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,8,power_law_1.01,0.34186880588531493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,8,power_law_1.01,0.37004799842834474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,8,power_law_1.01,0.39432320594787595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,8,power_law_1.01,0.4547264099121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,8,power_law_1.01,0.47135357856750487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,8,power_law_1.01,0.5663871765136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,8,power_law_1.01,0.645030403137207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,8,power_law_1.01,0.8116607666015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,64,power_law_1.2,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,8,power_law_1.01,0.8744447708129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,64,power_law_1.2,0.03697279989719391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,64,power_law_1.2,0.036652800440788266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,8,power_law_1.01,1.1291328430175782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,8,power_law_1.01,1.4503423690795898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,64,power_law_1.2,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,64,power_law_1.2,0.03874559998512268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,64,power_law_1.2,0.04115839898586273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,64,power_law_1.2,0.040531200170516965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,64,power_law_1.2,0.040531200170516965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,64,power_law_1.2,0.041222399473190306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,64,power_law_1.2,0.04080640077590943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,64,power_law_1.2,0.041536000370979306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,64,power_law_1.2,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,64,power_law_1.2,0.04204800128936768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,64,power_law_1.2,0.04268159866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,64,power_law_1.2,0.0473471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,64,power_law_1.2,0.0493120014667511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,64,power_law_1.2,0.05014399886131286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,64,power_law_1.2,0.055244797468185426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.2,0.062009602785110474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.2,0.07116159796714783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.2,0.08123520016670227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.2,0.10930559635162354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.2,0.13368959426879884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.2,0.1855039954185486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,16,balanced,0.044981335600217186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,16,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,16,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,16,balanced,0.03790933390458425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,16,balanced,0.03934400031963984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,16,balanced,0.04185600082079569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,16,balanced,0.04345066845417023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,16,balanced,0.043605332573254905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,16,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,16,balanced,0.04350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,16,balanced,0.041722665230433144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,16,balanced,0.043578664461771645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,16,balanced,0.043791999419530235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,16,balanced,0.044533332188924156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,16,balanced,0.047637333472569786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,16,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,16,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,16,balanced,0.05193600058555603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,16,balanced,0.05208533505598704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,16,balanced,0.06423999865849812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,16,balanced,0.06227200229962667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,16,balanced,0.08135466774304707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,16,balanced,0.08829866846402486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,16,balanced,0.1144586702187856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,16,balanced,0.12888532876968384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,16,balanced,0.1688693364461263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,16,balanced,0.20182400941848755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,4,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,4,balanced,0.05421333511670431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,4,balanced,0.05603733162085215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,4,balanced,0.05820266902446747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,4,balanced,0.06896533568700154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,4,balanced,0.09289066990216573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,4,balanced,0.09310932954152425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,4,balanced,0.09367466966311137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,4,balanced,0.09556800127029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,4,balanced,0.09732799728711446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,4,balanced,0.09700799981753032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,4,balanced,0.09910933176676433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,4,balanced,0.10117333134015401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,4,balanced,0.10382399956385295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,4,balanced,0.10894399881362915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,4,balanced,0.11105066537857056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,4,balanced,0.11541333794593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,4,balanced,0.12603200475374857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,4,balanced,0.13217066725095114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,4,balanced,0.1549493372440338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,4,balanced,0.1764906644821167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,4,balanced,0.21732266743977866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,4,balanced,0.25441600879033405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,4,balanced,0.3561066786448161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,4,balanced,0.44937066237131756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,4,balanced,0.6387413342793783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,4,balanced,0.8389653364817301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.2,0.21690239906311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.2,0.3250432014465332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.2,0.4769792079925537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,16,balanced,0.039936001102129616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,16,balanced,0.04062933226426443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,16,balanced,0.041493333876132965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,16,balanced,0.04553066690762838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,16,balanced,0.05996266504128774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,16,balanced,0.0757173349459966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,16,balanced,0.07816533247629802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,16,balanced,0.07732800145943959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,16,balanced,0.0783786674340566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,16,balanced,0.07801066835721333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,16,balanced,0.08051733175913493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,16,balanced,0.07966400186220805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,16,balanced,0.0806826651096344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,16,balanced,0.08135466774304707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,16,balanced,0.08726933598518372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,16,balanced,0.08650133013725281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,16,balanced,0.09298666318257649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,16,balanced,0.09717866778373718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,16,balanced,0.09984532992045085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,16,balanced,0.11105599999427795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,16,balanced,0.11944533387819926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,16,balanced,0.14193600416183472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,16,balanced,0.16166933377583823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,16,balanced,0.20376533269882202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,16,balanced,0.2398080031077067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,16,balanced,0.31861333052317303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,16,balanced,0.40347735087076825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,128,power_law_1.01,0.02908799946308136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,128,power_law_1.01,0.028889599442481994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,128,power_law_1.01,0.02901119887828827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,128,power_law_1.01,0.031436800956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,32,4,power_law_1.2,0.07263360023498536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,32,4,power_law_1.2,0.08097280263900757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,32,4,power_law_1.2,0.07964159846305847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,32,4,power_law_1.2,0.09804800152778625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,32,4,power_law_1.2,0.09959679841995239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,32,4,power_law_1.2,0.11608959436416626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,128,power_law_1.01,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,32,4,power_law_1.2,0.1226688027381897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,128,power_law_1.01,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,32,4,power_law_1.2,0.12347520589828491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,128,power_law_1.01,0.033888000249862674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,32,4,power_law_1.2,0.12347520589828491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,128,power_law_1.01,0.033907198905944826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,32,4,power_law_1.2,0.12700799703598023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,128,power_law_1.01,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,128,power_law_1.01,0.03505280017852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,128,power_law_1.01,0.03573760092258453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,128,power_law_1.01,0.03800320029258728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,128,power_law_1.01,0.03859840035438537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,128,power_law_1.01,0.04328320026397705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,128,power_law_1.01,0.0486272007226944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,32,4,power_law_1.2,0.1305408000946045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,32,4,power_law_1.2,0.13134080171585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,32,4,power_law_1.2,0.13351039886474608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,32,4,power_law_1.2,0.1344383955001831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,32,4,power_law_1.2,0.14283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,32,4,power_law_1.2,0.14180480241775512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,32,4,power_law_1.2,0.15072640180587768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,128,power_law_1.01,0.04655359983444214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,32,4,power_law_1.2,0.16914559602737428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,128,power_law_1.01,0.05214080214500427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,32,4,power_law_1.2,0.18465919494628907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,128,power_law_1.01,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,32,4,power_law_1.2,0.2143359899520874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,128,power_law_1.01,0.06723200082778931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,32,4,power_law_1.2,0.25358080863952637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,128,power_law_1.01,0.08398720026016235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,32,4,power_law_1.2,0.3155263900756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,128,power_law_1.01,0.09508479833602905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,32,4,power_law_1.2,0.3760576009750366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,128,power_law_1.01,0.13779200315475465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,32,4,power_law_1.2,0.5442304134368896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,32,4,power_law_1.2,0.6814271926879882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,32,4,power_law_1.2,1.0328767776489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,32,4,power_law_1.2,1.175590419769287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,128,power_law_1.01,0.16514559984207153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,128,power_law_1.01,0.2314687967300415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,128,power_law_1.01,0.3080575942993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,128,power_law_1.01,0.4701056003570557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,128,power_law_1.01,0.6194111824035644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,4,power_law_1.01,0.05085440278053284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,4,power_law_1.01,0.05370879769325256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,4,power_law_1.01,0.0679744005203247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,4,power_law_1.01,0.07772160172462464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,4,power_law_1.01,0.08305280208587647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,4,power_law_1.01,0.09267200231552124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,4,power_law_1.01,0.10537600517272949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,4,power_law_1.01,0.1050879955291748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,4,power_law_1.01,0.1071552038192749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,4,power_law_1.01,0.1081279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,4,power_law_1.01,0.11229439973831176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,4,power_law_1.01,0.11943039894104004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,4,power_law_1.01,0.11854079961776734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,4,power_law_1.01,0.13415039777755738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,4,power_law_1.01,0.14239360094070436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,4,power_law_1.01,0.1499008059501648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,4,power_law_1.01,0.15322879552841187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,4,power_law_1.01,0.1729920029640198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,4,power_law_1.01,0.19171199798583985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,4,power_law_1.01,0.23033599853515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,4,power_law_1.01,0.27674241065979005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,4,power_law_1.01,0.35189759731292725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,4,power_law_1.01,0.44722561836242675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,4,power_law_1.01,0.6101952075958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,4,power_law_1.01,0.7723584175109863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,4,power_law_1.01,1.0775296211242675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,4,power_law_1.01,1.4235967636108398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,4,power_law_1.2,0.05047680139541626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,4,power_law_1.2,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,4,power_law_1.2,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,2,power_law_1.2,0.03108479976654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,2,power_law_1.2,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,4,power_law_1.2,0.0669376015663147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,2,power_law_1.2,0.03178240060806274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,2,power_law_1.2,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,4,power_law_1.2,0.0812287986278534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,2,power_law_1.2,0.033024001121521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,2,power_law_1.2,0.03492479920387268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,4,power_law_1.2,0.09055359959602356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,2,power_law_1.2,0.03583360016345978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,2,power_law_1.2,0.03585279881954193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,4,power_law_1.2,0.09797760248184204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,2,power_law_1.2,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,2,power_law_1.2,0.03787519931793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,4,power_law_1.2,0.1062656044960022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,2,power_law_1.2,0.037920001149177554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,4,power_law_1.2,0.10498559474945068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,4,power_law_1.2,0.10528000593185424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,4,power_law_1.2,0.11657600402832032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,4,power_law_1.2,0.11519360542297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,4,power_law_1.2,0.11605759859085082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,2,power_law_1.2,0.028275200724601747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,4,power_law_1.2,0.12452479600906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,2,power_law_1.2,0.049004799127578734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,4,power_law_1.2,0.13939839601516724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,2,power_law_1.2,0.06272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,2,power_law_1.2,0.09319679737091065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,2,power_law_1.2,0.12615679502487182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,2,power_law_1.2,0.13814400434494017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,2,power_law_1.2,0.19405440092086793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,2,power_law_1.2,0.041075199842453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,2,power_law_1.2,0.20103681087493896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,2,power_law_1.2,0.04261760115623474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,2,power_law_1.2,0.20821120738983154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,4,power_law_1.2,0.14280320405960084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,2,power_law_1.2,0.0425024002790451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,2,power_law_1.2,0.21695361137390137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,4,power_law_1.2,0.14856959581375123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,2,power_law_1.2,0.22572801113128663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,4,power_law_1.2,0.17423360347747802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,2,power_law_1.2,0.21994879245758056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,2,power_law_1.2,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,2,power_law_1.2,0.23223679065704345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,2,power_law_1.2,0.24814081192016602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,2,power_law_1.2,0.051500797271728516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,2,power_law_1.2,0.2413503885269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,2,power_law_1.2,0.24908800125122071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,2,power_law_1.2,0.05752320289611816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,2,power_law_1.2,0.26197760105133056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,2,power_law_1.2,0.28755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,4,power_law_1.2,0.19397120475769042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.2,0.29653120040893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,4,power_law_1.2,0.24102399349212647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.2,0.6717631816864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.2,0.7559423923492432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.2,0.5019199848175049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,2,power_law_1.2,0.06913920044898987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.2,0.5801919937133789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.2,0.7263423919677734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,4,power_law_1.2,0.2796351909637451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,2,power_law_1.2,0.07310720086097718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.2,0.839027214050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,4,power_law_1.2,0.373523211479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,4,power_law_1.2,0.4584512233734131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.2,1.1034687995910644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,4,power_law_1.2,0.632806396484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.2,1.3565695762634278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,4,power_law_1.2,0.7822207927703857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,2,power_law_1.2,0.09485440254211426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,4,power_law_1.2,1.1735296249389648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,4,power_law_1.2,1.65546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,2,power_law_1.2,0.11527680158615113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,2,power_law_1.2,0.15009280443191528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,2,power_law_1.2,0.16965759992599488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,2,power_law_1.2,0.2560767889022827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,2,power_law_1.2,0.3254143953323364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,16,power_law_1.01,0.08287360072135926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,16,power_law_1.01,0.08424320220947265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,16,power_law_1.01,0.07580800056457519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,16,power_law_1.01,0.08262400031089782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,16,power_law_1.01,0.07810559868812561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,16,power_law_1.01,0.08217599987983704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,16,power_law_1.01,0.07995520234107971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,16,power_law_1.01,0.08107519745826722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,16,power_law_1.01,0.08149759769439698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,16,power_law_1.01,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,16,power_law_1.01,0.08419839739799499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,16,power_law_1.01,0.08186240196228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,16,power_law_1.01,0.08583679795265198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,16,power_law_1.01,0.08581119775772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,16,power_law_1.01,0.08945279717445373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,16,power_law_1.01,0.08780800104141236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,16,power_law_1.01,0.09700480103492737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,16,power_law_1.01,0.10935039520263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.01,0.11919360160827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.01,0.13905919790267945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.01,0.16412160396575928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.01,0.1998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.01,0.2468928098678589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.01,0.3242496013641357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.01,0.3990080118179321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.01,0.6630784034729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.01,0.7692031860351562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,64,power_law_1.2,0.0630400002002716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,64,power_law_1.2,0.06130560040473938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,64,power_law_1.2,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,64,power_law_1.2,0.06508799791336059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,64,power_law_1.2,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,64,power_law_1.2,0.06611199975013733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,64,power_law_1.2,0.06687999963760376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,64,power_law_1.2,0.06677119731903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,64,power_law_1.2,0.07226240038871765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,64,power_law_1.2,0.07386879920959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,64,power_law_1.2,0.07475839853286743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,64,power_law_1.2,0.08165760040283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,64,power_law_1.2,0.08090239763259888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,64,power_law_1.2,0.07700480222702026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,64,power_law_1.2,0.08798720240592957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,64,power_law_1.2,0.09505919814109802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,64,power_law_1.2,0.09182720184326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,64,power_law_1.2,0.11263999938964844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,64,power_law_1.2,0.12273279428482056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,64,power_law_1.2,0.14419200420379638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,64,power_law_1.2,0.15918079614639283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,64,power_law_1.2,0.21427841186523439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,64,power_law_1.2,0.26817920207977297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,64,power_law_1.2,0.41103358268737794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,64,power_law_1.2,0.5857920169830322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,64,power_law_1.2,0.7992767810821533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,2,power_law_1.2,0.48830718994140626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,64,power_law_1.2,1.0763456344604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,1,power_law_1.2,0.10160640478134156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,1,power_law_1.2,0.11103359460830689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,1,power_law_1.2,0.12842240333557128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,1,power_law_1.2,0.17357439994812013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,1,power_law_1.2,0.20787200927734376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,1,power_law_1.2,0.2504319906234741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,1,power_law_1.2,0.3224384069442749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,1,power_law_1.2,0.3254143953323364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,1,power_law_1.2,0.36116480827331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,1,power_law_1.2,0.3458175897598267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,1,power_law_1.2,0.36739840507507326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,1,power_law_1.2,0.35936000347137453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,1,power_law_1.2,0.37299840450286864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,1,power_law_1.2,0.3834496021270752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,1,power_law_1.2,0.40849919319152833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,1,power_law_1.2,0.40616960525512696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,1,power_law_1.2,0.4338367938995361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,1,power_law_1.2,0.4878335952758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.2,0.5284480094909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.2,0.6313727855682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.2,0.7143487930297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.2,0.915385627746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.2,1.1102720260620118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.2,1.4897279739379883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.2,1.886662483215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,1,power_law_1.2,2.66243839263916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,1,power_law_1.2,3.429849624633789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,2,power_law_1.2,0.5684800148010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,4,power_law_1.01,0.08471680283546448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,4,power_law_1.01,0.10676480531692505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,4,power_law_1.01,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,4,power_law_1.01,0.1520959973335266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,4,power_law_1.01,0.15949440002441406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,4,power_law_1.01,0.1760256052017212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,4,power_law_1.01,0.2110975980758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,4,power_law_1.01,0.1911936044692993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,4,power_law_1.01,0.20094079971313478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,4,power_law_1.01,0.1966976046562195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,4,power_law_1.01,0.20213758945465088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,4,power_law_1.01,0.20909440517425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,4,power_law_1.01,0.2036992073059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,4,power_law_1.01,0.216377592086792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,4,power_law_1.01,0.21370880603790282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,4,power_law_1.01,0.21825919151306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,4,power_law_1.01,0.2254080057144165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,4,power_law_1.01,0.2367039918899536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.01,0.25925118923187257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.01,0.2847424030303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.01,0.3466752052307129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.01,0.44016637802124026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.01,0.5142528057098389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.01,0.7024831771850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.01,0.8715328216552735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.01,1.1835071563720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.01,1.5311424255371093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,32,power_law_1.2,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,32,power_law_1.2,0.030502399802207945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,32,power_law_1.2,0.028934401273727418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,32,power_law_1.2,0.029683199524879456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,32,power_law_1.2,0.03000960052013397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,32,power_law_1.2,0.031244799494743347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,32,power_law_1.2,0.03162240087985992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,32,power_law_1.2,0.031590399146080014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,32,power_law_1.2,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,32,power_law_1.2,0.03253760039806366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,32,power_law_1.2,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,32,power_law_1.2,0.03578880131244659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,32,power_law_1.2,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,32,power_law_1.2,0.03736959993839264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,32,power_law_1.2,0.04041599929332733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,32,power_law_1.2,0.04145280122756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,32,power_law_1.2,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,32,power_law_1.2,0.04561919867992401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,32,power_law_1.2,0.04930559992790222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,32,power_law_1.2,0.0637503981590271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,32,power_law_1.2,0.0688256025314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,32,power_law_1.2,0.0926144003868103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,32,power_law_1.2,0.10449919700622559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,32,power_law_1.2,0.14488960504531861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,32,power_law_1.2,0.17264000177383423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,32,power_law_1.2,0.3028671979904175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,128,power_law_1.2,0.08340479731559754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,128,power_law_1.2,0.08171520233154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,128,power_law_1.2,0.0821120023727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,128,power_law_1.2,0.08858240246772767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,128,power_law_1.2,0.08866559863090515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,128,power_law_1.2,0.09266560077667237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,128,power_law_1.2,0.09589759707450866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,128,power_law_1.2,0.09390079975128174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,128,power_law_1.2,0.09489920139312744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,128,power_law_1.2,0.09388800263404846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,128,power_law_1.2,0.0889024019241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,128,power_law_1.2,0.09183359742164612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,128,power_law_1.2,0.09436799883842469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,128,power_law_1.2,0.08829439878463745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,128,power_law_1.2,0.09519360065460206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,128,power_law_1.2,0.09238399863243103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,128,power_law_1.2,0.09639679789543151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,128,power_law_1.2,0.10412800312042236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,128,power_law_1.2,0.13495680093765258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,128,power_law_1.2,0.15208959579467773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,128,power_law_1.2,0.18186240196228026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,128,power_law_1.2,0.24425599575042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,128,power_law_1.2,0.29989120960235593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,128,power_law_1.2,0.39809279441833495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,128,power_law_1.2,0.5531839847564697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,128,power_law_1.2,0.8239232063293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,128,power_law_1.2,1.193017578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,8,balanced,0.03986666599909464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,8,balanced,0.04145599901676178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,8,balanced,0.039733332892258964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,8,balanced,0.04052799940109253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,8,balanced,0.040421334405740104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,8,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,8,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,8,balanced,0.04177066683769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,8,balanced,0.041493333876132965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,8,balanced,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,8,balanced,0.04137066751718521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,8,balanced,0.04350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,8,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,8,balanced,0.04410133262475332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,8,balanced,0.049957334995269775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,8,balanced,0.0499893327554067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,8,balanced,0.049653331438700356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,8,balanced,0.05769599974155426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,8,balanced,0.061205332477887474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,8,balanced,0.07806399961312611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,8,balanced,0.08688533306121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,8,balanced,0.10745599865913391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,8,balanced,0.11991467078526814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,8,balanced,0.15043200055758157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,8,balanced,0.18620800971984863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,8,balanced,0.24764800071716309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,8,balanced,0.305786669254303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,32,power_law_1.2,0.34156160354614257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,2,power_law_1.01,0.03685120046138764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,2,power_law_1.01,0.04669440090656281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,2,power_law_1.01,0.05189759731292724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,2,power_law_1.01,0.06810240149497986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,2,power_law_1.01,0.0822272002696991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,2,power_law_1.01,0.1057088017463684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,2,power_law_1.01,0.11407999992370606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,2,power_law_1.01,0.11094399690628051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,2,power_law_1.01,0.11792000532150268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,2,power_law_1.01,0.11864320039749146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,2,power_law_1.01,0.12399359941482543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,2,power_law_1.01,0.12420480251312256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,2,power_law_1.01,0.1219648003578186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,2,power_law_1.01,0.13392000198364257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,2,power_law_1.01,0.1429952025413513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,2,power_law_1.01,0.1461824059486389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,2,power_law_1.01,0.1518720030784607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,2,power_law_1.01,0.16957440376281738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,2,power_law_1.01,0.17605760097503662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,2,power_law_1.01,0.20368640422821044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,2,power_law_1.01,0.20814719200134277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,2,power_law_1.01,0.25484800338745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,2,power_law_1.01,0.2920063972473145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,32,power_law_1.2,0.09939200282096863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,32,power_law_1.2,0.09937919974327088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,32,power_law_1.2,0.09635199904441834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,32,power_law_1.2,0.0964031994342804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,32,power_law_1.2,0.08819839954376221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,32,power_law_1.2,0.08903679847717286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,32,power_law_1.2,0.09183359742164612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,2,power_law_1.01,0.3619456052780151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,32,power_law_1.2,0.09638400077819824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,2,power_law_1.01,0.4798975944519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,32,power_law_1.2,0.09876480102539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,2,power_law_1.01,0.6274752140045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,32,power_law_1.2,0.09984639883041382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,2,power_law_1.01,0.7536384105682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,32,power_law_1.2,0.09834880232810975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,32,power_law_1.2,0.10086400508880615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,32,power_law_1.2,0.10701440572738648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,32,power_law_1.2,0.1124608039855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,32,power_law_1.2,0.12357120513916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,32,power_law_1.2,0.12424960136413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,32,power_law_1.2,0.132096004486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,32,power_law_1.2,0.1553536057472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,32,power_law_1.2,0.1571071982383728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,32,power_law_1.2,0.19561599493026732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,32,power_law_1.2,0.2292479991912842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,32,power_law_1.2,0.29103360176086424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,32,power_law_1.2,0.34410240650177004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,32,power_law_1.2,0.45998082160949705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,32,power_law_1.2,0.6607423782348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,32,power_law_1.2,0.8537728309631347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,32,power_law_1.2,1.0749183654785157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,2,power_law_1.01,0.15251840353012086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,2,power_law_1.01,0.2170624017715454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,2,power_law_1.01,0.2737663984298706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,2,power_law_1.01,0.3727744102478027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,2,power_law_1.01,0.4234111785888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,2,power_law_1.01,0.5995711803436279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,2,power_law_1.01,0.6532288074493409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,2,power_law_1.01,0.6181439876556396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,2,power_law_1.01,0.6561791896820068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,2,power_law_1.01,0.6650944232940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,2,power_law_1.01,0.6918079853057861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,2,power_law_1.01,0.6689280033111572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,2,power_law_1.01,0.6558464050292969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,2,power_law_1.01,0.6803647994995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,2,power_law_1.01,0.6794559955596924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,2,power_law_1.01,0.6980671882629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,2,power_law_1.01,0.7224192142486572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,2,power_law_1.01,0.780620813369751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,8,power_law_1.01,0.06231679916381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.01,0.8145343780517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,8,power_law_1.01,0.0725823998451233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.01,0.9776448249816895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.01,1.1032704353332519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.01,1.3291199684143067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.01,1.5695808410644532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.01,2.109094429016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,8,power_law_1.01,0.07175679802894593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.01,2.4734336853027346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,8,power_law_1.01,0.0833728015422821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,8,power_law_1.01,0.08402559757232667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.01,3.3140289306640627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,8,power_law_1.01,0.09085440039634704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,8,power_law_1.01,0.08913279771804809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,8,power_law_1.01,0.0927232027053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.01,4.202700805664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,8,power_law_1.01,0.09176959991455078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,8,power_law_1.01,0.09136639833450318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,8,power_law_1.01,0.09234560132026673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,8,power_law_1.01,0.09561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,8,power_law_1.01,0.0971455991268158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,8,power_law_1.01,0.0997376024723053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,8,power_law_1.01,0.10990719795227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,8,power_law_1.01,0.11056640148162841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,8,power_law_1.01,0.11432960033416747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,8,power_law_1.01,0.13074560165405275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,8,power_law_1.01,0.14344320297241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,8,power_law_1.01,0.1712767958641052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,8,power_law_1.01,0.19325439929962157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,8,power_law_1.01,0.2642431974411011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,8,power_law_1.01,0.33854720592498777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,8,power_law_1.01,0.45384960174560546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,8,power_law_1.01,0.6196479797363281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,8,power_law_1.01,0.8920448303222657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,8,power_law_1.01,1.0887295722961425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,64,power_law_1.01,0.06116480231285095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,64,power_law_1.01,0.060140800476074216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,64,power_law_1.01,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,64,power_law_1.01,0.062483197450637816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,64,power_law_1.01,0.058374398946762086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,64,power_law_1.01,0.05886719822883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,64,power_law_1.01,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,64,power_law_1.01,0.06238080263137817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,64,power_law_1.01,0.0630400002002716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,64,power_law_1.01,0.062118399143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,64,power_law_1.01,0.0627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,64,power_law_1.01,0.06308479905128479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,64,power_law_1.01,0.06629760265350342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,64,power_law_1.01,0.06440960168838501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,64,power_law_1.01,0.06901760101318359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,64,power_law_1.01,0.0683456003665924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,64,power_law_1.01,0.07037439942359924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,64,power_law_1.01,0.08147839903831482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,64,power_law_1.01,0.08583679795265198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,64,power_law_1.01,0.09967359900474548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,64,power_law_1.01,0.11527680158615113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,64,power_law_1.01,0.139136004447937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,64,power_law_1.01,0.1652672052383423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,64,power_law_1.01,0.2280639886856079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,64,power_law_1.01,0.27752320766448973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,64,power_law_1.01,0.40014081001281737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,64,power_law_1.01,0.49234561920166015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,4,power_law_1.2,0.020390400290489198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,4,power_law_1.2,0.022937600314617158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,4,power_law_1.2,0.025062400102615356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,4,power_law_1.2,0.029177600145339967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,4,power_law_1.2,0.037747201323509214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,4,power_law_1.2,0.036595198512077334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,4,power_law_1.2,0.04524160027503967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,4,power_law_1.2,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,4,power_law_1.2,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,4,power_law_1.2,0.04888960123062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,4,power_law_1.2,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,4,power_law_1.2,0.05612800121307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,4,power_law_1.2,0.05817599892616272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,4,power_law_1.2,0.05480960011482239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,4,power_law_1.2,0.06197119951248169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,4,power_law_1.2,0.06320639848709106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,4,power_law_1.2,0.07557119727134705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,4,power_law_1.2,0.08734080195426941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,4,power_law_1.2,0.08795520067214965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,4,power_law_1.2,0.11047680377960205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,4,power_law_1.2,0.13800959587097167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,4,power_law_1.2,0.1603775978088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,4,power_law_1.2,0.1859328031539917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,4,power_law_1.2,0.23860480785369872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,4,power_law_1.2,0.28996479511260986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,4,power_law_1.2,0.4030144214630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,4,power_law_1.2,0.5181439876556396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,16,balanced,0.052186667919158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,16,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,16,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,16,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,16,balanced,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,16,balanced,0.05282666782538096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,16,balanced,0.053770666321118675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,16,balanced,0.05621333420276642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,16,balanced,0.05588266750176748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,16,balanced,0.0580266664425532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,16,balanced,0.05804799993832906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,16,balanced,0.06006933252016703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,16,balanced,0.05992533266544342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,16,balanced,0.06204266846179962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,16,balanced,0.06477866570154826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,16,balanced,0.064751997590065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,16,balanced,0.06845866640408833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,16,balanced,0.0746559997399648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,16,balanced,0.08038933575153351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,16,balanced,0.09642133116722107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,16,balanced,0.1053706705570221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,16,balanced,0.1357439955075582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,16,balanced,0.15528000394503275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,16,balanced,0.20085867245992026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,16,balanced,0.2545013427734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,16,balanced,0.3524426619211833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,16,balanced,0.4291733503341675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,4,power_law_1.01,0.021331200003623964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,4,power_law_1.01,0.02346239984035492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,4,power_law_1.01,0.02539519965648651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,4,power_law_1.01,0.027353599667549133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,4,power_law_1.01,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,4,power_law_1.01,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,4,power_law_1.01,0.04311679899692535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,4,power_law_1.01,0.04377599954605103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,4,power_law_1.01,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,4,power_law_1.01,0.04916479885578155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,4,power_law_1.01,0.04896639883518219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,4,power_law_1.01,0.049184000492095946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,4,power_law_1.01,0.057920002937316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,4,power_law_1.01,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,4,power_law_1.01,0.05859839916229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,4,power_law_1.01,0.09311360120773315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,4,power_law_1.01,0.09461119771003723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,4,power_law_1.01,0.10451200008392333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,4,power_law_1.01,0.10339839458465576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,4,power_law_1.01,0.122489595413208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,4,power_law_1.01,0.12637439966201783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,4,power_law_1.01,0.1487231969833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,4,power_law_1.01,0.17626240253448486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,4,power_law_1.01,0.2380608081817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,4,power_law_1.01,0.29628159999847414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,4,power_law_1.01,0.4158143997192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,4,power_law_1.01,0.5055744171142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,4,power_law_1.2,0.05018240213394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,4,power_law_1.2,0.05203840136528015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,4,power_law_1.2,0.05605120062828064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,4,power_law_1.2,0.06284800171852112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,4,power_law_1.2,0.07082880139350892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,4,power_law_1.2,0.08208000063896179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,4,power_law_1.2,0.08268799781799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,4,power_law_1.2,0.08652160167694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,4,power_law_1.2,0.08844159841537476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,4,power_law_1.2,0.08815360069274902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,4,power_law_1.2,0.09121919870376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,4,power_law_1.2,0.09263359904289245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,4,power_law_1.2,0.09715200066566468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,4,power_law_1.2,0.1065600037574768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,4,power_law_1.2,0.11238399744033814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,4,power_law_1.2,0.11918079853057861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,4,power_law_1.2,0.12313599586486816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,4,power_law_1.2,0.14019839763641356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,4,power_law_1.2,0.15362559556961058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,4,power_law_1.2,0.18718719482421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,4,power_law_1.2,0.24021759033203124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,4,power_law_1.2,0.3155776023864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,4,power_law_1.2,0.36702721118927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,4,power_law_1.2,0.49409918785095214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,4,power_law_1.2,0.6431680202484131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,4,power_law_1.2,0.9706048011779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,4,power_law_1.2,1.1844160079956054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,4,power_law_1.2,0.04888319969177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,4,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,4,power_law_1.2,0.0576960027217865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,4,power_law_1.2,0.063372802734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,4,power_law_1.2,0.06926720142364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,4,power_law_1.2,0.07477759718894958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,4,power_law_1.2,0.08473600149154663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,4,power_law_1.2,0.09382399916648865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,4,power_law_1.2,0.09791359901428223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,4,power_law_1.2,0.09372159838676453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,4,power_law_1.2,0.0998207986354828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,4,power_law_1.2,0.10487040281295776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,4,power_law_1.2,0.11393920183181763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,4,power_law_1.2,0.11352959871292115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,4,power_law_1.2,0.12243839502334594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,4,power_law_1.2,0.13047679662704467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,4,power_law_1.2,0.1438976049423218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,4,power_law_1.2,0.17121280431747438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,4,power_law_1.2,0.1794111967086792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,4,power_law_1.2,0.22177278995513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,4,power_law_1.2,0.2682111978530884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,4,power_law_1.2,0.36136960983276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,4,power_law_1.2,0.4472064018249512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,4,power_law_1.2,0.5938687801361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,4,power_law_1.2,0.8389120101928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,4,power_law_1.2,1.2217408180236817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,4,power_law_1.2,1.5675392150878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,32,power_law_1.01,0.04848000109195709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,32,power_law_1.01,0.0465472012758255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,32,power_law_1.01,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,32,power_law_1.01,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,32,power_law_1.01,0.04555520117282867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,32,power_law_1.01,0.047225600481033324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,32,power_law_1.01,0.04789760112762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,32,power_law_1.01,0.04930559992790222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,32,power_law_1.01,0.049728000164031984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,32,power_law_1.01,0.05047680139541626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,32,power_law_1.01,0.051475197076797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,32,power_law_1.01,0.052832001447677614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,32,power_law_1.01,0.053862398862838744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,32,power_law_1.01,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,32,power_law_1.01,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,32,power_law_1.01,0.06169599890708923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,32,power_law_1.01,0.0641215980052948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,32,power_law_1.01,0.07480319738388061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,32,power_law_1.01,0.08358399868011475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,32,power_law_1.01,0.1128000020980835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,32,power_law_1.01,0.13894399404525756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,32,power_law_1.01,0.16956160068511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,32,power_law_1.01,0.2099519968032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,32,power_law_1.01,0.28764801025390624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,32,power_law_1.01,0.3632767915725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,32,power_law_1.01,0.5292928218841553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,32,power_law_1.01,0.6632319927215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,8,power_law_1.01,0.02974080145359039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,8,power_law_1.01,0.030675199627876282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,8,power_law_1.01,0.028364801406860353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,8,power_law_1.01,0.028352001309394838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,8,power_law_1.01,0.028998398780822755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,8,power_law_1.01,0.02868480086326599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,8,power_law_1.01,0.02993920147418976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,8,power_law_1.01,0.030099201202392577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,8,power_law_1.01,0.03091840147972107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,8,power_law_1.01,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,8,power_law_1.01,0.03146879971027374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,8,power_law_1.01,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,8,power_law_1.01,0.032416000962257385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,8,power_law_1.01,0.03282560110092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,8,power_law_1.01,0.03772160112857818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,8,power_law_1.01,0.03758719861507416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,8,power_law_1.01,0.04220159947872162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,8,power_law_1.01,0.04864639937877655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,8,power_law_1.01,0.05370240211486817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,8,power_law_1.01,0.06382079720497132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,8,power_law_1.01,0.07120000123977661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,8,power_law_1.01,0.08884479999542236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,8,power_law_1.01,0.11174399852752685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,8,power_law_1.01,0.14277759790420533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,2,balanced,0.05381333331267039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,2,balanced,0.054042667150497437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,2,balanced,0.06228266656398773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,2,balanced,0.08095466593901317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,2,balanced,0.11317333579063416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,2,balanced,0.17593065897623697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,2,balanced,0.23693333069483438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,2,balanced,0.23941334088643393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,2,balanced,0.242682675520579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,2,balanced,0.2465440034866333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,2,balanced,0.24247467517852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,2,balanced,0.24459733565648398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,2,balanced,0.2516053318977356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,2,balanced,0.2490826646486918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,2,balanced,0.25939200321833294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,2,balanced,0.264629324277242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,2,balanced,0.27482666571935016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,2,balanced,0.3035200039545695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,2,balanced,0.3223946690559387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,2,balanced,0.36821866035461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,2,balanced,0.42261334260304767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,2,balanced,0.5024426778157552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,2,balanced,0.6110399961471558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,2,balanced,0.7838826974232992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,2,balanced,1.021621306737264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,2,balanced,1.387162685394287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,2,balanced,1.8277546564737956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,32,8,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,32,8,balanced,0.043866669138272606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,32,8,balanced,0.043706665436426796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,32,8,balanced,0.047637333472569786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,32,8,balanced,0.049626668294270836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,32,8,balanced,0.05677866439024607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,32,8,balanced,0.05624000231424967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,32,8,balanced,0.05788266658782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,32,8,balanced,0.05604266623655955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,32,8,balanced,0.0598880002895991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,32,8,balanced,0.05821333328882853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,32,8,balanced,0.06025599936644236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,32,8,balanced,0.06169066826502482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,32,8,balanced,0.06403199831644694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,32,8,balanced,0.06840533514817555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,32,8,balanced,0.07142400244871776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,32,8,balanced,0.07786133388678233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,32,8,balanced,0.09073066711425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,32,8,balanced,0.09473066528638203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,32,8,balanced,0.11133866508801778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,32,8,balanced,0.12353600064913432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,32,8,balanced,0.16012799739837646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,32,8,balanced,0.19333332777023315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,32,8,balanced,0.26157333453496295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,32,8,balanced,0.3163253267606099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,32,8,balanced,0.4472053448359172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,32,8,balanced,0.568447987238566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,8,power_law_1.01,0.1678015947341919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,8,power_law_1.01,0.2502847909927368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,8,power_law_1.01,0.3262079954147339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,2,balanced,0.058133333921432495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,2,balanced,0.060085331400235496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,2,balanced,0.06525333225727081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,2,balanced,0.08726400136947632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,2,balanced,0.11398399869600932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,2,balanced,0.1665173371632894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,2,balanced,0.1657546659310659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,2,balanced,0.16909867525100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,2,balanced,0.16977065801620483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,2,balanced,0.17231466372807822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,2,balanced,0.17383466164271036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,2,balanced,0.1776533325513204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,2,balanced,0.1796320080757141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,2,balanced,0.18319465716679892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,2,balanced,0.18890666961669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,2,balanced,0.1941173275311788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,2,balanced,0.19641600052515665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,2,balanced,0.21588265895843506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,2,balanced,0.2334666649500529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,2,balanced,0.26658133665720624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,2,balanced,0.2999839981396993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,2,balanced,0.36878399054209393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,2,balanced,0.43698132038116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,2,balanced,0.6135199864705404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,2,balanced,0.744757334391276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,2,balanced,1.0876106421152751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,2,balanced,1.382032076517741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,4,balanced,0.07406400144100189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,4,balanced,0.09684800108273824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,4,balanced,0.13806933164596558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,4,balanced,0.21563732624053955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,4,balanced,0.3601706822713216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,4,balanced,0.3781013488769531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,4,balanced,0.37541866302490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,4,balanced,0.3820799986521403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,4,balanced,0.3928426504135132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,4,balanced,0.37590932846069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,4,balanced,0.3882400194803874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,4,balanced,0.3901760180791219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,4,balanced,0.39845867951711017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,4,balanced,0.39531199137369794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,4,balanced,0.3943466742833455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,4,balanced,0.3884906768798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,4,balanced,0.39851200580596924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,4,balanced,0.404421329498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,4,balanced,0.3856853246688843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,4,balanced,0.39050666491190594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,4,balanced,0.40029335021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,4,balanced,0.7285919984181722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,4,balanced,0.7040106455485026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,4,balanced,1.0163413683573406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,4,balanced,1.3710719744364421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,4,balanced,1.519861380259196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,4,balanced,2.21232000986735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,1,power_law_1.2,0.051283198595047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,1,power_law_1.2,0.08140159845352173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,1,power_law_1.2,0.1336192011833191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,1,power_law_1.2,0.18689919710159303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,1,power_law_1.2,0.24717440605163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,1,power_law_1.2,0.4249919891357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,1,power_law_1.2,0.4635903835296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,1,power_law_1.2,0.48470401763916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,1,power_law_1.2,0.5036799907684326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,1,power_law_1.2,0.5169792175292969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,1,power_law_1.2,0.5697792053222657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,1,power_law_1.2,0.5886335849761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,1,power_law_1.2,0.5634175777435303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,1,power_law_1.2,0.6019839763641357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,1,power_law_1.2,0.6220672130584717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,1,power_law_1.2,0.6041535854339599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,1,power_law_1.2,0.6259071826934814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,1,power_law_1.2,0.6753407955169678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.2,0.7253695964813233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.2,1.6222784042358398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.2,1.8339391708374024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.2,1.111513614654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.2,1.2294976234436035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.2,1.5233280181884765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.2,1.811667251586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.2,2.405612754821777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.2,2.9298175811767577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,1,power_law_1.01,0.020524799823760986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,1,power_law_1.01,0.019385600090026857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,1,power_law_1.01,0.019756799936294554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,1,power_law_1.01,0.019852800667285918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,1,power_law_1.01,0.021555200219154358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,1,power_law_1.01,0.02338559925556183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,1,power_law_1.01,0.02374400049448013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,1,power_law_1.01,0.024166400730609893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,1,power_law_1.01,0.024268800020217897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,1,power_law_1.01,0.024531200528144836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,1,power_law_1.01,0.025772801041603087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,1,power_law_1.01,0.026348799467086792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,1,power_law_1.01,0.026630398631095887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,1,power_law_1.01,0.028723201155662535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,1,power_law_1.01,0.05236480236053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,1,power_law_1.01,0.05296639800071716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,1,power_law_1.01,0.0542464017868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,1,power_law_1.01,0.04382080137729645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.01,0.04687359929084778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.01,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.01,0.07531520128250122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.01,0.0993664026260376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.01,0.12398079633712769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.01,0.17059199810028075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.01,0.20936961174011232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.01,0.30604801177978513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.01,0.39341440200805666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,16,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,16,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,16,balanced,0.03409066547950109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,16,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,16,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,16,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,16,balanced,0.03732266773780187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,16,balanced,0.03757333258787791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,16,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,16,balanced,0.037605332831541695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,16,balanced,0.03938666731119156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,16,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,16,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,16,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,16,balanced,0.04285866518815359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,16,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,16,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,16,balanced,0.04805333415667216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,16,balanced,0.04797333478927612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,16,balanced,0.05213866631189982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,16,balanced,0.05621333420276642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,16,balanced,0.06620799998442332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,16,balanced,0.07303466896216075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,16,balanced,0.09492266178131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,16,balanced,0.12128000458081563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,16,balanced,0.15653333067893982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,16,balanced,0.18976000944773355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,8,balanced,0.0517439991235733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,8,balanced,0.05586666862169901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,8,balanced,0.05807466804981232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,8,balanced,0.060229331254959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,8,balanced,0.07262933254241943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,8,balanced,0.09923199812571208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,8,balanced,0.122079998254776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,8,balanced,0.12171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,8,balanced,0.12357866764068604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,8,balanced,0.12583999832471213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,8,balanced,0.12574399511019388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,8,balanced,0.12756799658139548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,8,balanced,0.13019200166066489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,8,balanced,0.13152000308036804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,8,balanced,0.1372106671333313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,8,balanced,0.14191466569900513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,8,balanced,0.14481066664059958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,8,balanced,0.15555733442306519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,8,balanced,0.1674613356590271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,8,balanced,0.1906986633936564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,8,balanced,0.2080693244934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,8,balanced,0.25433067480723065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,8,balanced,0.30084266265233356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,8,balanced,0.39130667845408124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,8,balanced,0.5195466677347819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,8,balanced,0.6881493727366129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,8,balanced,0.8974613348642985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,4,power_law_1.01,0.039628800749778745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,4,power_law_1.01,0.037011200189590455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,4,power_law_1.01,0.0390720009803772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,4,power_law_1.01,0.041152000427246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,4,power_law_1.01,0.04689280092716217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,4,power_law_1.01,0.05615360140800476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,4,power_law_1.01,0.05623040199279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,4,power_law_1.01,0.05736320018768311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,4,power_law_1.01,0.05852800011634827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,4,power_law_1.01,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,4,power_law_1.01,0.061414402723312375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,4,power_law_1.01,0.06231039762496948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,4,power_law_1.01,0.06473600268363952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,4,power_law_1.01,0.06680319905281067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,4,power_law_1.01,0.07384960055351257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,4,power_law_1.01,0.07622399926185608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,4,power_law_1.01,0.07926399707794189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,4,power_law_1.01,0.08729599714279175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,4,power_law_1.01,0.10839680433273316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,4,power_law_1.01,0.12360320091247559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,4,power_law_1.01,0.13023359775543214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,4,power_law_1.01,0.17570559978485106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,4,power_law_1.01,0.2036736011505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,4,power_law_1.01,0.30519680976867675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,4,power_law_1.01,0.3713792085647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,4,power_law_1.01,0.5460415840148926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,4,power_law_1.01,0.7048768043518067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,1,balanced,0.021136000752449036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,1,balanced,0.02179199953873952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,1,balanced,0.025263999899228413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,1,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,1,balanced,0.054757331808408104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,1,balanced,0.08649067083994548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,1,balanced,0.08969066540400188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,1,balanced,0.09161600470542908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,1,balanced,0.09513599673906963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,1,balanced,0.09684266646703084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,1,balanced,0.09947733084360759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,1,balanced,0.10307199756304423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,1,balanced,0.10564800103505452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,1,balanced,0.1064906617005666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,1,balanced,0.10972799857457478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,1,balanced,0.11354133486747742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,1,balanced,0.12329600254694621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,1,balanced,0.14416000247001648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,1,balanced,0.15412267049153647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,1,balanced,0.19305066267649332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,1,balanced,0.2184106707572937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,1,balanced,0.2913866639137268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,1,balanced,0.2967519958813985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,1,balanced,0.41017599900563556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,1,balanced,0.522815982500712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,1,balanced,0.7507572968800863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,1,balanced,1.027125358581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,128,balanced,0.05249066650867462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,128,balanced,0.05413866539796194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,128,balanced,0.04797866443792979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,128,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,128,balanced,0.048138668139775596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,128,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,128,balanced,0.05217066903909048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,128,balanced,0.05251200000445048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,128,balanced,0.052101333936055504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,128,balanced,0.05392533540725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,128,balanced,0.05291733145713806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,128,balanced,0.05420266588528951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,128,balanced,0.05610666672388712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,128,balanced,0.05598400036493937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,128,balanced,0.06208533545335134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,128,balanced,0.06215466558933258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,128,balanced,0.06406933565934499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,128,balanced,0.07049066821734111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,128,balanced,0.07434666653474171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,128,balanced,0.08692266543706258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,128,balanced,0.10097066561381023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,128,balanced,0.1272426644961039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,128,balanced,0.14502933621406555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,128,balanced,0.18360000848770142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,128,balanced,0.22880532344182333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,128,balanced,0.30427734057108563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,128,balanced,0.3803679943084717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,4,power_law_1.01,0.07094399929046631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,4,power_law_1.01,0.09767040014266967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,4,power_law_1.01,0.14614399671554565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,4,power_law_1.01,0.15126399993896483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,4,power_law_1.01,0.1890239953994751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,4,power_law_1.01,0.2042367935180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,4,power_law_1.01,0.25240321159362794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,4,power_law_1.01,0.3042623996734619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,4,power_law_1.01,0.2832319974899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,4,power_law_1.01,0.28615679740905764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,4,power_law_1.01,0.3126784086227417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,4,power_law_1.01,0.3091264009475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,4,power_law_1.01,0.3107072114944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,4,power_law_1.01,0.3267519950866699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,4,power_law_1.01,0.3240384101867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,4,power_law_1.01,0.33889920711517335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,4,power_law_1.01,0.34268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,4,power_law_1.01,0.3914495944976807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,4,power_law_1.01,0.41999359130859376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,4,power_law_1.01,0.4908480167388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,4,power_law_1.01,0.5181759834289551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,4,power_law_1.01,0.6399424076080322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,4,power_law_1.01,0.7688831806182861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,4,power_law_1.01,0.9882176399230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,4,power_law_1.01,1.1222271919250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,4,power_law_1.01,1.5297344207763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,4,power_law_1.01,1.9272832870483398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,2,power_law_1.01,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,2,power_law_1.01,0.05103999972343445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,2,power_law_1.01,0.0666815996170044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,2,power_law_1.01,0.08313599824905396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,2,power_law_1.01,0.10289920568466186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,2,power_law_1.01,0.14443520307540894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,2,power_law_1.01,0.14721920490264892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,2,power_law_1.01,0.15368959903717042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,2,power_law_1.01,0.15949440002441406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,2,power_law_1.01,0.16620800495147706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,2,power_law_1.01,0.16918400526046753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,2,power_law_1.01,0.1696768045425415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,2,power_law_1.01,0.17630720138549805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,2,power_law_1.01,0.17440639734268187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,2,power_law_1.01,0.1855039954185486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,2,power_law_1.01,0.19486720561981202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,2,power_law_1.01,0.20572800636291505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,2,power_law_1.01,0.22239360809326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.01,0.2386176109313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.01,0.2844480037689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.01,0.33968000411987304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.01,0.3845247983932495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.01,0.4333695888519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.01,0.5685311794281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.01,0.7066751956939697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.01,0.9141183853149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.01,1.1468352317810058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,4,power_law_1.2,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,4,power_law_1.2,0.10629760026931763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,4,power_law_1.2,0.11150720119476318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,4,power_law_1.2,0.13667839765548706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,4,power_law_1.2,0.14275840520858765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,4,power_law_1.2,0.17913600206375122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,4,power_law_1.2,0.19438719749450684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,4,power_law_1.2,0.2075648069381714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,4,power_law_1.2,0.19556479454040526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,4,power_law_1.2,0.20067200660705567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,4,power_law_1.2,0.20330240726470947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,4,power_law_1.2,0.20617599487304689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,4,power_law_1.2,0.21850240230560303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,4,power_law_1.2,0.22176640033721923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,4,power_law_1.2,0.22047359943389894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,128,power_law_1.01,0.05100799798965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,128,power_law_1.01,0.054016000032424925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,128,power_law_1.01,0.04068480134010315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,128,power_law_1.01,0.04140160083770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,128,power_law_1.01,0.041356799006462094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,128,power_law_1.01,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,128,power_law_1.01,0.044012799859046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,128,power_law_1.01,0.044486400485038755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,128,power_law_1.01,0.04477440118789673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,128,power_law_1.01,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,128,power_law_1.01,0.045542401075363156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,128,power_law_1.01,0.046374401450157164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,128,power_law_1.01,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,128,power_law_1.01,0.04847359955310822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,128,power_law_1.01,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,128,power_law_1.01,0.05464959740638733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,128,power_law_1.01,0.057574397325515746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,128,power_law_1.01,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,128,power_law_1.01,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,128,power_law_1.01,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,128,power_law_1.01,0.09381759762763978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,128,power_law_1.01,0.12344319820404052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,128,power_law_1.01,0.14752639532089235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,128,power_law_1.01,0.20262401103973388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,128,power_law_1.01,0.2534656047821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,128,power_law_1.01,0.36043519973754884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,128,power_law_1.01,0.44478721618652345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,1,power_law_1.2,0.09723520278930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,1,power_law_1.2,0.11977599859237671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,1,power_law_1.2,0.16135040521621705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,1,power_law_1.2,0.21166720390319824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,1,power_law_1.2,0.25068800449371337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,1,power_law_1.2,0.38714239597320554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,1,power_law_1.2,0.39886720180511476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,1,power_law_1.2,0.41161599159240725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,1,power_law_1.2,0.4180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,1,power_law_1.2,0.4298111915588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,1,power_law_1.2,0.4373568058013916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,1,power_law_1.2,0.44641919136047364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,1,power_law_1.2,0.4480127811431885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,1,power_law_1.2,0.4663680076599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,1,power_law_1.2,0.48069119453430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,1,power_law_1.2,0.48899202346801757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,1,power_law_1.2,0.5119296073913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,1,power_law_1.2,0.5677120208740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.2,0.6276415824890137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.2,0.7261439800262451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.2,0.8380800247192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.2,1.0721599578857421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.2,1.288576030731201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.2,1.724550437927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.2,2.1718656539916994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.2,3.057708740234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.2,3.9383296966552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,8,power_law_1.01,0.03253760039806366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,8,power_law_1.01,0.03141759932041168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,8,power_law_1.01,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,8,power_law_1.01,0.03399679958820343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,8,power_law_1.01,0.034406399726867674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,8,power_law_1.01,0.03644160032272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,8,power_law_1.01,0.0364544004201889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,8,power_law_1.01,0.037299200892448425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,2,balanced,0.05004799862702688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,2,balanced,0.05080533524354299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,2,balanced,0.06127466758092245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,2,balanced,0.08142399787902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,2,balanced,0.11529067158699036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,2,balanced,0.17148266235987344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,2,balanced,0.17493865887324014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,2,balanced,0.17358932892481485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,2,balanced,0.17633599042892456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,2,balanced,0.17806400855382284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,2,balanced,0.18167465925216675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,2,balanced,0.18398400147755942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,2,balanced,0.18929600715637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,2,balanced,0.19290133317311606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,2,balanced,0.200490673383077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,2,balanced,0.20586667458216348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,2,balanced,0.23043199380238852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,2,balanced,0.2582026720046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,2,balanced,0.275765339533488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,2,balanced,0.32233067353566486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,2,balanced,0.36658668518066406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,2,balanced,0.4983839988708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,2,balanced,0.5765440066655477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,2,balanced,0.821722666422526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,2,balanced,1.006719986597697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,2,balanced,1.4647092819213867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,2,balanced,1.8749759991963704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,8,power_law_1.01,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,8,power_law_1.01,0.03852800130844116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,8,power_law_1.01,0.03900800049304962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,32,power_law_1.01,0.11897599697113037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,32,power_law_1.01,0.10974080562591552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,32,power_law_1.01,0.11281919479370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,32,power_law_1.01,0.11395839452743531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,32,power_law_1.01,0.1147007942199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,32,power_law_1.01,0.10858880281448365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,32,power_law_1.01,0.11276799440383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,32,power_law_1.01,0.11026560068130493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,32,power_law_1.01,0.11531519889831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,32,power_law_1.01,0.11831040382385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,32,power_law_1.01,0.12389119863510131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,32,power_law_1.01,0.1245568037033081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,32,power_law_1.01,0.12863359451293946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,32,power_law_1.01,0.11191680431365966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,32,power_law_1.01,0.13386240005493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,32,power_law_1.01,0.13763840198516847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,32,power_law_1.01,0.1443711996078491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,32,power_law_1.01,0.1667456030845642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,32,power_law_1.01,0.1760256052017212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,32,power_law_1.01,0.23466880321502687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,32,power_law_1.01,0.262444806098938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,32,power_law_1.01,0.32046079635620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,32,power_law_1.01,0.36438400745391847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,32,power_law_1.01,0.4614272117614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,32,power_law_1.01,0.5291967868804932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,32,power_law_1.01,0.7756735801696777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,32,power_law_1.01,0.9212224006652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,8,power_law_1.01,0.04141440093517303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,8,power_law_1.01,0.041920000314712526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,4,power_law_1.01,0.020147199928760528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,4,power_law_1.01,0.02494720071554184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,4,power_law_1.01,0.02852480113506317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,4,power_law_1.01,0.03948799967765808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,4,power_law_1.01,0.044409599900245664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,4,power_law_1.01,0.04501760005950928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,4,power_law_1.01,0.051923197507858274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,4,power_law_1.01,0.052153599262237546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,4,power_law_1.01,0.055846399068832396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,4,power_law_1.01,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,4,power_law_1.01,0.05947520136833191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,4,power_law_1.01,0.05991680026054382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,4,power_law_1.01,0.061260801553726194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,4,power_law_1.01,0.06110720038414001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,4,power_law_1.01,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,4,power_law_1.01,0.07274240255355835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,4,power_law_1.01,0.0751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,4,power_law_1.01,0.11527680158615113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,4,power_law_1.01,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,4,power_law_1.01,0.12842240333557128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,4,power_law_1.01,0.13154560327529907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,4,power_law_1.01,0.13171199560165406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,4,power_law_1.01,0.1467519998550415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,4,power_law_1.01,0.18302079439163207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,4,power_law_1.01,0.22444159984588624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,16,power_law_1.01,0.08840960264205933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,4,power_law_1.01,0.30327041149139405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,16,power_law_1.01,0.1218176007270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,4,power_law_1.01,0.37804160118103025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,16,power_law_1.01,0.11756160259246826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,16,power_law_1.01,0.1180351972579956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,16,power_law_1.01,0.12017920017242431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,16,power_law_1.01,0.115174400806427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,16,power_law_1.01,0.11731200218200684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,16,power_law_1.01,0.1146623969078064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,16,power_law_1.01,0.11939200162887573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,16,power_law_1.01,0.11835520267486573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,16,power_law_1.01,0.12170239686965942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,16,power_law_1.01,0.12515840530395508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,16,power_law_1.01,0.12221440076828002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,16,power_law_1.01,0.1263360023498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,16,power_law_1.01,0.14205440282821655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,16,power_law_1.01,0.14094079732894899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,16,power_law_1.01,0.14181120395660402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,16,power_law_1.01,0.1630079984664917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,16,power_law_1.01,0.1718016028404236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,16,power_law_1.01,0.19338239431381227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,16,power_law_1.01,0.21831040382385253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,16,power_law_1.01,0.2881407976150513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,16,power_law_1.01,0.34898560047149657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,16,power_law_1.01,0.4721216201782227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,16,power_law_1.01,0.6439231872558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,16,power_law_1.01,0.9009407997131348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,16,power_law_1.01,1.1540608406066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,8,power_law_1.01,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,8,power_law_1.01,0.053420799970626834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,8,power_law_1.01,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,8,power_law_1.01,0.06602879762649536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,2,power_law_1.01,0.024166400730609893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,2,power_law_1.01,0.02222079932689667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,2,power_law_1.01,0.021695999801158904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,2,power_law_1.01,0.025337600708007814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,2,power_law_1.01,0.030105599761009218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,2,power_law_1.01,0.030681601166725157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,2,power_law_1.01,0.02961919903755188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,2,power_law_1.01,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,2,power_law_1.01,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,2,power_law_1.01,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,2,power_law_1.01,0.036313599348068236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,2,power_law_1.01,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,2,power_law_1.01,0.07560319900512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,2,power_law_1.01,0.09095680117607116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,2,power_law_1.01,0.06471679806709289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,2,power_law_1.01,0.05658239722251892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,2,power_law_1.01,0.06287360191345215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,2,power_law_1.01,0.08257920145988465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.01,0.09915519952774048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.01,0.13782399892807007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.01,0.17155200242996216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.01,0.2388607978820801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.01,0.29921278953552244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.01,0.4353792190551758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,2,power_law_1.2,0.045900800824165346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,2,power_law_1.2,0.04814079999923706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,2,power_law_1.2,0.04921599924564361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,2,power_law_1.2,0.062105602025985716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,2,power_law_1.2,0.0642624020576477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,2,power_law_1.2,0.06529279947280883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,2,power_law_1.2,0.06522240042686463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,2,power_law_1.2,0.06607999801635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,2,power_law_1.2,0.06613759994506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,2,power_law_1.2,0.06601600050926208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,2,power_law_1.2,0.06683520078659058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,2,power_law_1.2,0.0746944010257721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,2,power_law_1.2,0.078438401222229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,8,power_law_1.01,0.07660800218582153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,2,power_law_1.2,0.08762239813804626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,2,power_law_1.2,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,2,power_law_1.2,0.09505919814109802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,8,power_law_1.01,0.09564800262451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,2,power_law_1.2,0.10240639448165893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,8,power_law_1.01,0.12003840208053589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,2,power_law_1.2,0.12814079523086547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,2,power_law_1.2,0.14550399780273438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.01,0.5625279903411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,2,power_law_1.2,0.19335039854049682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,2,power_law_1.2,0.23891839981079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,2,power_law_1.2,0.33187839984893797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.01,0.8289536476135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,8,power_law_1.2,0.035872000455856326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,8,power_law_1.2,0.03560320138931274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,8,power_law_1.2,0.03283840119838714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,8,power_law_1.2,0.03375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,8,power_law_1.2,0.037651199102401736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,2,power_law_1.2,0.4530303955078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,8,power_law_1.2,0.037555199861526486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,8,power_law_1.2,0.03680639863014221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,8,power_law_1.2,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,8,power_law_1.2,0.03772160112857818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,8,power_law_1.2,0.03860479891300202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,8,power_law_1.2,0.03895680010318756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,8,power_law_1.2,0.04009599983692169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,8,power_law_1.2,0.04196479916572571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,8,power_law_1.2,0.04503679871559143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,8,power_law_1.2,0.050406402349472045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,8,power_law_1.2,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,8,power_law_1.2,0.053395199775695804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,8,power_law_1.2,0.06212480068206787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,8,power_law_1.2,0.05989760160446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,8,power_law_1.2,0.07823359966278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,8,power_law_1.2,0.07946239709854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,8,power_law_1.2,0.09416319727897644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,8,power_law_1.2,0.10901119709014892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,8,power_law_1.2,0.14357119798660278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,8,power_law_1.01,0.15000319480895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,2,power_law_1.2,0.6190656185150146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,8,power_law_1.2,0.187225604057312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,16,balanced,0.023189333577950794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,16,balanced,0.021253332495689392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,16,balanced,0.021984001000722248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,16,balanced,0.023034666975339253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,16,balanced,0.023402666052182514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,16,balanced,0.025493333737055462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,16,balanced,0.027162666122118633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,16,balanced,0.027477333943049114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,16,balanced,0.029077333708604176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,16,balanced,0.02918400118748347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,16,balanced,0.029258665939172108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,16,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,16,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,16,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,16,balanced,0.04141333450873693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,16,balanced,0.04358933369318644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,16,balanced,0.05985599756240845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,16,balanced,0.06822933256626129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,16,balanced,0.07607999940713246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,8,power_law_1.01,0.21829121112823485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,16,balanced,0.08902933200200398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,16,balanced,0.10233066479365031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,16,balanced,0.12798399726549783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,16,balanced,0.16221867005030313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,16,balanced,0.2158506711324056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,16,balanced,0.26524800062179565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,16,balanced,0.392192006111145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,16,balanced,0.5025333166122437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,8,power_law_1.2,0.2695039987564087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,8,power_law_1.2,0.31979520320892335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,8,power_law_1.01,0.28531200885772706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,8,power_law_1.01,0.4329216003417969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,2,power_law_1.2,0.7742208003997803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,8,power_law_1.01,0.5951680183410645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,8,power_law_1.01,0.8886272430419921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.01,1.0908672332763671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,2,power_law_1.2,1.2875519752502442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,8,power_law_1.01,1.2688703536987305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,2,power_law_1.2,1.4564736366271973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,64,balanced,0.029205332199732464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,64,balanced,0.029301332930723827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,64,balanced,0.027306665976842243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,64,balanced,0.029680001238981884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,64,balanced,0.03175999969244003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,64,balanced,0.03164800008138021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,64,balanced,0.03223466624816259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,64,balanced,0.03379199902216593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,64,balanced,0.03339733431736628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,64,balanced,0.03348266581694285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,64,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,64,balanced,0.035818666219711304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,64,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,64,balanced,0.035599999129772186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,64,balanced,0.03972266614437103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,64,balanced,0.03980266551176707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,64,balanced,0.04182933270931244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,64,balanced,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,64,balanced,0.04599999884764353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,64,balanced,0.05205333232879639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,64,balanced,0.054117331902186074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,64,balanced,0.06451199948787689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,64,balanced,0.07394666473070781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,64,balanced,0.09479999542236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,64,balanced,0.11256532867749532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,64,balanced,0.15149333079655966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,64,balanced,0.19530133406321207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,4,power_law_1.2,0.15905280113220216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,4,power_law_1.2,0.20239360332489015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,4,power_law_1.2,0.23297278881072997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,4,power_law_1.2,0.3170111894607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,4,power_law_1.2,0.3478528022766113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,4,power_law_1.2,0.4838399887084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,4,power_law_1.2,0.5061759948730469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,4,power_law_1.2,0.5209152221679687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,4,power_law_1.2,0.5424831867218017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,4,power_law_1.2,0.5434175968170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,4,power_law_1.2,0.5251264095306396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,4,power_law_1.2,0.5695231914520263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,4,power_law_1.2,0.5869760036468505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,4,power_law_1.2,0.603379201889038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,4,power_law_1.2,0.6227392196655274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,4,power_law_1.2,0.6296512126922608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,4,power_law_1.2,0.6808383941650391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,4,power_law_1.2,0.7690688133239746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,4,power_law_1.2,0.8098879814147949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,4,power_law_1.2,0.8892864227294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,4,power_law_1.2,1.0075263977050781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,4,power_law_1.2,1.2988927841186524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,4,power_law_1.2,1.3065152168273926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,4,power_law_1.2,1.5236991882324218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,4,power_law_1.2,2.399545669555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,4,power_law_1.2,3.002604866027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,4,power_law_1.2,3.5974273681640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,8,power_law_1.2,0.05483520030975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,8,power_law_1.2,0.05411199927330017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,8,power_law_1.2,0.03390080034732819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,8,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,8,power_law_1.2,0.03466239869594574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,8,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,8,power_law_1.2,0.04077439904212952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,8,power_law_1.2,0.04217599928379059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,8,power_law_1.2,0.04772480130195618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,8,power_law_1.2,0.04747520089149475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,8,power_law_1.2,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,8,power_law_1.2,0.05288959741592407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,8,power_law_1.2,0.0525439977645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,8,power_law_1.2,0.05132160186767578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,8,power_law_1.2,0.06284160017967225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,8,power_law_1.2,0.0643455982208252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,8,power_law_1.2,0.07626240253448487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,8,power_law_1.2,0.09633920192718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,8,power_law_1.2,0.10286719799041748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,8,power_law_1.2,0.12711039781570435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,8,power_law_1.2,0.13980159759521485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,8,power_law_1.2,0.20169599056243898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,8,power_law_1.2,0.25131518840789796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,8,power_law_1.2,0.37205119132995607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,128,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,128,balanced,0.04844266672929128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,128,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,128,balanced,0.05220800141493479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,128,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,128,balanced,0.05420800050099691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,128,balanced,0.059402664502461754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,16,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,16,balanced,0.04177600145339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,16,balanced,0.04156800111134847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,16,balanced,0.04191466669241587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,16,balanced,0.04383466641108195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,16,balanced,0.04598933458328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,128,balanced,0.05826666454474131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,16,balanced,0.046469335754712425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,128,balanced,0.062133332093556724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,16,balanced,0.04764799773693085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,16,balanced,0.04832000037034353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,16,balanced,0.047637333472569786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,16,balanced,0.047983999053637184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,16,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,16,balanced,0.05166399975617727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,16,balanced,0.05197333296140035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,16,balanced,0.054085334142049156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,16,balanced,0.056186666091283165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,16,balanced,0.059631998340288796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,16,balanced,0.06411199768384297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,16,balanced,0.06810666620731354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,16,balanced,0.07841066519419353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,16,balanced,0.08690667152404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,16,balanced,0.10690133770306905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,16,balanced,0.1381439963976542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,16,balanced,0.17546133200327554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,16,balanced,0.2086133360862732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,128,balanced,0.059232001503308616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,16,balanced,0.27903467416763306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,16,balanced,0.3619999885559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,8,power_law_1.2,0.37136640548706057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,128,balanced,0.06412266691525777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,128,balanced,0.0746666689713796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,128,balanced,0.07601066430409749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,128,balanced,0.07027733325958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,1,balanced,0.04962133367856344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,1,balanced,0.06020799775918325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,1,balanced,0.08132799963156383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,1,balanced,0.11241599917411804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,1,balanced,0.17523199319839478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,1,balanced,0.20940266052881876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,1,balanced,0.2095359961191813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,1,balanced,0.21620800097783408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,1,balanced,0.21653334299723306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,128,balanced,0.08323733508586884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,128,balanced,0.08886933326721191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,128,balanced,0.0885813335577647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,128,balanced,0.1018506685892741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,128,balanced,0.10944533348083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,128,balanced,0.12813867131868997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,128,balanced,0.1421333352724711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,128,balanced,0.18315200010935465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,128,balanced,0.21658132473627725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,128,balanced,0.27983999252319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,1,balanced,0.21904534101486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,128,balanced,0.348037322362264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,1,balanced,0.22139199574788412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,1,balanced,0.22777599096298218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,128,balanced,0.4724586804707845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,1,balanced,0.2367466688156128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,1,balanced,0.23911466201146445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,1,balanced,0.24715199073155722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,1,balanced,0.25759466489156085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,1,balanced,0.27114667495091754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,1,balanced,0.3076053261756897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,1,balanced,0.33166933059692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,1,balanced,0.4471520185470581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,1,balanced,0.4696586529413859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,1,balanced,0.7010080019632975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,1,balanced,0.7647999922434489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,128,balanced,0.5956000089645386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,1,balanced,1.222442626953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,1,balanced,1.3875679969787598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,1,balanced,2.200010617574056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,1,balanced,2.6220107078552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,8,power_law_1.2,0.5937151908874512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,128,power_law_1.01,0.035078400373458864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,128,power_law_1.01,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,128,power_law_1.01,0.034585601091384886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,128,power_law_1.01,0.03701759874820709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,128,power_law_1.01,0.037062400579452516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,128,power_law_1.01,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,128,power_law_1.01,0.03829120099544525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,128,power_law_1.01,0.03763200044631958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,128,power_law_1.01,0.038412800431251524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,128,power_law_1.01,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,128,power_law_1.01,0.038771200180053714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,128,power_law_1.01,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,128,power_law_1.01,0.04010240137577057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,128,power_law_1.01,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,8,power_law_1.2,0.7681151866912842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,128,power_law_1.01,0.04304639995098114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,128,power_law_1.01,0.04545280039310455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,128,power_law_1.01,0.04527359902858734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,128,power_law_1.01,0.049235200881958006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.01,0.05333120226860046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.01,0.05975040197372437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.01,0.06943359971046448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.01,0.08581759929656982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.01,0.10335359573364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.01,0.13035520315170288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.01,0.1621888041496277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.01,0.24162559509277343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.01,0.3117120027542114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,2,balanced,0.06239999830722809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,2,balanced,0.06618133187294006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,2,balanced,0.07014399766921997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,2,balanced,0.08389866352081299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,2,balanced,0.1039626697699229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,2,balanced,0.160671999057134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,2,balanced,0.1613866686820984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,2,balanced,0.15757866700490317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,2,balanced,0.15846932927767435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,2,balanced,0.160970667997996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,2,balanced,0.1609280010064443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,2,balanced,0.16500266393025717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,2,balanced,0.1681333382924398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,2,balanced,0.17059199015299478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,2,balanced,0.17725332578023276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,2,balanced,0.18040533860524496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,2,balanced,0.1867039998372396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,2,balanced,0.20974934101104736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,2,balanced,0.2265066703160604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,2,balanced,0.26740266879399616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,2,balanced,0.3011786739031474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,2,balanced,0.39394664764404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,2,balanced,0.47304534912109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,2,balanced,0.6925386587778727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,2,balanced,0.8775573571523031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,2,balanced,1.2851520379384358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,2,balanced,1.657434622446696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,4,balanced,0.058277333776156105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,4,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,4,balanced,0.06833600004514058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,4,balanced,0.082805335521698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,4,balanced,0.1071626643339793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,4,balanced,0.16133333245913187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,4,balanced,0.1572480003039042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,4,balanced,0.15664533774058023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,4,balanced,0.1514133314291636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,4,balanced,0.15294933319091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,4,balanced,0.15392000476519266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,4,balanced,0.15686399737993875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,4,balanced,0.15633599956830344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,4,balanced,0.15786666671435037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,4,balanced,0.16459733247756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,4,balanced,0.1646560033162435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,4,balanced,0.16886399189631143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,4,balanced,0.18273067474365234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,4,balanced,0.19243733088175455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,4,balanced,0.2176533341407776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,4,balanced,0.23752532402674356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,4,balanced,0.28384000062942505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,4,balanced,0.3307466705640157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,4,balanced,0.4861493508021037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,4,balanced,0.5810720125834147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,4,balanced,0.8158400058746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,4,balanced,1.0600053469340007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,1,balanced,0.052101333936055504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,1,balanced,0.05190933247407278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,1,balanced,0.05932266513506571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,1,balanced,0.08489066362380981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,1,balanced,0.12201066811879475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,1,balanced,0.12433600425720215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,1,balanced,0.12797333796819052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,1,balanced,0.12782399853070578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,1,balanced,0.13024533788363138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,1,balanced,0.13076800107955933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,1,balanced,0.13292266925175986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,1,balanced,0.1374666690826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,1,balanced,0.1404906709988912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,1,balanced,0.14477866888046265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,1,balanced,0.15203733245531717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,1,balanced,0.15784533818562826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,1,balanced,0.16641599933306375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,1,balanced,0.19107200702031454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,1,balanced,0.21268266439437866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,1,balanced,0.2608906626701355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,1,balanced,0.30643200874328613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,1,balanced,0.4352000157038371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,1,balanced,0.5256213347117106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,1,balanced,0.7562133471171061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,1,balanced,0.9611999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,1,balanced,1.4081439971923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,2,power_law_1.01,0.023974399268627166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,2,power_law_1.01,0.03041279911994934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,2,power_law_1.01,0.041043201088905336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,2,power_law_1.01,0.0535103976726532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,2,power_law_1.01,0.06242560148239136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,2,power_law_1.01,0.07998719811439514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,2,power_law_1.01,0.08490880131721497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,2,power_law_1.01,0.08590720295906067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,1,balanced,1.8562347094217937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,2,power_law_1.01,0.09095039963722229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2688,1856,6,128,1,2,power_law_1.01,0.09088000059127807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2688,1856,6,128,1,2,power_law_1.01,0.09479680061340331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2688,1856,6,128,1,2,power_law_1.01,0.09919360280036926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2688,1856,6,128,1,2,power_law_1.01,0.10235519409179687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2688,1856,6,128,1,2,power_law_1.01,0.10949759483337403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2688,1856,6,128,1,2,power_law_1.01,0.11681920289993286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2688,1856,6,128,1,2,power_law_1.01,0.130348801612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2688,1856,6,128,1,2,power_law_1.01,0.11827199459075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,64,power_law_1.01,0.029843199253082275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,64,power_law_1.01,0.029043200612068176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,64,power_law_1.01,0.029420799016952513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,64,power_law_1.01,0.031206399202346802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,64,power_law_1.01,0.03246079981327057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,64,power_law_1.01,0.03367680013179779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,64,power_law_1.01,0.03406080007553101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,64,power_law_1.01,0.03437440097332001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,64,power_law_1.01,0.03516159951686859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,64,power_law_1.01,0.035436800122261046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,64,power_law_1.01,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,64,power_law_1.01,0.03654400110244751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,64,power_law_1.01,0.03713279962539673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,64,power_law_1.01,0.03896960020065308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,64,power_law_1.01,0.04529280066490173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,64,power_law_1.01,0.04601599872112274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,64,power_law_1.01,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,64,power_law_1.01,0.0551360011100769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,64,power_law_1.01,0.06602879762649536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,64,power_law_1.01,0.07917439937591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,64,power_law_1.01,0.09845119714736938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,64,power_law_1.01,0.12572159767150878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,64,power_law_1.01,0.1670464038848877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,64,power_law_1.01,0.21293439865112304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,64,power_law_1.01,0.3143359899520874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,64,power_law_1.01,0.47548799514770507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,64,power_law_1.01,0.5775680065155029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,32,balanced,0.03450666616360346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,32,balanced,0.0329066663980484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,32,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,32,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,32,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,32,balanced,0.0378560001651446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,32,balanced,0.037952000896135964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,32,balanced,0.039893334110577904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,32,balanced,0.03977599988381068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,32,balanced,0.03977599988381068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,32,balanced,0.039834665755430855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,32,balanced,0.04317333300908407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,32,balanced,0.04192533095677694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,32,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,32,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,32,balanced,0.05004799862702688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,32,balanced,0.05042133231957754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,32,balanced,0.054933334390322365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,32,balanced,0.058320000767707825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,32,balanced,0.06539733211199443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,32,balanced,0.07066666583220164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,32,balanced,0.09640000263849895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,32,balanced,0.10744532942771912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,32,balanced,0.13456533352533975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,32,balanced,0.15928000211715698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,32,balanced,0.22203733523686728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,32,balanced,0.27848533789316815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,16,power_law_1.01,0.026700800657272337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,16,power_law_1.01,0.025017601251602174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,16,power_law_1.01,0.024646399915218352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,16,power_law_1.01,0.025043201446533204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,16,power_law_1.01,0.025113600492477416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,16,power_law_1.01,0.02744320034980774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,16,power_law_1.01,0.027993598580360414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,16,power_law_1.01,0.02917119860649109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,16,power_law_1.01,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,16,power_law_1.01,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,16,power_law_1.01,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,16,power_law_1.01,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,16,power_law_1.01,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,16,power_law_1.01,0.03473280072212219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,16,power_law_1.01,0.03787519931793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,16,power_law_1.01,0.03893760144710541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,16,power_law_1.01,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,16,power_law_1.01,0.04831359982490539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,16,power_law_1.01,0.04859519898891449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,16,power_law_1.01,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,16,power_law_1.01,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,16,power_law_1.01,0.07577599883079529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,16,power_law_1.01,0.0874239981174469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,16,power_law_1.01,0.11726080179214478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,16,power_law_1.01,0.14023040533065795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,16,power_law_1.01,0.17976319789886475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,16,power_law_1.01,0.23006720542907716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,32,balanced,0.03370666752258936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,32,balanced,0.03536533315976461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,32,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,32,balanced,0.03733866661787033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,32,balanced,0.03737599899371465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,32,balanced,0.037290667494138084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,32,balanced,0.03873066604137421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,32,balanced,0.039349332451820374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,32,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,32,balanced,0.038762666285037994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,32,balanced,0.03931200007597605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,32,balanced,0.04029333343108495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,32,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,32,balanced,0.04171733558177948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,32,balanced,0.0440533310174942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,32,balanced,0.0454773356517156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,32,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,32,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,32,balanced,0.048986668388048805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,32,balanced,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,32,balanced,0.05808533231417338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,32,balanced,0.07023466626803081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,32,balanced,0.07601599891980489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,32,balanced,0.09525332848230998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,32,balanced,0.11693333586057027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,32,balanced,0.15096533298492432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,32,balanced,0.18831467628479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,8,power_law_1.2,0.027430400252342224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,8,power_law_1.2,0.02457599937915802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,8,power_law_1.2,0.024307200312614442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,8,power_law_1.2,0.025043201446533204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,8,power_law_1.2,0.027027198672294618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,8,power_law_1.2,0.02834559977054596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,8,power_law_1.2,0.028147199749946596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,8,power_law_1.2,0.03141759932041168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,8,power_law_1.2,0.03172479867935181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,8,power_law_1.2,0.03235200047492981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,8,power_law_1.2,0.03363839983940124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,8,power_law_1.2,0.03470079898834229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,8,power_law_1.2,0.03521920144557953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,8,power_law_1.2,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,8,power_law_1.2,0.04031359851360321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,8,power_law_1.2,0.04136959910392761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,8,power_law_1.2,0.04679679870605469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,8,power_law_1.2,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,8,power_law_1.2,0.0579584002494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,8,power_law_1.2,0.06349440217018128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,8,power_law_1.2,0.07141119837760926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,8,power_law_1.2,0.09259520173072815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,8,power_law_1.2,0.11514879465103149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,8,power_law_1.2,0.13794560432434083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,8,power_law_1.2,0.2183680057525635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,2,power_law_1.2,0.09224960207939148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,2,power_law_1.2,0.11491199731826782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,2,power_law_1.2,0.14773759841918946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,2,power_law_1.2,0.18730239868164061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,2,power_law_1.2,0.22770559787750244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,2,power_law_1.2,0.3799871921539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,2,power_law_1.2,0.3675519943237305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,2,power_law_1.2,0.379532790184021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,2,power_law_1.2,0.40687999725341795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,2,power_law_1.2,0.4144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,2,power_law_1.2,0.4180352210998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,2,power_law_1.2,0.44376959800720217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,2,power_law_1.2,0.4483071804046631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,2,power_law_1.2,0.4705215930938721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,2,power_law_1.2,0.4906496047973633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,2,power_law_1.2,0.5094336032867431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,2,power_law_1.2,0.5174464225769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,2,power_law_1.2,0.5834368228912353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,2,power_law_1.2,0.6354112148284912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,2,power_law_1.2,0.7581503868103028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,2,power_law_1.2,0.7975679874420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,2,power_law_1.2,1.0224512100219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,2,power_law_1.2,1.1385024070739747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,2,power_law_1.2,1.5488767623901367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,2,power_law_1.2,1.9505983352661134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,2,power_law_1.2,2.658297538757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,2,power_law_1.2,3.1823551177978517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,8,power_law_1.2,0.2548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,1,power_law_1.01,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,1,power_law_1.01,0.05381760001182556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,1,power_law_1.01,0.05930240154266357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,1,power_law_1.01,0.07240960001945496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,1,power_law_1.01,0.08842880129814149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,1,power_law_1.01,0.10078719854354859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,1,power_law_1.01,0.12041599750518799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,1,power_law_1.01,0.12413439750671387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,1,power_law_1.01,0.12866560220718384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,1,power_law_1.01,0.13071999549865723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,1,power_law_1.01,0.13809280395507811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,1,power_law_1.01,0.14720640182495118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,1,power_law_1.01,0.15240960121154784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,1,power_law_1.01,0.16187520027160646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,1,power_law_1.01,0.17185920476913452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,1,power_law_1.01,0.18236160278320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,1,power_law_1.01,0.18428800106048585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,1,power_law_1.01,0.22144639492034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,1,power_law_1.01,0.24724481105804444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,1,power_law_1.01,0.31395840644836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,1,power_law_1.01,0.3699520111083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,1,power_law_1.01,0.4917952060699463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,1,power_law_1.01,0.611353588104248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,1,power_law_1.01,0.8451392173767089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,1,power_law_1.01,1.0783040046691894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,1,power_law_1.01,1.5570816040039062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,1,power_law_1.01,2.0262144088745115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,8,power_law_1.2,0.3107264041900635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,8,power_law_1.01,0.03601279854774475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,8,power_law_1.01,0.03941119909286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,8,power_law_1.01,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,8,power_law_1.01,0.04264959990978241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,8,power_law_1.01,0.04344959855079651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,8,power_law_1.01,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,8,power_law_1.01,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,8,power_law_1.01,0.046035200357437134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,8,power_law_1.01,0.045126399397850035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,8,power_law_1.01,0.04542720019817352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,8,power_law_1.01,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,8,power_law_1.01,0.0474047988653183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,8,power_law_1.01,0.04878720045089722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,8,power_law_1.01,0.05237119793891907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,8,power_law_1.01,0.05603839755058289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,8,power_law_1.01,0.057683199644088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,8,power_law_1.01,0.06179839968681335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,8,power_law_1.01,0.07114239931106567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,8,power_law_1.01,0.08020480275154114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,8,power_law_1.01,0.09747840166091919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,8,power_law_1.01,0.10799360275268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,8,power_law_1.01,0.14536319971084594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,8,power_law_1.01,0.16628479957580566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,8,power_law_1.01,0.24208641052246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,8,power_law_1.01,0.32858240604400635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,8,power_law_1.01,0.4672255992889404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,8,power_law_1.01,0.6610623836517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,4,balanced,0.06622399886449178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,4,balanced,0.06434133152167003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,4,balanced,0.08356266220410664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,4,balanced,0.0869760016600291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,4,balanced,0.0885599950949351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,4,balanced,0.08851200342178345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,4,balanced,0.08886399865150452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,4,balanced,0.09090666969617207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,4,balanced,0.09000000357627869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,4,balanced,0.09029866258303325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,4,balanced,0.09018666545550029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,4,balanced,0.09973866740862529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,4,balanced,0.09884267052014668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,4,balanced,0.09975999593734741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,4,balanced,0.1011199951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,4,balanced,0.1033066709836324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,4,balanced,0.11114133397738139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,4,balanced,0.13301866253217062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,4,balanced,0.13914133111635843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,4,balanced,0.1874879995981852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,4,balanced,0.21544533967971802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,4,balanced,0.2985866665840149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,4,balanced,0.3705973227818807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,4,balanced,0.5214986801147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,4,balanced,0.6623680194218954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,4,balanced,0.9901866912841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,4,balanced,1.293392022450765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,1,power_law_1.2,0.0798143982887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,1,power_law_1.2,0.10584959983825684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,1,power_law_1.2,0.15003520250320435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,1,power_law_1.2,0.25096960067749025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,1,power_law_1.2,0.37011840343475344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,1,power_law_1.2,0.511187219619751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,1,power_law_1.2,0.6046080112457275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,1,power_law_1.2,0.8170175552368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,1,power_law_1.2,0.8521087646484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,1,power_law_1.2,0.8790783882141113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,1,power_law_1.2,0.9299072265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,1,power_law_1.2,0.9738944053649903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,1,power_law_1.2,1.0100607872009277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,1,power_law_1.2,1.0394368171691895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,32,power_law_1.01,0.08477439880371093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,32,power_law_1.01,0.0809216022491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,1,power_law_1.2,1.0773823738098145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,32,power_law_1.01,0.07845759987831116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,1,power_law_1.2,1.1496895790100097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,32,power_law_1.01,0.08352640271186829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,32,power_law_1.01,0.08180480003356934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,1,power_law_1.2,1.1671680450439452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,32,power_law_1.01,0.08321920037269592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,32,power_law_1.01,0.08524159789085388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,1,power_law_1.2,1.2932031631469727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,32,power_law_1.01,0.08754559755325317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,32,power_law_1.01,0.08798720240592957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,1,power_law_1.2,1.4004223823547364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,32,power_law_1.01,0.08817279934883118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,32,power_law_1.01,0.08740479946136474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,1,power_law_1.2,1.6196096420288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,32,power_law_1.01,0.09073920249938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,4,power_law_1.01,0.04751999974250794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,32,power_law_1.01,0.09214720129966736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,1,power_law_1.2,1.674118423461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,4,power_law_1.01,0.052697598934173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,32,power_law_1.01,0.09317759871482849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,1,power_law_1.2,2.0299583435058595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,4,power_law_1.01,0.05939199924468994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,4,power_law_1.01,0.06241919994354248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,1,power_law_1.2,2.128940773010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,4,power_law_1.01,0.07156479954719544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,32,power_law_1.01,0.10531840324401856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,4,power_law_1.01,0.08052480220794678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,32,power_law_1.01,0.10699520111083985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,1,power_law_1.2,2.6522111892700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,4,power_law_1.01,0.08359040021896362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,4,power_law_1.01,0.09114239811897278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,4,power_law_1.01,0.09412479996681214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,1,power_law_1.2,3.254374313354492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,4,power_law_1.01,0.09571840167045594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,4,power_law_1.01,0.0993664026260376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,32,power_law_1.01,0.10680960416793824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,4,power_law_1.01,0.1074560046195984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,32,power_law_1.01,0.12316800355911255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,1,power_law_1.2,4.181171035766601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,4,power_law_1.01,0.10771199464797973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,32,power_law_1.01,0.13114880323410033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,4,power_law_1.01,0.11198079586029053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,32,power_law_1.01,0.14725120067596437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,4,power_law_1.01,0.12543359994888306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,32,power_law_1.01,0.1808575987815857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,32,power_law_1.01,0.23071999549865724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,1,power_law_1.2,5.27955207824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,32,power_law_1.01,0.27217280864715576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,32,power_law_1.01,0.39214720726013186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,4,power_law_1.01,0.1282304048538208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,32,power_law_1.01,0.47986559867858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,4,power_law_1.01,0.14881279468536376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,32,power_law_1.01,0.7292096138000488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,32,power_law_1.01,0.9179519653320313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,4,power_law_1.01,0.17244160175323486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,4,power_law_1.01,0.1761664032936096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,4,power_law_1.01,0.2192768096923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,4,power_law_1.01,0.25478401184082033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,4,power_law_1.01,0.3389440059661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,4,power_law_1.01,0.4256703853607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,4,power_law_1.01,0.5696127891540528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,4,power_law_1.01,0.7843264102935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,4,power_law_1.01,1.0982144355773926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,4,power_law_1.01,1.4209024429321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,4,power_law_1.01,0.1516543984413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,4,power_law_1.01,0.2460479974746704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,4,power_law_1.01,0.472057580947876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,4,power_law_1.01,0.4039487838745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,4,power_law_1.01,0.6462912082672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,4,power_law_1.01,0.7943168163299561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,4,power_law_1.01,0.8249792098999024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,4,power_law_1.01,0.8169343948364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,4,power_law_1.01,1.012723159790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,4,power_law_1.01,1.1080960273742675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,4,power_law_1.01,1.0222720146179198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,4,power_law_1.01,0.9307007789611816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,4,power_law_1.01,1.0488575935363769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,4,power_law_1.01,0.9907072067260743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,4,power_law_1.01,1.093183994293213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,4,power_law_1.01,1.0701760292053222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,4,power_law_1.01,1.1124032020568848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,4,power_law_1.01,1.2099455833435058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,4,power_law_1.01,1.2677568435668944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,4,power_law_1.01,1.4353599548339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,4,power_law_1.01,1.4680512428283692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,4,power_law_1.01,1.7079423904418944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,4,power_law_1.01,1.7166719436645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,4,power_law_1.01,1.998201560974121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,4,power_law_1.01,2.4054143905639647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,4,power_law_1.01,2.9316032409667967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,4,power_law_1.01,3.97204475402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,32,1,power_law_1.01,0.05237119793891907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,32,1,power_law_1.01,0.05598719716072083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,32,1,power_law_1.01,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,32,1,power_law_1.01,0.08501759767532349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,32,1,power_law_1.01,0.11180800199508667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,32,1,power_law_1.01,0.12942080497741698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,32,1,power_law_1.01,0.1564288020133972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,32,1,power_law_1.01,0.16369919776916503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,32,1,power_law_1.01,0.1702015995979309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,32,1,power_law_1.01,0.17452160120010377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,32,1,power_law_1.01,0.18173439502716066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,32,1,power_law_1.01,0.19127680063247682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,32,1,power_law_1.01,0.19930880069732665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,32,1,power_law_1.01,0.20823678970336915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,32,1,power_law_1.01,0.21294078826904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,32,balanced,0.08186666667461395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,32,balanced,0.08148266871770223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,32,balanced,0.08270399769147237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,32,balanced,0.09971200426419576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,32,balanced,0.12158399820327759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,32,balanced,0.20164799690246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,32,balanced,0.26470400889714557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,32,balanced,0.26393065849939984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,32,balanced,0.26588799556096393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,32,balanced,0.25620800256729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,32,balanced,0.2648586630821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,32,balanced,0.2547626694043477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,32,balanced,0.26030399401982623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,32,balanced,0.2789333264033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,32,1,power_law_1.01,0.22103679180145264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,32,balanced,0.2688639958699544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,32,balanced,0.2633919914563497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,32,1,power_law_1.01,0.23235199451446534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,32,1,power_law_1.01,0.3127680063247681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,32,1,power_law_1.01,0.334879994392395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,32,1,power_law_1.01,0.40974721908569334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,32,1,power_law_1.01,0.46005759239196775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,32,1,power_law_1.01,0.6058559894561768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,32,balanced,0.28045332431793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,32,balanced,0.2749600013097127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,32,balanced,0.28707200288772583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,32,1,power_law_1.01,0.7471295833587647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,32,balanced,0.3014880021413167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,32,balanced,0.30188266436258954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,32,balanced,0.3269173304239909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,32,1,power_law_1.01,1.0301247596740724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,32,1,power_law_1.01,1.292563247680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,32,1,power_law_1.01,1.8295743942260743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,32,balanced,0.3500426610310872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,32,balanced,0.4110986789067586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,32,balanced,0.5928320089975992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,32,balanced,0.6644853353500366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,32,balanced,0.8782080014546713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,32,1,power_law_1.01,2.388614463806152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,8,power_law_1.2,0.08419839739799499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,8,power_law_1.2,0.08507519960403442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,8,power_law_1.2,0.0843455970287323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,8,power_law_1.2,0.0840831995010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,8,power_law_1.2,0.08391680121421814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,8,power_law_1.2,0.08807680010795593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,8,power_law_1.2,0.0981440007686615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,8,power_law_1.2,0.09618560075759888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,8,power_law_1.2,0.10168319940567017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,8,power_law_1.2,0.10647039413452149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,8,power_law_1.2,0.10061440467834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,8,power_law_1.2,0.11203839778900146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,8,power_law_1.2,0.11470719575881957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,8,power_law_1.2,0.11850240230560302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,8,power_law_1.2,0.13543039560317993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,8,power_law_1.2,0.14114559888839723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,8,power_law_1.2,0.1644863963127136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,8,power_law_1.2,0.19831680059432982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,8,power_law_1.2,0.26791679859161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,8,power_law_1.2,0.40514559745788575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,8,power_law_1.2,0.465715217590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,8,power_law_1.2,0.6476863861083985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,8,power_law_1.2,0.8619392395019532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,8,power_law_1.2,1.202899169921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,8,power_law_1.2,1.447596836090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,8,power_law_1.2,2.6824960708618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,4,balanced,0.02160533269246419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,4,balanced,0.021087999145189922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,4,balanced,0.021040000021457672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,4,balanced,0.02107733239730199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,4,balanced,0.021514666577180225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,4,balanced,0.021168000996112823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,4,balanced,0.025439999997615814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,4,balanced,0.025120000044504803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,64,power_law_1.2,0.11351679563522339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,4,balanced,0.05041066805521647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,4,balanced,0.049914668003718056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,4,balanced,0.049786667029062905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,64,power_law_1.2,0.11128959655761719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,64,power_law_1.2,0.1106943964958191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,64,power_law_1.2,0.1085055947303772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,64,power_law_1.2,0.10311679840087891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,64,power_law_1.2,0.10292479991912842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,4,balanced,0.04317333300908407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,4,balanced,0.041573333243529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,4,balanced,0.038592000802357994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,4,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,4,balanced,0.03941333293914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,4,balanced,0.043712000052134194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,4,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,4,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,4,balanced,0.08297599852085114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,4,balanced,0.1042133371035258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,4,balanced,0.1390720009803772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,4,balanced,0.1690773367881775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,64,power_law_1.2,0.11724799871444702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,4,balanced,0.2401706576347351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,4,balanced,0.3040213386217753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,64,power_law_1.2,0.11501439809799194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,4,balanced,0.4394773244857788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,64,power_law_1.2,0.11648000478744507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,4,balanced,0.5754986604054769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,64,power_law_1.2,0.11204479932785034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,64,power_law_1.2,0.11404160261154175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,64,power_law_1.2,0.10928640365600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,64,power_law_1.2,0.11772799491882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,64,power_law_1.2,0.11612800359725953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,64,power_law_1.2,0.11677440404891967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,64,power_law_1.2,0.1174015998840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,64,power_law_1.2,0.1265920042991638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,64,power_law_1.2,0.14936959743499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,64,power_law_1.2,0.1636672019958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,64,power_law_1.2,0.1978368043899536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,64,power_law_1.2,0.2255295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,64,power_law_1.2,0.29517440795898436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,64,power_law_1.2,0.36303999423980715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,64,power_law_1.2,0.45678081512451174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,64,power_law_1.2,0.6613887786865235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,64,power_law_1.2,1.0503168106079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,64,power_law_1.2,1.315334415435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,8,power_law_1.2,3.4085247039794924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,8,power_law_1.2,0.04188160002231598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,8,power_law_1.2,0.0491456001996994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,8,power_law_1.2,0.047865599393844604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,8,power_law_1.2,0.04896639883518219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,8,power_law_1.2,0.04959999918937683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,8,power_law_1.2,0.055174398422241214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,8,power_law_1.2,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,8,power_law_1.2,0.04766719937324524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,8,power_law_1.2,0.052908802032470705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,8,power_law_1.2,0.05135999917984009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,8,power_law_1.2,0.050496000051498416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,8,power_law_1.2,0.055129599571228025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,8,power_law_1.2,0.058278399705886844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,8,power_law_1.2,0.05857920050621033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,8,power_law_1.2,0.06733440160751343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,8,power_law_1.2,0.07134720087051391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,8,power_law_1.2,0.0750976026058197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,8,power_law_1.2,0.09330559968948364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,8,power_law_1.2,0.110316801071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,8,power_law_1.2,0.13159040212631226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,8,power_law_1.2,0.1560639977455139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,8,power_law_1.2,0.21113600730895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,8,power_law_1.2,0.2676287889480591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,4,power_law_1.2,0.07542399764060974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,8,power_law_1.2,0.3832000017166138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,4,power_law_1.2,0.09781119823455811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,4,power_law_1.2,0.12492159605026246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,4,power_law_1.2,0.14788479804992677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,4,power_law_1.2,0.18792959451675414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,4,power_law_1.2,0.21095681190490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,4,power_law_1.2,0.24467198848724364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,4,power_law_1.2,0.3153408050537109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,4,power_law_1.2,0.27365760803222655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,4,power_law_1.2,0.28932480812072753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,4,power_law_1.2,0.26184959411621095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,4,power_law_1.2,0.29132800102233886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,4,power_law_1.2,0.3148672103881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,4,power_law_1.2,0.29268479347229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,4,power_law_1.2,0.3245311975479126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,4,power_law_1.2,0.3266304016113281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,4,power_law_1.2,0.35324161052703856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,4,power_law_1.2,0.37687039375305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,4,power_law_1.2,0.41212158203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,4,power_law_1.2,0.47791361808776855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,4,power_law_1.2,0.49686398506164553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,4,power_law_1.2,0.6611584186553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,4,power_law_1.2,0.7239232063293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,4,power_law_1.2,0.9561344146728515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,2,power_law_1.01,0.01666560024023056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,2,power_law_1.01,0.016550399363040924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,2,power_law_1.01,0.016672000288963318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,4,power_law_1.2,1.047052764892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,2,power_law_1.01,0.017664000391960144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,2,power_law_1.01,0.01820800006389618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,2,power_law_1.01,0.018617600202560425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,8,power_law_1.2,0.49370880126953126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,2,power_law_1.01,0.019392000138759614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,2,power_law_1.01,0.01902720034122467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,4,power_law_1.2,1.4004927635192872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,2,power_law_1.01,0.020582400262355804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,2,power_law_1.01,0.020300799608230592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,2,power_law_1.01,0.02126079946756363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,4,power_law_1.2,1.735968017578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,2,power_law_1.01,0.02306559979915619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,2,power_law_1.01,0.027142399549484254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,2,power_law_1.01,0.025088000297546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,2,power_law_1.01,0.026335999369621277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,2,power_law_1.01,0.028774398565292358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,2,power_law_1.01,0.03351680040359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,2,power_law_1.01,0.03941119909286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.01,0.04120959937572479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,128,power_law_1.01,0.03808639943599701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,128,power_law_1.01,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,128,power_law_1.01,0.037651199102401736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.01,0.053011202812194826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,128,power_law_1.01,0.04113920032978058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,128,power_law_1.01,0.04142079949378967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,128,power_law_1.01,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,128,power_law_1.01,0.041331198811531064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,128,power_law_1.01,0.04119040071964264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,128,power_law_1.01,0.04184960126876831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,128,power_law_1.01,0.04140160083770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.01,0.059468799829483034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,128,power_law_1.01,0.04161919951438904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,128,power_law_1.01,0.04184960126876831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,128,power_law_1.01,0.04287999868392944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,128,power_law_1.01,0.042931199073791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,128,power_law_1.01,0.04545280039310455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,128,power_law_1.01,0.04785920083522797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,128,power_law_1.01,0.047417598962783816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,128,power_law_1.01,0.05381119847297668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,128,power_law_1.01,0.06054400205612183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,128,power_law_1.01,0.07022079825401306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,128,power_law_1.01,0.08670079708099365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.01,0.07839999794960022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,128,power_law_1.01,0.12002559900283813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,128,power_law_1.01,0.14231040477752685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,128,power_law_1.01,0.19964799880981446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,128,power_law_1.01,0.2324160099029541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,128,power_law_1.01,0.36537599563598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,128,power_law_1.01,0.48888959884643557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.01,0.09526399970054626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,8,power_law_1.2,0.7282368183135987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.01,0.12491519451141357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,8,power_law_1.2,0.8438207626342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.01,0.15845760107040405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.01,0.22247679233551027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.01,0.2861759901046753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,2,power_law_1.01,0.07314559817314148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,2,power_law_1.01,0.07848320007324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,2,power_law_1.01,0.09589120149612426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,2,power_law_1.01,0.11185280084609986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,2,power_law_1.01,0.13889919519424437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,2,power_law_1.01,0.1687872052192688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,2,power_law_1.01,0.1736191987991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,2,power_law_1.01,0.2069632053375244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,2,power_law_1.01,0.21831679344177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,2,power_law_1.01,0.21724159717559816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,2,power_law_1.01,0.22308480739593506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,2,power_law_1.01,0.221343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,4,power_law_1.01,0.06970239877700805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,4,power_law_1.01,0.0948415994644165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,4,power_law_1.01,0.11745920181274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,4,power_law_1.01,0.12853120565414428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,4,power_law_1.01,0.1574399948120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,4,power_law_1.01,0.17633919715881347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,4,power_law_1.01,0.2198784112930298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,4,power_law_1.01,0.22773120403289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,4,power_law_1.01,0.2095168113708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,4,power_law_1.01,0.21587839126586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,2,power_law_1.01,0.2321727991104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,4,power_law_1.01,0.22741758823394775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,2,power_law_1.01,0.23873279094696045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,4,power_law_1.01,0.23412480354309081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,4,power_law_1.01,0.23708798885345458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,4,power_law_1.01,0.2439807891845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,4,power_law_1.01,0.2557823896408081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,4,power_law_1.01,0.25715839862823486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,4,power_law_1.01,0.27898240089416504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,4,power_law_1.01,0.3216831922531128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,2,power_law_1.01,0.24786560535430907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,1,power_law_1.01,0.0748416006565094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,2,power_law_1.01,0.25418241024017335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,1,power_law_1.01,0.09086719751358033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,2,power_law_1.01,0.27143039703369143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,1,power_law_1.01,0.12564480304718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,2,power_law_1.01,0.2920768022537231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,2,power_law_1.01,0.325766396522522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,2,power_law_1.01,0.37655038833618165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,4,power_law_1.01,0.32919039726257326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,2,power_law_1.01,0.4347775936126709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,4,power_law_1.01,0.3911936044692993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,2,power_law_1.01,0.5517632007598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,4,power_law_1.01,0.4164735794067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,1,power_law_1.01,0.1525056004524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,4,power_law_1.01,0.5206016063690185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,1,power_law_1.01,0.18662400245666505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,1,power_law_1.01,0.25148160457611085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,4,power_law_1.01,0.6066368103027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,1,power_law_1.01,0.2645632028579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,1,power_law_1.01,0.2782912015914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,4,power_law_1.01,0.8243328094482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,1,power_law_1.01,0.2734015941619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,1,power_law_1.01,0.2735487937927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,2,power_law_1.01,0.6542272090911865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,4,power_law_1.01,1.0118911743164063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,1,power_law_1.01,0.2797951936721802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,2,power_law_1.01,0.8922304153442383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,1,power_law_1.01,0.2856192111968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,4,power_law_1.01,1.3563008308410645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,1,power_law_1.01,0.2858112096786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,2,power_law_1.01,1.1867263793945313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,1,power_law_1.01,0.29109759330749513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,4,power_law_1.01,1.8322944641113281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,1,power_law_1.01,0.3007359981536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,1,power_law_1.01,0.3020224094390869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,2,power_law_1.01,1.6281856536865233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,1,power_law_1.01,0.3200063943862915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,1,power_law_1.01,0.3686784029006958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,2,power_law_1.01,2.049260711669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.01,0.406496000289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.01,0.4985983848571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.01,0.5760255813598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.01,0.7432640075683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.01,0.9057344436645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.01,1.228223991394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.01,1.550387191772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.01,2.191801643371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.01,2.8355648040771486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,8,balanced,0.025978667040665943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,8,balanced,0.027615999182065327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,8,balanced,0.025546667476495106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,8,balanced,0.03291733314593633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,8,balanced,0.0499946673711141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,8,balanced,0.052144000927607216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,8,balanced,0.05266133447488149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,8,balanced,0.05236266553401947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,8,balanced,0.05219733218352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,8,balanced,0.05303466816743215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,8,balanced,0.05373866856098175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,8,balanced,0.05375466744105021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,8,balanced,0.05440000196297964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,8,balanced,0.059205333391825356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,8,balanced,0.06152533491452535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,8,balanced,0.062277331948280334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,8,balanced,0.06454933186372121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,8,balanced,0.06672533353169759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,8,balanced,0.07668266693751018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,8,balanced,0.08187733093897502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,8,balanced,0.08137600123882294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,8,balanced,0.09277333815892537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,8,balanced,0.10498666763305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,8,balanced,0.13276799519856772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,8,balanced,0.16337600350379944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,8,balanced,0.21029865741729736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,8,balanced,0.24804800748825073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,32,power_law_1.01,0.02677760124206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,32,power_law_1.01,0.03170560002326965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,32,power_law_1.01,0.025158399343490602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,32,power_law_1.01,0.02444159984588623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,32,power_law_1.01,0.026848000288009644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,32,power_law_1.01,0.027622398734092713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,32,power_law_1.01,0.030956798791885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,32,power_law_1.01,0.03219200074672699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,32,power_law_1.01,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,32,power_law_1.01,0.036735999584198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,32,power_law_1.01,0.04108799993991852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,32,power_law_1.01,0.039929598569869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,32,power_law_1.01,0.04552960097789764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,32,power_law_1.01,0.06820480227470398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,32,power_law_1.01,0.06151679754257202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,32,power_law_1.01,0.07674880027770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,32,power_law_1.01,0.08459519743919372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,128,balanced,0.01977066695690155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,128,balanced,0.019391999890406925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,32,power_law_1.01,0.09119359850883484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,32,power_law_1.01,0.1110975980758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,32,power_law_1.01,0.14055039882659912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,32,power_law_1.01,0.09314560294151306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,32,power_law_1.01,0.12227840423583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,32,power_law_1.01,0.15210239887237548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,128,balanced,0.019285333653291065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,32,power_law_1.01,0.1970944046974182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,32,power_law_1.01,0.25395200252532957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,32,power_law_1.01,0.37071359157562256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,32,power_law_1.01,0.4765439987182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,128,balanced,0.019776000330845516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,128,balanced,0.02347733328739802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,128,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,128,balanced,0.03152533372243246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,128,balanced,0.031498665610949196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,128,balanced,0.030586667358875275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,128,balanced,0.03165333221356074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,128,balanced,0.03158933420976003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,128,balanced,0.03148799886306127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,128,balanced,0.03156266609827677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,128,balanced,0.03164266546567281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,128,balanced,0.03577066709597906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,128,balanced,0.034815999368826546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,128,balanced,0.03398933261632919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,128,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,128,balanced,0.04171733558177948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,128,balanced,0.04543999830881754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,128,balanced,0.05621333420276642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,128,balanced,0.06770666440327962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,128,balanced,0.08302400012811025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,128,balanced,0.11149866382280986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,128,balanced,0.14004799723625183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,128,balanced,0.19592533508936563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,128,balanced,0.25332266092300415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,64,power_law_1.2,0.03415679931640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,64,power_law_1.2,0.032416000962257385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,64,power_law_1.2,0.0315200001001358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,64,power_law_1.2,0.03266560137271881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,64,power_law_1.2,0.032332798838615416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,64,power_law_1.2,0.033548799157142636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,64,power_law_1.2,0.033958399295806886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,64,power_law_1.2,0.034668800234794614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,64,power_law_1.2,0.034841600060462954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,64,power_law_1.2,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,64,power_law_1.2,0.038073599338531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,64,power_law_1.2,0.03953920006752014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,64,power_law_1.2,0.0419840008020401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,64,power_law_1.2,0.04322560131549835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,64,power_law_1.2,0.04812160134315491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,64,power_law_1.2,0.04989440143108368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,64,power_law_1.2,0.051097601652145386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,64,power_law_1.2,0.0636672019958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,64,power_law_1.2,0.07494400143623352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,64,power_law_1.2,0.09444479942321778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,64,power_law_1.2,0.11630079746246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,64,power_law_1.2,0.15473920106887817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,64,power_law_1.2,0.1815808057785034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,64,power_law_1.2,0.2619136095046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,64,power_law_1.2,0.304531192779541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,64,power_law_1.2,0.48168320655822755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,64,power_law_1.2,0.6203904151916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,1,power_law_1.2,0.10503040552139283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,1,power_law_1.2,0.11335680484771729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,1,power_law_1.2,0.12032639980316162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,1,power_law_1.2,0.14435839653015137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,1,power_law_1.2,0.1770624041557312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,1,power_law_1.2,0.2084736108779907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,1,power_law_1.2,0.2310271978378296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,1,power_law_1.2,0.27626240253448486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,1,power_law_1.2,0.28564479351043703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,1,power_law_1.2,0.29675519466400146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,1,power_law_1.2,0.3036799907684326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,1,power_law_1.2,0.32245121002197263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,1,power_law_1.2,0.3268671989440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,1,power_law_1.2,0.34182400703430177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,1,power_law_1.2,0.3573568105697632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,1,power_law_1.2,0.36627199649810793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,1,power_law_1.2,0.38474240303039553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,1,power_law_1.2,0.44972801208496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.2,0.490777587890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.2,0.6006400108337402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.2,0.6866687774658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.2,0.8890560150146485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.2,1.0823871612548828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.2,1.4751872062683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,1,power_law_1.2,1.8659008026123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,1,power_law_1.2,2.657222366333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,1,power_law_1.2,3.4208065032958985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,64,balanced,0.06281599899133046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,64,balanced,0.06201600035031637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,64,balanced,0.0633653352657954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,64,balanced,0.06650666892528534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,64,balanced,0.0687253326177597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,64,balanced,0.0949173370997111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,64,balanced,0.08916266759236653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,64,balanced,0.08876267075538635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,64,balanced,0.0946666697661082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,64,balanced,0.09548266728719075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,64,balanced,0.09271466732025146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,64,balanced,0.0995146632194519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,64,balanced,0.09923199812571208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,64,balanced,0.10095999638239543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,64,balanced,0.10636267066001892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,64,balanced,0.11077333490053813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,64,balanced,0.10806933045387268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,64,balanced,0.12171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,64,balanced,0.12686399618784586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,64,balanced,0.14436800281206766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,64,balanced,0.16099733114242554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,64,balanced,0.19965332746505737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,64,balanced,0.23482666412989298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,64,balanced,0.322160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,64,balanced,0.38896000385284424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,64,balanced,0.5410879850387573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,64,balanced,0.6940586566925049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,2,balanced,0.035775999228159584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,2,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,2,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,2,balanced,0.058229332168896995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,2,balanced,0.0855519970258077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,2,balanced,0.08458133538564046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,2,balanced,0.08481599887212117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,2,balanced,0.08657067020734151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,2,balanced,0.08819199601809184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,2,balanced,0.0879253347714742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,2,balanced,0.08933867017428081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,2,balanced,0.09053333600362141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,2,balanced,0.09431999921798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,2,balanced,0.09702933828035991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,2,balanced,0.10269866387049358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,2,balanced,0.10641066233317058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,2,balanced,0.11084266503651936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,2,balanced,0.12311999996503194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,2,balanced,0.13396267096201578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,2,balanced,0.16089600324630737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,2,balanced,0.18134399255116782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,2,balanced,0.23849600553512573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,2,balanced,0.28587732712427777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,2,balanced,0.4076106548309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,2,balanced,0.5054346720377604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,2,balanced,0.7325546741485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,2,balanced,0.9600479602813721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,8,balanced,0.05211733281612396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,8,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,8,balanced,0.06402133405208588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,8,balanced,0.0821919987599055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,8,balanced,0.10572800040245056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,8,balanced,0.15364799896876016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,8,balanced,0.14083199699719748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,8,balanced,0.14636266231536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,8,balanced,0.13784000277519226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,8,balanced,0.13486933708190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,8,balanced,0.1362613340218862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,8,balanced,0.14145599802335104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,8,balanced,0.1397279997666677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,8,balanced,0.13730667034784952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,8,balanced,0.14315199851989746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,8,balanced,0.15362133582433066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,8,balanced,0.1588159998257955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,8,balanced,0.1507306694984436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,8,balanced,0.16715733210245767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,8,balanced,0.17083734273910522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,8,balanced,0.18263999621073404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,8,balanced,0.21036267280578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,8,balanced,0.23599465688069662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,8,balanced,0.35197333494822186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,8,balanced,0.40565331776936847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,8,balanced,0.5735093355178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,8,balanced,0.7213919957478842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,16,power_law_1.2,0.0615231990814209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,16,power_law_1.2,0.07883520126342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,16,power_law_1.2,0.07384960055351257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,16,power_law_1.2,0.07652480006217957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,16,power_law_1.2,0.07564799785614014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,16,power_law_1.2,0.06647679805755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,16,power_law_1.2,0.07366399765014649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,16,power_law_1.2,0.0757311999797821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,16,power_law_1.2,0.07743359804153442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,16,power_law_1.2,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,16,power_law_1.2,0.07864959836006165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,16,power_law_1.2,0.07740799784660339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,16,power_law_1.2,0.08389120101928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,16,power_law_1.2,0.08284159898757934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,16,power_law_1.2,0.0917248010635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,16,power_law_1.2,0.09434880018234253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,16,power_law_1.2,0.10188800096511841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,16,power_law_1.2,0.11681280136108399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,16,power_law_1.2,0.11987839937210083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,16,power_law_1.2,0.14949120283126832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,16,power_law_1.2,0.16011519432067872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,16,power_law_1.2,0.19611519575119019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,16,power_law_1.2,0.23829760551452636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,16,power_law_1.2,0.33225600719451903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,16,power_law_1.2,0.45781121253967283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,16,power_law_1.2,0.5908736228942871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,16,power_law_1.2,0.7116543769836425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,4,balanced,0.025605333348115284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,4,balanced,0.02977599948644638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,4,balanced,0.04494399825731913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,4,balanced,0.06038933495680491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,4,balanced,0.08593599994977315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,4,balanced,0.14101333419481912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,4,balanced,0.14863999684651694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,4,balanced,0.15294933319091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,4,balanced,0.15979199608167013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,4,balanced,0.1669386625289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,4,balanced,0.15979199608167013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,4,balanced,0.16880534092585245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,4,balanced,0.17737066745758057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,4,balanced,0.1771679917971293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,4,balanced,0.18293867508570352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,4,balanced,0.18914133310317993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,4,balanced,0.23333332935969034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,4,balanced,0.24029332399368286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,4,balanced,0.2461013396581014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,4,balanced,0.2641706665356954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,4,balanced,0.368064006169637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,4,balanced,0.4126400152842204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,4,balanced,0.46144533157348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,4,balanced,0.5448960065841675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,4,balanced,0.6300586859385172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,4,balanced,1.0260426998138428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,4,balanced,1.2038613160451253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,2,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,2,balanced,0.035504000882307686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,2,balanced,0.03678400069475174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,2,balanced,0.037632000943024956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,2,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,2,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,2,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,2,balanced,0.038058665891488395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,2,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,2,balanced,0.039317332208156586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,2,balanced,0.03739733248949051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,2,balanced,0.03977066775163015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,2,balanced,0.03949866692225138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,2,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,2,balanced,0.045610666275024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,2,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,2,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,2,balanced,0.055957332253456116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,2,balanced,0.05874133110046387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,2,balanced,0.06644799808661143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,2,balanced,0.07843199868996938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,2,balanced,0.10281067093213399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,2,balanced,0.11793599526087443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,2,balanced,0.15650666753451029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,2,balanced,0.19447465737660727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,2,balanced,0.27270400524139404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,2,balanced,0.35126932462056476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,2,power_law_1.2,0.15102720260620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,2,power_law_1.2,0.2151360034942627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,2,power_law_1.2,0.2683072090148926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,2,power_law_1.2,0.38209919929504393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,2,power_law_1.2,0.4984000205993652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,2,power_law_1.2,0.5901951789855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,2,power_law_1.2,0.7798975944519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,2,power_law_1.2,0.772435188293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,2,power_law_1.2,0.7594111919403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,2,power_law_1.2,0.7881663799285888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,2,power_law_1.2,0.8584192276000977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,2,power_law_1.2,0.8809023857116699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,2,power_law_1.2,0.8676735877990722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,2,power_law_1.2,0.8896127700805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,2,power_law_1.2,0.8731072425842286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,2,power_law_1.2,0.9442560195922851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,2,power_law_1.2,0.9400639533996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,2,power_law_1.2,1.0313599586486817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.2,1.080179214477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.2,1.2010175704956054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,2,balanced,0.06843733290831248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,2,balanced,0.07267199953397115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,2,balanced,0.08577600121498108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,2,balanced,0.11730666955312093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,2,balanced,0.16708266735076904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,2,balanced,0.25195733706156415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,2,balanced,0.24651199579238892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,2,balanced,0.24597332874933878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,2,balanced,0.2408799926439921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,2,balanced,0.24810133377710977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.2,1.3450752258300782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.2,1.5774271965026856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.2,1.9040000915527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,2,balanced,0.2428213357925415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,2,balanced,0.24838932355244955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,2,balanced,0.24668800830841064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,2,balanced,0.2509653369585673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,2,balanced,0.2582933306694031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,2,balanced,0.26285866896311444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,2,balanced,0.26920533180236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,2,balanced,0.2911093235015869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,2,balanced,0.3054506580034892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,2,balanced,0.34142398834228516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,2,balanced,0.36987733840942383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,2,balanced,0.4533919890721639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,2,balanced,0.528170665105184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,2,balanced,0.7971733411153158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,2,balanced,0.946821371714274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,2,balanced,1.3471573193868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,2,balanced,1.7438133557637532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.2,2.3172607421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.2,2.7839040756225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.2,3.7628414154052736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.2,4.78845443725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,2,power_law_1.2,0.030950400233268737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,2,power_law_1.2,0.032102400064468385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,2,power_law_1.2,0.030937600135803222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,2,power_law_1.2,0.03438720107078552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,2,power_law_1.2,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,2,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,2,power_law_1.2,0.04293760061264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,2,power_law_1.2,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,2,power_law_1.2,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,2,power_law_1.2,0.047091200947761536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,2,power_law_1.2,0.04588159918785095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,2,power_law_1.2,0.053350400924682614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,2,power_law_1.2,0.052537602186203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,2,power_law_1.2,0.05319039821624756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,2,power_law_1.2,0.06305279731750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,2,power_law_1.2,0.06656000018119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,2,power_law_1.2,0.06881279945373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,2,power_law_1.2,0.0800000011920929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,2,power_law_1.2,0.09099519848823548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,2,power_law_1.2,0.11476479768753052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,2,power_law_1.2,0.14218239784240722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,2,power_law_1.2,0.1699903964996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,2,power_law_1.2,0.22096641063690187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,2,power_law_1.2,0.31215999126434324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,2,power_law_1.2,0.3627583980560303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,16,power_law_1.2,0.05144320130348205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,2,power_law_1.2,0.5012159824371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,16,power_law_1.2,0.06453120112419128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,16,power_law_1.2,0.0546176016330719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,16,power_law_1.2,0.04953599870204926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,16,power_law_1.2,0.05092480182647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,16,power_law_1.2,0.05113599896430969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,16,power_law_1.2,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,16,power_law_1.2,0.05413119792938233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,16,power_law_1.2,0.054636800289154054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,16,power_law_1.2,0.05516160130500793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,16,power_law_1.2,0.05720319747924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,16,power_law_1.2,0.05902720093727112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,16,power_law_1.2,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,16,power_law_1.2,0.0639360010623932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,16,power_law_1.2,0.07130240201950074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,32,power_law_1.01,0.023622399568557738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,32,power_law_1.01,0.020268799364566804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,32,power_law_1.01,0.020147199928760528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,32,power_law_1.01,0.020025600492954255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,32,power_law_1.01,0.020883199572563172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,32,power_law_1.01,0.02172800004482269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,32,power_law_1.01,0.022547200322151184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,16,power_law_1.2,0.07487360239028931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,32,power_law_1.01,0.022252799570560457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,16,power_law_1.2,0.07676159739494323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,32,power_law_1.01,0.023321600258350374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,16,power_law_1.2,0.09093760251998902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,32,power_law_1.01,0.023571200668811798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,16,power_law_1.2,0.10501760244369507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,32,power_law_1.01,0.02730880081653595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,16,power_law_1.2,0.13164160251617432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,32,power_law_1.01,0.02797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,16,power_law_1.2,0.159443199634552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,16,power_law_1.2,0.2085632085800171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,16,power_law_1.2,0.27124478816986086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,16,power_law_1.2,0.4039680004119873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,16,power_law_1.2,0.48705282211303713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,16,power_law_1.2,0.7305088043212891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,16,power_law_1.2,0.9968000411987304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,1,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,1,balanced,0.04743466774622599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,1,balanced,0.05658666789531708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,1,balanced,0.05787733197212219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,32,power_law_1.01,0.028288000822067262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,1,balanced,0.05830933153629303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,1,balanced,0.05829333265622457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,1,balanced,0.055946667989095054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,1,balanced,0.05978666742642721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,1,balanced,0.056661332647005715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,1,balanced,0.0561653325955073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,1,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,1,balanced,0.05603733162085215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,1,balanced,0.058602665861447654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,1,balanced,0.0601440022389094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,1,balanced,0.06230400005976359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,1,balanced,0.06422933439413707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,1,balanced,0.06806933383146922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,1,balanced,0.09011733531951904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,1,balanced,0.09469333291053772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,1,balanced,0.12171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,1,balanced,0.14073066910107931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,1,balanced,0.1887893279393514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,32,power_law_1.01,0.043398401141166686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,32,power_law_1.01,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,32,power_law_1.01,0.042719998955726625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,32,power_law_1.01,0.04568960070610047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,32,power_law_1.01,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,32,power_law_1.01,0.05310720205307007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,32,power_law_1.01,0.06273919939994813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,1,balanced,0.23438932498296103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,32,power_law_1.01,0.07179520130157471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,1,balanced,0.3307146628697713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,32,power_law_1.01,0.08910080194473266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,1,balanced,0.42260265350341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,32,power_law_1.01,0.10592000484466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,32,power_law_1.01,0.14977279901504517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,32,power_law_1.01,0.18466559648513795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,32,power_law_1.01,0.26201601028442384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,32,power_law_1.01,0.3468415975570679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,1,balanced,0.6081706682840983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,1,balanced,0.7982239723205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,2,power_law_1.2,0.7192383766174316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,4,power_law_1.2,0.10611200332641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,4,power_law_1.2,0.11694079637527466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,4,power_law_1.2,0.16607999801635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,4,power_law_1.2,0.18877439498901366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,4,power_law_1.2,0.20176639556884765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,4,power_law_1.2,0.27716479301452634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,4,power_law_1.2,0.2783936023712158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,4,power_law_1.2,0.28645761013031007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,4,power_law_1.2,0.29497599601745605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,4,power_law_1.2,0.29665279388427734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,4,power_law_1.2,0.3029247999191284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,4,power_law_1.2,0.30950400829315183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,4,power_law_1.2,0.305401611328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,4,power_law_1.2,0.30935039520263674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,4,power_law_1.2,0.3422976016998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,4,power_law_1.2,0.32972800731658936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,4,power_law_1.2,0.38409600257873533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,4,power_law_1.2,0.41558399200439455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,4,power_law_1.2,0.4224639892578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,4,power_law_1.2,0.5402431964874268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,4,power_law_1.2,0.6184639930725098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,4,power_law_1.2,0.7321536064147949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,4,power_law_1.2,0.7561600208282471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,4,power_law_1.2,0.9059776306152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,4,power_law_1.2,1.2132287979125977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,4,power_law_1.2,1.708198356628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,4,power_law_1.2,2.089804840087891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,4,power_law_1.01,0.02815999984741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,4,power_law_1.01,0.04831359982490539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,4,power_law_1.01,0.06028159856796265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,4,power_law_1.01,0.07768959999084472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,4,power_law_1.01,0.08967040181159973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,4,power_law_1.01,0.1059648036956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,4,power_law_1.01,0.12034560441970825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,4,power_law_1.01,0.13118720054626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,4,power_law_1.01,0.12900480031967163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,4,power_law_1.01,0.1362239956855774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,4,power_law_1.01,0.12966400384902954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,4,power_law_1.01,0.13621759414672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,4,power_law_1.01,0.14268800020217895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,4,power_law_1.01,0.14092799425125122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,4,power_law_1.01,0.1467967987060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,4,power_law_1.01,0.1494976043701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,4,power_law_1.01,0.16088320016860963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,4,power_law_1.01,0.16987520456314087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,32,power_law_1.01,0.059443199634552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.01,0.18190720081329345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,32,power_law_1.01,0.05783680081367493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,32,power_law_1.01,0.057171201705932616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,32,power_law_1.01,0.05790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,32,power_law_1.01,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,32,power_law_1.01,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,32,power_law_1.01,0.056383997201919556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.01,0.21649279594421386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,32,power_law_1.01,0.05999360084533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.01,0.27665278911590574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,32,power_law_1.01,0.059001600742340087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.01,0.2946176052093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.01,0.35155200958251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.01,0.43710718154907224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.01,0.5297023773193359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.01,0.768339204788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,32,power_law_1.01,0.05923839807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,32,power_law_1.01,0.05937280058860779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.01,0.9851648330688476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,32,power_law_1.01,0.06087039709091187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,32,power_law_1.01,0.06292480230331421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,32,power_law_1.01,0.06284160017967225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,32,power_law_1.01,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,32,power_law_1.01,0.06812800168991089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,4,power_law_1.01,0.10205440521240235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,32,power_law_1.01,0.0706496000289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,4,power_law_1.01,0.14056320190429689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,32,power_law_1.01,0.07863680124282837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,4,power_law_1.01,0.15358079671859742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,32,power_law_1.01,0.08370559811592101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,4,power_law_1.01,0.192467200756073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,32,power_law_1.01,0.09689599871635438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,4,power_law_1.01,0.22112638950347902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,32,power_law_1.01,0.11289600133895875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,4,power_law_1.01,0.25502719879150393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,32,power_law_1.01,0.139136004447937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,32,power_law_1.01,0.025043201446533204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,32,power_law_1.01,0.16586240530014038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,32,power_law_1.01,0.023980799317359924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,32,power_law_1.01,0.21863679885864257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,32,power_law_1.01,0.023686400055885314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,32,power_law_1.01,0.02319999933242798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,32,power_law_1.01,0.026111999154090883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,4,power_law_1.01,0.27715840339660647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,32,power_law_1.01,0.027993598580360414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,32,power_law_1.01,0.27469439506530763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,4,power_law_1.01,0.2825216054916382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,32,power_law_1.01,0.03678719997406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,4,power_law_1.01,0.28256640434265134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,32,power_law_1.01,0.37336320877075196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,32,power_law_1.01,0.03675520122051239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,4,power_law_1.01,0.2922879934310913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,32,power_law_1.01,0.50348801612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,32,power_law_1.01,0.05082880258560181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,4,power_law_1.01,0.2939071893692017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,32,power_law_1.01,0.05389440059661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,4,power_law_1.01,0.28295040130615234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,32,power_law_1.01,0.05259519815444946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,4,power_law_1.01,0.2863424062728882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,32,power_law_1.01,0.05655679702758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,4,power_law_1.01,0.29687039852142333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,32,power_law_1.01,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,4,power_law_1.01,0.3038399934768677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,32,power_law_1.01,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,4,power_law_1.01,0.31152000427246096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,32,power_law_1.01,0.0549888014793396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,4,power_law_1.01,0.3103679895401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,32,power_law_1.01,0.053932797908782956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,4,power_law_1.01,0.3341824054718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,32,power_law_1.01,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.01,0.35879039764404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,32,power_law_1.01,0.05934720039367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.01,0.4219200134277344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,32,power_law_1.01,0.05549439787864685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.01,0.4651072025299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.01,0.5651584148406983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.01,0.6661439895629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.01,0.8579775810241699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,32,power_law_1.01,0.06453120112419128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,32,power_law_1.01,0.07342079877853394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,32,power_law_1.01,0.08338559865951538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,32,power_law_1.01,0.10228480100631714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,32,power_law_1.01,0.14110080003738404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,32,power_law_1.01,0.18595839738845826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,32,power_law_1.01,0.25178239345550535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.01,1.0488384246826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,32,power_law_1.01,0.33052160739898684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.01,1.4722623825073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.01,1.8381824493408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,4,power_law_1.2,0.1516991972923279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,4,power_law_1.2,0.24033279418945314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,4,power_law_1.2,0.36727681159973147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,4,power_law_1.2,0.43462400436401366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,4,power_law_1.2,0.5894464015960693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,4,power_law_1.2,0.7304831981658936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,4,power_law_1.2,0.8191807746887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,4,power_law_1.2,1.0682944297790526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,4,power_law_1.2,0.9842240333557128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,4,power_law_1.2,0.898470401763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,4,power_law_1.2,0.856595230102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,4,power_law_1.2,0.989459228515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,4,power_law_1.2,1.1464896202087402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,4,power_law_1.2,1.1321151733398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,4,power_law_1.2,1.0227519989013671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,4,power_law_1.2,1.0836928367614747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,4,power_law_1.2,1.060166358947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,4,power_law_1.2,1.2424192428588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,4,power_law_1.2,1.2255104064941407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,4,power_law_1.2,1.5164480209350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,4,power_law_1.2,1.6077632904052734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,4,power_law_1.2,1.8887872695922852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,4,power_law_1.2,1.8433727264404296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,4,power_law_1.2,2.102560043334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,4,power_law_1.2,2.5548288345336916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,4,power_law_1.2,3.1470592498779295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,4,power_law_1.2,4.544121551513672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,32,power_law_1.2,0.03978880047798157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,32,power_law_1.2,0.031769600510597226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,32,power_law_1.2,0.029836800694465638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,32,power_law_1.2,0.031795200705528257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,32,power_law_1.2,0.03134720027446747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,32,power_law_1.2,0.03287039995193482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,32,power_law_1.2,0.03255040049552917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,32,power_law_1.2,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,32,power_law_1.2,0.03423359990119934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,32,power_law_1.2,0.03506560027599335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,32,power_law_1.2,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,32,power_law_1.2,0.03764480054378509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,32,power_law_1.2,0.03920640051364899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,32,power_law_1.2,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,32,power_law_1.2,0.04517120122909546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,32,power_law_1.2,0.04705919921398163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,32,power_law_1.2,0.048588800430297854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,32,power_law_1.2,0.05130239725112915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,32,power_law_1.2,0.05749760270118713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,32,power_law_1.2,0.06584960222244263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,32,power_law_1.2,0.0716543972492218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,32,power_law_1.2,0.09062399864196777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,32,power_law_1.2,0.12007679939270019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,32,power_law_1.2,0.1627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,32,power_law_1.2,0.1861631989479065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,32,power_law_1.2,0.25690879821777346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,8,balanced,0.047925333182017006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,8,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,8,balanced,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,8,balanced,0.04850666721661886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,8,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,8,balanced,0.050144001841545105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,8,balanced,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,8,balanced,0.04807466765244802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,8,balanced,0.04984533290068308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,8,balanced,0.049728001157442726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,8,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,8,balanced,0.04800533254941305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,8,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,8,balanced,0.04964800179004669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,8,balanced,0.05217066903909048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,8,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,8,balanced,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,8,balanced,0.05665066838264465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,8,balanced,0.060421332716941833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,8,balanced,0.06620799998442332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,8,balanced,0.07453866799672444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,8,balanced,0.08780266841252644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,32,power_law_1.2,0.3528831958770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,8,balanced,0.09744000434875488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,8,balanced,0.13091733058293661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,8,balanced,0.15004799763361612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,8,balanced,0.2039146622021993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,8,balanced,0.2593173384666443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,128,power_law_1.01,0.08198400139808655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,128,power_law_1.01,0.07994239926338195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,128,power_law_1.01,0.08184319734573364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,128,power_law_1.01,0.081004798412323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,128,power_law_1.01,0.08021759986877441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,128,power_law_1.01,0.08401920199394226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,128,power_law_1.01,0.0790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,128,power_law_1.01,0.08579199910163879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,128,power_law_1.01,0.08353919982910156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,128,power_law_1.01,0.08521599769592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,128,power_law_1.01,0.08238080143928528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,128,power_law_1.01,0.08736640214920044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,128,power_law_1.01,0.08892800211906433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,128,power_law_1.01,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,128,power_law_1.01,0.091430401802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,128,power_law_1.01,0.09133440256118774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,128,power_law_1.01,0.09847679734230042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,128,power_law_1.01,0.10338560342788697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,128,power_law_1.01,0.11648000478744507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,128,power_law_1.01,0.14055039882659912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,128,power_law_1.01,0.15947519540786742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,128,power_law_1.01,0.20495998859405518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,128,power_law_1.01,0.24498560428619384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,128,power_law_1.01,0.3333568096160889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,128,power_law_1.01,0.4239359855651855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,128,power_law_1.01,0.6035967826843261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,128,power_law_1.01,0.7655295848846435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,4,power_law_1.01,0.10073599815368653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,4,power_law_1.01,0.11328639984130859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,4,power_law_1.01,0.11274880170822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,4,power_law_1.01,0.13774720430374146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,4,power_law_1.01,0.1387519955635071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,4,power_law_1.01,0.14149760007858275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,4,power_law_1.01,0.16828160285949706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,4,power_law_1.01,0.17605760097503662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,4,power_law_1.01,0.21115520000457763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,4,power_law_1.01,0.222707200050354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,4,power_law_1.01,0.16673920154571534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,4,power_law_1.01,0.17533440589904786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,4,power_law_1.01,0.19902080297470093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,4,power_law_1.01,0.19645440578460693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,4,power_law_1.01,0.2107072114944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,4,power_law_1.01,0.21854720115661622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,4,power_law_1.01,0.23383679389953613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,4,power_law_1.01,0.32515840530395507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,4,power_law_1.01,0.40494718551635744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,4,power_law_1.01,0.5404032230377197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,4,power_law_1.01,0.5547584056854248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,4,power_law_1.01,0.8241600036621094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,4,power_law_1.01,1.1370559692382813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,4,power_law_1.01,1.4400192260742188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,4,power_law_1.01,1.8291072845458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,4,power_law_1.01,0.045535999536514285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,4,power_law_1.01,0.04542079865932465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,4,power_law_1.01,0.04434559941291809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,4,power_law_1.01,0.0466623991727829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,4,power_law_1.01,0.0494592010974884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,4,power_law_1.01,0.05273600220680237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,4,power_law_1.01,0.07434880137443542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,4,power_law_1.01,0.06661120057106018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,4,power_law_1.01,0.08908159732818603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,4,power_law_1.01,0.0793727993965149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,4,power_law_1.01,0.07456640005111695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,4,power_law_1.01,0.1337407946586609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,4,power_law_1.01,0.126636803150177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,4,power_law_1.01,0.10022399425506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,4,power_law_1.01,0.1449728012084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,4,power_law_1.01,0.12485760450363159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,4,power_law_1.01,0.16082559823989867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,4,power_law_1.01,0.2109055995941162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.01,0.27384960651397705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.01,0.38744320869445803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.01,0.5182271957397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.01,0.7444928169250489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,128,power_law_1.2,0.11570559740066529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,128,power_law_1.2,0.040006399154663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,128,power_law_1.2,0.040505599975585935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,128,power_law_1.2,0.0420415997505188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,128,power_law_1.2,0.043628799915313723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,128,power_law_1.2,0.04417920112609863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,128,power_law_1.2,0.04533120095729828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,128,power_law_1.2,0.04561919867992401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,128,power_law_1.2,0.04567680060863495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,128,power_law_1.2,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,128,power_law_1.2,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,4,power_law_1.01,3.067897605895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,128,power_law_1.2,0.04745599925518036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.01,0.9885567665100098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,128,power_law_1.2,0.050220799446105954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,128,power_law_1.2,0.052211201190948485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,128,power_law_1.2,0.05767040252685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,128,power_law_1.2,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,128,power_law_1.2,0.06328319907188415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,128,power_law_1.2,0.0722815990447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,128,power_law_1.2,0.07758079767227173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,128,power_law_1.2,0.09287679791450501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,128,power_law_1.2,0.1090880036354065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,128,power_law_1.2,0.15418239831924438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,128,power_law_1.2,0.17396479845046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,128,power_law_1.2,0.24606719017028808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,128,power_law_1.2,0.30087039470672605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.01,1.4629440307617188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,128,power_law_1.2,0.4475776195526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,128,power_law_1.2,0.5970560073852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.01,1.9194559097290038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,4,power_law_1.01,3.408787155151367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.01,2.8763200759887697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.01,3.804127883911133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,2,power_law_1.01,0.03597440123558045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,2,power_law_1.01,0.05006719827651977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,2,power_law_1.01,0.06412799954414368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,2,power_law_1.01,0.07328640222549439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,2,power_law_1.01,0.08090239763259888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,2,power_law_1.01,0.11549439430236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,2,power_law_1.01,0.11436799764633179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,2,power_law_1.01,0.12188800573348998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,2,power_law_1.01,0.12102400064468384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,2,power_law_1.01,0.12172800302505493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,2,power_law_1.01,0.12560640573501586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,2,power_law_1.01,0.1295232057571411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,2,power_law_1.01,0.13401600122451782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,2,power_law_1.01,0.13737599849700927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,2,power_law_1.01,0.1443519949913025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,2,power_law_1.01,0.1488319993019104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,2,power_law_1.01,0.15870720148086548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,2,power_law_1.01,0.17303680181503295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,2,power_law_1.01,0.1893952012062073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,2,power_law_1.01,0.23289599418640136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,2,power_law_1.01,0.2532671928405762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,2,power_law_1.01,0.34314239025115967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,2,power_law_1.01,0.3965951919555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,2,power_law_1.01,0.5272704124450683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,2,power_law_1.01,0.6291903972625732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,2,power_law_1.01,0.952569580078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,2,power_law_1.01,1.2108991622924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,1,power_law_1.2,0.019545599818229675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,1,power_law_1.2,0.02234880030155182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,1,power_law_1.2,0.028678399324417115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,1,power_law_1.2,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,1,power_law_1.2,0.045363199710845944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,1,power_law_1.2,0.061292797327041626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,1,power_law_1.2,0.06673920154571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,1,power_law_1.2,0.06606720089912414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,1,power_law_1.2,0.06948480010032654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,1,power_law_1.2,0.07196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,1,power_law_1.2,0.07480319738388061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,1,power_law_1.2,0.07549440264701843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,1,power_law_1.2,0.07841280102729797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,1,power_law_1.2,0.0856768012046814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,1,power_law_1.2,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,1,power_law_1.2,0.08372480273246766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,1,power_law_1.2,0.09173120260238647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,1,power_law_1.2,0.09911680221557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.2,0.10814080238342286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.2,0.2086848020553589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.2,0.2378432035446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.2,0.20739200115203857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.2,0.2326335906982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.2,0.2947391986846924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.2,0.35650560855865476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,4,power_law_1.01,0.018092800676822663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,4,power_law_1.01,0.019398400187492372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,4,power_law_1.01,0.019443200528621675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.2,0.4803199768066406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,4,power_law_1.01,0.02064639925956726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,4,power_law_1.01,0.0208064004778862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,4,power_law_1.01,0.02237440049648285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,4,power_law_1.01,0.02279040068387985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,4,power_law_1.01,0.023219199478626253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,4,power_law_1.01,0.024307200312614442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,4,power_law_1.01,0.02462719976902008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,4,power_law_1.01,0.02565760016441345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,4,power_law_1.01,0.027084800601005554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,4,power_law_1.01,0.028460800647735596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,4,power_law_1.01,0.02771199941635132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,4,power_law_1.01,0.02961280047893524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,4,power_law_1.01,0.0304639995098114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,4,power_law_1.01,0.03309440016746521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,4,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,4,power_law_1.01,0.057555198669433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,4,power_law_1.01,0.06893439888954163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,4,power_law_1.01,0.06111999750137329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,4,power_law_1.01,0.07519360184669495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,4,power_law_1.01,0.0862272024154663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,4,power_law_1.01,0.11052160263061524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,4,power_law_1.01,0.145196795463562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,4,power_law_1.01,0.1932479977607727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,4,power_law_1.01,0.23006079196929932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.2,0.6220992088317872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,2,power_law_1.2,0.2050368070602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,2,power_law_1.2,0.2503232002258301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,8,power_law_1.01,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,2,power_law_1.2,0.24595201015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,8,power_law_1.01,0.03678080141544342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,2,power_law_1.2,0.3942528009414673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,8,power_law_1.01,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,2,power_law_1.2,0.3923840045928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,8,power_law_1.01,0.03722879886627197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,8,power_law_1.01,0.03808639943599701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,2,power_law_1.2,0.38675200939178467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,8,power_law_1.01,0.038700801134109494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,2,power_law_1.2,0.38216960430145264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,2,power_law_1.2,0.40692481994628904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,2,power_law_1.2,0.4208960056304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,2,power_law_1.2,0.4262847900390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,2,power_law_1.2,0.4402624130249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,2,power_law_1.2,0.46299519538879397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,8,power_law_1.01,0.039289599657058714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,2,power_law_1.2,0.4464384078979492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,8,power_law_1.01,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,8,power_law_1.01,0.040608000755310056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,2,power_law_1.2,0.4820352077484131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,8,power_law_1.01,0.042182400822639465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,8,power_law_1.01,0.042771199345588685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,2,power_law_1.2,0.5325823783874511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,2,power_law_1.2,0.575648021697998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,2,power_law_1.2,0.590822410583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,8,power_law_1.01,0.04482559859752655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,8,power_law_1.01,0.04991360008716583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,8,power_law_1.01,0.052275198698043826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,8,power_law_1.01,0.05836799740791321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,2,power_law_1.2,0.7159552097320556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,8,power_law_1.01,0.06279680132865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,8,power_law_1.01,0.06651520133018493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,2,power_law_1.2,0.7465919971466064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,8,power_law_1.01,0.07968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.01,0.08948479890823365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.01,0.11363840103149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.01,0.13197439908981323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,2,power_law_1.2,0.9873727798461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.01,0.17821439504623413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,2,power_law_1.2,1.3058239936828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.01,0.212556791305542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.01,0.3027647972106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,2,power_law_1.2,1.6264768600463868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.01,0.3661632061004639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,2,power_law_1.2,2.4608127593994142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,2,power_law_1.2,3.3401214599609377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.01,0.5621056079864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,2,power_law_1.2,4.048038482666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.01,0.7583424091339112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,2,power_law_1.2,6.925433349609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,4,power_law_1.2,0.049568000435829165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,4,power_law_1.2,0.054604798555374146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,4,power_law_1.2,0.07104640007019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,4,power_law_1.2,0.08172799944877625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,4,power_law_1.2,0.09488639831542969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,4,power_law_1.2,0.10617599487304688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,4,power_law_1.2,0.11100800037384033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,4,power_law_1.2,0.13003519773483277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,4,power_law_1.2,0.12192000150680542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,4,power_law_1.2,0.12846720218658447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,4,power_law_1.2,0.13325439691543578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,4,power_law_1.2,0.13689600229263305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,4,power_law_1.2,0.14113919734954833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,4,power_law_1.2,0.14680320024490356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,4,power_law_1.2,0.16433919668197633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,4,power_law_1.2,0.17155840396881103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,4,power_law_1.2,0.1844480037689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,4,power_law_1.2,0.21552000045776368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,4,power_law_1.2,0.2271104097366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,4,power_law_1.2,0.2735039949417114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,4,power_law_1.2,0.31895039081573484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,4,power_law_1.2,0.4231872081756592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,4,power_law_1.2,0.5365439891815186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,4,power_law_1.2,0.7548992156982421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,4,power_law_1.2,0.8891136169433593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,4,power_law_1.2,1.2342144012451173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,4,power_law_1.2,1.6492992401123048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,2,power_law_1.2,7.627769470214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,2,power_law_1.2,0.10058879852294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,2,power_law_1.2,0.1491711974143982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,2,power_law_1.2,0.19020160436630248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,2,power_law_1.2,0.2671360015869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,2,power_law_1.2,0.3731584072113037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,2,power_law_1.2,0.46149120330810545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,2,power_law_1.2,0.6370816230773926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,2,power_law_1.2,0.6866687774658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,2,power_law_1.2,0.7007359981536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,2,power_law_1.2,0.6850048065185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,2,power_law_1.2,0.6902207851409912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,2,power_law_1.2,0.7184512138366699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,2,power_law_1.2,0.7542848110198974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,2,power_law_1.2,0.7685632228851318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,2,power_law_1.2,0.8103360176086426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,2,power_law_1.2,0.8546431541442872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,2,power_law_1.2,0.8893376350402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,2,power_law_1.2,0.99749755859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,2,power_law_1.2,0.9971391677856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,1,power_law_1.2,0.07034239768981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,2,power_law_1.2,1.1310400009155273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,1,power_law_1.2,0.07975040078163147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,1,power_law_1.2,0.09171199798583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,2,power_law_1.2,1.1673855781555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,1,power_law_1.2,0.1373695969581604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,1,power_law_1.2,0.13617279529571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,2,power_law_1.2,1.3085311889648437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,1,power_law_1.2,0.13976320028305053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,1,power_law_1.2,0.14274560213088988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,2,power_law_1.2,1.5565759658813476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,1,power_law_1.2,0.14394240379333495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,1,power_law_1.2,0.1435647964477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,1,power_law_1.2,0.14306559562683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,1,power_law_1.01,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,2,power_law_1.2,2.014886474609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,1,power_law_1.01,0.031436800956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,1,power_law_1.01,0.03294720053672791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,1,power_law_1.01,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,1,power_law_1.01,0.04723840057849884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,1,power_law_1.01,0.07043840289115906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,1,power_law_1.01,0.0624064028263092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,1,power_law_1.01,0.07180799841880799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,1,power_law_1.2,0.1462656021118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,1,power_law_1.01,0.06124160289764404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,1,power_law_1.01,0.06668800115585327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,1,power_law_1.2,0.154585599899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,1,power_law_1.01,0.07430400252342224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,1,power_law_1.01,0.07065600156784058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,1,power_law_1.2,0.1602944016456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,1,power_law_1.01,0.07616639733314515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,2,power_law_1.2,2.3911359786987303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,1,power_law_1.2,0.16713600158691405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,1,power_law_1.2,0.18297599554061889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,2,power_law_1.2,3.058867263793945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,1,power_law_1.2,0.18735359907150267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,1,power_law_1.01,0.08840320110321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,1,power_law_1.2,0.19998719692230224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,2,power_law_1.2,4.002092742919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,1,power_law_1.01,0.17504639625549318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,1,power_law_1.01,0.20602879524230958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,1,power_law_1.01,0.25589759349823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,1,power_law_1.01,0.13777920007705688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,1,power_law_1.2,0.24455680847167968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.01,0.16340479850769044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.2,0.2834815979003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.01,0.23029758930206298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.2,0.36034560203552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.01,0.2923903942108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.2,0.44686079025268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.01,0.41716480255126953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.2,0.6101759910583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.01,0.5298751831054688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.01,0.7832128047943115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.2,0.7802112102508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.01,1.0219903945922852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.2,1.1205632209777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.2,1.4492799758911132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.01,1.5140800476074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.01,2.0053504943847655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,4,power_law_1.01,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,4,power_law_1.01,0.035743999481201175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,4,power_law_1.01,0.03544319868087768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,4,power_law_1.01,0.03783040046691895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,4,power_law_1.01,0.03813120126724243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,4,power_law_1.01,0.04234879910945892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,4,power_law_1.01,0.04206719994544983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,4,power_law_1.01,0.043647998571395875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,4,power_law_1.01,0.04568319916725159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,1,power_law_1.2,0.04968959987163544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,4,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,1,power_law_1.2,0.05826560258865356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,4,power_law_1.01,0.05023360252380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,4,power_law_1.01,0.05264000296592712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,1,power_law_1.2,0.07704960107803345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,4,power_law_1.01,0.057196801900863646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,4,power_law_1.01,0.05534719824790955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,4,power_law_1.01,0.061504000425338747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,4,power_law_1.01,0.06533120274543762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,4,power_law_1.01,0.06815999746322632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,4,power_law_1.01,0.07750399708747864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,4,power_law_1.01,0.09105920195579528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,4,power_law_1.01,0.1141759991645813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,4,power_law_1.01,0.13690240383148194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,4,power_law_1.01,0.18802560567855836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,4,power_law_1.01,0.2326591968536377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,4,power_law_1.01,0.3475967884063721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,4,power_law_1.01,0.43546881675720217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,4,power_law_1.01,0.5827136039733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,4,power_law_1.01,0.8217344284057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,1,power_law_1.2,0.09874560236930847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,1,power_law_1.2,0.11864320039749146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,1,power_law_1.2,0.17247359752655028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,1,power_law_1.2,0.1896448016166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,1,power_law_1.2,0.19533439874649047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,1,power_law_1.2,0.2005311965942383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,1,power_law_1.2,0.20991361141204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,1,power_law_1.2,0.21463680267333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,1,power_law_1.2,0.22614400386810302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,1,power_law_1.2,0.23772799968719482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,1,power_law_1.2,0.2445375919342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,1,power_law_1.2,0.25512959957122805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,1,power_law_1.2,0.2643136024475098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,1,power_law_1.2,0.2877631902694702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,1,power_law_1.2,0.3376703977584839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,1,power_law_1.2,0.38258559703826905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,1,power_law_1.2,0.48090238571166993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,1,power_law_1.2,0.5576767921447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,1,power_law_1.2,0.735814380645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,1,power_law_1.2,0.9755135536193847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,1,power_law_1.2,1.3027327537536622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,1,power_law_1.2,1.6921472549438477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.2,2.130956840515137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,1,power_law_1.2,2.331622314453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,1,power_law_1.2,2.9395263671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.2,2.791289520263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,16,power_law_1.01,0.026630398631095887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,16,power_law_1.01,0.02273920029401779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,16,power_law_1.01,0.020953600108623505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,16,power_law_1.01,0.021408000588417055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,16,power_law_1.01,0.022144000232219695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,16,power_law_1.01,0.02305919975042343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,16,power_law_1.01,0.023839999735355378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,16,power_law_1.01,0.024639999866485594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,16,power_law_1.01,0.02426239997148514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,16,power_law_1.01,0.02622080147266388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,16,power_law_1.01,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,16,power_law_1.01,0.029606398940086365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,16,power_law_1.01,0.03155199885368347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,16,power_law_1.01,0.05413119792938233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,16,power_law_1.01,0.054585599899291994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,16,power_law_1.01,0.05668479800224304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,16,power_law_1.01,0.05260800123214722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,16,power_law_1.01,0.05288959741592407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,16,power_law_1.01,0.054841601848602296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,16,power_law_1.01,0.06595199704170226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,16,power_law_1.01,0.0769599974155426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,16,power_law_1.01,0.10393600463867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,16,power_law_1.01,0.12462079524993896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,16,power_law_1.01,0.15944960117340087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,16,power_law_1.01,0.20899839401245118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,16,power_law_1.01,0.28257920742034914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,16,power_law_1.01,0.38629119396209716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,1,power_law_1.2,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,1,power_law_1.2,0.06854400038719177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,1,power_law_1.2,0.111571204662323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,1,power_law_1.2,0.17943040132522584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,1,power_law_1.2,0.246726393699646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,1,power_law_1.2,0.3508991956710815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,1,power_law_1.2,0.486681604385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,1,power_law_1.2,0.5295551776885986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,1,power_law_1.2,0.557593584060669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,1,power_law_1.2,0.582047986984253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,1,power_law_1.2,0.6092480182647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,1,power_law_1.2,0.6488959789276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,1,power_law_1.2,0.6661183834075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,1,power_law_1.2,0.7272064208984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,1,power_law_1.2,0.6860991954803467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,1,power_law_1.2,0.7017727851867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,1,power_law_1.2,0.7776768207550049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,1,power_law_1.2,0.7742911815643311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,1,power_law_1.2,0.8603648185729981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,1,power_law_1.2,0.8775232315063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,1,power_law_1.2,0.937446403503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,1,power_law_1.2,2.322585678100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,1,power_law_1.2,2.764780807495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,1,power_law_1.2,1.5612607955932618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,1,power_law_1.2,1.7950271606445312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,1,power_law_1.2,2.250655937194824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,1,power_law_1.2,2.8548864364624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,1,power_law_1.2,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,1,power_law_1.2,0.05184000134468079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,1,power_law_1.2,0.07739520072937012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,1,power_law_1.2,0.10290559530258178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,1,power_law_1.2,0.1400383949279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,1,power_law_1.2,0.2290879964828491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,1,power_law_1.2,0.24506239891052245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,1,power_law_1.2,0.2680448055267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,1,power_law_1.2,0.26238079071044923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,1,power_law_1.2,0.28501760959625244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,1,power_law_1.2,0.3016063928604126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,1,power_law_1.2,0.32006399631500243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,1,power_law_1.2,0.2957632064819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,1,power_law_1.2,0.3282815933227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,1,power_law_1.2,0.3368767976760864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,1,power_law_1.2,0.31703040599822996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,1,power_law_1.2,0.33746559619903566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,1,power_law_1.2,0.35228800773620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.2,0.3891648054122925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.2,0.8606399536132813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.2,0.983456039428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.2,0.5790847778320313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.2,0.7091584205627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.2,0.8294655799865722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.2,0.9947327613830567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.2,1.3642239570617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.2,1.6753215789794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,128,balanced,0.05426133175690969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,128,balanced,0.06620799998442332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,128,balanced,0.0496319979429245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,128,balanced,0.050101334849993386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,128,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,128,balanced,0.053904001911481224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,128,balanced,0.053861334919929504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,128,balanced,0.054560000697771706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,128,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,128,balanced,0.05593599875768026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,128,balanced,0.05527999997138977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,128,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,128,balanced,0.05829866727193197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,128,balanced,0.05769066512584686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,128,balanced,0.06400533517201741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,128,balanced,0.06413333117961884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,128,balanced,0.06603200236956279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,128,balanced,0.07250133156776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,128,balanced,0.0784746656815211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,128,balanced,0.09474133451779683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,128,balanced,0.11023466785748799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,128,balanced,0.13404800494511923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,128,balanced,0.1497599979241689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,128,balanced,0.1922666629155477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,128,balanced,0.23237866163253784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,128,balanced,0.3156053423881531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,128,balanced,0.39432533582051593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,32,balanced,0.03472000112136205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,32,balanced,0.030645333230495453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,32,balanced,0.0313226655125618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,32,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,32,balanced,0.037861332297325134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,32,balanced,0.038202665746212006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,32,balanced,0.03745600084463755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,32,balanced,0.039621333281199135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,32,balanced,0.03953066716591517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,32,balanced,0.03925333420435587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,32,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,32,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,32,balanced,0.0415040006240209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,32,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,32,balanced,0.044591998060544334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,32,balanced,0.047397335370381675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,32,balanced,0.049365331729253135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,32,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,32,balanced,0.057855998476346336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,32,balanced,0.0764213353395462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,32,balanced,0.08052800099054973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,32,balanced,0.09956266482671101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,32,balanced,0.11362666885058086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,32,balanced,0.14637333154678345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,32,balanced,0.1798293391863505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,32,balanced,0.24344533681869507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,32,balanced,0.3020053307215373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,2,power_law_1.01,0.0502016007900238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,2,power_law_1.01,0.07591040134429931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,2,power_law_1.01,0.10889600515365601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,2,power_law_1.01,0.1440000057220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,2,power_law_1.01,0.19642879962921142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,2,power_law_1.01,0.26477439403533937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,2,power_law_1.01,0.27683200836181643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,2,power_law_1.01,0.28209919929504396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,2,power_law_1.01,0.295468807220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,2,power_law_1.01,0.29877119064331054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,2,power_law_1.01,0.30933120250701907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,2,power_law_1.01,0.3165695905685425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,2,power_law_1.01,0.3384063959121704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,2,power_law_1.01,0.3168960094451904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,2,power_law_1.01,0.3415488004684448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,2,power_law_1.01,0.3671103954315186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,2,power_law_1.01,0.3735935926437378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,2,power_law_1.01,0.43534078598022463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.01,0.4227903842926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.01,0.5084735870361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.01,0.6013631820678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.01,0.6604159832000732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.01,0.7732480049133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.01,0.97128324508667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.01,1.2073344230651855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.01,1.705580711364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.01,2.01525764465332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,1,power_law_1.2,0.10890239477157593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,1,power_law_1.2,0.12035839557647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,1,power_law_1.2,0.15182080268859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,1,power_law_1.2,0.17022080421447755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,1,power_law_1.2,0.1844928026199341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,1,power_law_1.2,0.20582399368286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,1,power_law_1.2,0.24196479320526124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,1,power_law_1.2,0.24978559017181395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,1,power_law_1.2,0.2590399980545044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,1,power_law_1.2,0.2622720003128052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,1,power_law_1.2,0.2684351921081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,1,power_law_1.2,0.2754688024520874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,1,power_law_1.2,0.2816256046295166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,1,power_law_1.2,0.2922175884246826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,1,power_law_1.2,0.2940864086151123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,1,power_law_1.2,0.3027967929840088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,1,power_law_1.2,0.31283841133117674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,1,power_law_1.2,0.3482752084732056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,1,power_law_1.2,0.3832704067230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,1,power_law_1.2,0.4538815975189209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,1,power_law_1.2,0.5376319885253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,16,power_law_1.01,0.04391680061817169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,16,power_law_1.01,0.06354560256004334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,16,power_law_1.01,0.05931519865989685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,16,power_law_1.01,0.060249602794647215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,16,power_law_1.01,0.06234239935874939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,16,power_law_1.01,0.05493119955062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,16,power_law_1.01,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,16,power_law_1.01,0.05973759889602661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,16,power_law_1.01,0.060889601707458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,16,power_law_1.01,0.06265599727630615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,16,power_law_1.01,0.06222720146179199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,16,power_law_1.01,0.06467840075492859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,16,power_law_1.01,0.06684160232543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,16,power_law_1.01,0.06967679858207702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,16,power_law_1.01,0.07468159794807434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,16,power_law_1.01,0.07690240144729614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,16,power_law_1.01,0.07693439722061157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,16,power_law_1.01,0.08600320219993592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,16,power_law_1.01,0.08727679848670959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,16,power_law_1.01,0.10295039415359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,16,power_law_1.01,0.11267199516296386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,16,power_law_1.01,0.14392319917678834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,16,power_law_1.01,0.16929279565811156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,16,power_law_1.01,0.20770559310913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,16,power_law_1.01,0.26924800872802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,16,power_law_1.01,0.37813758850097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,16,power_law_1.01,0.4682112216949463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,8,power_law_1.01,0.038700801134109494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,8,power_law_1.01,0.04870400130748749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,8,power_law_1.01,0.05002239942550659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,8,power_law_1.01,0.05307520031929016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,8,power_law_1.01,0.05446400046348572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,32,power_law_1.01,0.10229120254516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,8,power_law_1.01,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,8,power_law_1.01,0.057843202352523805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,32,power_law_1.01,0.10329600572586059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,8,power_law_1.01,0.05777279734611511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,32,power_law_1.01,0.10367360115051269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,8,power_law_1.01,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,8,power_law_1.01,0.05979520082473755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,32,power_law_1.01,0.10343680381774903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,8,power_law_1.01,0.06020479798316956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,8,power_law_1.01,0.063673597574234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,32,power_law_1.01,0.10247679948806762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,8,power_law_1.01,0.06592000126838685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,32,power_law_1.01,0.09744639992713929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,8,power_law_1.01,0.06719359755516052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,32,power_law_1.01,0.10079360008239746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,32,power_law_1.01,0.09920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,32,power_law_1.01,0.10243200063705445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,32,power_law_1.01,0.10388480424880982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,32,power_law_1.01,0.10365439653396606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,32,power_law_1.01,0.10629760026931763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,32,power_law_1.01,0.10730880498886108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,32,power_law_1.01,0.11076480150222778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,32,power_law_1.01,0.11759359836578369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,32,power_law_1.01,0.12014080286026001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,32,power_law_1.01,0.1281216025352478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,32,power_law_1.01,0.149510395526886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,32,power_law_1.01,0.16552319526672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,8,power_law_1.01,0.07257599830627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,32,power_law_1.01,0.20147840976715087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,8,power_law_1.01,0.07633919715881347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,8,power_law_1.01,0.0809216022491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,8,power_law_1.01,0.09303039908409119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,8,power_law_1.01,0.11166720390319824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,8,power_law_1.01,0.12448639869689941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,8,power_law_1.01,0.13819520473480223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,8,power_law_1.01,0.19003520011901856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,8,power_law_1.01,0.22755839824676513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,32,power_law_1.01,0.2217087984085083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,8,power_law_1.01,0.3217600107192993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,32,power_law_1.01,0.28395519256591795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,8,power_law_1.01,0.38253440856933596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,32,power_law_1.01,0.2757312059402466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,8,power_law_1.01,0.5554751873016357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,32,power_law_1.01,0.37115519046783446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,8,power_law_1.01,0.707200002670288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,32,power_law_1.01,0.46941437721252444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,32,power_law_1.01,0.674067211151123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,32,power_law_1.01,0.7484608173370362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,8,balanced,0.05529066423575083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,8,balanced,0.05579733351866404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,8,balanced,0.05599466462930044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,8,balanced,0.062314664324124656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,8,balanced,0.07245866457621257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,8,balanced,0.0923466682434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,8,balanced,0.09224533041318257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,8,balanced,0.09196799993515015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,8,balanced,0.09512533744176228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,8,balanced,0.09285333752632141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,8,balanced,0.0939466655254364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,8,balanced,0.09700266520182292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,8,balanced,0.09733866651852925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,8,balanced,0.10019200046857198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,8,balanced,0.10857599973678589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,8,balanced,0.10939733187357585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,8,balanced,0.11347732941309611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,8,balanced,0.12983999649683634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,8,balanced,0.13370133439699808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,8,balanced,0.14934399724006653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,8,balanced,0.1726293365160624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,8,balanced,0.2132373253504435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,8,balanced,0.25961599747339886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,8,balanced,0.37222933769226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,8,balanced,0.4639253218968709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,8,balanced,0.6769333680470785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,8,balanced,0.8744853337605795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,16,power_law_1.2,0.03270399868488312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,16,power_law_1.2,0.031174400448799135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,16,power_law_1.2,0.031379199028015135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,16,power_law_1.2,0.03356800079345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,16,power_law_1.2,0.03333120048046112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,16,power_law_1.2,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,16,power_law_1.2,0.035897600650787356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,16,power_law_1.2,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,16,power_law_1.2,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,16,power_law_1.2,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,16,power_law_1.2,0.03682560026645661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,16,power_law_1.2,0.039212799072265624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,16,power_law_1.2,0.03916159868240356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,16,power_law_1.2,0.041631999611854556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,16,power_law_1.2,0.04442239999771118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,16,power_law_1.2,0.046009600162506104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,16,power_law_1.2,0.05178239941596985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,16,power_law_1.2,0.05619840025901794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,16,power_law_1.2,0.06595199704170226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,16,power_law_1.2,0.09052799940109253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,16,power_law_1.2,0.10664960145950317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,16,power_law_1.2,0.13865599632263184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,16,power_law_1.2,0.18421759605407714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,16,power_law_1.2,0.2767807960510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,8,power_law_1.01,0.02263679951429367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,8,power_law_1.01,0.02958720028400421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,8,power_law_1.01,0.02963840067386627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,8,power_law_1.01,0.034227201342582704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,8,power_law_1.01,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,8,power_law_1.01,0.04116480052471161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,8,power_law_1.01,0.05060480237007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,8,power_law_1.01,0.0500544011592865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,8,power_law_1.01,0.049395200610160825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2688,1856,6,128,1,8,power_law_1.01,0.0738431990146637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2688,1856,6,128,1,8,power_law_1.01,0.06413440108299255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2688,1856,6,128,1,8,power_law_1.01,0.06548479795455933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,16,power_law_1.2,0.32826240062713624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,16,power_law_1.2,0.5901055812835694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,16,power_law_1.2,0.8082112312316895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,4,balanced,0.07700266440709432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,4,balanced,0.07647466659545898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,4,balanced,0.08767466743787129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,4,balanced,0.10703999797503154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,4,balanced,0.13742400209108988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,4,balanced,0.18553600708643594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,4,balanced,0.1856000026067098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,4,balanced,0.1843573252360026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,4,balanced,0.1876159906387329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,4,balanced,0.18687466780344644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,4,balanced,0.18870933850606283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,4,balanced,0.1930453379948934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,4,balanced,0.19445333878199259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,4,balanced,0.1971893310546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,4,balanced,0.20589866240819296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,4,balanced,0.20504534244537354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,4,balanced,0.21074666579564413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,4,balanced,0.22722667455673218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,4,balanced,0.23666665951410928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,4,balanced,0.2711519996325175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,4,balanced,0.3022666573524475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,4,balanced,0.36503998438517254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,4,balanced,0.4288426637649536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,4,balanced,0.6024693250656128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,4,balanced,0.725605328877767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,4,balanced,1.0380533536275227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,1,power_law_1.2,0.06935679912567139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,4,balanced,1.3544212977091472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,1,power_law_1.2,0.08538879752159119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,1,power_law_1.2,0.1137279987335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,1,power_law_1.2,0.16455039978027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,1,power_law_1.2,0.22389121055603028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,1,power_law_1.2,0.28545920848846434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,1,power_law_1.2,0.41548800468444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,1,power_law_1.2,0.42849278450012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,1,power_law_1.2,0.45555839538574217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,1,power_law_1.2,0.4654848098754883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,1,power_law_1.2,0.47843198776245116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,1,power_law_1.2,0.5017343997955322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,1,power_law_1.2,0.5105216026306152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,1,power_law_1.2,0.5289728164672851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,1,power_law_1.2,0.5551743984222413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,1,power_law_1.2,0.5799680233001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,1,power_law_1.2,0.60316162109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,1,power_law_1.2,0.6927296161651612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,1,power_law_1.2,0.6945087909698486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,1,power_law_1.2,0.8094783782958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,1,power_law_1.2,0.8339839935302734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,1,power_law_1.2,1.0034111976623534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,1,power_law_1.2,1.1663807868957519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,1,power_law_1.2,1.5078656196594238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,1,power_law_1.2,1.8327295303344726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,1,power_law_1.2,2.4737407684326174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,1,power_law_1.2,3.1506303787231444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,2,power_law_1.01,0.07563520073890687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,2,power_law_1.01,0.0818560004234314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,2,power_law_1.01,0.09340159893035889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,2,power_law_1.01,0.10189440250396728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,2,power_law_1.01,0.11961599588394164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,2,power_law_1.01,0.14312319755554198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,2,power_law_1.01,0.1535423994064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,2,power_law_1.01,0.17662080526351928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,2,power_law_1.01,0.1796223998069763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,2,power_law_1.01,0.18038400411605834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,2,power_law_1.01,0.17850879430770875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,2,power_law_1.01,0.18472319841384888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,2,power_law_1.01,0.19861760139465331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,2,power_law_1.01,0.20102400779724122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,2,power_law_1.01,0.20904960632324218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,2,power_law_1.01,0.21774721145629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,2,power_law_1.01,0.23255679607391358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,2,power_law_1.01,0.2550784111022949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,2,power_law_1.01,0.2863424062728882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,2,power_law_1.01,0.3382719993591309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,2,power_law_1.01,0.39083518981933596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,2,power_law_1.01,0.5172160148620606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,2,power_law_1.01,0.6372288227081299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,2,power_law_1.01,0.8963264465332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,2,power_law_1.01,1.0892095565795898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,2,power_law_1.01,1.5737215995788574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,2,power_law_1.01,2.0182464599609373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,1,balanced,0.09891200065612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,1,balanced,0.15187199910481772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,1,balanced,0.25697600841522217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,1,balanced,0.4763840039571126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,1,balanced,0.8962240219116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,1,balanced,1.1087360382080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,1,balanced,1.1132266521453857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,1,balanced,1.11900798479716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,1,balanced,1.1237066586812336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,1,balanced,1.1279306411743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,1,balanced,1.1332106590270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,1,balanced,1.141434669494629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,1,balanced,1.1470879713694255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,1,balanced,1.156213363011678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,1,balanced,1.1780160268147786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,1,balanced,1.1838826338450115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,1,balanced,1.211194674173991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,1,balanced,1.3702239990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,1,balanced,1.2905920346577961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,1,balanced,1.5823520024617512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,1,balanced,1.4755627314249675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,1,balanced,1.8758452733357747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,1,balanced,1.8665653864542644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,1,balanced,2.936410586039225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,1,balanced,3.04097588857015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,1,balanced,5.335594813028972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,1,balanced,5.583712259928386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,64,power_law_1.01,0.08284800052642823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,64,power_law_1.01,0.08288639783859253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,64,power_law_1.01,0.08014079928398132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,64,power_law_1.01,0.08458240032196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,64,power_law_1.01,0.07843199968338013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,64,power_law_1.01,0.08640639781951905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,64,power_law_1.01,0.08335360288619995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,64,power_law_1.01,0.08272640109062195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,64,power_law_1.01,0.08570880293846131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,64,power_law_1.01,0.0852288007736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,64,power_law_1.01,0.08280320167541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,64,power_law_1.01,0.08567039966583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,1,balanced,0.1025279959042867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,64,power_law_1.01,0.08700799942016602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,64,power_law_1.01,0.08805760145187377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,64,power_law_1.01,0.09354239702224731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,1,balanced,0.1072160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,1,balanced,0.11899200081825256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,64,power_law_1.01,0.0869055986404419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,1,balanced,0.14366933703422546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,1,balanced,0.19374932845433554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,1,balanced,0.28244266907374066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,1,balanced,0.36846399307250977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,1,balanced,0.3667573531468709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,1,balanced,0.3732373317082723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,1,balanced,0.3758293390274048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,1,balanced,0.3752799828847249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,1,balanced,0.38239999612172443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,64,power_law_1.01,0.09635840058326721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,1,balanced,0.38900800546010333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,64,power_law_1.01,0.10880000591278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,1,balanced,0.3945600191752116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,1,balanced,0.4060320059458415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,64,power_law_1.01,0.11733119487762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,64,power_law_1.01,0.14136960506439208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,64,power_law_1.01,0.15520000457763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,64,power_law_1.01,0.20316801071166993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,64,power_law_1.01,0.24362239837646485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,1,balanced,0.41581865151723224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,64,power_law_1.01,0.33121280670166015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,64,balanced,0.061066667238871254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,64,balanced,0.06039999922116598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,64,balanced,0.06098133325576782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,64,power_law_1.01,0.4041855812072754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,64,power_law_1.01,0.5650176048278809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,64,power_law_1.01,0.7798975944519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,1,balanced,0.4298826853434245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,1,balanced,0.47925865650177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,64,balanced,0.06419733166694641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,64,balanced,0.06712000072002411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,64,balanced,0.08658666412035625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,64,balanced,0.09608532985051473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,64,balanced,0.10057066877683003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,64,balanced,0.08780800302823384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,64,balanced,0.08965333302815755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,64,balanced,0.09504533807436626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,64,balanced,0.09066667159398396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,64,balanced,0.0926080048084259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,64,balanced,0.09186666210492452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,64,balanced,0.0974079966545105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,64,balanced,0.09675733248392741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,64,balanced,0.09711466232935588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,64,balanced,0.10495466987291972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,64,balanced,0.10707199573516846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,64,balanced,0.11905599633852641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,1,balanced,0.5165813366572062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,1,balanced,0.597760001818339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,1,balanced,0.672922690709432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,1,balanced,0.8428053061167399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,1,balanced,1.017407973607381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,1,balanced,1.3985279401143391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,64,balanced,0.12753066420555115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,64,balanced,0.14870933691660562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,1,balanced,1.8857173919677734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,64,balanced,0.1694773236910502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,64,balanced,0.21358933051427206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,64,balanced,0.25172267357508343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,64,balanced,0.35124266147613525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,1,balanced,2.6570399602254233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,64,balanced,0.4328266779581706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,1,balanced,3.521477381388346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,64,power_law_1.01,0.1017408013343811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,64,power_law_1.01,0.09879680275917054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,64,power_law_1.01,0.09816960096359253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,64,power_law_1.01,0.10404479503631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,64,power_law_1.01,0.10477440357208252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,64,power_law_1.01,0.10616960525512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,64,power_law_1.01,0.10944000482559205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,64,power_law_1.01,0.1155392050743103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,64,power_law_1.01,0.11311999559402466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,64,power_law_1.01,0.11962239742279053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,64,power_law_1.01,0.11403520107269287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,64,power_law_1.01,0.11737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,64,power_law_1.01,0.1235967993736267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,64,power_law_1.01,0.11923199892044067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,64,power_law_1.01,0.13096319437026976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,64,power_law_1.01,0.12726399898529053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,64,power_law_1.01,0.140774405002594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,64,power_law_1.01,0.15879039764404296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,64,power_law_1.01,0.1720576047897339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,64,power_law_1.01,0.22274560928344728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,64,power_law_1.01,0.26068480014801027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,64,power_law_1.01,0.35199360847473143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,64,power_law_1.01,0.3944256067276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,64,power_law_1.01,0.546131181716919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,64,power_law_1.01,0.6763455867767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,64,power_law_1.01,1.0111616134643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,64,power_law_1.01,1.396735954284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,64,power_law_1.2,0.08275840282440186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,64,power_law_1.2,0.08036479949951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,64,power_law_1.2,0.07745280265808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,64,power_law_1.2,0.07952640056610108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,64,power_law_1.2,0.08072320222854615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,64,power_law_1.2,0.07102720141410827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,64,power_law_1.2,0.07745919823646545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,64,power_law_1.2,0.08271359801292419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,64,power_law_1.2,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,64,power_law_1.2,0.08210560083389282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,64,power_law_1.2,0.0800704002380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,64,power_law_1.2,0.08034560084342957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,64,power_law_1.2,0.08526080250740051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,64,power_law_1.2,0.08594560027122497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,64,power_law_1.2,0.09602559804916382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,64,power_law_1.2,0.09471359848976135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,64,power_law_1.2,0.10069760084152221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,64,power_law_1.2,0.11354240179061889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,64,power_law_1.2,0.12222080230712891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,64,power_law_1.2,0.1470720052719116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,64,power_law_1.2,0.1703935980796814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,64,power_law_1.2,0.2111680030822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,64,power_law_1.2,0.2556159973144531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,64,power_law_1.2,0.35844480991363525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,64,power_law_1.2,0.4684288024902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,64,power_law_1.2,0.6632063865661622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,64,power_law_1.2,0.8734975814819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,32,balanced,0.02145066608985265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,32,balanced,0.02332799881696701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,32,balanced,0.021018666525681812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,32,balanced,0.02126399924357732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,32,balanced,0.02311466634273529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,32,balanced,0.02508266766866048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,32,balanced,0.027242665489514668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,32,balanced,0.02661866694688797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,32,balanced,0.02957333376010259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,32,balanced,0.02921066681543986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,32,balanced,0.031498665610949196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,32,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,32,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,32,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,32,balanced,0.06155733267466227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,32,balanced,0.060906668504079185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,32,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,32,balanced,0.06190933287143707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,32,balanced,0.07428800066312154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,32,balanced,0.09121066331863403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,32,balanced,0.10825600226720174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,32,balanced,0.1406880021095276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,32,balanced,0.17432000239690146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,32,balanced,0.23695466915766397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,32,balanced,0.30558933814366657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,32,balanced,0.44511465231577557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,32,balanced,0.5738133192062378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,2,power_law_1.01,0.01966720074415207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,2,power_law_1.01,0.019200000166893005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,2,power_law_1.01,0.018681600689888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,2,power_law_1.01,0.018515199422836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,2,power_law_1.01,0.0191551998257637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,2,power_law_1.01,0.020294399559497835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,2,power_law_1.01,0.026015999913215637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,2,power_law_1.01,0.024326400458812715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,2,power_law_1.01,0.02810879945755005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,2,power_law_1.01,0.02683520019054413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,2,power_law_1.01,0.027807998657226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,2,power_law_1.01,0.029465600848197937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,2,power_law_1.01,0.03333759903907776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,2,power_law_1.01,0.034969601035118106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,2,power_law_1.01,0.04456959962844849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,2,power_law_1.01,0.04264959990978241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,2,power_law_1.01,0.04430719912052154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,2,power_law_1.01,0.05742719769477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.01,0.06572160124778748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.01,0.08613119721412658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,4,power_law_1.2,0.019871999323368073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,4,power_law_1.2,0.024275200068950654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,4,power_law_1.2,0.026092800498008727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,4,power_law_1.2,0.03738240003585815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,4,power_law_1.2,0.0377344012260437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,4,power_law_1.2,0.04659200012683869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.01,0.10785280466079712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,4,power_law_1.2,0.05240960121154785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,4,power_law_1.2,0.05141119956970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,4,power_law_1.2,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,4,power_law_1.2,0.0538752019405365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,4,power_law_1.2,0.05804160237312317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,4,power_law_1.2,0.05989120006561279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,4,power_law_1.2,0.05939199924468994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,4,power_law_1.2,0.06050559878349304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,4,power_law_1.2,0.07375360131263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,4,power_law_1.2,0.07972480058670044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,4,power_law_1.2,0.0753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,4,power_law_1.2,0.1195520043373108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.01,0.14596480131149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,4,power_law_1.2,0.13304959535598754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,4,power_law_1.2,0.130457603931427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,4,power_law_1.2,0.127347195148468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,4,power_law_1.2,0.1366976022720337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,4,power_law_1.2,0.15646719932556152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,4,power_law_1.2,0.19827200174331666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,4,power_law_1.2,0.23482880592346192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,4,power_law_1.2,0.318342399597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,4,power_law_1.2,0.4295616149902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.01,0.17908480167388915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.01,0.25471360683441163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.01,0.32760961055755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.01,0.4734079837799072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.01,0.6202623844146729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,1,power_law_1.2,0.03594880104064942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,1,power_law_1.2,0.035641598701477054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,1,power_law_1.2,0.03727999925613403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,1,power_law_1.2,0.043750399351119997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,1,power_law_1.2,0.04968959987163544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,1,power_law_1.2,0.06601600050926208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,1,power_law_1.2,0.06825600266456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,1,power_law_1.2,0.06863359808921814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,1,power_law_1.2,0.07195519804954528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,1,power_law_1.2,0.07256320118904114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,1,power_law_1.2,0.07708799839019775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,1,power_law_1.2,0.08199040293693542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,1,power_law_1.2,0.0846015989780426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,1,power_law_1.2,0.08810880184173583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,1,power_law_1.2,0.09442560076713562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,1,power_law_1.2,0.0989184021949768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,1,power_law_1.2,0.10810240507125854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,1,power_law_1.2,0.13137919902801515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,1,power_law_1.2,0.15312000513076782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,1,power_law_1.2,0.19708160161972046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,1,power_law_1.2,0.23519999980926515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,1,power_law_1.2,0.324236798286438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,1,power_law_1.2,0.4093632221221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,1,power_law_1.2,0.5865856170654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,1,power_law_1.2,0.7619840145111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,1,power_law_1.2,1.1180031776428223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,1,power_law_1.2,1.4599743843078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,32,balanced,0.056261335810025535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,32,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,32,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,32,balanced,0.0687306672334671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,32,balanced,0.08453333377838135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,32,balanced,0.11740799744923909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,32,balanced,0.11822932958602905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,32,balanced,0.116565336783727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,32,balanced,0.12008532881736755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,32,balanced,0.11829333504041036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,32,balanced,0.11955733100573222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,32,balanced,0.12505066394805908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,32,balanced,0.12408533692359924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,32,balanced,0.12739200393358865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,32,balanced,0.14655466874440512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,32,balanced,0.14678399761517844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,32,balanced,0.14958932995796204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,32,balanced,0.1625493367513021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,32,balanced,0.16687999169031778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,32,balanced,0.1819360057512919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,32,balanced,0.1966773271560669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,32,balanced,0.2288586695988973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,32,balanced,0.2549813389778137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,32,balanced,0.33060266574223834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,32,balanced,0.3823733329772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,32,balanced,0.5280746618906657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,32,balanced,0.6667146682739258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,32,power_law_1.2,0.10703999996185302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,32,power_law_1.2,0.10089600086212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,32,power_law_1.2,0.0932416021823883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,32,power_law_1.2,0.09407359957695008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,32,power_law_1.2,0.09198079705238342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,32,power_law_1.2,0.08039039969444275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,32,power_law_1.2,0.09028480052947999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,32,power_law_1.2,0.09256319999694824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,32,power_law_1.2,0.09371520280838012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,32,power_law_1.2,0.08935040235519409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,32,power_law_1.2,0.0950655996799469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,32,power_law_1.2,0.09339519739151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,32,power_law_1.2,0.10191359519958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,32,power_law_1.2,0.10202879905700683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,32,power_law_1.2,0.10782079696655274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,32,power_law_1.2,0.1109120011329651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,32,power_law_1.2,0.12494720220565796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,32,power_law_1.2,0.1344256043434143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,32,power_law_1.2,0.14421119689941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,32,power_law_1.2,0.18856960535049438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,32,power_law_1.2,0.20941441059112548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,32,power_law_1.2,0.2506239891052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,32,power_law_1.2,0.279532790184021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,8,balanced,0.05190399785836538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,32,power_law_1.2,0.3616640090942383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,8,balanced,0.07268266876538594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,8,balanced,0.09923199812571208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,8,balanced,0.14973866939544678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,32,power_law_1.2,0.47777280807495115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,8,balanced,0.2494666576385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,8,balanced,0.44150400161743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,8,balanced,0.44379734992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,8,balanced,0.44196800390879315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,8,balanced,0.4503360191980998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,8,balanced,0.4476906855901082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,8,balanced,0.4485493501027425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,8,balanced,0.44938135147094727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,8,balanced,0.4524373213450114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,8,balanced,0.4588213364283244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,8,balanced,0.4647839864095052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,8,balanced,0.46458133061726886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,32,power_law_1.2,0.7579967975616455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,8,balanced,0.46931199232737225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,8,balanced,0.4774080117543538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,32,power_law_1.2,0.7863743782043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,8,balanced,0.49026668071746826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,8,balanced,0.5078826745351156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,8,balanced,0.5363253355026245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,64,power_law_1.01,0.025171199440956117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,8,balanced,0.571669340133667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,64,power_law_1.01,0.026291200518608095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,8,balanced,0.6066453456878662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,64,power_law_1.01,0.02677760124206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,8,balanced,0.7541813055674235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,64,power_law_1.01,0.02725760042667389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,8,power_law_1.2,0.04601599872112274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,8,balanced,0.7804426352183024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,64,power_law_1.01,0.027212798595428467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,8,power_law_1.2,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,64,power_law_1.01,0.027718400955200194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,8,balanced,1.104149341583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,64,power_law_1.01,0.028019198775291444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,64,power_law_1.01,0.03112959861755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,8,balanced,1.211957295735677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,64,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,64,power_law_1.01,0.03375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,64,power_law_1.01,0.03482879996299744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,64,power_law_1.01,0.03568640053272247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,64,power_law_1.01,0.03603839874267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,64,power_law_1.01,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,64,power_law_1.01,0.036627200245857236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,64,power_law_1.01,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,64,power_law_1.01,0.03896960020065308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,64,power_law_1.01,0.04232319891452789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,64,power_law_1.01,0.04466559886932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,64,power_law_1.01,0.05004159808158874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,64,power_law_1.01,0.051596802473068235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,64,power_law_1.01,0.060211199522018435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,64,power_law_1.01,0.07098879814147949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,64,power_law_1.01,0.09403520226478576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,8,power_law_1.2,0.04635519981384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,64,power_law_1.01,0.11645439863204957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,8,power_law_1.2,0.04513919949531555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,64,power_law_1.01,0.1603008031845093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,8,power_law_1.2,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,64,power_law_1.01,0.1938752055168152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,8,power_law_1.2,0.04714879989624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,8,power_law_1.2,0.047295999526977536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,8,power_law_1.2,0.04843519926071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,8,power_law_1.2,0.047942399978637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,8,power_law_1.2,0.04824959933757782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,8,power_law_1.2,0.04968959987163544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,8,power_law_1.2,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,8,power_law_1.2,0.05214080214500427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,8,power_law_1.2,0.05294079780578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,8,power_law_1.2,0.05744640231132507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,8,power_law_1.2,0.058771198987960814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,8,power_law_1.2,0.06224640011787415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,8,power_law_1.2,0.07144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,8,power_law_1.2,0.08042880296707153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,8,power_law_1.2,0.09438719749450683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,8,power_law_1.2,0.11089279651641845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,8,power_law_1.2,0.13663359880447387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,8,power_law_1.2,0.18035199642181396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,8,power_law_1.2,0.2508415937423706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,8,power_law_1.2,0.349068808555603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,8,power_law_1.2,0.5127295970916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,8,power_law_1.2,0.6284800052642823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,2,balanced,0.058101331194241844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,2,balanced,0.058970664938290916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,2,balanced,0.06400000055631001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,2,balanced,0.08062933385372162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,2,balanced,0.10769599676132202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,2,balanced,0.10834133625030518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,2,balanced,0.10894399881362915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,2,balanced,0.10931733250617981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,2,balanced,0.11105599999427795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,2,balanced,0.11239467064539592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,2,balanced,0.11185600360234578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,2,balanced,0.11771733562151591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,2,balanced,0.1174773375193278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,2,balanced,0.1190826694170634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,2,balanced,0.11998933553695679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,2,balanced,0.12229866782824199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,2,balanced,0.12379200259844463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,2,balanced,0.13312533497810364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,2,balanced,0.14069333672523499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,2,balanced,0.15858667095502219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,2,balanced,0.17243733008702597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,2,balanced,0.23976532618204752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,2,balanced,0.27993067105611164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,2,balanced,0.39770666758219403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,2,balanced,0.518618663152059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,2,balanced,0.7409119606018066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,2,balanced,0.971717357635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,2,power_law_1.01,0.049593600630760196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,2,power_law_1.01,0.07544320225715637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,2,power_law_1.01,0.11253119707107544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,2,power_law_1.01,0.1752832055091858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,2,power_law_1.01,0.2342463970184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,2,power_law_1.01,0.27320959568023684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,2,power_law_1.01,0.3653055906295776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,2,power_law_1.01,0.36364800930023194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,2,power_law_1.01,0.36456320285797117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,2,power_law_1.01,0.3913088083267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,2,power_law_1.01,0.4058112144470215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,2,power_law_1.01,0.41294078826904296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,2,power_law_1.01,0.43796482086181643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,2,power_law_1.01,0.42595200538635253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,2,power_law_1.01,0.44428157806396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,2,power_law_1.01,0.4624512195587158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,2,power_law_1.01,0.45803518295288087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,2,power_law_1.01,0.524671983718872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.01,0.5873536109924317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.01,0.7315135955810547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.01,0.680620813369751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.01,0.8319104194641114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.01,0.9651840209960938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.01,1.1552639961242677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.01,1.3690239906311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.01,1.7794944763183593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.01,2.211846351623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,32,power_law_1.01,0.05045120120048523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,32,power_law_1.01,0.05128960013389587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,32,power_law_1.01,0.05003520250320435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,32,power_law_1.01,0.05278720259666443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,32,power_law_1.01,0.05280640125274658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,32,power_law_1.01,0.05371519923210144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,32,power_law_1.01,0.05456640124320984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,32,power_law_1.01,0.05511040091514587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,32,power_law_1.01,0.055199998617172244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,128,power_law_1.2,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,32,power_law_1.01,0.05553280115127564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,128,power_law_1.2,0.039673599600791934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,128,power_law_1.2,0.03960959911346436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,128,power_law_1.2,0.04222719967365265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,32,power_law_1.01,0.05597440004348755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,128,power_law_1.2,0.04249599874019623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,32,power_law_1.01,0.05761920213699341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,128,power_law_1.2,0.044947201013565065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,32,power_law_1.01,0.0588096022605896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,32,power_law_1.01,0.05937920212745666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,128,power_law_1.2,0.04531840085983276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,32,power_law_1.01,0.06618239879608154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,128,power_law_1.2,0.04649600088596344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,128,power_law_1.2,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,32,power_law_1.01,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,128,power_law_1.2,0.047513601183891294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,64,balanced,0.05327466626962026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,32,power_law_1.01,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,128,power_law_1.2,0.049267199635505673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,32,power_law_1.01,0.08035200238227844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,128,power_law_1.2,0.05135999917984009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,128,power_law_1.2,0.05011199712753296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,128,power_law_1.2,0.05489280223846436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,128,power_law_1.2,0.06154879927635193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,128,power_law_1.2,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,64,balanced,0.053904001911481224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,64,balanced,0.05402666827042898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,128,power_law_1.2,0.06495360136032105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,128,power_law_1.2,0.07588480114936828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,128,power_law_1.2,0.08134400248527526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,128,power_law_1.2,0.10080000162124633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,128,power_law_1.2,0.11828479766845704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,64,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,128,power_law_1.2,0.1694975972175598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,64,balanced,0.05816000203291575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,64,balanced,0.06451199948787689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,64,balanced,0.06607999900976817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,64,balanced,0.0679253339767456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,128,power_law_1.2,0.21639680862426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,64,balanced,0.06657066444555919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,64,balanced,0.06646400193373363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,64,balanced,0.06876799960931142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,64,balanced,0.06846400101979573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,64,balanced,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,64,balanced,0.07260799904664357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,64,balanced,0.08070933322111766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,64,balanced,0.07840000092983246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,64,balanced,0.08690133690834045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,64,balanced,0.09821333487828572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,64,balanced,0.11317333579063416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.01,0.09164800047874451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,64,balanced,0.13822399576505026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,64,balanced,0.1583093305428823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.01,0.1122431993484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,64,balanced,0.20493332544962564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,64,balanced,0.24452799558639526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.01,0.12320640087127685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,128,power_law_1.2,0.3247936010360718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,64,balanced,0.3290666739145915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.01,0.15765119791030885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,64,balanced,0.41731735070546466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,128,power_law_1.2,0.38360960483551027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,64,balanced,0.5846773386001587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.01,0.19023360013961793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,64,balanced,0.7552266915639242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,128,power_law_1.2,0.6674496173858643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,128,power_law_1.2,0.8370944023132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.01,0.24960639476776122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.01,0.32835841178894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.01,0.4277184009552002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.01,0.5924479961395264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,32,balanced,0.050245334704717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,32,balanced,0.05009066561857859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,32,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,32,balanced,0.0562666654586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,32,balanced,0.05596266686916351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,32,balanced,0.06830400228500366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,32,balanced,0.06819733480612437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,32,balanced,0.0684746652841568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,32,balanced,0.06612266600131989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,32,balanced,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,32,balanced,0.06987200180689494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,32,balanced,0.06881066660086314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,32,balanced,0.07086933155854543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,32,balanced,0.07259733478228252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,32,balanced,0.0767146646976471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,32,balanced,0.07881066699822743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,32,balanced,0.08071466783682506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,32,balanced,0.09314666191736858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,32,balanced,0.10513599713643391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,32,balanced,0.12582932909329733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,32,balanced,0.14058132966359457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,32,balanced,0.17706666390101114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,32,balanced,0.20851733287175497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,32,balanced,0.28249067068099976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,32,balanced,0.3484373490015666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,32,balanced,0.49287466208140057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,32,balanced,0.6265706618626913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,8,balanced,0.05291733145713806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,8,balanced,0.055829331278800964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,8,balanced,0.058703998724619545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,8,balanced,0.0603359987338384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,8,balanced,0.06650666892528534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,8,balanced,0.07978666822115581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,8,balanced,0.09073600172996521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,8,balanced,0.09061333537101746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,8,balanced,0.09161600470542908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,8,balanced,0.09309867024421692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,8,balanced,0.09271466732025146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,8,balanced,0.09542399644851685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,8,balanced,0.09706133604049683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,8,balanced,0.09916266798973083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,8,balanced,0.1032373309135437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,8,balanced,0.1062506635983785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,8,balanced,0.11342400312423706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,8,balanced,0.12178132931391399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,8,balanced,0.13396267096201578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,8,balanced,0.1613653302192688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,8,balanced,0.1777013341585795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,8,balanced,0.22171199321746826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,8,balanced,0.27727999289830524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,8,balanced,0.35970131556193036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,8,balanced,0.4790453513463338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,8,balanced,0.6479626496632894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,8,balanced,0.850810686747233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,32,power_law_1.01,0.021907199919223786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,32,power_law_1.01,0.021753600239753722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,32,power_law_1.01,0.02194560021162033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,32,power_law_1.01,0.02200320065021515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,32,power_law_1.01,0.021792000532150267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,32,power_law_1.01,0.023104000091552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,32,power_law_1.01,0.02794879972934723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,32,power_law_1.01,0.02885119915008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,32,power_law_1.01,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,32,power_law_1.01,0.05309439897537231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,32,power_law_1.01,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,32,power_law_1.01,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,32,power_law_1.01,0.04503679871559143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,32,power_law_1.01,0.04434559941291809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,32,power_law_1.01,0.04615040123462677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,32,power_law_1.01,0.04714879989624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,32,power_law_1.01,0.0497408002614975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,32,power_law_1.01,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.01,0.058182400465011594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.01,0.06903679966926575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.01,0.07656319737434387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.01,0.09815679788589478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.01,0.12111999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.01,0.167302405834198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.01,0.2011647939682007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.01,0.2818495988845825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.01,0.35824639797210694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,32,balanced,0.021509334444999695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,32,balanced,0.025231999655564625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,32,balanced,0.023050665855407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,32,balanced,0.023071999351183575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,32,balanced,0.02516799916823705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,32,balanced,0.02532800038655599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,32,balanced,0.028245332340399425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,8,power_law_1.2,0.02948479950428009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,8,power_law_1.2,0.04749439954757691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,8,power_law_1.2,0.053862398862838744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,8,power_law_1.2,0.06152960062026978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,8,power_law_1.2,0.07260800004005433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,32,balanced,0.027274665733178455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,32,balanced,0.0313226655125618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,32,balanced,0.031008000175158184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,8,power_law_1.2,0.07857279777526856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,32,balanced,0.03326933334271113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,32,balanced,0.03919466584920883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,32,balanced,0.036346666514873505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,32,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,32,balanced,0.06198399762312571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,32,balanced,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,32,balanced,0.059978668888409935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,32,balanced,0.06968533496061961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,32,balanced,0.08282666901747386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,32,balanced,0.09982933600743611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,32,balanced,0.1169706682364146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,32,balanced,0.15296000242233276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,8,power_law_1.2,0.06881279945373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,32,balanced,0.19294933478037515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,8,power_law_1.2,0.07158399820327759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,32,balanced,0.26558399200439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,32,balanced,0.3429386615753174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,8,power_law_1.2,0.07896320223808288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,32,balanced,0.5061706701914469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,8,power_law_1.2,0.0800320029258728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,32,balanced,0.6663680076599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,8,power_law_1.2,0.07976319789886474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,8,power_law_1.2,0.0987775981426239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,8,power_law_1.2,0.09760640263557434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,8,power_law_1.2,0.09022079706192017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,8,power_law_1.2,0.15614720582962036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,8,power_law_1.2,0.16136959791183472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,8,power_law_1.2,0.16792320013046264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,8,power_law_1.2,0.14753279685974122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.2,0.1532096028327942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.2,0.1732159972190857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.2,0.17952640056610109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.2,0.213753604888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.2,0.2583168029785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.2,0.3287168025970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.2,0.4196800231933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.2,0.5352255821228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.2,0.7936704158782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,32,power_law_1.2,0.05772799849510193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,32,power_law_1.2,0.053887999057769774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,32,power_law_1.2,0.051846402883529666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,32,power_law_1.2,0.05389440059661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,32,power_law_1.2,0.05437440276145935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,32,power_law_1.2,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,32,power_law_1.2,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,32,power_law_1.2,0.059084802865982056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,32,power_law_1.2,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,32,power_law_1.2,0.05939199924468994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,32,power_law_1.2,0.06117759943008423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,32,power_law_1.2,0.06424959897994995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,32,power_law_1.2,0.06578559875488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,32,power_law_1.2,0.0669376015663147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,32,power_law_1.2,0.0775488018989563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,32,power_law_1.2,0.077920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,32,power_law_1.2,0.08731520175933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,32,power_law_1.2,0.10545920133590699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,32,power_law_1.2,0.10893440246582031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,32,power_law_1.2,0.1360576033592224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,32,power_law_1.2,0.16135040521621705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,32,power_law_1.2,0.2106112003326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,32,power_law_1.2,0.253657603263855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,32,power_law_1.2,0.3778111934661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,32,power_law_1.2,0.4993279933929443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,32,power_law_1.2,0.7534592151641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,32,power_law_1.2,1.0433728218078613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,8,power_law_1.01,0.070169597864151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,8,power_law_1.01,0.09618560075759888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,8,power_law_1.01,0.12201600074768067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,8,power_law_1.01,0.1124735951423645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,8,power_law_1.01,0.13207679986953735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,8,power_law_1.01,0.14551680088043212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,8,power_law_1.01,0.15128320455551147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,8,power_law_1.01,0.15714559555053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,8,power_law_1.01,0.14281599521636962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,8,power_law_1.01,0.13601280450820924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,8,power_law_1.01,0.1493183970451355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,8,power_law_1.01,0.146560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,8,power_law_1.01,0.1566655993461609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,8,power_law_1.01,0.158188796043396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,8,power_law_1.01,0.1592128038406372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,8,power_law_1.01,0.16672639846801757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,8,power_law_1.01,0.1729920029640198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,8,power_law_1.01,0.1893887996673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,8,power_law_1.01,0.20526719093322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,8,power_law_1.01,0.2500992059707642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,8,power_law_1.01,0.27355520725250243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,8,power_law_1.01,0.35662078857421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,8,power_law_1.01,0.3983295917510986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,8,power_law_1.01,0.5073344230651855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,8,power_law_1.01,0.6588543891906739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,8,power_law_1.01,0.9085375785827636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,8,power_law_1.01,1.1705792427062989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,2,power_law_1.01,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,2,power_law_1.01,0.07296640276908875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,2,power_law_1.01,0.0870848000049591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,2,power_law_1.01,0.11562240123748779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,2,power_law_1.01,0.14120320081710816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,2,power_law_1.01,0.16866559982299806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,2,power_law_1.01,0.2145983934402466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,2,power_law_1.01,0.21734399795532228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,2,power_law_1.01,0.22091519832611084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,2,power_law_1.01,0.23135359287261964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,2,power_law_1.01,0.2346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,2,power_law_1.01,0.2403007984161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,2,power_law_1.01,0.2459007978439331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,2,power_law_1.01,0.2533440113067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,2,power_law_1.01,0.26299519538879396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,2,power_law_1.01,0.28361599445343016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,2,power_law_1.01,0.29732480049133303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,2,power_law_1.01,0.32561919689178465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,2,power_law_1.01,0.3474751949310303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,2,power_law_1.01,0.40752639770507815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,2,power_law_1.01,0.4664576053619385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,2,power_law_1.01,0.5786303997039794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,2,power_law_1.01,0.6781311988830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,2,power_law_1.01,0.8940223693847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,2,power_law_1.01,1.1752960205078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,2,power_law_1.01,1.6023040771484376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,2,power_law_1.01,1.9114879608154296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,4,power_law_1.01,0.047193598747253415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,4,power_law_1.01,0.05128960013389587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,4,power_law_1.01,0.06021760106086731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,4,power_law_1.01,0.07366399765014649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,4,power_law_1.01,0.08103039860725403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,4,power_law_1.01,0.08641279935836792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,4,power_law_1.01,0.1018496036529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,4,power_law_1.01,0.09837440252304078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,4,power_law_1.01,0.09973120093345642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,4,power_law_1.01,0.10035200119018554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,4,power_law_1.01,0.10713599920272827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,4,power_law_1.01,0.1096127986907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,4,power_law_1.01,0.11374720335006713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,16,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,4,power_law_1.01,0.11893119812011718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,16,balanced,0.04791999856630961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,16,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,4,power_law_1.01,0.12780799865722656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,16,balanced,0.05409599840641022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,1,balanced,0.0562666654586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,1,balanced,0.059989333152770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,1,balanced,0.06825066606203715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,1,balanced,0.10223467151323955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,1,balanced,0.1459999978542328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,1,balanced,0.22438400983810425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,1,balanced,0.29839466015497845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,1,balanced,0.2996693253517151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,1,balanced,0.3030186692873637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,1,balanced,0.3057653307914734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,1,balanced,0.3094080090522766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,16,balanced,0.05590933561325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,16,balanced,0.07260266443093617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,16,balanced,0.07415999968846639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,16,balanced,0.07311466832955678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,16,balanced,0.07584533095359802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,16,balanced,0.073253333568573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,16,balanced,0.07605333129564922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,16,balanced,0.07828799883524577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,16,balanced,0.07784533500671387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,16,balanced,0.0792799989382426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,16,balanced,0.08413333694140117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,16,balanced,0.08450667063395183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,16,balanced,0.08689066767692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,16,balanced,0.09633066256841023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,16,balanced,0.09878399968147278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,16,balanced,0.11211733023325603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,16,balanced,0.12447999914487202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,16,balanced,0.14860799908638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,16,balanced,0.1732693314552307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,16,balanced,0.23357866207758585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,16,balanced,0.2800106604894002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,1,balanced,0.3171839912732442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,1,balanced,0.3230453332265218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,1,balanced,0.3285173376401265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,1,balanced,0.3389013210932414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,1,balanced,0.3466613292694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,1,balanced,0.36774933338165283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,1,balanced,0.4187413454055786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,1,balanced,0.4538666804631551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,4,power_law_1.01,0.13552000522613525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,1,balanced,0.5352426767349243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,4,power_law_1.01,0.14101120233535766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,1,balanced,0.6031680107116699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,16,balanced,0.3904800017674764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,4,power_law_1.01,0.1595199942588806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,1,balanced,0.7729120254516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,16,balanced,0.5030986467997233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,4,power_law_1.01,0.1769600033760071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,1,balanced,0.9749600092569987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,4,power_law_1.01,0.2129472017288208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,1,balanced,1.297866662343343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,4,power_law_1.01,0.2572223901748657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,4,power_law_1.01,0.3192512035369873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,1,balanced,1.7473492622375488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,4,power_law_1.01,0.39800961017608644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,1,balanced,2.4789439837137857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,4,power_law_1.01,0.5457791805267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,4,power_law_1.01,0.7015999794006348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,1,balanced,3.247472127278646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,4,power_law_1.01,0.9497023582458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,4,power_law_1.01,1.3140671730041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,64,balanced,0.05212266743183136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,64,balanced,0.05230933427810669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,64,balanced,0.05397333204746246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,64,balanced,0.0561653325955073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,64,balanced,0.06217599908510844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,64,balanced,0.0804799993832906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,64,balanced,0.0806879997253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,64,balanced,0.08099733293056488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,64,balanced,0.08303466439247131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,64,balanced,0.08132799963156383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,64,balanced,0.08273066580295563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,64,balanced,0.08844266335169475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,64,balanced,0.08885332942008972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,64,balanced,0.08751466870307922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,64,balanced,0.09873599807421367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,64,balanced,0.09941333532333374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,64,balanced,0.10085333387056987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,64,balanced,0.11109333237012227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,64,balanced,0.11537599563598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,64,balanced,0.12662399808565775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,64,balanced,0.14012266198794046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,64,balanced,0.17108267545700073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,64,balanced,0.19714132944742838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,64,balanced,0.266154666741689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,64,balanced,0.3123040000597636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,64,balanced,0.4126933415730794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,64,balanced,0.5296639998753866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,8,balanced,0.05146666864554087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,8,balanced,0.051872000098228455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,8,balanced,0.05376000205675761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,8,balanced,0.05726933479309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,8,balanced,0.06407999992370605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,8,balanced,0.08063466846942902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,8,balanced,0.08212266862392426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,8,balanced,0.08064533273379008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,8,balanced,0.08304533362388611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,8,balanced,0.0806933343410492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,8,balanced,0.0836906631787618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,8,balanced,0.08513066172599792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,8,balanced,0.08463467160860698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,8,balanced,0.08718933661778767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,8,balanced,0.09224533041318257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,8,balanced,0.092357337474823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,8,balanced,0.09628799557685852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,8,balanced,0.10346666971842448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,8,balanced,0.10867200295130412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,8,balanced,0.12404800454775493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,8,balanced,0.1402079959710439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,8,balanced,0.16592533389727274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,8,balanced,0.1926986575126648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,8,balanced,0.2630186676979065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,8,balanced,0.32267733414967853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,8,balanced,0.46727466583251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,8,balanced,0.6052800019582113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,4,power_law_1.2,0.04731520116329193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,4,power_law_1.2,0.047788798809051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,4,power_law_1.2,0.054099202156066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,4,power_law_1.2,0.057158398628234866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,4,power_law_1.2,0.062483197450637816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,4,power_law_1.2,0.06846079826354981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,4,power_law_1.2,0.07340160012245178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,4,power_law_1.2,0.07468159794807434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,4,power_law_1.2,0.07468799948692321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,4,power_law_1.2,0.0766975998878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,4,power_law_1.2,0.08052480220794678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,4,power_law_1.2,0.08146560192108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,4,power_law_1.2,0.08465920090675354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,4,power_law_1.2,0.08618239760398864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,4,power_law_1.2,0.09550719857215881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,4,power_law_1.2,0.09806079864501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,4,power_law_1.2,0.10550400018692016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,4,power_law_1.2,0.12003840208053589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,4,power_law_1.2,0.13388160467147828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,4,power_law_1.2,0.17178879976272582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,4,power_law_1.2,0.2188800096511841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,4,power_law_1.2,0.2783423900604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,4,power_law_1.2,0.3658751964569092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,4,power_law_1.2,0.44741120338439944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,4,power_law_1.2,0.621395206451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,4,power_law_1.2,0.8315391540527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,1,balanced,0.06551999847094218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,1,balanced,0.0930560032526652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,1,balanced,0.13915200034777322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,1,balanced,0.14325867096583048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,1,balanced,0.1455839971701304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,1,balanced,0.1455519994099935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,1,balanced,0.14662399888038635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,1,balanced,0.14831466476122537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,1,balanced,0.14884799718856812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,1,balanced,0.14863999684651694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,1,balanced,0.15130133430163065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,1,balanced,0.1527359982331594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,1,balanced,0.15737066666285196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,1,balanced,0.16296533743540445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,1,balanced,0.1715893348058065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,1,balanced,0.1772693395614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,1,balanced,0.18517333269119263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,1,balanced,0.2272533377011617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,1,balanced,0.2434613307317098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,1,balanced,0.3114666740099589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,1,balanced,0.38126933574676514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,1,balanced,0.5165813366572062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,1,balanced,0.6600159804026285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,1,balanced,0.9338613351186117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,1,balanced,1.199280023574829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,8,power_law_1.2,0.060166400671005246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,8,power_law_1.2,0.06707199811935424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,8,power_law_1.2,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,8,power_law_1.2,0.07329919934272766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,8,power_law_1.2,0.07709439992904663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,8,power_law_1.2,0.0884607970714569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,8,power_law_1.2,0.08781440258026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,8,power_law_1.2,0.08819199800491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,1,balanced,1.7054133415222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,8,power_law_1.2,0.08985599875450134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,1,balanced,2.2776106198628745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,8,power_law_1.2,0.08450559973716736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,8,power_law_1.2,0.0886080026626587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,8,power_law_1.2,0.08974720239639282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,8,power_law_1.2,0.09353600144386291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,8,power_law_1.2,0.09924479722976684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,8,power_law_1.2,0.10506240129470826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,8,power_law_1.2,0.10781439542770385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,8,power_law_1.2,0.12167680263519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,8,power_law_1.2,0.14215680360794067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,8,power_law_1.2,0.15333759784698486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,8,power_law_1.2,0.19356800317764283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,8,power_law_1.2,0.20157439708709718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,8,power_law_1.2,0.2575103998184204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,8,power_law_1.2,0.31861119270324706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,8,power_law_1.2,0.4574592113494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,8,power_law_1.2,0.6064127922058106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,8,power_law_1.2,0.8565376281738282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,8,power_law_1.2,1.056588840484619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,4,power_law_1.2,1.0236991882324218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,2,power_law_1.01,0.08977919816970825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,2,power_law_1.01,0.1344831943511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,2,power_law_1.01,0.19066879749298096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,2,power_law_1.01,0.2601855993270874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,2,power_law_1.01,0.3612735986709595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,2,power_law_1.01,0.38759679794311525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,2,power_law_1.01,0.5749184131622315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,2,power_law_1.01,0.5624256134033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,2,power_law_1.01,0.5778240203857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,2,power_law_1.01,0.5895040035247803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,2,power_law_1.01,0.595359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,2,power_law_1.01,0.6230527877807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,2,power_law_1.01,0.6684864044189454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,2,power_law_1.01,0.6499263763427734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,2,power_law_1.01,0.6991551876068115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,2,power_law_1.01,0.7206592082977294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,2,power_law_1.01,0.7284992218017579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,2,power_law_1.01,0.7982079982757568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,2,power_law_1.01,0.831993579864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,2,power_law_1.01,0.8958847999572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,2,power_law_1.01,0.938316822052002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,2,power_law_1.01,1.0680576324462892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,2,power_law_1.01,1.197913646697998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,2,power_law_1.01,1.5236479759216308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,2,power_law_1.01,1.733830451965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,2,power_law_1.01,2.2824256896972654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,2,power_law_1.01,2.924083137512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,2,power_law_1.2,0.025305598974227905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,2,power_law_1.2,0.025446400046348572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,2,power_law_1.2,0.03027839958667755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,2,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,2,power_law_1.2,0.04085119962692261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,2,power_law_1.2,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,2,power_law_1.2,0.05126399993896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,2,power_law_1.2,0.05467519760131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,2,power_law_1.2,0.05494400262832642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,2,power_law_1.2,0.05742080211639404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,2,power_law_1.2,0.05745919942855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,2,power_law_1.2,0.05991680026054382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,2,power_law_1.2,0.061478400230407716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,2,power_law_1.2,0.06796799898147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,2,power_law_1.2,0.07828480005264282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,2,power_law_1.2,0.07859200239181519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,2,power_law_1.2,0.08157439827919007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,2,power_law_1.2,0.09180160164833069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,2,power_law_1.2,0.09664639830589294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,2,power_law_1.2,0.11708799600601197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,2,power_law_1.2,0.135315203666687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,2,power_law_1.2,0.18454400300979615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,2,power_law_1.2,0.20855679512023925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,2,power_law_1.2,0.28752000331878663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,2,power_law_1.2,0.38925440311431886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,2,power_law_1.2,0.548576021194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,8,power_law_1.01,0.04810880124568939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,8,power_law_1.01,0.06422399878501892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,8,power_law_1.01,0.0685696005821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,8,power_law_1.01,0.07310720086097718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,8,power_law_1.01,0.07891200184822082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,8,power_law_1.01,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,8,power_law_1.01,0.07530879974365234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,8,power_law_1.01,0.07681279778480529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,8,power_law_1.01,0.08013439774513245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,8,power_law_1.01,0.08279680013656616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,8,power_law_1.01,0.08083199858665466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,8,power_law_1.01,0.08551679849624634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,2,power_law_1.2,0.7260735988616943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,8,power_law_1.01,0.08807680010795593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,8,power_law_1.01,0.09036800265312195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,8,power_law_1.01,0.09760640263557434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,8,power_law_1.01,0.0998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,8,power_law_1.01,0.10366079807281495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,8,power_law_1.01,0.11839359998703003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,8,power_law_1.01,0.12369279861450196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,8,power_law_1.01,0.14417279958724977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,8,power_law_1.01,0.16428159475326537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,8,power_law_1.01,0.20937600135803222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,8,power_law_1.01,0.24686079025268554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,8,power_law_1.01,0.3346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,8,power_law_1.01,0.4256256103515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,8,power_law_1.01,0.5846208095550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,8,power_law_1.01,0.7703167915344238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,64,power_law_1.2,0.08909440040588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,64,power_law_1.2,0.08554880023002624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,64,power_law_1.2,0.08550400137901307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,64,power_law_1.2,0.08981119990348815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,64,power_law_1.2,0.08158720135688782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,64,power_law_1.2,0.08942080140113831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,64,power_law_1.2,0.09123200178146362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,64,power_law_1.2,0.09317119717597962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,64,power_law_1.2,0.09023360013961793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,64,power_law_1.2,0.09226880073547364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,64,power_law_1.2,0.09356160163879394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,64,power_law_1.2,0.09635840058326721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,64,power_law_1.2,0.09608319997787476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,64,power_law_1.2,0.09719039797782898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,64,power_law_1.2,0.10195200443267823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,64,power_law_1.2,0.1033471941947937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,64,power_law_1.2,0.11121920347213746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,64,power_law_1.2,0.13623039722442626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,64,power_law_1.2,0.14561280012130737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,64,power_law_1.2,0.17739520072937012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,64,power_law_1.2,0.20817921161651612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,64,power_law_1.2,0.27987840175628664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,64,power_law_1.2,0.3429120063781738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,64,power_law_1.2,0.47188482284545896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,64,power_law_1.2,0.6268479824066162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,64,power_law_1.2,0.8835519790649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,64,power_law_1.2,1.1667200088500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,2,power_law_1.2,0.07742080092430115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,2,power_law_1.2,0.10060800313949585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,2,power_law_1.2,0.11930240392684936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,2,power_law_1.2,0.15813759565353394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,2,power_law_1.2,0.19409919977188111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,2,power_law_1.2,0.24674561023712158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,2,power_law_1.2,0.2740799903869629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,2,power_law_1.2,0.35202560424804685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,2,power_law_1.2,0.34752640724182127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,2,power_law_1.2,0.35797119140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,2,power_law_1.2,0.38514559268951415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,2,power_law_1.2,0.3812031984329224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,2,power_law_1.2,0.39349119663238524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,2,power_law_1.2,0.3996608018875122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,2,power_law_1.2,0.42513279914855956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,2,power_law_1.2,0.4303296089172363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,2,power_law_1.2,0.4478784084320068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,2,power_law_1.2,0.5039936065673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,2,power_law_1.2,0.5474112033843994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,2,power_law_1.2,0.6522240161895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,2,power_law_1.2,0.695743989944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,2,power_law_1.2,0.8466943740844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,2,power_law_1.2,1.0655232429504395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,2,power_law_1.2,1.346771240234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,2,power_law_1.2,1.676313591003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,2,power_law_1.2,2.2100799560546873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,2,power_law_1.2,2.714579200744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,32,8,power_law_1.2,0.02215680032968521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,32,8,power_law_1.2,0.023366400599479677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,32,8,power_law_1.2,0.02459519952535629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,32,8,power_law_1.2,0.025913599133491515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,32,8,power_law_1.2,0.027692800760269164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,32,8,power_law_1.2,0.02797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,32,8,power_law_1.2,0.030054399371147157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,32,8,power_law_1.2,0.031142398715019226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,32,8,power_law_1.2,0.03212159872055054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,32,8,power_law_1.2,0.03166080117225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,32,8,power_law_1.2,0.0339711993932724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,32,8,power_law_1.2,0.04111360013484955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,32,8,power_law_1.2,0.04174720048904419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,32,8,power_law_1.2,0.04177280068397522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,32,8,power_law_1.2,0.06535680294036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,2,power_law_1.01,0.098854398727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,32,8,power_law_1.2,0.06748800277709961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,32,8,power_law_1.2,0.07045120000839233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,2,power_law_1.01,0.11596159934997559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,32,8,power_law_1.2,0.07616000175476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,2,power_law_1.01,0.1313599944114685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,2,power_law_1.01,0.1685696005821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,2,power_law_1.01,0.19631359577178956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,2,power_law_1.01,0.22017920017242432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,2,power_law_1.01,0.2676160097122192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,2,power_law_1.01,0.26632959842681886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,2,power_law_1.01,0.2718847990036011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,2,power_law_1.01,0.27039361000061035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,2,power_law_1.01,0.2783040046691895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,32,8,power_law_1.2,0.08031359910964966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,2,power_law_1.01,0.2930111885070801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,32,8,power_law_1.2,0.09234560132026673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,2,power_law_1.01,0.29522559642791746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,32,8,power_law_1.2,0.1001471996307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,2,power_law_1.01,0.3012415885925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,2,power_law_1.01,0.31454079151153563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,2,power_law_1.01,0.32269439697265623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,2,power_law_1.01,0.32318079471588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,2,power_law_1.01,0.379423999786377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,32,8,power_law_1.2,0.12580480575561523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,2,power_law_1.01,0.4015615940093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,32,8,power_law_1.2,0.144652795791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,2,power_law_1.01,0.46975998878479003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,2,power_law_1.01,0.5344768047332764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,2,power_law_1.01,0.6692543983459472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,2,power_law_1.01,0.8155967712402343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,32,8,power_law_1.2,0.20634241104125978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,2,power_law_1.01,1.115065574645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,32,8,power_law_1.2,0.24501121044158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,32,8,power_law_1.2,0.3504447937011719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,2,power_law_1.01,1.4245696067810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,32,8,power_law_1.2,0.454636812210083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,32,power_law_1.2,0.028467199206352232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,32,power_law_1.2,0.027807998657226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,32,power_law_1.2,0.0278656005859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,32,power_law_1.2,0.030124801397323608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,32,power_law_1.2,0.030745598673820495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,32,power_law_1.2,0.031763198971748355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,32,power_law_1.2,0.032179200649261476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,32,power_law_1.2,0.0324864000082016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,32,power_law_1.2,0.03421440124511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,32,power_law_1.2,0.03407999873161316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,32,power_law_1.2,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,32,power_law_1.2,0.036006399989128114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,32,power_law_1.2,0.03641600012779236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,32,power_law_1.2,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,32,power_law_1.2,0.04446719884872437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,32,power_law_1.2,0.04312320053577423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,32,power_law_1.2,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,32,power_law_1.2,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,32,power_law_1.2,0.05607680082321167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,32,power_law_1.2,0.07546240091323853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,32,power_law_1.2,0.0969215989112854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,32,power_law_1.2,0.13420799970626832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,32,power_law_1.2,0.1537600040435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,32,power_law_1.2,0.21759359836578368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,32,power_law_1.2,0.29950718879699706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,32,power_law_1.2,0.4939136028289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,2,power_law_1.01,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,2,power_law_1.01,0.07976959943771363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,2,power_law_1.01,0.09111040234565734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,2,power_law_1.01,0.10661760568618775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,2,power_law_1.01,0.12254079580307006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,2,power_law_1.01,0.1539903998374939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,2,power_law_1.01,0.1590399980545044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,2,power_law_1.01,0.15749119520187377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,2,power_law_1.01,0.167084801197052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,2,power_law_1.01,0.16813440322875978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,2,power_law_1.01,0.17367680072784425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,2,power_law_1.01,0.17325439453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,2,power_law_1.01,0.17824000120162964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,2,power_law_1.01,0.1803712010383606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,16,power_law_1.2,0.05025280117988586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,2,power_law_1.01,0.18643200397491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,16,power_law_1.2,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,2,power_law_1.01,0.18880000114440917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,16,power_law_1.2,0.04625279903411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,16,power_law_1.2,0.047839999198913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,2,power_law_1.01,0.19724160432815552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,16,power_law_1.2,0.04805760085582733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,2,power_law_1.01,0.21370880603790282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,16,power_law_1.2,0.04883840084075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,16,power_law_1.2,0.050393599271774295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,2,power_law_1.01,0.24247679710388184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,16,power_law_1.2,0.052083200216293334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,2,power_law_1.01,0.28590080738067625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,16,power_law_1.2,0.05329279899597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,16,power_law_1.2,0.053273600339889524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,2,power_law_1.01,0.3377984046936035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,16,power_law_1.2,0.05610880255699158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,16,power_law_1.2,0.05783680081367493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,2,power_law_1.01,0.45919361114501955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,16,power_law_1.2,0.058956801891326904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,16,power_law_1.2,0.06100479960441589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,2,power_law_1.01,0.5484543800354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,16,power_law_1.2,0.06798719763755798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,16,power_law_1.2,0.0703935980796814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,2,power_law_1.01,0.8490688323974609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,16,power_law_1.2,0.07145599722862243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,16,power_law_1.2,0.08831359744071961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,2,power_law_1.01,1.02107515335083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,16,power_law_1.2,0.09774079918861389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,16,power_law_1.2,0.12398719787597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,2,power_law_1.01,1.5088255882263184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,16,power_law_1.2,0.15982719659805297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,16,power_law_1.2,0.21934080123901367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,16,power_law_1.2,0.245580792427063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,2,power_law_1.01,2.1053823471069335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,16,power_law_1.2,0.35111041069030763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,16,power_law_1.2,0.45903358459472654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,16,power_law_1.2,0.7330431938171387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,16,power_law_1.2,0.8479616165161132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,32,power_law_1.2,0.5983104228973388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,4,balanced,0.022053333620230358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,4,balanced,0.0232640008131663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,4,balanced,0.025040000677108765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,4,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,4,balanced,0.051269332567850746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,4,balanced,0.07444799939791362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,4,balanced,0.07654933134714763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,4,balanced,0.0786186655362447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,4,balanced,0.08062399923801422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,4,balanced,0.081386665503184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,4,balanced,0.08634666601816814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,1,power_law_1.01,0.08664320111274719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,4,balanced,0.08567999800046285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,4,balanced,0.08827733000119527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,4,balanced,0.09643200039863586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,1,power_law_1.01,0.10629760026931763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,1,power_law_1.01,0.12744959592819213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,1,power_law_1.01,0.1574399948120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,1,power_law_1.01,0.181004798412323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,1,power_law_1.01,0.24822399616241456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,1,power_law_1.01,0.26754560470581057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,4,balanced,0.09869333108266194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,1,power_law_1.01,0.27435519695281985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,4,balanced,0.10726400216420491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,4,balanced,0.1311840017636617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,4,balanced,0.13524799545605978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,4,balanced,0.14150933424631754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,4,balanced,0.15667733550071716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,4,balanced,0.22807466983795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,4,balanced,0.25783467292785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,4,balanced,0.28573334217071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,4,balanced,0.34030401706695557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,4,balanced,0.3965760072072347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,4,balanced,0.6272799968719482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,1,power_law_1.01,0.28309121131896975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,4,balanced,0.7368960380554199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,1,power_law_1.01,0.2850559949874878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,1,power_law_1.01,0.2877376079559326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,1,power_law_1.01,0.2953408002853394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,1,power_law_1.01,0.2994623899459839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,1,power_law_1.01,0.30750079154968263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,1,power_law_1.01,0.30687999725341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,1,power_law_1.01,0.31601920127868655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,1,power_law_1.01,0.3319360017776489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,1,power_law_1.01,0.374835205078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,1,power_law_1.01,0.41451520919799806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,1,power_law_1.01,0.5060800075531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,1,power_law_1.01,0.5970240116119385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,1,power_law_1.01,0.7813504219055176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,1,power_law_1.01,0.9688960075378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,1,power_law_1.01,1.3408639907836915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,1,power_law_1.01,1.7123136520385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,1,power_law_1.01,2.45743350982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,1,power_law_1.01,3.2099647521972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,2,power_law_1.2,0.060550397634506224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,2,power_law_1.2,0.06625919938087463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,2,power_law_1.2,0.07402880191802978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,2,power_law_1.2,0.0866815984249115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,2,power_law_1.2,0.09397119879722596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,2,power_law_1.2,0.12205439805984497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,2,power_law_1.2,0.12671999931335448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,2,power_law_1.2,0.12992000579833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,2,power_law_1.2,0.13286399841308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,2,power_law_1.2,0.1351807951927185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,2,power_law_1.2,0.13891839981079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,2,power_law_1.2,0.1413823962211609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,2,power_law_1.2,0.14520319700241088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,2,power_law_1.2,0.1512768030166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,2,power_law_1.2,0.1598207950592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,2,power_law_1.2,0.1636672019958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,2,power_law_1.2,0.177401602268219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,2,power_law_1.2,0.20276479721069335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,2,power_law_1.2,0.22907519340515137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,2,power_law_1.2,0.28241279125213625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,2,power_law_1.2,0.3328511953353882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,2,power_law_1.2,0.43340158462524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,2,power_law_1.2,0.6041535854339599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,2,power_law_1.2,0.777184009552002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,2,power_law_1.2,1.017369556427002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,2,power_law_1.2,1.4250944137573243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,2,power_law_1.2,1.9961536407470704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,16,balanced,0.05299733579158783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,16,balanced,0.052576000491778054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,16,balanced,0.054048001766204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,16,balanced,0.06468800206979115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,16,balanced,0.08184533317883809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,16,balanced,0.1114026705423991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,16,balanced,0.11349333326021831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,16,balanced,0.11148800452550252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,16,balanced,0.11487467090288798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,16,balanced,0.1130506694316864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,16,balanced,0.11338667074839275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,16,balanced,0.11542399724324544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,16,balanced,0.11571199695269267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,16,balanced,0.11806933085123698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,16,balanced,0.12845866878827414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,16,balanced,0.12893333037694296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,16,balanced,0.13197333614031473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,16,balanced,0.1513759990533193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,16,balanced,0.15626666943232217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,16,balanced,0.17139732837677002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,16,balanced,0.1857866644859314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,16,balanced,0.2166879971822103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,16,balanced,0.24579733610153198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,16,balanced,0.3288319905598958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,16,balanced,0.39026133219401044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,16,balanced,0.5478613376617432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,16,balanced,0.7082346280415853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,32,1,power_law_1.01,0.05518720149993896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,32,1,power_law_1.01,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,32,1,power_law_1.01,0.06593919992446899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,32,1,power_law_1.01,0.08119680285453797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,32,1,power_law_1.01,0.10113919973373413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,32,1,power_law_1.01,0.11653759479522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,32,1,power_law_1.01,0.14055039882659912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,32,1,power_law_1.01,0.14489599466323852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,32,1,power_law_1.01,0.1498304009437561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,32,1,power_law_1.01,0.1550655961036682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,32,1,power_law_1.01,0.16255999803543092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,32,1,power_law_1.01,0.17313920259475707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,32,1,power_law_1.01,0.18037760257720947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,32,1,power_law_1.01,0.19765119552612304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,32,1,power_law_1.01,0.2138688087463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,32,1,power_law_1.01,0.22799360752105713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,32,1,power_law_1.01,0.22464001178741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,32,1,power_law_1.01,0.27137279510498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,32,1,power_law_1.01,0.30773119926452636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,32,1,power_law_1.01,0.395084810256958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,32,1,power_law_1.01,0.4697535991668701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,32,1,power_law_1.01,0.6332863807678223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,32,1,power_law_1.01,0.7924799919128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,32,1,power_law_1.01,1.0954943656921388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,32,1,power_law_1.01,1.4038144111633302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,32,1,power_law_1.01,2.0016895294189454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,32,1,power_law_1.01,2.6248512268066406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,4,power_law_1.01,0.09247360229492188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,4,power_law_1.01,0.11911040544509888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,4,power_law_1.01,0.11943680047988892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,4,power_law_1.01,0.14331519603729248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,4,power_law_1.01,0.15532159805297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,4,power_law_1.01,0.19964799880981446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,4,power_law_1.01,0.19553920030593872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,4,power_law_1.01,0.19371520280838012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,4,power_law_1.01,0.19913599491119385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,4,power_law_1.01,0.20055038928985597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,4,power_law_1.01,0.20446720123291015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,4,power_law_1.01,0.21636478900909423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,4,power_law_1.01,0.22133760452270507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,4,power_law_1.01,0.22688639163970947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,4,power_law_1.01,0.23957760334014894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,4,power_law_1.01,0.2579263925552368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,4,power_law_1.01,0.27475199699401853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,4,power_law_1.01,0.3141632080078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,4,power_law_1.01,0.3637120008468628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,4,power_law_1.01,0.45211520195007326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,4,power_law_1.01,0.5063680171966553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,4,power_law_1.01,0.6757887840270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,4,power_law_1.01,0.775980806350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,4,power_law_1.01,1.0884480476379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,4,power_law_1.01,1.3744832038879395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,4,power_law_1.01,2.012281608581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,4,power_law_1.01,2.53940486907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,4,power_law_1.01,0.03348479866981506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,4,power_law_1.01,0.033817601203918454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,4,power_law_1.01,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,4,power_law_1.01,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,4,power_law_1.01,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,4,power_law_1.01,0.03656960129737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,4,power_law_1.01,0.037651199102401736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,4,power_law_1.01,0.037459200620651244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,4,power_law_1.01,0.03914879858493805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,4,power_law_1.01,0.038815999031066896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,4,power_law_1.01,0.03925119936466217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,4,power_law_1.01,0.04085760116577149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,4,power_law_1.01,0.04202240109443665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,4,power_law_1.01,0.04165120124816894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,4,power_law_1.01,0.0466623991727829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,4,power_law_1.01,0.04893440008163452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,4,power_law_1.01,0.0500927984714508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,4,power_law_1.01,0.05780479907989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.01,0.06931840181350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.01,0.08023679852485657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.01,0.08647040128707886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.01,0.12546559572219848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.01,0.15712640285491944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.01,0.19953919649124147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.01,0.23521919250488282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,16,power_law_1.2,0.03343360126018524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,16,power_law_1.2,0.03308799862861633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,16,power_law_1.2,0.028998398780822755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,16,power_law_1.2,0.030086401104927062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,16,power_law_1.2,0.03126400113105774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,16,power_law_1.2,0.03288959860801697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,16,power_law_1.2,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,16,power_law_1.2,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,16,power_law_1.2,0.03312639892101288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,16,power_law_1.2,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,16,power_law_1.2,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,16,power_law_1.2,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,16,power_law_1.2,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,16,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,16,power_law_1.2,0.04376319944858551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,16,power_law_1.2,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,16,power_law_1.2,0.04772480130195618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,16,power_law_1.2,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,16,power_law_1.2,0.06570879817008972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,16,power_law_1.2,0.0754368007183075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,16,power_law_1.2,0.0850048005580902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,16,power_law_1.2,0.10269440412521362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,16,power_law_1.2,0.12658560276031494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,16,power_law_1.2,0.16621439456939696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.01,0.33512320518493655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,16,power_law_1.2,0.21589760780334472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,16,power_law_1.2,0.3153280019760132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.01,0.5249087810516357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,2,power_law_1.2,0.07656959891319275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,2,power_law_1.2,0.0824512004852295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,2,power_law_1.2,0.08871039748191833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,2,power_law_1.2,0.10279680490493774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,2,power_law_1.2,0.11854720115661621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,2,power_law_1.2,0.13938560485839843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,2,power_law_1.2,0.14840320348739625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,2,power_law_1.2,0.16911360025405883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,2,power_law_1.2,0.17431679964065552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,16,power_law_1.2,0.41361279487609864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,2,power_law_1.2,0.18439040184020997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,2,power_law_1.2,0.18592640161514282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,2,power_law_1.2,0.1896191954612732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,2,power_law_1.2,0.19171199798583985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,2,power_law_1.2,0.19735679626464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,2,power_law_1.2,0.21530239582061766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,2,power_law_1.2,0.21389439105987548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,2,power_law_1.2,0.2301055908203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,2,power_law_1.2,0.25783040523529055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,2,power_law_1.2,0.28570239543914794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,2,power_law_1.2,0.3549247980117798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,2,power_law_1.2,0.40266880989074705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,2,power_law_1.2,0.5291840076446533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,2,power_law_1.2,0.6518847942352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,2,power_law_1.2,0.8745280265808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,2,power_law_1.2,1.1135616302490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,2,power_law_1.2,1.565721607208252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,2,power_law_1.2,2.0728128433227537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,8,power_law_1.2,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,8,power_law_1.2,0.03278079926967621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,8,power_law_1.2,0.030995199084281923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,8,power_law_1.2,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,16,power_law_1.01,0.048607999086380006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,8,power_law_1.2,0.03356800079345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,16,power_law_1.01,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,16,power_law_1.01,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,16,power_law_1.01,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,16,power_law_1.01,0.05329279899597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,16,power_law_1.01,0.05280640125274658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,16,power_law_1.01,0.052691197395324706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,8,power_law_1.2,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,16,power_law_1.01,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,8,power_law_1.2,0.03578880131244659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,8,power_law_1.2,0.03583360016345978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,8,power_law_1.2,0.0369024008512497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,8,power_law_1.2,0.03648000061511993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,8,power_law_1.2,0.037510401010513304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,8,power_law_1.2,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,8,power_law_1.2,0.04002560079097748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,8,power_law_1.2,0.04270080029964447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,8,power_law_1.2,0.045484799146652224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,16,power_law_1.01,0.05482239723205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,8,power_law_1.2,0.05010560154914856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,16,power_law_1.01,0.05464320182800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,8,power_law_1.2,0.053395199775695804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,16,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,16,power_law_1.01,0.057068800926208495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,16,power_law_1.01,0.058790397644042966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,16,power_law_1.01,0.059654402732849124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,16,power_law_1.01,0.0648576021194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,16,power_law_1.01,0.06581760048866273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,16,power_law_1.01,0.07066879868507385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,8,power_law_1.2,0.059468799829483034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,16,power_law_1.01,0.08001919984817504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,8,power_law_1.2,0.07189760208129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,16,power_law_1.01,0.0864192008972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,8,power_law_1.2,0.08566399812698364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,16,power_law_1.01,0.10513919591903687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,8,power_law_1.2,0.09794560074806213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,16,power_law_1.01,0.11600639820098876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,16,power_law_1.01,0.15317120552062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,16,power_law_1.01,0.17511039972305298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,16,power_law_1.01,0.24554240703582764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,16,power_law_1.01,0.303385591506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,8,power_law_1.2,0.12747520208358765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,8,power_law_1.2,0.15432319641113282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,16,power_law_1.01,0.4087103843688965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,16,power_law_1.01,0.516486406326294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,8,power_law_1.2,0.21890559196472167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,8,power_law_1.2,0.2892224073410034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,16,power_law_1.01,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,16,power_law_1.01,0.07639039754867553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,16,power_law_1.01,0.08834559917449951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,16,power_law_1.01,0.08769279718399048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,16,power_law_1.01,0.08895999789237977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,16,power_law_1.01,0.09381759762763978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,16,power_law_1.01,0.08384640216827392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,16,power_law_1.01,0.08807039856910706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,16,power_law_1.01,0.08930559754371643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,16,power_law_1.01,0.0833407998085022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,16,power_law_1.01,0.0920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,16,power_law_1.01,0.09263359904289245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,8,power_law_1.2,0.3742847919464111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,16,power_law_1.01,0.09246079921722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,16,power_law_1.01,0.09783679842948914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,16,power_law_1.01,0.10610560178756714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,16,power_law_1.01,0.10751359462738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,16,power_law_1.01,0.1176192045211792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,16,power_law_1.01,0.13354239463806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,16,power_law_1.01,0.14341119527816773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,16,power_law_1.01,0.17494399547576905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,16,power_law_1.01,0.20023679733276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,16,power_law_1.01,0.2597248077392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,16,power_law_1.01,0.3151488065719604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,16,power_law_1.01,0.4348480224609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,16,power_law_1.01,0.5292352199554443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,16,power_law_1.01,0.7549439907073975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,16,power_law_1.01,0.9792256355285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,8,power_law_1.2,0.4722879886627197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,2,power_law_1.01,0.0739135980606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,2,power_law_1.01,0.08506879806518555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,2,power_law_1.01,0.1056704044342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,2,power_law_1.01,0.1336575984954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,2,power_law_1.01,0.14033279418945313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,2,power_law_1.01,0.1496448040008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,2,power_law_1.01,0.1512895941734314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,2,power_law_1.01,0.14718719720840454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,2,power_law_1.01,0.15509120225906373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,2,power_law_1.01,0.1632192015647888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,2,power_law_1.01,0.1617792010307312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,2,power_law_1.01,0.16777600049972535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,2,power_law_1.01,0.1788607954978943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,2,power_law_1.01,0.18061439990997313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,2,power_law_1.01,0.19867520332336425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,2,power_law_1.01,0.20878078937530517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,2,power_law_1.01,0.24087040424346923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,2,power_law_1.01,0.2761791944503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,2,power_law_1.01,0.3255167961120605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,2,power_law_1.01,0.39390718936920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,2,power_law_1.01,0.488646411895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,2,power_law_1.01,0.656435203552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,2,power_law_1.01,0.9201984405517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,2,power_law_1.01,1.1654720306396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,2,power_law_1.01,1.5088383674621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,2,power_law_1.01,1.9780672073364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,8,power_law_1.01,0.04700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,8,power_law_1.01,0.05082240104675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,8,power_law_1.01,0.0490880012512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,8,power_law_1.01,0.04755200147628784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,8,power_law_1.01,0.05021439790725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,8,power_law_1.01,0.04880639910697937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,8,power_law_1.01,0.05090559720993042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,8,power_law_1.01,0.0514240026473999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,8,power_law_1.01,0.05129600167274475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,8,power_law_1.01,0.05247359871864319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,8,power_law_1.01,0.05249279737472534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,8,power_law_1.01,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,8,power_law_1.01,0.05504000186920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,8,power_law_1.01,0.05632640123367309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,8,power_law_1.01,0.06346880197525025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,8,power_law_1.01,0.06307200193405152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,8,power_law_1.01,0.06812800168991089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,8,power_law_1.01,0.08222079873085023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,8,power_law_1.01,0.08491520285606384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,8,power_law_1.01,0.10001920461654663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,8,power_law_1.01,0.12001279592514039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,8,power_law_1.01,0.14477440118789672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,8,power_law_1.01,0.17665280103683473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,8,power_law_1.01,0.23725440502166747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,8,power_law_1.01,0.2902591943740845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,8,power_law_1.01,0.4028223991394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,8,power_law_1.01,0.5209856033325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,2,power_law_1.01,3.131328010559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,4,balanced,0.029466666281223297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,4,balanced,0.02924266705910365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,4,balanced,0.03325333446264267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,4,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,4,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,4,balanced,0.04552533229192098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,4,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,4,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,4,balanced,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,4,balanced,0.04557333389918009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,4,balanced,0.04602666695912679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,4,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,4,balanced,0.050026665131251015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,4,balanced,0.0498986691236496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,4,balanced,0.05509866774082184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,4,balanced,0.057861333092053734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,4,balanced,0.059877331058184304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,4,balanced,0.07041599849859874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,4,balanced,0.07026133437951405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,4,balanced,0.09822932879130046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,4,balanced,0.11751466989517212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,4,balanced,0.15843733151753744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,4,balanced,0.19324799378712973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,4,balanced,0.2773866653442383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,4,balanced,0.3439840078353882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,128,power_law_1.2,0.04016000032424927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,4,balanced,0.5031253496805826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,128,power_law_1.2,0.04035199880599975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,4,balanced,0.6451840003331503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,128,power_law_1.2,0.04002560079097748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,128,power_law_1.2,0.0431551992893219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,128,power_law_1.2,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,128,power_law_1.2,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,128,power_law_1.2,0.047065600752830505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,128,power_law_1.2,0.04848000109195709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,128,power_law_1.2,0.04933759868144989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,128,power_law_1.2,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,128,power_law_1.2,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,128,power_law_1.2,0.055430400371551516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,128,power_law_1.2,0.051507198810577394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,128,power_law_1.2,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,128,power_law_1.2,0.06199679970741272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,128,power_law_1.2,0.06216959953308106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,128,power_law_1.2,0.06696320176124573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,128,power_law_1.2,0.07596799731254578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,128,power_law_1.2,0.08401280045509338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,128,power_law_1.2,0.10499199628829955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,128,power_law_1.2,0.13459839820861816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,128,power_law_1.2,0.18367999792099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,128,power_law_1.2,0.23303680419921874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,128,power_law_1.2,0.3528768062591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,128,power_law_1.2,0.4200448036193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,128,power_law_1.2,0.7061503887176513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,128,power_law_1.2,0.9336959838867187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,2,power_law_1.2,0.06429439783096313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,2,power_law_1.2,0.0802944004535675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,2,power_law_1.2,0.08969600200653076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,2,power_law_1.2,0.12353919744491577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,2,power_law_1.2,0.14419200420379638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,2,power_law_1.2,0.16913919448852538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,2,power_law_1.2,0.24625279903411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,2,power_law_1.2,0.2457535982131958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,2,power_law_1.2,0.24591360092163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,2,power_law_1.2,0.25486719608306885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,2,power_law_1.2,0.26041600704193113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,2,power_law_1.2,0.26474239826202395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,2,power_law_1.2,0.2680959939956665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,2,power_law_1.2,0.2827199935913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,2,power_law_1.2,0.2994112014770508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,2,power_law_1.2,0.3067903995513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,2,power_law_1.2,0.32164480686187746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,2,power_law_1.2,0.365011191368103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,2,power_law_1.2,0.36458239555358884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,2,power_law_1.2,0.4176703929901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,2,power_law_1.2,0.4302527904510498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,2,power_law_1.2,0.5168896198272706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,2,power_law_1.2,0.604908800125122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,2,power_law_1.2,0.733900785446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,2,power_law_1.2,0.927462387084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,2,power_law_1.2,1.239475154876709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,2,power_law_1.2,1.6054784774780273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,1,balanced,0.1011253297328949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,1,balanced,0.11135466893513997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,1,balanced,0.1372053325176239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,1,balanced,0.18389866749445596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,1,balanced,0.27058132489522296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,1,balanced,0.4408266544342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,1,balanced,0.4321333169937134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,1,balanced,0.43423465887705487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,1,balanced,0.43426668643951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,1,balanced,0.43062933286031085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,1,balanced,0.42846401532491046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,1,balanced,0.4408693313598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,1,balanced,0.44233067830403644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,1,balanced,0.4424213171005249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,1,balanced,0.4513653516769409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,1,balanced,0.46234134833017987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,1,balanced,0.47516798973083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,1,balanced,0.5174506505330404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,1,balanced,0.5532960096995035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,1,balanced,0.6286133527755737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,1,balanced,0.6943253676096598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,1,balanced,0.8508480389912924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,1,balanced,1.0064906279246013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,1,balanced,1.529813289642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,1,balanced,1.834928035736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,1,balanced,2.662207921346029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,1,balanced,3.4683891932169595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,4,power_law_1.2,0.04480000138282776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,4,power_law_1.2,0.04293760061264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,4,power_law_1.2,0.04694400131702423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,4,power_law_1.2,0.05189759731292724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,4,power_law_1.2,0.05752320289611816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,4,power_law_1.2,0.059552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,4,power_law_1.2,0.05996800065040588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,4,power_law_1.2,0.06273279786109924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,4,power_law_1.2,0.06053119897842407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,4,power_law_1.2,0.061638402938842776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,4,power_law_1.2,0.061267197132110596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,4,power_law_1.2,0.06503040194511414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,4,power_law_1.2,0.06421759724617004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,4,power_law_1.2,0.06609920263290406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,4,power_law_1.2,0.07062399983406067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,4,power_law_1.2,0.07470719814300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,4,power_law_1.2,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,4,power_law_1.2,0.08615040183067321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,4,power_law_1.2,0.09809280037879944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,4,power_law_1.2,0.12232320308685303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,4,power_law_1.2,0.13537280559539794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,4,power_law_1.2,0.18840960264205933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,4,power_law_1.2,0.22564480304718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,4,power_law_1.2,0.27527039051055907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,4,power_law_1.2,0.4105855941772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,4,power_law_1.2,0.5350656032562255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,4,power_law_1.2,0.798137617111206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,1,balanced,0.0784800002972285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,1,balanced,0.12173866232236226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,1,balanced,0.2025173306465149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,1,balanced,0.37406933307647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,1,balanced,0.7124533653259277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,1,balanced,0.7166240215301514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,1,balanced,0.7210666338602701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,1,balanced,0.7181813716888428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,1,balanced,0.7194666862487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,1,balanced,0.7220586935679117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,1,balanced,0.7282880147298177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,1,balanced,0.7317386468251547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,1,balanced,0.7380853494008383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,1,balanced,0.7460266749064127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,1,balanced,0.7541013558705648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,1,balanced,0.7630026340484619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,1,balanced,0.7767893473307291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,1,balanced,0.8045547008514404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,16,power_law_1.2,0.042412799596786496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,1,balanced,0.8577067057291666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,16,power_law_1.2,0.03738240003585815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,16,power_law_1.2,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,16,power_law_1.2,0.03614720106124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,16,power_law_1.2,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,16,power_law_1.2,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,16,power_law_1.2,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,16,power_law_1.2,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,16,power_law_1.2,0.03749760091304779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,16,power_law_1.2,0.03720319867134094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,16,power_law_1.2,0.06949120163917541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,1,balanced,0.899994691212972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,16,power_law_1.2,0.06744319796562195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,1,balanced,0.995136022567749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,16,power_law_1.2,0.06883199810981751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,1,balanced,1.1148107051849365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,16,power_law_1.2,0.07384960055351257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,1,balanced,1.1999839941660564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,16,power_law_1.2,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,16,power_law_1.2,0.07339519858360291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,16,power_law_1.2,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,16,power_law_1.2,0.07611520290374756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,16,power_law_1.2,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,16,power_law_1.2,0.0766207993030548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,16,power_law_1.2,0.04314239919185638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,16,power_law_1.2,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,16,power_law_1.2,0.07831040024757385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,16,power_law_1.2,0.047884801030159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,16,power_law_1.2,0.07720320224761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,16,power_law_1.2,0.050425601005554196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,16,power_law_1.2,0.054028797149658206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,16,power_law_1.2,0.06144639849662781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,1,balanced,1.7349440256754558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,16,power_law_1.2,0.06465280055999756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,16,power_law_1.2,0.07579519748687744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,1,balanced,1.8490932782491047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,16,power_law_1.2,0.07832319736480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,16,power_law_1.2,0.0979200005531311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,1,balanced,2.604778607686361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,16,power_law_1.2,0.10982400178909302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,16,power_law_1.2,0.07959679961204529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,16,power_law_1.2,0.1424191951751709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,16,power_law_1.2,0.08084480166435241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,1,balanced,3.3713601430257163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,16,power_law_1.2,0.1811903953552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,16,power_law_1.2,0.2351680040359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,16,power_law_1.2,0.08849920034408569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,16,power_law_1.2,0.09239680171012879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,16,power_law_1.2,0.09864959716796876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,16,power_law_1.2,0.09962239861488342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,16,power_law_1.2,0.10720000267028809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,16,power_law_1.2,0.11316479444503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,16,power_law_1.2,0.12775039672851562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,16,power_law_1.2,0.15396480560302733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,16,power_law_1.2,0.1950719952583313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,16,power_law_1.2,0.24913280010223388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,16,power_law_1.2,0.32734079360961915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,16,power_law_1.2,0.45357441902160645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,16,power_law_1.2,0.6900864124298096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,16,power_law_1.2,0.9588671684265136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,16,power_law_1.2,0.3074944019317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,16,power_law_1.2,1.4274304389953614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,8,power_law_1.2,0.05363839864730835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,8,power_law_1.2,0.05395200252532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,8,power_law_1.2,0.061689597368240354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,8,power_law_1.2,0.06283519864082336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,8,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,8,power_law_1.2,0.07217280268669128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,8,power_law_1.2,0.07192320227622986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,8,power_law_1.2,0.07034879922866821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,8,power_law_1.2,0.07166079878807068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,8,power_law_1.2,0.07409279942512512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,8,power_law_1.2,0.074099200963974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,8,power_law_1.2,0.07761279940605163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,8,power_law_1.2,0.07862399816513062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,8,power_law_1.2,0.08672000169754028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,8,power_law_1.2,0.0908735990524292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,8,power_law_1.2,0.09651200175285339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,8,power_law_1.2,0.10560640096664428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,8,power_law_1.2,0.11559679508209228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,8,power_law_1.2,0.13173120021820067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,8,power_law_1.2,0.15360000133514404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,8,power_law_1.2,0.13865599632263184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,8,power_law_1.2,0.16684800386428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,8,power_law_1.2,0.19319039583206177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,8,power_law_1.2,0.2199104070663452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,8,power_law_1.2,0.2819904088973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,8,power_law_1.2,0.4286911964416504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,8,power_law_1.2,0.5645887851715088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,4,power_law_1.01,0.05588480234146118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,4,power_law_1.01,0.07381119728088378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,4,power_law_1.01,0.08268799781799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,4,power_law_1.01,0.0976256012916565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,4,power_law_1.01,0.11185280084609986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,4,power_law_1.01,0.11539839506149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,4,power_law_1.01,0.14225280284881592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,4,power_law_1.01,0.13984639644622804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,4,power_law_1.01,0.13565440177917482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,4,power_law_1.01,0.13524479866027833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,4,power_law_1.01,0.149452805519104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,4,power_law_1.01,0.15443840026855468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,4,power_law_1.01,0.16058239936828614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,4,power_law_1.01,0.15411200523376464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,4,power_law_1.01,0.16551680564880372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,4,power_law_1.01,0.17268480062484742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,4,power_law_1.01,0.17994879484176635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,4,power_law_1.01,0.20460159778594972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,4,power_law_1.01,0.23029758930206298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,4,power_law_1.01,0.25505919456481935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,4,power_law_1.01,0.2987648010253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,4,power_law_1.01,0.3483776092529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,4,power_law_1.01,0.4373631954193115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,4,power_law_1.01,0.559116792678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,4,power_law_1.01,0.678329610824585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,4,power_law_1.01,0.9746944427490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,4,power_law_1.01,1.2476032257080079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,1,power_law_1.2,0.13420159816741944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,1,power_law_1.2,0.21441919803619386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,4,power_law_1.2,0.044095999002456664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,4,power_law_1.2,0.03495680093765259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,4,power_law_1.2,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,4,power_law_1.2,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,4,power_law_1.2,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,1,power_law_1.2,0.37920000553131106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,4,power_law_1.2,0.03665919899940491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,1,power_law_1.2,0.6429503917694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,4,power_law_1.2,0.039315199851989745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,4,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,1,power_law_1.2,0.9653311729431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,1,power_law_1.2,1.2679295539855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,4,power_law_1.2,0.041440001130104064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,4,power_law_1.2,0.042905598878860474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,4,power_law_1.2,0.04076800048351288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,16,16,balanced,0.0489279975493749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,16,16,balanced,0.048122664292653404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,16,16,balanced,0.0491946687301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,16,16,balanced,0.052373334765434265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,16,16,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,16,16,balanced,0.05871466795603434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,16,16,balanced,0.060319999853769936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,16,16,balanced,0.06018666426340739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,16,16,balanced,0.06233066817124685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,16,16,balanced,0.06168533364931742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,1,power_law_1.2,1.9110912322998046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,16,16,balanced,0.06252266466617584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,4,power_law_1.2,0.0462336003780365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,16,16,balanced,0.06565333406130473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,16,16,balanced,0.06516799827416737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,4,power_law_1.2,0.04723840057849884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,16,16,balanced,0.06773333251476288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,1,power_law_1.2,2.0513727188110353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,4,power_law_1.2,0.051545602083206174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,4,power_law_1.2,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,4,power_law_1.2,0.06625279784202576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,16,16,balanced,0.07667733232180278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,16,16,balanced,0.07851199805736542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,1,power_law_1.2,2.128428840637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,16,16,balanced,0.0827946662902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,1,power_law_1.2,2.1929599761962892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,4,power_law_1.2,0.06787840127944947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,1,power_law_1.2,2.256211280822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,16,16,balanced,0.09557333588600159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,16,16,balanced,0.10398933291435242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,16,16,balanced,0.11529599626859029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,1,power_law_1.2,2.3191104888916017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,16,16,balanced,0.13167466719945273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,16,16,balanced,0.162800004084905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,16,16,balanced,0.1909546653429667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,4,power_law_1.2,0.07778559923171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,1,power_law_1.2,2.3733312606811525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,1,power_law_1.2,2.4821760177612306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,16,16,balanced,0.25870933135350543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,1,power_law_1.2,2.5307008743286135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,16,16,balanced,0.31940799951553345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,16,16,balanced,0.45844801266988117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,16,16,balanced,0.5960426727930704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,4,power_law_1.2,0.0886847972869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,1,power_law_1.2,2.6242752075195312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,4,power_law_1.2,0.12012799978256225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,1,power_law_1.2,2.8452735900878907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,1,power_law_1.2,3.064838409423828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,1,power_law_1.2,3.1263168334960936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,4,power_law_1.2,0.1240447998046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,1,power_law_1.2,3.4934593200683595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,1,power_law_1.2,3.337055969238281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,1,power_law_1.2,3.7846401214599608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,1,power_law_1.2,4.419174575805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,1,power_law_1.2,5.134431838989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,1,power_law_1.2,5.832633590698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,1,power_law_1.2,7.773677062988281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,4,power_law_1.2,0.16047999858856202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,1,power_law_1.2,9.222560119628906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,4,power_law_1.2,0.1941696047782898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,4,power_law_1.2,0.2476799964904785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,4,power_law_1.2,0.3502336025238037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,4,power_law_1.2,0.45308799743652345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,2,power_law_1.2,0.06459519863128663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,2,power_law_1.2,0.07120000123977661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,2,power_law_1.2,0.0753216028213501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,2,power_law_1.2,0.1016319990158081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,2,power_law_1.2,0.11510399580001832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,2,power_law_1.2,0.12683520317077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,2,power_law_1.2,0.16380159854888915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,2,power_law_1.2,0.1621440052986145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,2,power_law_1.2,0.17531520128250122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,2,power_law_1.2,0.17325439453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,2,power_law_1.2,0.17900799512863158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,2,power_law_1.2,0.18357759714126587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,2,power_law_1.2,0.1827839970588684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,2,power_law_1.2,0.19550080299377443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,2,power_law_1.2,0.20108799934387206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,2,power_law_1.2,0.20729598999023438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,2,power_law_1.2,0.22563838958740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,2,power_law_1.2,0.2434175968170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.2,0.27777919769287107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.2,0.3272831916809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.2,0.37329280376434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.2,0.4881599903106689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.2,0.6017087936401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.2,0.8042112350463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.2,0.9802304267883301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.2,1.4946751594543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.2,1.9359935760498046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,4,power_law_1.2,0.6364672183990479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,8,power_law_1.01,0.0315200001001358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,8,power_law_1.01,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,8,power_law_1.01,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,8,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,8,power_law_1.01,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,8,power_law_1.01,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,8,power_law_1.01,0.03456639945507049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,8,power_law_1.01,0.035174399614334106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,8,power_law_1.01,0.03585279881954193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,8,power_law_1.01,0.03769600093364715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,8,power_law_1.01,0.03779839873313904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,8,power_law_1.01,0.039289599657058714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,8,power_law_1.01,0.040908798575401306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,8,power_law_1.01,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,8,power_law_1.01,0.050323200225830075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,8,power_law_1.01,0.056550401449203494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,8,power_law_1.01,0.0627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,8,power_law_1.01,0.07341439723968506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,8,power_law_1.01,0.08718720078468323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,8,power_law_1.01,0.1245311975479126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,8,power_law_1.01,0.1686784029006958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,8,power_law_1.01,0.2184704065322876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,8,power_law_1.01,0.26138880252838137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,8,power_law_1.01,0.41598081588745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,8,power_law_1.01,0.6116415977478027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,1,balanced,0.06006933252016703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,1,balanced,0.06829866766929626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,1,balanced,0.08378666639328003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,1,balanced,0.11853866775830586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,1,balanced,0.18083200852076212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,1,balanced,0.28946133454640705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,1,balanced,0.3978293339411418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,1,balanced,0.4000053405761719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,1,balanced,0.40272001425425213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,1,balanced,0.40458667278289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,1,balanced,0.4099786678949992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,1,balanced,0.41626131534576416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,1,balanced,0.42159998416900635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,1,balanced,0.4292213519414266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,1,balanced,0.4395466645558675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,1,balanced,0.44818135102589923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,1,balanced,0.4681919813156128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,1,balanced,0.5279839833577474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,1,balanced,0.5829013188680013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,1,balanced,0.6699786980946859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,1,balanced,0.7431306838989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,1,balanced,0.9142560164133707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,1,balanced,1.1284693082173665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,1,balanced,1.4652533531188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,8,power_law_1.01,0.9549375534057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,1,balanced,1.940922737121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,1,balanced,2.637450695037842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,1,balanced,3.5218238830566406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,8,power_law_1.01,1.2462271690368651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,32,power_law_1.01,0.139193594455719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,32,power_law_1.01,0.1692031979560852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,32,power_law_1.01,0.16941440105438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,32,power_law_1.01,0.17031680345535277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,32,power_law_1.01,0.17004159688949586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,32,power_law_1.01,0.15527679920196533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,32,power_law_1.01,0.13359999656677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,32,power_law_1.01,0.14131200313568115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,32,power_law_1.01,0.14344320297241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,32,power_law_1.01,0.15387519598007202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,32,power_law_1.01,0.14336639642715454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,32,power_law_1.01,0.1553663969039917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,32,power_law_1.01,0.1572864055633545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,32,power_law_1.01,0.16085120439529418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,32,power_law_1.01,0.148307204246521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,32,power_law_1.01,0.1630784034729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,32,power_law_1.01,0.16038399934768677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,32,power_law_1.01,0.1698240041732788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,32,power_law_1.01,0.1784767985343933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,32,power_law_1.01,0.21752960681915284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,32,power_law_1.01,0.24769918918609618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,32,power_law_1.01,0.29974400997161865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,32,power_law_1.01,0.36003201007843016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,32,power_law_1.01,0.5076543807983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,32,power_law_1.01,0.5650688171386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,32,power_law_1.01,0.8390912055969239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,32,power_law_1.01,1.0210368156433105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,128,power_law_1.01,0.047219198942184445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,128,power_law_1.01,0.04547840058803558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,128,power_law_1.01,0.04602240025997162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,128,power_law_1.01,0.04936319887638092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,128,power_law_1.01,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,128,power_law_1.01,0.052083200216293334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,128,power_law_1.01,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,128,power_law_1.01,0.05580800175666809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,128,power_law_1.01,0.05846400260925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,128,power_law_1.01,0.05813120007514953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,128,power_law_1.01,0.05999360084533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,128,power_law_1.01,0.06380800008773804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,128,power_law_1.01,0.06663680076599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,128,power_law_1.01,0.06428800225257873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,128,power_law_1.01,0.07403519749641418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,128,power_law_1.01,0.07953280210494995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,128,power_law_1.01,0.08191999793052673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,128,power_law_1.01,0.09583359956741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,128,power_law_1.01,0.10836479663848878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,128,power_law_1.01,0.13063039779663085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,128,power_law_1.01,0.15023360252380372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,128,power_law_1.01,0.1924415946006775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,128,power_law_1.01,0.2305919885635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,128,power_law_1.01,0.31305599212646484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,128,power_law_1.01,0.4262335777282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,128,power_law_1.01,0.6080512046813965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,128,power_law_1.01,0.820678424835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,1,power_law_1.2,0.07091839909553528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,1,power_law_1.2,0.09360640048980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,1,power_law_1.2,0.10207359790802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,1,power_law_1.2,0.12015999555587768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,1,power_law_1.2,0.1315840005874634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,1,power_law_1.2,0.17363200187683106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,1,power_law_1.2,0.1852031946182251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,1,power_law_1.2,0.18814719915390016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,1,power_law_1.2,0.19306880235671997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,1,power_law_1.2,0.1969215989112854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,1,power_law_1.2,0.19913599491119385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,1,power_law_1.2,0.2053056001663208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,1,power_law_1.2,0.21020801067352296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,1,power_law_1.2,0.2167680025100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,1,power_law_1.2,0.21838719844818116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,1,power_law_1.2,0.22560639381408693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,1,power_law_1.2,0.2380352020263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,1,power_law_1.2,0.28142719268798827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.2,0.3229055881500244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,16,balanced,0.045824001232783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,16,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.2,0.4091519832611084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.2,0.492844820022583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.2,0.6617983818054199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,16,balanced,0.04660266637802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,16,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.2,0.81778564453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,16,balanced,0.049957334995269775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,16,balanced,0.05605866511662801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,16,balanced,0.056234667698542275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,16,balanced,0.056474665800730385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,16,balanced,0.05615466833114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,16,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,16,balanced,0.056143999099731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,16,balanced,0.057946667075157166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,16,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,16,balanced,0.0603413333495458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,16,balanced,0.06639466683069865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,16,balanced,0.06628266473611195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,16,balanced,0.06840533514817555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,16,balanced,0.07457600037256877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,16,balanced,0.07874666651089986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,16,balanced,0.09098133444786072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,16,balanced,0.0993386705716451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,16,balanced,0.122789333264033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,16,balanced,0.14965333541234335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,16,balanced,0.20560000340143839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,16,balanced,0.2508479952812195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,16,balanced,0.3340373436609904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,16,balanced,0.4267253478368123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.2,1.1623295783996581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.2,1.4955072402954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,8,power_law_1.01,0.09113600254058837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,8,power_law_1.01,0.12541439533233642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,8,power_law_1.01,0.11119359731674194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,8,power_law_1.01,0.1253376007080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,8,power_law_1.01,0.1323456048965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,8,power_law_1.01,0.13571200370788575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,8,power_law_1.01,0.13585280179977416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,8,power_law_1.01,0.13967360258102418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,8,power_law_1.01,0.1383679986000061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,8,power_law_1.01,0.14156800508499146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,8,power_law_1.01,0.14336639642715454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,8,power_law_1.01,0.14439040422439575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,8,power_law_1.01,0.14810240268707275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,8,power_law_1.01,0.15056639909744263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,8,power_law_1.01,0.15872639417648315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,8,power_law_1.01,0.15901440382003784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,8,power_law_1.01,0.17061760425567626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,8,power_law_1.01,0.18486399650573732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,8,power_law_1.01,0.20171520709991456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,8,power_law_1.01,0.23325440883636475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.2,2.162246322631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,8,power_law_1.01,0.2758271932601929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,8,power_law_1.01,0.34174718856811526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,8,power_law_1.01,0.44821758270263673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,8,power_law_1.01,0.6135871887207032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,8,power_law_1.01,0.8603775978088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,8,power_law_1.01,1.1316415786743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,8,power_law_1.01,1.5888575553894042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.2,2.8543935775756837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,4,balanced,0.019120000302791595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,4,balanced,0.020010666300853092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,4,balanced,0.020874666670958202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,4,balanced,0.024330665667851765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,4,balanced,0.029365333418051403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,4,balanced,0.03145600110292435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,4,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,4,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,4,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,4,balanced,0.039674667020638786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,4,balanced,0.04032533367474874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,4,balanced,0.05156800150871277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,4,balanced,0.04901866614818573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,4,balanced,0.051967998345692955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,4,balanced,0.06388799846172333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,4,balanced,0.06435733536879222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,4,balanced,0.07011199990908305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,4,balanced,0.07960533102353413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,4,balanced,0.08778132994969685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,4,balanced,0.09336533149083455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,4,balanced,0.09893866380055745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,4,balanced,0.10725333293279012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,4,balanced,0.11819199721018474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,4,balanced,0.1886133352915446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,4,balanced,0.2057973345120748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,4,balanced,0.29285866022109985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,4,balanced,0.38149865468343097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,32,1,power_law_1.2,0.020371200144290925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,32,1,power_law_1.2,0.021804800629615782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,32,1,power_law_1.2,0.024831999838352204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,32,1,power_law_1.2,0.030796799063682555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,32,1,power_law_1.2,0.04012799859046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,32,1,power_law_1.2,0.058156800270080564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,32,1,power_law_1.2,0.07363200187683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,32,1,power_law_1.2,0.07722880244255066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,32,1,power_law_1.2,0.07938560247421264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,32,1,power_law_1.2,0.08352000117301941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,32,1,power_law_1.2,0.08380799889564514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,32,1,power_law_1.2,0.08675839900970458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,32,1,power_law_1.2,0.08967679738998413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,32,1,power_law_1.2,0.09276800155639649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,32,1,power_law_1.2,0.09698560237884521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,32,1,power_law_1.2,0.10085120201110839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,32,1,power_law_1.2,0.10902400016784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,32,1,power_law_1.2,0.13025280237197875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.2,0.14384640455245973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.2,0.18042240142822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.2,0.21487998962402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.2,0.2885632038116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.2,0.36559998989105225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.2,0.45171198844909666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.2,0.52008957862854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,32,1,power_law_1.2,0.6970367908477784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,32,1,power_law_1.2,0.8750656127929688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,64,power_law_1.01,0.07600640058517456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,64,power_law_1.01,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,64,power_law_1.01,0.07101439833641052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,64,power_law_1.01,0.07518079876899719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,64,power_law_1.01,0.07502719759941101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,64,power_law_1.01,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,64,power_law_1.01,0.0732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,64,power_law_1.01,0.07838079929351807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,64,power_law_1.01,0.07893120050430298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,64,power_law_1.01,0.07822719812393189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,64,power_law_1.01,0.08893439769744874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,64,power_law_1.01,0.09166079759597778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,64,power_law_1.01,0.0926912009716034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,64,power_law_1.01,0.08775039911270141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,64,power_law_1.01,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,64,power_law_1.01,0.10371199846267701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,64,power_law_1.01,0.10779520273208618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,64,power_law_1.01,0.11621760129928589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,64,power_law_1.01,0.12649600505828856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,64,power_law_1.01,0.14928640127182008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,64,power_law_1.01,0.1639680027961731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,64,power_law_1.01,0.21062400341033935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,64,power_law_1.01,0.2648832082748413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,64,power_law_1.01,0.34983038902282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,64,power_law_1.01,0.4604800224304199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,64,power_law_1.01,0.677510404586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,64,power_law_1.01,0.860531234741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,16,power_law_1.2,0.051046401262283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,16,power_law_1.2,0.04664320051670075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,16,power_law_1.2,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,16,power_law_1.2,0.0456063985824585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,16,power_law_1.2,0.045638400316238406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,16,power_law_1.2,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,16,power_law_1.2,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,16,power_law_1.2,0.04984320104122162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,16,power_law_1.2,0.04896639883518219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,16,power_law_1.2,0.04986239969730377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,16,power_law_1.2,0.04976640045642853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,16,power_law_1.2,0.051583999395370485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,16,power_law_1.2,0.05061759948730469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,16,power_law_1.2,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,16,power_law_1.2,0.058310401439666745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,16,power_law_1.2,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,16,power_law_1.2,0.061350399255752565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,16,power_law_1.2,0.07186560034751892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.2,0.07646719813346863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.2,0.09434239864349366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.2,0.11227519512176513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,32,balanced,0.04342933495839437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,32,balanced,0.04200000067551931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.2,0.13106559514999389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,32,balanced,0.04343999922275543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,32,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,32,balanced,0.04794133206208547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,32,balanced,0.05643199880917867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,32,balanced,0.07250666618347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,32,balanced,0.0707946668068568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,32,balanced,0.07172800103823344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,32,balanced,0.0724533349275589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,32,balanced,0.07306666672229767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,32,balanced,0.07614933451016744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,32,balanced,0.07860266665617625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,32,balanced,0.07828799883524577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,32,balanced,0.08619733651479085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,32,balanced,0.08738133311271667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,32,balanced,0.09428800145785014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,32,balanced,0.10520000259081523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,32,balanced,0.11517332990964253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,32,balanced,0.1365653375784556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,32,balanced,0.15451733271280924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,32,balanced,0.19753066698710123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,32,balanced,0.24055999517440796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,32,balanced,0.3220213254292806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,32,balanced,0.4061973492304484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,32,balanced,0.5716266632080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,32,balanced,0.7326613267262777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.2,0.16636159420013427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.2,0.2284480094909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.2,0.2901952028274536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.2,0.3745215892791748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.2,0.5225152015686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,8,power_law_1.01,0.046412798762321475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,8,power_law_1.01,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,8,power_law_1.01,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,8,power_law_1.01,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,8,power_law_1.01,0.047513601183891294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,8,power_law_1.01,0.048256000876426695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,8,power_law_1.01,0.04756479859352112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,8,power_law_1.01,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,8,power_law_1.01,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,8,power_law_1.01,0.046911999583244324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,8,power_law_1.01,0.046438398957252505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,8,power_law_1.01,0.04892799854278564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,8,power_law_1.01,0.049465599656105044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,8,power_law_1.01,0.04955520033836365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,8,power_law_1.01,0.05611519813537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,8,power_law_1.01,0.055129599571228025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,8,power_law_1.01,0.061292797327041626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,8,power_law_1.01,0.06806399822235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,8,power_law_1.01,0.07939839959144593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,8,power_law_1.01,0.10101759433746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,8,power_law_1.01,0.11847039461135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,8,power_law_1.01,0.15667840242385864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,8,power_law_1.01,0.18410880565643312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,8,power_law_1.01,0.25571839809417723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,8,power_law_1.01,0.3430847883224487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,32,8,power_law_1.2,0.016230399906635284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,32,8,power_law_1.2,0.016595199704170227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,32,8,power_law_1.2,0.016505600512027742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,32,8,power_law_1.2,0.017132799327373504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,32,8,power_law_1.2,0.017401599884033205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,32,8,power_law_1.2,0.018323199450969697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,32,8,power_law_1.2,0.017843200266361235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,32,8,power_law_1.2,0.020204800367355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,32,8,power_law_1.2,0.021004800498485566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,32,8,power_law_1.2,0.020710399746894835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,32,8,power_law_1.2,0.028908801078796387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,32,8,power_law_1.2,0.029542401432991028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,32,8,power_law_1.2,0.030112001299858093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,32,8,power_law_1.2,0.0296640008687973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,32,8,power_law_1.2,0.030297601222991945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,32,8,power_law_1.2,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,32,8,power_law_1.2,0.032416000962257385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,32,8,power_law_1.2,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,32,8,power_law_1.2,0.040940800309181215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,32,8,power_law_1.2,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,32,8,power_law_1.2,0.04917759895324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,8,power_law_1.01,0.5174784183502197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,32,8,power_law_1.2,0.06867200136184692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,32,8,power_law_1.2,0.07858560085296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,32,8,power_law_1.2,0.10629119873046874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,8,power_law_1.01,0.641926383972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,32,8,power_law_1.2,0.13579519987106323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,8,power_law_1.2,0.05915520191192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,8,power_law_1.2,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,8,power_law_1.2,0.06033920049667359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,8,power_law_1.2,0.06310399770736694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,8,power_law_1.2,0.0615231990814209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,8,power_law_1.2,0.060012799501419065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,8,power_law_1.2,0.05902720093727112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,8,power_law_1.2,0.05952640175819397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,8,power_law_1.2,0.055283200740814206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,8,power_law_1.2,0.05464320182800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,8,power_law_1.2,0.05011839866638183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,8,power_law_1.2,0.05749760270118713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,8,power_law_1.2,0.05971840023994446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,8,power_law_1.2,0.05589119791984558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,8,power_law_1.2,0.07182080149650574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,8,power_law_1.2,0.0754047989845276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,8,power_law_1.2,0.0825984001159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,8,power_law_1.2,0.10812159776687622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.2,0.12420480251312256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.2,0.15028480291366578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,8,power_law_1.2,0.01870719939470291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,8,power_law_1.2,0.018265600502490997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,8,power_law_1.2,0.019500799477100372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,8,power_law_1.2,0.019487999379634857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,8,power_law_1.2,0.019289599359035493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,32,8,power_law_1.2,0.1875391960144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,8,power_law_1.2,0.020812800526618956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,8,power_law_1.2,0.022336000204086305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,8,power_law_1.2,0.02280319929122925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.2,0.1999295949935913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,8,power_law_1.2,0.027673599123954774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,8,power_law_1.2,0.02776319980621338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,8,power_law_1.2,0.028044798970222475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,8,power_law_1.2,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,8,power_law_1.2,0.04392960071563721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,8,power_law_1.2,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,8,power_law_1.2,0.04398080110549927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,8,power_law_1.2,0.04412800073623657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,8,power_law_1.2,0.044275200366973876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,8,power_law_1.2,0.046988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,8,power_law_1.2,0.050432002544403075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,8,power_law_1.2,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,8,power_law_1.2,0.06565120220184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,8,power_law_1.2,0.08704000115394592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,8,power_law_1.2,0.10232319831848144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,8,power_law_1.2,0.13615360260009765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.2,0.271347188949585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,8,power_law_1.2,0.1720896005630493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.2,0.35165441036224365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,32,8,power_law_1.2,0.23423359394073487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,8,power_law_1.2,0.23592960834503174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,4,power_law_1.2,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,4,power_law_1.2,0.033580800890922545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,4,power_law_1.2,0.031916800141334536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,4,power_law_1.2,0.03201279938220978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,4,power_law_1.2,0.032518398761749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,4,power_law_1.2,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.2,0.5062784194946289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,4,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,4,power_law_1.2,0.03434880077838898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,4,power_law_1.2,0.035334399342536925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,4,power_law_1.2,0.035545599460601804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,4,power_law_1.2,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,4,power_law_1.2,0.037171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,4,power_law_1.2,0.038073599338531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,4,power_law_1.2,0.039743998646736146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,4,power_law_1.2,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,4,power_law_1.2,0.05448319911956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,4,power_law_1.2,0.053574401140213015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,4,power_law_1.2,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,4,power_law_1.2,0.0717248022556305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,2,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,2,balanced,0.05182399849096934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,2,balanced,0.07561066746711731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,2,balanced,0.07663466533025105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,2,balanced,0.07667199770609538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,2,balanced,0.0784746656815211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,2,balanced,0.07799999912579854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,4,power_law_1.2,0.08796160221099854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,2,balanced,0.07662400106589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,2,balanced,0.07930666704972585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,2,balanced,0.07876266539096832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,4,power_law_1.2,0.10142079591751099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,2,balanced,0.07939200103282928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,2,balanced,0.08079466720422109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,4,power_law_1.2,0.13490560054779052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,2,balanced,0.08079466720422109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,2,balanced,0.08656533559163411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,2,balanced,0.09029866258303325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,2,balanced,0.09111467003822327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,2,balanced,0.10027199983596802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,2,balanced,0.10708266496658325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,2,balanced,0.11726933717727661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,2,balanced,0.14155200123786926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,2,balanced,0.1681600014368693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,2,balanced,0.2227840026219686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,2,balanced,0.2643413345019023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,2,balanced,0.35604266325632733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,2,balanced,0.42738668123881024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,2,balanced,0.6203519900639852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,2,balanced,0.7878986994425455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.2,0.6615295886993409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,8,power_law_1.2,0.29650559425354006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,4,power_law_1.2,0.14858239889144897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,4,power_law_1.2,0.20101120471954345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,4,power_law_1.2,0.24211199283599855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.2,0.9529279708862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,4,power_law_1.2,0.3635456085205078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.2,1.2865344047546388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,4,power_law_1.01,0.03898240029811859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,4,power_law_1.01,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,4,power_law_1.01,0.031193599104881287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,4,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,4,power_law_1.01,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,4,power_law_1.01,0.03403519988059998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,4,power_law_1.01,0.03400320112705231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,4,power_law_1.01,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,4,power_law_1.01,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,4,power_law_1.01,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,4,power_law_1.01,0.036582401394844054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,4,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,4,power_law_1.01,0.03865599930286408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,4,power_law_1.01,0.040659201145172116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,4,power_law_1.01,0.04838399887084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,4,power_law_1.01,0.05107839703559876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,4,power_law_1.01,0.05324159860610962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,4,power_law_1.01,0.06276479959487916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,4,power_law_1.01,0.06785920262336731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,4,power_law_1.01,0.09378560185432434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,4,power_law_1.01,0.10561920404434204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,4,power_law_1.01,0.13160959482192994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,4,power_law_1.2,0.4295807838439941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,4,power_law_1.01,0.16556799411773682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,4,power_law_1.01,0.22830719947814943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,4,power_law_1.01,0.25884160995483396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,2,power_law_1.01,0.04097920060157776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,2,power_law_1.01,0.044012799859046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,2,power_law_1.01,0.043942400813102724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,2,power_law_1.01,0.05472639799118042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,2,power_law_1.01,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,2,power_law_1.01,0.05761280059814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,2,power_law_1.01,0.05947520136833191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,2,power_law_1.01,0.06302080154418946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,2,power_law_1.01,0.06351360082626342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,2,power_law_1.01,0.06299520134925843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,2,power_law_1.01,0.06405119895935059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,2,power_law_1.01,0.06445440053939819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,2,power_law_1.01,0.066048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,2,power_law_1.01,0.0754368007183075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,2,power_law_1.01,0.08000640273094177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,1,power_law_1.01,0.09267200231552124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,1,power_law_1.01,0.13625600337982177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,1,power_law_1.01,0.2132159948348999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,1,power_law_1.01,0.3790015935897827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,1,power_law_1.01,0.564793586730957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,2,power_law_1.01,0.0840448021888733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,2,power_law_1.01,0.09870719909667969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,1,power_law_1.01,0.7204160213470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,4,power_law_1.01,0.3531455993652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,1,power_law_1.01,1.0220159530639648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,2,power_law_1.01,0.10867840051651001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,1,power_law_1.01,1.092908763885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,1,power_law_1.01,1.0923968315124513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,1,power_law_1.01,1.1232383728027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,1,power_law_1.01,1.162623977661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,2,power_law_1.01,0.12225279808044434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,1,power_law_1.01,1.2040448188781738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,1,power_law_1.01,1.2011391639709472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,2,power_law_1.01,0.14859520196914672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,1,power_law_1.01,1.265062427520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,1,power_law_1.01,1.2914624214172363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,1,power_law_1.01,1.3147392272949219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,1,power_law_1.01,1.3693696022033692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,2,power_law_1.01,0.19315199851989745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,1,power_law_1.01,1.4912832260131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,1,power_law_1.01,1.537440013885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,1,power_law_1.01,1.7511552810668944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,1,power_law_1.01,1.7060928344726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,1,power_law_1.01,1.9164543151855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,1,power_law_1.01,2.234956741333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,2,power_law_1.01,0.25851519107818605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,1,power_law_1.01,2.7460351943969727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,2,power_law_1.01,0.30001280307769773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,1,power_law_1.01,3.185087966918945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,1,power_law_1.01,4.132371139526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,1,power_law_1.01,5.049318313598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,32,8,power_law_1.01,0.021747200191020964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,32,8,power_law_1.01,0.022918400168418885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,32,8,power_law_1.01,0.02444159984588623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,32,8,power_law_1.01,0.025036799907684325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,32,8,power_law_1.01,0.026790401339530943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,32,8,power_law_1.01,0.02844800055027008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,32,8,power_law_1.01,0.030656000971794127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,32,8,power_law_1.01,0.031737598776817325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,32,8,power_law_1.01,0.032153600454330446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,32,8,power_law_1.01,0.0324288010597229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,32,8,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,32,8,power_law_1.01,0.041228801012039185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,32,8,power_law_1.01,0.04035840034484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,32,8,power_law_1.01,0.04162560105323791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,32,8,power_law_1.01,0.0645695984363556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,32,8,power_law_1.01,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,32,8,power_law_1.01,0.06842880249023438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,32,8,power_law_1.01,0.07311999797821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,32,8,power_law_1.01,0.08068479895591736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,32,8,power_law_1.01,0.0917568027973175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,32,8,power_law_1.01,0.09920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,32,8,power_law_1.01,0.1219648003578186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,32,8,power_law_1.01,0.13430399894714357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,32,8,power_law_1.01,0.19159679412841796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,2,power_law_1.01,0.41354880332946775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,32,8,power_law_1.01,0.2381056070327759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,32,8,power_law_1.01,0.3221888065338135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,32,8,power_law_1.01,0.41515522003173827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,4,power_law_1.01,0.4612864017486572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,2,power_law_1.01,0.5569151878356934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,2,power_law_1.01,0.7565951824188233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,2,power_law_1.01,1.038924789428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,2,power_law_1.01,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,2,power_law_1.01,0.05288959741592407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,2,power_law_1.01,0.05722879767417908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,2,power_law_1.01,0.06420480012893677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,4,power_law_1.01,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,2,power_law_1.01,0.07000319957733155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,4,power_law_1.01,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,2,power_law_1.01,0.07946239709854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,4,power_law_1.01,0.037510401010513304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,2,power_law_1.01,0.07936000227928161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,4,power_law_1.01,0.04059520065784454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,2,power_law_1.01,0.0838591992855072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,4,power_law_1.01,0.040992000699043275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,2,power_law_1.01,0.0831488013267517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,4,power_law_1.01,0.04328320026397705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,4,power_law_1.01,0.04356479942798615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,4,power_law_1.01,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,4,power_law_1.01,0.04312959909439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,4,power_law_1.01,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,4,power_law_1.01,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,2,power_law_1.01,0.0838271975517273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,2,power_law_1.01,0.08721280097961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,2,power_law_1.01,0.0893119990825653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,2,power_law_1.01,0.09155840277671815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,2,power_law_1.01,0.09479680061340331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,2,power_law_1.01,0.09353600144386291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,2,power_law_1.01,0.09324799776077271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,2,power_law_1.01,0.09900799989700318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,2,power_law_1.01,0.11045759916305542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,2,power_law_1.01,0.12412159442901612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,4,power_law_1.01,0.05200639963150024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,2,power_law_1.01,0.1505344033241272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,4,power_law_1.01,0.05742080211639404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,4,power_law_1.01,0.06169599890708923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,2,power_law_1.01,0.17277439832687377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,2,power_law_1.01,0.21532158851623534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,2,power_law_1.01,0.30100479125976565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,2,power_law_1.01,0.4149631977081299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,2,power_law_1.01,0.5546879768371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,2,power_law_1.01,0.7977920055389405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,2,power_law_1.01,1.039577579498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,4,power_law_1.01,0.07762560248374939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,4,power_law_1.01,0.07317119836807251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,4,power_law_1.01,0.09724799990653991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,4,power_law_1.01,0.11164799928665162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,4,power_law_1.01,0.12520960569381714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,4,power_law_1.01,0.1705471992492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,4,power_law_1.01,0.20473599433898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,4,power_law_1.01,0.3047039985656738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,4,power_law_1.01,0.4432191848754883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,4,power_law_1.01,0.6319551944732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,4,power_law_1.01,0.8371711730957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,64,power_law_1.01,0.01608320027589798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,64,power_law_1.01,0.017100800573825837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,64,power_law_1.01,0.0159168004989624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,64,power_law_1.01,0.016102400422096253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,64,power_law_1.01,0.01706880033016205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,64,power_law_1.01,0.01963520050048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,64,power_law_1.01,0.01950719952583313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,64,power_law_1.01,0.025971201062202454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,64,power_law_1.01,0.02531839907169342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,64,power_law_1.01,0.025484800338745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,64,power_law_1.01,0.02587519884109497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,64,power_law_1.01,0.026015999913215637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,64,power_law_1.01,0.026310399174690247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,64,power_law_1.01,0.02685439884662628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,64,power_law_1.01,0.027993598580360414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,64,power_law_1.01,0.028300800919532777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,64,power_law_1.01,0.029158401489257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,64,power_law_1.01,0.031667199730873105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,64,power_law_1.01,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,64,power_law_1.01,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,64,power_law_1.01,0.041536000370979306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,64,power_law_1.01,0.04686720073223114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,64,power_law_1.01,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,64,power_law_1.01,0.06933760046958923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,64,power_law_1.01,0.08584960103034973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,64,power_law_1.01,0.1204800009727478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,64,power_law_1.01,0.14199039936065674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,16,balanced,0.049925332268079124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,16,balanced,0.05087999999523163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,16,balanced,0.05118933320045471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,16,balanced,0.06115200122197469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,16,balanced,0.0804799993832906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,16,balanced,0.10946133732795715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,16,balanced,0.10872532924016316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,16,balanced,0.10925867160161336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,16,balanced,0.10949333508809407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,16,balanced,0.11077333490053813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,16,balanced,0.10940266648928325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,16,balanced,0.112527996301651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,16,balanced,0.1141866644223531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,16,balanced,0.11468266447385152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,16,balanced,0.12114666899045308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,16,balanced,0.12153066198031108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,16,balanced,0.12398933370908101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,16,balanced,0.1334933340549469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,16,balanced,0.1414346694946289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,16,balanced,0.15693333745002747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,16,balanced,0.1725813349088033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,16,balanced,0.19910933574040732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,16,balanced,0.22915732860565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,16,balanced,0.29710932572682697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,16,balanced,0.3586719830830892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,16,balanced,0.48880000909169513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,16,balanced,0.6181973218917847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,2,balanced,0.05486399928728739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,2,balanced,0.055071999629338585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,2,balanced,0.057328000664711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,2,balanced,0.06820266445477803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,2,balanced,0.08246933420499165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,2,balanced,0.08346133430798848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,2,balanced,0.08409600456555684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,2,balanced,0.08528533577919006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,2,balanced,0.08498666683832805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,2,balanced,0.08686932921409607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,2,balanced,0.08629866441090901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,2,balanced,0.091648002465566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,2,balanced,0.09087999661763509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,2,balanced,0.09290666381518047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,2,balanced,0.09445333480834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,2,balanced,0.09578133622805278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,2,balanced,0.09770666559537251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,2,balanced,0.10712533195813496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,2,balanced,0.11347200473149617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,2,balanced,0.13064000010490417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,2,balanced,0.14363200465838113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,2,balanced,0.19729065895080566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,2,balanced,0.23566933472951254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,2,balanced,0.34043200810750324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,2,balanced,0.4380906820297241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,2,balanced,0.6288640101750692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,2,balanced,0.8226933479309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,4,power_law_1.01,1.1977919578552245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,4,power_law_1.01,1.630726432800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,2,balanced,0.03995199998219808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,2,balanced,0.04603200157483419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,2,balanced,0.05613866448402405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,2,balanced,0.05081599950790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,2,balanced,0.053914666175842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,2,balanced,0.053898667295773826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,2,balanced,0.049679999550183616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,2,balanced,0.05825066566467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,2,balanced,0.0540533314148585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,2,balanced,0.053504000107447304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,2,balanced,0.050069332122802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,2,balanced,0.053114667534828186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,2,balanced,0.05343466500441233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,2,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,2,balanced,0.053818667928377785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,2,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,2,balanced,0.05795200169086456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,2,balanced,0.07505066692829132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,2,balanced,0.07868266602357228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,2,balanced,0.09622933467229207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,2,balanced,0.11124266187349956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,2,balanced,0.14401066303253174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,2,balanced,0.17500799894332886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,2,balanced,0.24142932891845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,2,balanced,0.306549330552419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,2,balanced,0.43275201320648193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,2,balanced,0.5633386770884196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,128,power_law_1.2,0.02855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,128,power_law_1.2,0.02810879945755005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,128,power_law_1.2,0.028787198662757873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,128,power_law_1.2,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,128,power_law_1.2,0.032627201080322264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,128,power_law_1.2,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,128,power_law_1.2,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,8,power_law_1.2,0.09581440091133117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,8,power_law_1.2,0.14282239675521852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,8,power_law_1.2,0.1811519980430603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,8,power_law_1.2,0.22228479385375977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,8,power_law_1.2,0.2553728103637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,8,power_law_1.2,0.28846719264984133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,8,power_law_1.2,0.30707199573516847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,8,power_law_1.2,0.32126080989837646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,8,power_law_1.2,0.2631360054016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,8,power_law_1.2,0.2968064069747925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,128,power_law_1.2,0.03389439880847931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,8,power_law_1.2,0.2961280107498169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,8,power_law_1.2,0.2824640035629272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,8,power_law_1.2,0.28612480163574217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,8,power_law_1.2,0.2796544075012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,8,power_law_1.2,0.3309184074401855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,8,power_law_1.2,0.31395199298858645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,128,power_law_1.2,0.035436800122261046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,8,power_law_1.2,0.3525183916091919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,8,power_law_1.2,0.3824383974075317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,8,power_law_1.2,0.3971839904785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,8,power_law_1.2,0.48592000007629393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,8,power_law_1.2,0.5047872066497803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,64,power_law_1.01,0.018777599930763243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,64,power_law_1.01,0.01820800006389618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,8,power_law_1.2,0.6095551967620849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,64,power_law_1.01,0.01886720061302185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,64,power_law_1.01,0.018515199422836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,8,power_law_1.2,0.6653759956359864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,64,power_law_1.01,0.019648000597953796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,64,power_law_1.01,0.02287999987602234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,8,power_law_1.2,0.9008319854736329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,128,power_law_1.2,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,64,power_law_1.01,0.022937600314617158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,8,power_law_1.2,0.9644479751586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,64,power_law_1.01,0.03245440125465393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,64,power_law_1.01,0.03256320059299469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,8,power_law_1.2,1.2405183792114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,8,power_law_1.2,1.4895359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,128,power_law_1.2,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,64,power_law_1.01,0.032339200377464294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,64,power_law_1.01,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,64,power_law_1.01,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,64,power_law_1.01,0.03210879862308502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,64,power_law_1.01,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,64,power_law_1.01,0.033632001280784606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,128,power_law_1.2,0.0378495991230011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,64,power_law_1.01,0.03447040021419525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,128,power_law_1.2,0.03980799913406372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,64,power_law_1.01,0.035648000240325925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,64,power_law_1.01,0.03978880047798157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.01,0.04251520037651062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.01,0.04853120148181915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.01,0.05380480289459229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.01,0.07475839853286743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.01,0.09054080247879029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,128,power_law_1.2,0.04352000057697296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,128,power_law_1.2,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,128,power_law_1.2,0.04967679977416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,128,power_law_1.2,0.056620800495147706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,128,power_law_1.2,0.06652799844741822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,128,power_law_1.2,0.07554559707641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.01,0.11297919750213622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,128,power_law_1.2,0.10410239696502685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.01,0.13742719888687133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,128,power_law_1.2,0.12336000204086303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.01,0.20570240020751954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.01,0.2750976085662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,128,power_law_1.2,0.19623039960861205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,128,power_law_1.2,0.2249919891357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,128,power_law_1.2,0.36515839099884034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,128,power_law_1.2,0.4858687877655029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,128,power_law_1.2,0.6964032173156738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,64,power_law_1.2,0.01828480064868927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,64,power_law_1.2,0.01791999936103821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,64,power_law_1.2,0.02083200067281723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,64,power_law_1.2,0.019193600118160247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,64,power_law_1.2,0.02131839990615845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,32,power_law_1.01,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,32,power_law_1.01,0.029177600145339967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,32,power_law_1.01,0.02813439965248108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,32,power_law_1.01,0.02900480031967163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,32,power_law_1.01,0.030553600192070006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,32,power_law_1.01,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,32,power_law_1.01,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,32,power_law_1.01,0.031699201464653014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,32,power_law_1.01,0.03162879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,32,power_law_1.01,0.032102400064468385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,32,power_law_1.01,0.03346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,32,power_law_1.01,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,32,power_law_1.01,0.03497599959373474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,32,power_law_1.01,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,32,power_law_1.01,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,32,power_law_1.01,0.041407999396324155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,32,power_law_1.01,0.04136320054531097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,32,power_law_1.01,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,32,power_law_1.01,0.048614400625228885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,32,power_law_1.01,0.051910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,32,power_law_1.01,0.0637503981590271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,32,power_law_1.01,0.08561919927597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,32,power_law_1.01,0.0991424024105072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,32,power_law_1.01,0.12347520589828491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,32,power_law_1.01,0.1596992015838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,32,power_law_1.01,0.22028160095214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,32,power_law_1.01,0.32755839824676514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,128,power_law_1.2,1.1653887748718261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,8,power_law_1.01,0.05539199709892273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,8,power_law_1.01,0.060812801122665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,8,power_law_1.01,0.07549440264701843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,8,power_law_1.01,0.07608320116996765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,8,power_law_1.01,0.08215680122375488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,8,power_law_1.01,0.09282559752464295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,8,power_law_1.01,0.09968000054359435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,8,power_law_1.01,0.10016000270843506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,8,power_law_1.01,0.10259840488433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,8,power_law_1.01,0.10468480587005616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,8,power_law_1.01,0.09961599707603455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,8,power_law_1.01,0.09901440143585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,8,power_law_1.01,0.10101120471954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,8,power_law_1.01,0.10408320426940917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,8,power_law_1.01,0.11487360000610351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,8,power_law_1.01,0.11936639547348023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,8,power_law_1.01,0.12574080228805543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,8,power_law_1.01,0.14705920219421387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,8,power_law_1.01,0.1604159951210022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,8,power_law_1.01,0.18976000547409058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,8,power_law_1.01,0.20229120254516603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,8,power_law_1.01,0.26007680892944335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,8,power_law_1.01,0.3189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,8,power_law_1.01,0.4312255859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,8,power_law_1.01,0.5546239852905274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,8,power_law_1.01,0.7922944068908692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,8,power_law_1.01,1.0247167587280273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,2,power_law_1.01,0.06599680185317994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,2,power_law_1.01,0.08325120210647582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,2,power_law_1.01,0.08261119723320007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,2,power_law_1.01,0.09730560183525086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,2,power_law_1.01,0.09735680222511292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,2,power_law_1.01,0.09815040230751038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,2,power_law_1.01,0.09847040176391601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,2,power_law_1.01,0.09924479722976684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,2,power_law_1.01,0.09934719800949096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,2,power_law_1.01,0.09975680112838745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,2,power_law_1.01,0.10107519626617431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,2,power_law_1.01,0.11892479658126831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,2,power_law_1.01,0.11970560550689698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,2,power_law_1.01,0.12604160308837892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,2,power_law_1.01,0.12999680042266845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,2,power_law_1.01,0.13733760118484498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,2,power_law_1.01,0.145632004737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,2,power_law_1.01,0.18394240140914916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,2,power_law_1.01,0.19774080514907838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,2,power_law_1.01,0.271724796295166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,2,power_law_1.01,0.36033918857574465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,2,power_law_1.01,0.4859839916229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,2,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,2,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,2,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,2,balanced,0.05147199829419454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,2,balanced,0.07393066585063934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,2,balanced,0.08661333719889323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,2,balanced,0.08730133374532063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,2,balanced,0.08624533812204997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,2,balanced,0.08726400136947632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,2,balanced,0.0876586635907491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,2,balanced,0.08691199620564778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,2,balanced,0.08827199538548787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,2,balanced,0.08692800005276997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,2,balanced,0.09102400143941243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,2,power_law_1.01,0.6112576007843018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,2,balanced,0.09578667084376018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,2,balanced,0.09654399752616882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,2,balanced,0.09871466954549153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,2,balanced,0.10828800002733867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,2,balanced,0.11412800351778667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,2,balanced,0.13994133472442627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,2,balanced,0.14672533671061197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,2,balanced,0.20029866695404053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,2,balanced,0.21010132630666098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,2,balanced,0.29339200258255005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,2,balanced,0.3357706864674886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,2,balanced,0.4835786819458008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,2,balanced,0.582640012105306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,8,balanced,0.021333334346612293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,8,balanced,0.022122666239738464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,8,balanced,0.024458666642506916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,8,balanced,0.02515200028816859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,8,balanced,0.0271519993742307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,8,balanced,0.04144000013669332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,8,balanced,0.04414399961630503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,8,balanced,0.04442666471004486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,8,balanced,0.045610666275024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,8,balanced,0.04751466711362203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,8,balanced,0.0492799977461497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,8,balanced,0.05187733471393585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,8,balanced,0.05331199864546458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,8,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,8,balanced,0.06228266656398773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,8,balanced,0.06594666838645935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,8,balanced,0.07099199791749318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,8,balanced,0.08533866206804912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,8,balanced,0.11789333820343018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,8,balanced,0.15914133191108704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,8,balanced,0.14115200440088907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,8,balanced,0.17337600390116373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,2,power_law_1.01,0.7895808219909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,8,balanced,0.20164799690246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,8,balanced,0.25753066937128705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,8,balanced,0.31653332710266113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,8,balanced,0.4760426680246989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,8,balanced,0.5959039926528931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,2,power_law_1.01,1.0225919723510741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,2,power_law_1.01,1.5429632186889648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,2,power_law_1.01,1.9366016387939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,1,balanced,0.04833599925041199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,1,balanced,0.05186666548252106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,1,balanced,0.061887999375661217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,1,balanced,0.09073066711425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,1,balanced,0.12878400087356567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,1,balanced,0.20005865891774496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,1,balanced,0.20400534073511759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,1,balanced,0.20084265867869058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,1,balanced,0.2042186657587687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,1,balanced,0.20484266678492227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,1,balanced,0.20706133047739664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,1,balanced,0.21434134244918823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,1,balanced,0.21846399704615274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,1,balanced,0.2205386757850647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,1,balanced,0.23010667165120444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,1,balanced,0.23775466283162436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,1,balanced,0.24922666947046915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,1,balanced,0.2802666624387105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,1,balanced,0.3069973389307658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,1,balanced,0.36561067899068195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,1,balanced,0.4151466687520345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,1,balanced,0.5579946835835775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,1,balanced,0.6423360109329224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,1,balanced,0.9147786299387614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,1,balanced,1.128165324529012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,1,balanced,1.61844269434611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,1,balanced,2.088592052459717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,4,balanced,0.044154668847719826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,4,balanced,0.04404266675313314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,4,balanced,0.05619200070699056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,4,balanced,0.0724373310804367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,4,balanced,0.11019200086593628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,4,balanced,0.11558399597803752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,4,balanced,0.11105599999427795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,4,balanced,0.10941867033640544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,4,balanced,0.10803199807802837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,4,balanced,0.10499733686447144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,4,balanced,0.10497066378593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,4,balanced,0.10380799571673076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,4,balanced,0.10877866546312968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,4,balanced,0.1095306674639384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,4,balanced,0.11083199580510457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,4,balanced,0.11564800143241882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,4,balanced,0.11183466513951619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,4,balanced,0.11940800150235494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,4,balanced,0.12212266524632771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,4,balanced,0.12823999921480814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,4,balanced,0.13879467050234476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,4,balanced,0.21385065714518228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,4,balanced,0.22343466679255167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,4,balanced,0.32153065999348956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,4,balanced,0.35679999987284344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,4,balanced,0.5257173379262289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,4,balanced,0.6501226822535197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,1,power_law_1.01,0.07046399712562561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,1,power_law_1.01,0.0790336012840271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,1,power_law_1.01,0.10506880283355713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,1,power_law_1.01,0.13808000087738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,1,power_law_1.01,0.13815679550170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,1,power_law_1.01,0.14227839708328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,1,power_law_1.01,0.14257279634475709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,1,power_law_1.01,0.14151040315628052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,1,power_law_1.01,0.14275840520858765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,1,power_law_1.01,0.1433727979660034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,1,power_law_1.01,0.14624639749526977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,1,power_law_1.01,0.15976959466934204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,1,power_law_1.01,0.15544960498809815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,1,power_law_1.01,0.16586240530014038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,1,power_law_1.01,0.1822335958480835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,1,power_law_1.01,0.1876736044883728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,1,power_law_1.01,0.20058879852294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,1,power_law_1.01,0.2425407886505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.01,0.2843071937561035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.01,0.3675839900970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.01,0.4497407913208008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.01,0.5967167854309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.01,0.7828671932220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.01,1.096992015838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.01,1.4580863952636718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.01,2.1158336639404296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.01,2.7526655197143555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,32,balanced,0.04819199939568838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,32,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,32,balanced,0.045791998505592346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,32,balanced,0.05240533252557119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,32,balanced,0.0767626663049062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,32,balanced,0.08459200461705525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,32,balanced,0.08573333422342937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,32,balanced,0.08517866333325703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,32,balanced,0.08699733018875122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,32,balanced,0.08687999844551086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,32,balanced,0.08752533793449402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,32,balanced,0.08935466408729553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,32,balanced,0.0916373332341512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,32,balanced,0.09256533781687419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,32,balanced,0.09833066662152608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,32,balanced,0.099781334400177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,32,balanced,0.10191999872525533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,32,balanced,0.11573333541552226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,32,balanced,0.11694399515787761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,32,balanced,0.14245866735776266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,32,balanced,0.1449066698551178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,32,balanced,0.19047999382019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,32,balanced,0.20758400360743204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,32,balanced,0.3123093247413635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,32,balanced,0.33218665917714435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,32,balanced,0.4790079991022746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,32,balanced,0.5886880159378052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,1,balanced,0.10068266590436299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,1,balanced,0.15252799789110819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,1,balanced,0.25625065962473553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,1,balanced,0.4694826602935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,1,balanced,0.8857653141021729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,1,balanced,1.1010666688283284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,1,balanced,1.1068639755249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,1,balanced,1.1088053385416667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,1,balanced,1.1151093641916912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,1,balanced,1.1205920378367107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,1,balanced,1.1255146662394206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,1,balanced,1.1330880324045818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,1,balanced,1.132698694864909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,1,balanced,1.1461013158162434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,1,balanced,1.1621546745300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,1,balanced,1.17030930519104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,1,balanced,1.19269863764445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,1,balanced,1.3392586708068848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,1,balanced,1.2669333616892497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,1,balanced,1.5222187042236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,1,balanced,1.4236052831013997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,1,balanced,1.8192960421244304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,1,balanced,1.720586617787679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,1,balanced,2.747621218363444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,1,balanced,2.752042770385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,1,balanced,4.54095999399821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,1,balanced,5.030149459838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,16,power_law_1.2,0.10473599433898925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,16,power_law_1.2,0.11155840158462524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,16,power_law_1.2,0.11384320259094238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,16,power_law_1.2,0.10455039739608765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,16,power_law_1.2,0.11038719415664673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,16,power_law_1.2,0.10314879417419434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,16,power_law_1.2,0.10952320098876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,16,power_law_1.2,0.11390719413757325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,16,power_law_1.2,0.1151360034942627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,16,power_law_1.2,0.11861759424209595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,16,power_law_1.2,0.12003840208053589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,16,power_law_1.2,0.12383359670639038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,16,power_law_1.2,0.125491201877594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,16,power_law_1.2,0.12899839878082275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,16,power_law_1.2,0.13498239517211913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,16,power_law_1.2,0.14392319917678834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,16,power_law_1.2,0.1406399965286255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,16,power_law_1.2,0.17152639627456664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,16,power_law_1.2,0.21505279541015626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,16,power_law_1.2,0.2769407987594604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,16,power_law_1.2,0.2726399898529053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,16,power_law_1.2,0.3720319986343384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,16,power_law_1.2,0.3979520082473755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,16,power_law_1.2,0.5410624027252198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,16,power_law_1.2,0.8007743835449219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,16,power_law_1.2,1.0241727828979492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,8,power_law_1.2,0.11537280082702636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,8,power_law_1.2,0.11850240230560302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,8,power_law_1.2,0.11624319553375244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,8,power_law_1.2,0.11943039894104004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,8,power_law_1.2,0.11855360269546508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,8,power_law_1.2,0.11766400337219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,8,power_law_1.2,0.11298559904098511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,16,power_law_1.2,1.3936767578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,8,power_law_1.2,0.11052160263061524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,8,power_law_1.2,0.11073280572891235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,8,power_law_1.2,0.10469759702682495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,8,power_law_1.2,0.09095039963722229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,8,power_law_1.2,0.13470720052719115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,8,power_law_1.2,0.14058239459991456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,8,power_law_1.2,0.11797120571136474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,8,power_law_1.2,0.17160320281982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,8,power_law_1.2,0.14790400266647338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,8,power_law_1.2,0.18337279558181763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,8,power_law_1.2,0.25063679218292234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.2,0.29889280796051027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.2,0.40436477661132814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.2,0.562227201461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.2,0.7586304187774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,64,balanced,0.06052266558011373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,64,balanced,0.05416533350944519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,64,balanced,0.04557866851488749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,64,balanced,0.0476746658484141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,64,balanced,0.04818666477998098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,64,balanced,0.05046399931112925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,64,balanced,0.05009600023428599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,64,balanced,0.05202133456865946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.2,0.9199359893798829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,64,balanced,0.050426666935284935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,64,balanced,0.051781331499417625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,64,balanced,0.051957334081331887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,64,balanced,0.05402133365472158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,64,balanced,0.05315199991067251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,64,balanced,0.05609600245952606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,64,balanced,0.059450666109720864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,64,balanced,0.05852266649405161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,64,balanced,0.06203199923038483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,64,balanced,0.0681279997030894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,64,balanced,0.07083733379840851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,64,balanced,0.0848533312479655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,64,balanced,0.0930560032526652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,64,balanced,0.11334932843844096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,64,balanced,0.1320266624291738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,64,balanced,0.16692266861597696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,64,balanced,0.20295999447504678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,64,balanced,0.2643413345019023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,64,balanced,0.32765867312749225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.2,1.4849023818969727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.2,2.003596878051758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.2,2.7661760330200194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.2,3.5694656372070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,2,power_law_1.01,0.08823680281639099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,2,power_law_1.01,0.09909120202064514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,2,power_law_1.01,0.10122239589691162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,2,power_law_1.01,0.13585920333862306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,2,power_law_1.01,0.140774405002594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,2,power_law_1.01,0.14242559671401978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,2,power_law_1.01,0.14216959476470947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,2,power_law_1.01,0.1430400013923645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,2,power_law_1.01,0.14366719722747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,2,power_law_1.01,0.1435647964477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,2,power_law_1.01,0.1446336030960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,2,power_law_1.01,0.15907200574874877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,2,power_law_1.01,0.1662976026535034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,2,power_law_1.01,0.18065279722213745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,2,power_law_1.01,0.1863168001174927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,2,power_law_1.01,0.19587199687957763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,2,power_law_1.01,0.21342079639434813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,2,power_law_1.01,0.25800960063934325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,2,power_law_1.01,0.3234623908996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,2,power_law_1.01,0.3999232053756714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,2,power_law_1.01,0.4805823802947998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,2,power_law_1.01,0.6399807929992676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,2,power_law_1.01,0.8724479675292969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,2,power_law_1.01,1.1744256019592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,32,power_law_1.2,0.02134400010108948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,32,power_law_1.2,0.021254399418830873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,32,power_law_1.2,0.0215488001704216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,32,power_law_1.2,0.021823999285697938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,32,power_law_1.2,0.020979200303554536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,32,power_law_1.2,0.02250880002975464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,2,power_law_1.01,1.754003143310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,32,power_law_1.2,0.028543999791145323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,32,power_law_1.2,0.027692800760269164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,32,power_law_1.2,0.05330560207366943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,32,power_law_1.2,0.05311999917030334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,32,power_law_1.2,0.053913599252700804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,32,power_law_1.2,0.04550400078296661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,32,power_law_1.2,0.045203199982643126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,32,power_law_1.2,0.045900800824165346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,32,power_law_1.2,0.044121599197387694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,32,power_law_1.2,0.0474368005990982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,32,power_law_1.2,0.04827519953250885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,32,power_law_1.2,0.0525439977645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.2,0.05811840295791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.2,0.07032960057258605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.2,0.08043519854545593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.2,0.09836800098419189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.2,0.12004480361938477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.2,0.1806272029876709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.2,0.22705280780792236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.2,0.293887996673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.2,0.40560002326965333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,2,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,2,balanced,0.03789866715669632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,2,balanced,0.052485331892967224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,2,balanced,0.07630399862925212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,2,balanced,0.12988266348838806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,2,balanced,0.12743999560674033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,2,balanced,0.1279253363609314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,2,balanced,0.12754133343696594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,2,balanced,0.12982933719952902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,2,balanced,0.12981333335240683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,2,balanced,0.1318666636943817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,2,balanced,0.13524799545605978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,2,balanced,0.1346986691157023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,2,balanced,0.13928000132242838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,2,balanced,0.14834133783976236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,2,balanced,0.14879467089970908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,2,balanced,0.1544373333454132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,2,balanced,0.16582399606704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,2,power_law_1.01,2.5665088653564454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,2,balanced,0.18194133043289185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,2,balanced,0.20635199546813965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,2,balanced,0.23134400447209677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,2,balanced,0.3015039960543315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,2,balanced,0.3407893180847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,2,balanced,0.4764853318532308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,2,balanced,0.5840746561686198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,2,balanced,0.8168319861094157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,2,balanced,1.0622506936391194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,2,power_law_1.01,2.9891008377075194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,32,2,power_law_1.2,0.04986239969730377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,32,2,power_law_1.2,0.05369600057601929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,32,2,power_law_1.2,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,32,2,power_law_1.2,0.06669440269470214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,32,2,power_law_1.2,0.07676159739494323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,32,2,power_law_1.2,0.09425920248031616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,32,2,power_law_1.2,0.1133247971534729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,32,2,power_law_1.2,0.11139839887619019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,32,2,power_law_1.2,0.11537920236587525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,32,2,power_law_1.2,0.11707520484924316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,32,2,power_law_1.2,0.11894400119781494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,32,2,power_law_1.2,0.12188800573348998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,32,2,power_law_1.2,0.12670079469680787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,32,2,power_law_1.2,0.12929279804229737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,32,2,power_law_1.2,0.139628803730011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,32,2,power_law_1.2,0.14634239673614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,32,2,power_law_1.2,0.15441919565200807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,32,2,power_law_1.2,0.18032000064849854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,32,2,power_law_1.2,0.19987200498580932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,32,2,power_law_1.2,0.24824318885803223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,32,2,power_law_1.2,0.29037439823150635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,32,2,power_law_1.2,0.37621119022369387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,32,2,power_law_1.2,0.48444161415100095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,32,2,power_law_1.2,0.6656064033508301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,32,2,power_law_1.2,0.8230079650878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,32,2,power_law_1.2,1.1482943534851073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,32,2,power_law_1.2,1.4839743614196776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,1,power_law_1.01,0.05723519921302796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,1,power_law_1.01,0.061964797973632815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,1,power_law_1.01,0.07155839800834655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,1,power_law_1.01,0.09484800100326538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,1,power_law_1.01,0.12090239524841309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,1,power_law_1.01,0.14315520524978637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,1,power_law_1.01,0.17666560411453247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,1,power_law_1.01,0.18699519634246825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,1,power_law_1.01,0.19263360500335694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,1,power_law_1.01,0.19571199417114257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,1,power_law_1.01,0.2019711971282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,1,power_law_1.01,0.21304960250854493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,1,power_law_1.01,0.21935360431671141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,1,power_law_1.01,0.23887360095977783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,1,power_law_1.01,0.24988160133361817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,1,power_law_1.01,0.26365439891815184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,1,power_law_1.01,0.26394240856170653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,1,power_law_1.01,0.31283841133117674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,1,power_law_1.01,0.38853759765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,1,power_law_1.01,0.4641280174255371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,1,power_law_1.01,0.5185664176940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,1,power_law_1.01,0.6802175998687744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,1,power_law_1.01,0.842137622833252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,1,power_law_1.01,1.1470272064208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,1,power_law_1.01,1.4869376182556153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,1,power_law_1.01,2.1146303176879884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,1,power_law_1.01,2.7628671646118166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,256,power_law_1.2,0.05184000134468079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,256,power_law_1.2,0.0637440025806427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,256,power_law_1.2,0.04053759872913361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,256,power_law_1.2,0.04033919870853424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,256,power_law_1.2,0.04185599982738495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,256,power_law_1.2,0.04227199852466583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,256,power_law_1.2,0.04325119853019714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,256,power_law_1.2,0.044435200095176694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,256,power_law_1.2,0.04410240054130554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,256,power_law_1.2,0.04506239891052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,256,power_law_1.2,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,256,power_law_1.2,0.04636160135269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,256,power_law_1.2,0.04723840057849884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,256,power_law_1.2,0.04925439953804016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,256,power_law_1.2,0.05399680137634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,256,power_law_1.2,0.05541120171546936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,256,power_law_1.2,0.05922560095787048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,256,power_law_1.2,0.06789759993553161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,256,power_law_1.2,0.07277439832687378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,256,power_law_1.2,0.08918399810791015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,256,power_law_1.2,0.10131839513778687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,256,power_law_1.2,0.13779200315475465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,256,power_law_1.2,0.16531840562820435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,256,power_law_1.2,0.2433919906616211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,256,power_law_1.2,0.2913408041000366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,256,power_law_1.2,0.4411327838897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,256,power_law_1.2,0.5539328098297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,32,power_law_1.2,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,32,power_law_1.2,0.06580479741096497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,32,power_law_1.2,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,32,power_law_1.2,0.0699455976486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,32,power_law_1.2,0.06878719925880432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,32,power_law_1.2,0.06898559927940369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,32,power_law_1.2,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,128,balanced,0.055455997586250305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,128,balanced,0.055813332398732506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,128,balanced,0.055973331133524575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,32,power_law_1.2,0.07087360024452209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,128,balanced,0.07388266424338023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,128,balanced,0.08110933502515157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,128,balanced,0.08189866443475087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,32,power_law_1.2,0.07457919716835022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,32,power_law_1.2,0.07294719815254211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,32,power_law_1.2,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,128,balanced,0.08479467034339905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,128,balanced,0.08775466680526733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,128,balanced,0.08505066235860188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,128,balanced,0.08714666962623596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,128,balanced,0.08915733297665913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,128,balanced,0.09102933605511983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,128,balanced,0.09322667121887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,128,balanced,0.09342400232950847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,128,balanced,0.10175466537475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,32,power_law_1.2,0.07796480059623719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,128,balanced,0.10347200433413188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,128,balanced,0.10976533095041911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,32,power_law_1.2,0.07849599719047547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,128,balanced,0.12346133589744568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,128,balanced,0.13449600338935852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,128,balanced,0.15962666273117065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,128,balanced,0.18371200561523438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,128,balanced,0.2585600018501282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,128,balanced,0.26472532749176025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,128,balanced,0.33480532964070636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,128,balanced,0.4000106652577718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,32,power_law_1.2,0.07964159846305847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,128,balanced,0.5589973529179891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,128,balanced,0.6999146938323975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,16,power_law_1.01,0.01706880033016205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,16,power_law_1.01,0.01621759980916977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,32,power_law_1.2,0.09770879745483399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,32,power_law_1.2,0.09788159728050232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,32,power_law_1.2,0.09793279767036438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,32,power_law_1.2,0.11486079692840576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,32,power_law_1.2,0.12392959594726563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,32,power_law_1.2,0.14543999433517457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,32,power_law_1.2,0.17403520345687867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,32,power_law_1.2,0.21771519184112548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,32,power_law_1.2,0.2704576015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,16,power_law_1.01,0.016473600268363954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,16,power_law_1.01,0.016467200219631196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,32,power_law_1.2,0.40021119117736814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,16,power_law_1.01,0.017209599912166595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,16,power_law_1.01,0.017446400225162507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,32,power_law_1.2,0.5242496013641358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,16,power_law_1.01,0.01849599927663803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,16,power_law_1.01,0.018156799674034118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,16,power_law_1.01,0.01932159960269928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,16,power_law_1.01,0.019462400674819948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,16,power_law_1.01,0.02142080068588257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,16,power_law_1.01,0.022086399793624877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,16,power_law_1.01,0.022694399952888487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,16,power_law_1.01,0.029651200771331786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,16,power_law_1.01,0.029414400458335876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,16,power_law_1.01,0.02919679880142212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,16,power_law_1.01,0.03116160035133362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,16,power_law_1.01,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,16,power_law_1.01,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,16,power_law_1.01,0.04178560078144074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,16,power_law_1.01,0.04382719993591309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,16,power_law_1.01,0.04935039877891541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,16,power_law_1.01,0.06027519702911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,16,power_law_1.01,0.07919999957084656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,16,power_law_1.01,0.09711359739303589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,16,power_law_1.01,0.12732800245285034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,16,power_law_1.01,0.16779520511627197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,32,power_law_1.2,0.8320063591003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,32,power_law_1.2,1.0565055847167968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,64,power_law_1.01,0.04741120040416717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,64,power_law_1.01,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,64,power_law_1.01,0.044838398694992065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,64,power_law_1.01,0.04769920110702515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,64,power_law_1.01,0.04808320105075836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,2,power_law_1.01,0.018464000523090364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,64,power_law_1.01,0.051910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,64,power_law_1.01,0.05121279954910278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,64,power_law_1.01,0.05260800123214722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,64,power_law_1.01,0.05308160185813904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,64,power_law_1.01,0.05307520031929016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,64,power_law_1.01,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,2,power_law_1.01,0.01828480064868927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,2,power_law_1.01,0.018086400628089905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,64,power_law_1.01,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,2,power_law_1.01,0.018873600661754607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,2,power_law_1.01,0.02019840031862259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,64,power_law_1.01,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,2,power_law_1.01,0.021478399634361267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,2,power_law_1.01,0.021631999313831328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,2,power_law_1.01,0.02271360009908676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,2,power_law_1.01,0.02388480007648468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,2,power_law_1.01,0.02364159971475601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,2,power_law_1.01,0.024223999679088594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,2,power_law_1.01,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,2,power_law_1.01,0.04222080111503601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,64,power_law_1.01,0.06116480231285095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,2,power_law_1.01,0.042284798622131345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,64,power_law_1.01,0.06972159743309021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,2,power_law_1.01,0.03654400110244751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,64,power_law_1.01,0.07311999797821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,2,power_law_1.01,0.03353599905967712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,2,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,2,power_law_1.01,0.046342399716377256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,64,power_law_1.01,0.0725823998451233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,64,power_law_1.01,0.08777599930763244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,64,power_law_1.01,0.0997376024723053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,64,power_law_1.01,0.11846400499343872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.01,0.048256000876426695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,64,power_law_1.01,0.13279999494552613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,64,power_law_1.01,0.17123839855194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.01,0.06282879710197449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.01,0.07544959783554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,64,power_law_1.01,0.20590078830718994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,64,power_law_1.01,0.2722111940383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,64,power_law_1.01,0.37056000232696534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,64,power_law_1.01,0.548473596572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,64,power_law_1.01,0.6837887763977051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.01,0.09641600251197815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.01,0.12034560441970825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.01,0.16401920318603516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,4,power_law_1.01,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,4,power_law_1.01,0.03439359962940216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,4,power_law_1.01,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,4,power_law_1.01,0.04267520010471344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,4,power_law_1.01,0.043558400869369504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,4,power_law_1.01,0.05188480019569397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,4,power_law_1.01,0.05445759892463684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,4,power_law_1.01,0.05416319966316223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,4,power_law_1.01,0.05486720204353333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,4,power_law_1.01,0.055289602279663085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,4,power_law_1.01,0.05486720204353333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,4,power_law_1.01,0.05818880200386047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,4,power_law_1.01,0.05774719715118408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,4,power_law_1.01,0.060249602794647215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,4,power_law_1.01,0.06641280055046081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,4,power_law_1.01,0.066975998878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,4,power_law_1.01,0.06924160122871399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.01,0.2073280096054077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,4,power_law_1.01,0.0773248016834259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,4,power_law_1.01,0.0782144010066986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,4,power_law_1.01,0.08995839953422546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,4,power_law_1.01,0.0997759997844696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,4,power_law_1.01,0.1229248046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,4,power_law_1.01,0.1459391951560974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,4,power_law_1.01,0.17987200021743774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,4,power_law_1.01,0.22056961059570312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,4,power_law_1.01,0.31642239093780516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,4,power_law_1.01,0.38705921173095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.01,0.29523839950561526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.01,0.381113600730896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,64,power_law_1.01,0.08620160222053527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,64,power_law_1.01,0.08221439719200134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,64,power_law_1.01,0.08208000063896179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,64,power_law_1.01,0.08488320112228394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,64,power_law_1.01,0.0836031973361969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,64,power_law_1.01,0.08477439880371093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,64,power_law_1.01,0.08472959995269776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,64,power_law_1.01,0.08507519960403442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,64,power_law_1.01,0.08609920144081115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,64,power_law_1.01,0.08266239762306213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,64,power_law_1.01,0.08507519960403442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,64,power_law_1.01,0.08811519742012024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,64,power_law_1.01,0.08869119882583618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,64,power_law_1.01,0.09007359743118286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,64,power_law_1.01,0.09600639939308167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,64,power_law_1.01,0.09484800100326538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,64,power_law_1.01,0.10060160160064698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,64,power_law_1.01,0.10845439434051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,64,power_law_1.01,0.11713279485702514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,64,power_law_1.01,0.13182079792022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,64,power_law_1.01,0.1500864028930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,64,power_law_1.01,0.18385920524597169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,64,power_law_1.01,0.22392959594726564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,64,power_law_1.01,0.3103935956954956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,64,power_law_1.01,0.3739903926849365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,64,power_law_1.01,0.5152768135070801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,64,power_law_1.01,0.6684927940368652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,2,power_law_1.01,0.027475199103355406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,2,power_law_1.01,0.04981119930744171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,2,power_law_1.01,0.06585599780082703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,2,power_law_1.01,0.1014143943786621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,2,power_law_1.01,0.12766720056533815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,2,power_law_1.01,0.15959680080413818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,2,power_law_1.01,0.1892351984977722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,2,power_law_1.01,0.20000638961791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,2,power_law_1.01,0.2119744062423706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,2,power_law_1.01,0.20762240886688232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,2,power_law_1.01,0.22542719841003417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,2,power_law_1.01,0.22306559085845948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,2,power_law_1.01,0.22303359508514403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,2,power_law_1.01,0.23865599632263185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,2,power_law_1.01,0.2374336004257202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,2,power_law_1.01,0.2492608070373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,2,power_law_1.01,0.2492608070373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,2,power_law_1.01,0.2753472089767456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.01,0.28446719646453855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.01,0.627129602432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.01,0.7338111877441407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.01,0.5017727851867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.01,0.5292416095733643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.01,0.672979211807251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.01,0.7811264038085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.01,1.0315391540527343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.01,1.3052864074707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,16,power_law_1.2,0.05927680134773254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,16,power_law_1.2,0.06551039814949036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,16,power_law_1.2,0.06110079884529114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,16,power_law_1.2,0.06840959787368775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,16,power_law_1.2,0.06837120056152343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,16,power_law_1.2,0.0582144021987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,16,power_law_1.2,0.06936320066452026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,16,power_law_1.2,0.06688640117645264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,16,power_law_1.2,0.07134079933166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,16,power_law_1.2,0.06970880031585694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,16,power_law_1.2,0.06999679803848266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,16,power_law_1.2,0.06897280216217042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,16,power_law_1.2,0.07304319739341736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,16,power_law_1.2,0.07845119833946228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,16,power_law_1.2,0.08478080034255982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,16,power_law_1.2,0.08989440202713013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,16,power_law_1.2,0.0920960009098053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,16,power_law_1.2,0.10583679676055908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,16,power_law_1.2,0.10839680433273316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,16,power_law_1.2,0.1257472038269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,16,power_law_1.2,0.1494271993637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,16,power_law_1.2,0.187718403339386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,16,power_law_1.2,0.22382080554962158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,16,power_law_1.2,0.3029887914657593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,16,power_law_1.2,0.39229440689086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,16,power_law_1.2,0.5354688167572021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,16,power_law_1.2,0.6823103904724122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,2,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,2,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,2,balanced,0.04167999823888143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,2,balanced,0.04163199911514918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,2,balanced,0.043509334325790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,2,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,2,balanced,0.0431573341290156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,2,balanced,0.04558933277924856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,2,balanced,0.04570133487383524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,2,balanced,0.04548266530036926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,2,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,2,balanced,0.0473333348830541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,2,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,2,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,2,balanced,0.060821334520975746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,2,balanced,0.06569600105285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,2,balanced,0.0689386675755183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,2,balanced,0.0806826651096344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,2,balanced,0.08686400453249614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,2,balanced,0.10547733306884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,2,balanced,0.12986133495966592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,2,balanced,0.1644373337427775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,2,balanced,0.1848426659901937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,2,balanced,0.2553706765174866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,2,balanced,0.32443199555079144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,2,balanced,0.45820800463358563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,2,balanced,0.5919466813405355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,16,balanced,0.043968002001444496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,16,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,16,balanced,0.044010668992996216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,16,balanced,0.04983466863632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,16,balanced,0.05603733162085215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,16,balanced,0.07901866734027863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,16,balanced,0.09582933783531189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,16,balanced,0.09409067034721375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,16,balanced,0.09479999542236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,16,balanced,0.09574400385220845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,16,balanced,0.09716266393661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,16,balanced,0.10050666332244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,16,balanced,0.10259200135866801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,16,balanced,0.10322133700052898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,16,balanced,0.11137066284815471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,16,balanced,0.11135466893513997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,16,balanced,0.12006400028864543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,16,balanced,0.13274666666984558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,16,balanced,0.14258666833241782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,16,balanced,0.16643733779589334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,16,balanced,0.1886826753616333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,16,balanced,0.2340373396873474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,16,balanced,0.2839306592941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,16,balanced,0.3718080123265584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,16,balanced,0.4715413252512614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,16,balanced,0.6529013315836588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,16,balanced,0.8436533610026041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,1,power_law_1.2,0.07672960162162781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,1,power_law_1.2,0.10339839458465576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,1,power_law_1.2,0.1500864028930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,1,power_law_1.2,0.22449278831481934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,1,power_law_1.2,0.31154561042785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,1,power_law_1.2,0.4002495765686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,1,power_law_1.2,0.6238592147827149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,1,power_law_1.2,0.6391168117523194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,1,power_law_1.2,0.6783040046691895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,1,power_law_1.2,0.6757184028625488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,1,power_law_1.2,0.7069952011108398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,1,power_law_1.2,0.7378367900848388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,1,power_law_1.2,0.7484992027282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,1,power_law_1.2,0.7893248081207276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,1,power_law_1.2,0.8195648193359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,1,power_law_1.2,0.8380672454833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,1,power_law_1.2,0.8826815605163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,1,power_law_1.2,0.9862208366394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,1,power_law_1.2,1.0049280166625976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,1,power_law_1.2,1.1446911811828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,1,power_law_1.2,1.2105792045593262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,1,power_law_1.2,1.437388801574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,1,power_law_1.2,1.662611198425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,1,power_law_1.2,2.0851072311401366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,1,power_law_1.2,2.4875839233398436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,1,power_law_1.2,3.3563968658447267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,1,power_law_1.2,4.194022369384766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,1,power_law_1.01,0.07109760046005249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,1,power_law_1.01,0.09246079921722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,1,power_law_1.01,0.11548160314559937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,1,power_law_1.01,0.15823999643325806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,1,power_law_1.01,0.16179840564727782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,1,power_law_1.01,0.16224639415740966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,1,power_law_1.01,0.16353280544281007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,1,power_law_1.01,0.1646399974822998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,1,power_law_1.01,0.16433279514312743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,1,power_law_1.01,0.1642303943634033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,1,power_law_1.01,0.16417280435562134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,1,power_law_1.01,0.16876159906387328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,1,power_law_1.01,0.1697343945503235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,1,power_law_1.01,0.1765504002571106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,1,power_law_1.01,0.19809279441833497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,1,power_law_1.01,0.20131840705871581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,1,power_law_1.01,0.21198079586029053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,1,power_law_1.01,0.24478719234466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.01,0.29167358875274657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.01,0.40023040771484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.01,0.4882815837860107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.01,0.6391424179077149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,2,balanced,0.023306667804718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,2,balanced,0.02513599892457326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,2,balanced,0.03254933406909307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,2,balanced,0.04892266790072123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,2,balanced,0.07320000231266022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.01,0.8206912040710449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,2,balanced,0.07452799876530965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,2,balanced,0.07725333174069722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,2,balanced,0.08059733112653096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,2,balanced,0.07989866534868877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,2,balanced,0.08206399778525035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,2,balanced,0.08422399560610454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,2,balanced,0.09490133325258891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,2,balanced,0.09432533383369446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,2,balanced,0.09264000256856282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,2,balanced,0.12644267082214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,2,balanced,0.13111467162768045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,2,balanced,0.12338133653004964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,2,balanced,0.18054932355880737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,2,balanced,0.2037066618601481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,2,balanced,0.2046026587486267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,2,balanced,0.21268266439437866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,2,balanced,0.21345599492390951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,2,balanced,0.22043200333913168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,2,balanced,0.3784373203913371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,2,balanced,0.3832746744155884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,2,balanced,0.5571360190709432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,2,balanced,0.7309119701385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.01,1.1851840019226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.01,1.5512960433959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,32,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,32,balanced,0.046336000164349876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,32,balanced,0.04709866642951965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,32,balanced,0.04770666857560476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,32,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,32,balanced,0.0516480008761088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,32,balanced,0.05249600112438202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,32,balanced,0.05177066723505656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,32,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,32,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,32,balanced,0.05373333394527435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,32,balanced,0.05416533350944519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,32,balanced,0.05602666735649109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,32,balanced,0.05750399827957153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,32,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,32,balanced,0.062074666221936546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,32,balanced,0.06432533264160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,32,balanced,0.06842133402824402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,32,balanced,0.07218133409818013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,32,balanced,0.08207466701666515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,32,balanced,0.09056533376375835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,32,balanced,0.10939199725786845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,32,balanced,0.1243839959303538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,32,balanced,0.17520000537236533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,32,balanced,0.2066133419672648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,32,balanced,0.28092799584070843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,32,balanced,0.35950398445129395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,32,power_law_1.01,0.02928000092506409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,32,power_law_1.01,0.027423998713493346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,32,power_law_1.01,0.027270400524139406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,32,power_law_1.01,0.029388800263404846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,32,power_law_1.01,0.03000960052013397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,32,power_law_1.01,0.03128319978713989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,32,power_law_1.01,0.03201279938220978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,32,power_law_1.01,0.03188480138778686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,32,power_law_1.01,0.03315840065479279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.01,2.243328094482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,32,power_law_1.01,0.032671999931335446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,32,power_law_1.01,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,32,power_law_1.01,0.034272000193595886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,32,power_law_1.01,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,32,power_law_1.01,0.034995201230049136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,32,power_law_1.01,0.03900800049304962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,32,power_law_1.01,0.03991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,32,power_law_1.01,0.043033599853515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,32,power_law_1.01,0.04594559967517853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,32,power_law_1.01,0.05169919729232788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,32,power_law_1.01,0.06149759888648987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,32,power_law_1.01,0.06996480226516724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,32,power_law_1.01,0.09418240189552307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,32,power_law_1.01,0.11723519563674926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,32,power_law_1.01,0.1588096022605896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,32,power_law_1.01,0.20968320369720458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,32,power_law_1.01,0.2783360004425049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,8,power_law_1.2,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,8,power_law_1.2,0.04880000054836273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,8,power_law_1.2,0.04927999973297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,8,power_law_1.2,0.05306879878044128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,8,power_law_1.2,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,8,power_law_1.2,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,8,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,32,power_law_1.01,0.38442881107330323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,8,power_law_1.2,0.05656319856643677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,8,power_law_1.2,0.05868800282478333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,8,power_law_1.2,0.05988479852676391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,8,power_law_1.2,0.06382719874382019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,8,power_law_1.2,0.06566399931907654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,8,power_law_1.2,0.06954240202903747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,8,power_law_1.2,0.07081599831581116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,8,power_law_1.2,0.07570559978485107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,8,power_law_1.2,0.08191999793052673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,8,power_law_1.2,0.08843520283699036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,8,power_law_1.2,0.10331519842147827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,8,power_law_1.2,0.11719679832458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,8,power_law_1.2,0.151910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,8,power_law_1.2,0.18001919984817505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,8,power_law_1.2,0.2503743886947632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,8,power_law_1.2,0.2960639953613281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,8,power_law_1.2,0.45287041664123534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,8,power_law_1.2,0.551251220703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,8,power_law_1.2,0.7560832023620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.01,2.947475242614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,8,power_law_1.2,1.0927743911743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,32,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,32,balanced,0.05705066521962484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,32,balanced,0.06669333577156067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,32,balanced,0.09101866682370503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,32,balanced,0.10830400387446086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,32,balanced,0.11002666751543681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,32,balanced,0.11008000373840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,32,balanced,0.11084266503651936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,32,balanced,0.11056533455848694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,32,balanced,0.11109333237012227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,32,balanced,0.11145066221555074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,32,balanced,0.11494933565457661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,32,balanced,0.11512000362078349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,32,balanced,0.11636267105738322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,32,balanced,0.12185600399971008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,32,balanced,0.12342933813730876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,32,balanced,0.12681066989898682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,32,balanced,0.14873600006103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,32,balanced,0.14972800016403198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,32,balanced,0.1906879941622416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,32,balanced,0.1945706605911255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,32,balanced,0.28179200490315753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,32,balanced,0.24830933411916098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,32,balanced,0.3323199947675069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,32,balanced,0.3792639970779419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,32,balanced,0.5209546486536661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,32,balanced,0.6196373303731283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,16,power_law_1.2,0.12414079904556274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,16,power_law_1.2,0.12382719516754151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,16,power_law_1.2,0.1265727996826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,16,power_law_1.2,0.11676160097122193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,16,power_law_1.2,0.11744639873504639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,16,power_law_1.2,0.1310528039932251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,16,power_law_1.2,0.131494402885437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,16,power_law_1.2,0.13254400491714477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,16,power_law_1.2,0.13715840578079225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,16,power_law_1.2,0.1367616057395935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,16,power_law_1.2,0.137881600856781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,16,power_law_1.2,0.13806719779968263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,16,power_law_1.2,0.13667839765548706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,16,power_law_1.2,0.14012160301208496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,16,power_law_1.2,0.15287040472030639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,16,power_law_1.2,0.15728000402450562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,16,power_law_1.2,0.16537599563598632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,16,power_law_1.2,0.1859007954597473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,16,power_law_1.2,0.20662400722503663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,16,power_law_1.2,0.24048640727996826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,16,power_law_1.2,0.3101567983627319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,16,power_law_1.2,0.4291520118713379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,16,power_law_1.2,0.49404158592224123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,16,power_law_1.2,0.7487743854522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,8,balanced,0.05022933085759481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,8,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,8,balanced,0.05633600056171417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,8,balanced,0.06384533147017162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,8,balanced,0.07222933570543925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,8,balanced,0.09522133072217305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,8,balanced,0.09617066383361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,8,balanced,0.09488532940546672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,8,balanced,0.09679466485977173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,8,balanced,0.09788266817728679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,8,balanced,0.0937013328075409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,8,balanced,0.10001599788665771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,8,balanced,0.09841066598892212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,8,balanced,0.0957973301410675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,8,balanced,0.10368000467618306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,8,balanced,0.10295466581980388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,8,balanced,0.11052800218264262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,8,balanced,0.11784533659617107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,8,balanced,0.12549333771069845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,16,power_law_1.2,0.962451171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,8,balanced,0.14590400457382202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,8,balanced,0.1632426679134369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,8,balanced,0.1981226603190104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,8,balanced,0.231930673122406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,8,balanced,0.32473599910736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,8,balanced,0.39603734016418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,8,balanced,0.5564479827880859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,8,balanced,0.7197706699371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,16,power_law_1.2,1.4643327713012695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,16,power_law_1.2,2.0373567581176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,16,balanced,0.0341386670867602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,16,balanced,0.034074666599432625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,16,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,16,balanced,0.037471999724706016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,16,balanced,0.04015466570854187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,16,balanced,0.04154133299986521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,16,balanced,0.04153066625197729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,16,balanced,0.04186133543650309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,16,balanced,0.04182399809360504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,16,balanced,0.04144000013669332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,16,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,16,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,16,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,16,balanced,0.044853334625562034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,16,balanced,0.04738133152325948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,16,balanced,0.04854399959246317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,16,balanced,0.05000533163547516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,16,balanced,0.05247466762860616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,16,balanced,0.05533333122730255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,16,balanced,0.062261333068211876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,16,balanced,0.06835199892520905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,16,balanced,0.09308266639709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,16,balanced,0.10585600137710571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,16,balanced,0.13317867120107016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,16,balanced,0.1627893348534902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,16,balanced,0.21289066473642984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,16,balanced,0.2613813281059265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,64,balanced,0.03757333258787791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,64,balanced,0.03646933287382126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,64,balanced,0.0373279998699824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,64,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,64,balanced,0.0429013321797053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,64,balanced,0.04309333364168803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,64,balanced,0.04196799794832865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,64,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,64,balanced,0.04248533149560293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,64,balanced,0.04284800092379252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,64,balanced,0.04381333291530609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,64,balanced,0.04186666508515676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,64,balanced,0.045850664377212524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,64,balanced,0.04690133531888326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,64,balanced,0.04605866471926371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,64,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,64,balanced,0.0499893327554067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,64,balanced,0.05230399966239929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,64,balanced,0.05203199883302053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,64,balanced,0.05825600028038025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,64,balanced,0.06033066908518473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,64,balanced,0.07367999851703644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,64,balanced,0.0821919987599055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,64,balanced,0.10739200313886006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,64,balanced,0.12708266576131186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,64,balanced,0.16449066996574402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,64,balanced,0.20566399892171225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,4,balanced,0.04711466530958811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,4,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,4,balanced,0.050106664498647056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,4,balanced,0.054085334142049156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,4,balanced,0.05995733539263407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,4,balanced,0.07941866914431255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,4,balanced,0.07875200112660725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,4,balanced,0.0827893316745758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,4,balanced,0.08453333377838135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,4,balanced,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,4,balanced,0.08110933502515157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,4,balanced,0.08866666754086812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,4,balanced,0.0895306666692098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,4,balanced,0.09602133433024089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,4,balanced,0.10275733470916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,4,balanced,0.10731200377146403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,4,balanced,0.11656000216801961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,4,balanced,0.13328533371289572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,4,balanced,0.14757333199183145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,4,balanced,0.176256000995636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,4,balanced,0.20046399037043253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,4,balanced,0.2813546657562256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,4,balanced,0.33241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,4,balanced,0.484282652537028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,4,balanced,0.6103519996007284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,4,balanced,0.8869919776916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,4,balanced,1.1643893718719482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,2,power_law_1.2,0.05804799795150757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,2,power_law_1.2,0.0728767991065979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,2,power_law_1.2,0.09132800102233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,2,power_law_1.2,0.12243200540542602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,2,power_law_1.2,0.1541632056236267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,2,power_law_1.2,0.1773695945739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,2,power_law_1.2,0.21453440189361572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,2,power_law_1.2,0.25998079776763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,2,power_law_1.2,0.27841920852661134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,2,power_law_1.2,0.27847681045532224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,2,power_law_1.2,0.2962368011474609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,2,power_law_1.2,0.29637119770050047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,2,power_law_1.2,0.3077440023422241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,2,power_law_1.2,0.3146559953689575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,2,power_law_1.2,0.3326656103134155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,2,power_law_1.2,0.34993278980255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,2,power_law_1.2,0.36687359809875486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,2,power_law_1.2,0.40575361251831055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,2,power_law_1.2,0.44851198196411135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,2,power_law_1.2,0.512499189376831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,2,power_law_1.2,0.5733439922332764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,2,power_law_1.2,0.7197696208953858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,2,power_law_1.2,0.8558208465576171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,2,power_law_1.2,1.1441856384277345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,2,power_law_1.2,1.325772762298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,2,power_law_1.2,1.7107967376708983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,2,power_law_1.2,2.247385597229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,16,power_law_1.2,0.025542399287223815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,16,power_law_1.2,0.0233024001121521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,16,power_law_1.2,0.0239424005150795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,16,power_law_1.2,0.02512640058994293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,16,power_law_1.2,0.02600319981575012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,16,power_law_1.2,0.02778240144252777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,16,power_law_1.2,0.02858240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,16,power_law_1.2,0.03273600041866302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,16,power_law_1.2,0.03442560136318207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,16,power_law_1.2,0.03498879969120026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,16,power_law_1.2,0.036294400691986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,16,power_law_1.2,0.03722879886627197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,16,power_law_1.2,0.037350401282310486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,16,power_law_1.2,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,16,power_law_1.2,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,16,power_law_1.2,0.041094401478767396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,16,power_law_1.2,0.04542720019817352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,16,power_law_1.2,0.05085440278053284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,16,power_law_1.2,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,16,power_law_1.2,0.05873280167579651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,16,power_law_1.2,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,16,power_law_1.2,0.09541119933128357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,128,power_law_1.01,0.021196800470352172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,128,power_law_1.01,0.021055999398231506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,16,power_law_1.2,0.09760000109672547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,128,power_law_1.01,0.021964800357818604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,128,power_law_1.01,0.024908800423145295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,128,power_law_1.01,0.03182719945907593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,128,power_law_1.01,0.05100160241127014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,128,power_law_1.01,0.048467200994491574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,128,power_law_1.01,0.048537600040435794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,128,power_law_1.01,0.04925439953804016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,128,power_law_1.01,0.04979200065135956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,128,power_law_1.01,0.049491199851036075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,128,power_law_1.01,0.04958719909191132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,128,power_law_1.01,0.04849919974803925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,128,power_law_1.01,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,128,power_law_1.01,0.042387199401855466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,128,power_law_1.01,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,128,power_law_1.01,0.048076799511909483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,128,power_law_1.01,0.05144320130348205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,128,power_law_1.01,0.06460800170898437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,128,power_law_1.01,0.07677440047264099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,128,power_law_1.01,0.09060479998588562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,128,power_law_1.01,0.12574080228805543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,128,power_law_1.01,0.15343999862670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,16,power_law_1.2,0.14177279472351073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,128,power_law_1.01,0.21575040817260743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,128,power_law_1.01,0.2518656015396118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,128,power_law_1.01,0.38815999031066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,128,power_law_1.01,0.497651195526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,16,power_law_1.2,0.17763199806213378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,2,power_law_1.01,0.09985920190811157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,2,power_law_1.01,0.14883840084075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,2,power_law_1.01,0.24737920761108398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,2,power_law_1.01,0.3290112018585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,2,power_law_1.01,0.47674241065979006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,2,power_law_1.01,0.6425727844238281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,2,power_law_1.01,0.6735936164855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,2,power_law_1.01,0.8742079734802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,2,power_law_1.01,0.9102399826049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,2,power_law_1.01,0.9538047790527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,2,power_law_1.01,1.000767993927002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,2,power_law_1.01,1.0204031944274903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,2,power_law_1.01,1.0087871551513672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,2,power_law_1.01,1.0295295715332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,2,power_law_1.01,1.045785617828369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,2,power_law_1.01,1.1057600021362304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,2,power_law_1.01,1.1576767921447755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,2,power_law_1.01,1.2432191848754883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,2,power_law_1.01,1.3182080268859864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,2,power_law_1.01,1.4666175842285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,2,power_law_1.01,1.5411456108093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,2,power_law_1.01,1.8604032516479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,2,power_law_1.01,1.7662912368774415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,2,power_law_1.01,2.229292869567871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,2,power_law_1.01,2.6567039489746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,2,power_law_1.01,3.3385665893554686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,2,power_law_1.01,4.473977661132812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,4,power_law_1.01,0.0702015995979309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,4,power_law_1.01,0.0869055986404419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,4,power_law_1.01,0.09157119989395142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,4,power_law_1.01,0.10484479665756226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,4,power_law_1.01,0.11391359567642212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,4,power_law_1.01,0.1221824049949646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,4,power_law_1.01,0.13228800296783447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,4,power_law_1.01,0.1363711953163147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,4,power_law_1.01,0.1397503972053528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,4,power_law_1.01,0.1350592017173767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,4,power_law_1.01,0.13875839710235596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,4,power_law_1.01,0.14088319540023803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,4,power_law_1.01,0.14665600061416625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,4,power_law_1.01,0.15280640125274658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,4,power_law_1.01,0.15581439733505248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,4,power_law_1.01,0.1570623993873596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,4,power_law_1.01,0.16387840509414672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,4,power_law_1.01,0.17870719432830812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,4,power_law_1.01,0.19545600414276124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,4,power_law_1.01,0.2430783987045288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,4,power_law_1.01,0.28845438957214353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,4,power_law_1.01,0.37648639678955076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,4,power_law_1.01,0.44618239402771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,4,power_law_1.01,0.6877247810363769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,4,power_law_1.01,0.8363391876220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,4,power_law_1.01,1.11844482421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,4,power_law_1.01,1.5118144035339356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,16,power_law_1.2,0.20758399963378907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,16,power_law_1.2,0.34941439628601073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,2,power_law_1.2,0.08624640107154846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,2,power_law_1.2,0.11486719846725464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,2,power_law_1.2,0.13787519931793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,2,power_law_1.2,0.1860159993171692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,2,power_law_1.2,0.23930881023406983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,2,power_law_1.2,0.24787840843200684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,2,power_law_1.2,0.3914175987243652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,2,power_law_1.2,0.40253438949584963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,2,power_law_1.2,0.39562239646911623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,2,power_law_1.2,0.41473917961120604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,2,power_law_1.2,0.4315199851989746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,2,power_law_1.2,0.455840015411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,2,power_law_1.2,0.44065279960632325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,2,power_law_1.2,0.44954237937927244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,2,power_law_1.2,0.4712639808654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,2,power_law_1.2,0.4753407955169678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,2,power_law_1.2,0.5185472011566162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,2,power_law_1.2,0.5521599769592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,2,power_law_1.2,0.5918528079986572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,2,power_law_1.2,0.6764287948608398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,2,power_law_1.2,0.7446015834808349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,2,power_law_1.2,0.9086591720581054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,2,power_law_1.2,1.0516927719116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,2,power_law_1.2,1.4024255752563477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,2,power_law_1.2,1.6718975067138673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,2,power_law_1.2,2.3386112213134767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,2,power_law_1.2,2.897100830078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,256,power_law_1.2,0.15687040090560914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,256,power_law_1.2,0.060063999891281125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,256,power_law_1.2,0.05792639851570129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,256,power_law_1.2,0.05371519923210144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,256,power_law_1.2,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,256,power_law_1.2,0.056358402967453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,256,power_law_1.2,0.05637120008468628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,256,power_law_1.2,0.05723519921302796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,256,power_law_1.2,0.05694720149040222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,256,power_law_1.2,0.05739520192146301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,256,power_law_1.2,0.06572800278663635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,256,power_law_1.2,0.07221760153770447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,256,power_law_1.2,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,256,power_law_1.2,0.07859839797019959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,256,power_law_1.2,0.08602880239486695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,256,power_law_1.2,0.08909440040588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,256,power_law_1.2,0.09775999784469605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,256,power_law_1.2,0.12450560331344604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,256,power_law_1.2,0.12215679883956909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,256,power_law_1.2,0.15395840406417846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,256,power_law_1.2,0.17019519805908204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,256,power_law_1.2,0.22584960460662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,256,power_law_1.2,0.24150400161743163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,256,power_law_1.2,0.37625598907470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,256,power_law_1.2,0.47605118751525877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,256,power_law_1.2,0.7086592197418213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,256,power_law_1.2,0.8084159851074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,64,balanced,0.021221332252025604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,64,balanced,0.01930133377512296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,64,balanced,0.023007998863856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,64,balanced,0.020986666282018025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,64,balanced,0.023056000471115112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,16,power_law_1.01,0.03660799860954285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,16,power_law_1.01,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,16,power_law_1.01,0.03410559892654419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,16,power_law_1.01,0.03580799996852875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,16,power_law_1.01,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,16,power_law_1.01,0.03873279988765717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,16,power_law_1.01,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,16,power_law_1.01,0.03903999924659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,16,power_law_1.01,0.03928320109844208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,16,power_law_1.01,0.03919999897480011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,16,power_law_1.01,0.03991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,16,power_law_1.01,0.04291839897632599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,16,power_law_1.01,0.04366079866886139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,16,power_law_1.01,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,16,power_law_1.01,0.04928640127182007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,16,power_law_1.01,0.050937598943710326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,16,power_law_1.01,0.06094080209732056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,16,power_law_1.01,0.0658623993396759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,16,power_law_1.01,0.07444480061531067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,16,power_law_1.01,0.09437440037727356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,16,power_law_1.01,0.10298880338668823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,16,power_law_1.01,0.1646399974822998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,16,power_law_1.01,0.17855360507965087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,16,power_law_1.01,0.2600320100784302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,16,power_law_1.01,0.33703041076660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,16,power_law_1.01,0.5164159774780274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,16,power_law_1.01,0.620140790939331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,1,power_law_1.01,0.018438400328159334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,1,power_law_1.01,0.018508799374103546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,1,power_law_1.01,0.01886720061302185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,1,power_law_1.01,0.018739199638366698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,1,power_law_1.01,0.01966080069541931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,1,power_law_1.01,0.021107199788093566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,1,power_law_1.01,0.02041600048542023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,1,power_law_1.01,0.021254399418830873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,1,power_law_1.01,0.021792000532150267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,1,power_law_1.01,0.021792000532150267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,1,power_law_1.01,0.022963200509548188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,1,power_law_1.01,0.02492160052061081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,1,power_law_1.01,0.0255295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,1,power_law_1.01,0.027289599180221558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,1,power_law_1.01,0.05050879716873169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,1,power_law_1.01,0.051660799980163576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,1,power_law_1.01,0.053286397457122804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,1,power_law_1.01,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.01,0.046342399716377256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.01,0.054585599899291994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.01,0.062457597255706786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,32,power_law_1.2,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,32,power_law_1.2,0.04880639910697937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,32,power_law_1.2,0.04693120121955872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,32,power_law_1.2,0.049292799830436704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.01,0.08652160167694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,32,power_law_1.2,0.04990079998970032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,32,power_law_1.2,0.048844799399375916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,32,power_law_1.2,0.050809597969055174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,32,power_law_1.2,0.05120000243186951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,32,power_law_1.2,0.05144960284233093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,32,power_law_1.2,0.05211520195007324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,32,power_law_1.2,0.05226879715919495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,32,power_law_1.2,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,32,power_law_1.2,0.055430400371551516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,32,power_law_1.2,0.056576001644134524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,32,power_law_1.2,0.06225919723510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,32,power_law_1.2,0.06159359812736511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,32,power_law_1.2,0.06563839912414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,32,power_law_1.2,0.07414399981498718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,32,power_law_1.2,0.07831680178642272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,32,power_law_1.2,0.09415680170059204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,32,power_law_1.2,0.10491520166397095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,32,power_law_1.2,0.13279999494552613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,32,power_law_1.2,0.16230399608612062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,32,power_law_1.2,0.22410240173339843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,32,power_law_1.2,0.28652799129486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,32,power_law_1.2,0.4160768032073975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,32,power_law_1.2,0.4946176052093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.01,0.09995520114898682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.01,0.1353279948234558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.01,0.17191040515899658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,1,balanced,0.02372266600529353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,1,balanced,0.023141334454218548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,1,balanced,0.023007998863856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,1,balanced,0.023045333723227184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,1,balanced,0.02369066576162974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,1,balanced,0.027034667630990345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,1,balanced,0.031845333675543465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,1,balanced,0.031727999448776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,1,balanced,0.04042666653792063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,1,balanced,0.03594133257865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,1,balanced,0.037834666669368744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,1,balanced,0.04304533203442892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,1,balanced,0.04823466638724009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,1,balanced,0.050399998823801674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,1,balanced,0.06942399839560191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,1,balanced,0.058378666639328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,1,balanced,0.0660159985224406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,1,balanced,0.08574933807055156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,1,balanced,0.10146133104960124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,1,balanced,0.14018133282661438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,1,balanced,0.17466133832931519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,1,balanced,0.24477332830429077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,1,balanced,0.30300267537434894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,1,balanced,0.4368693431218465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,1,balanced,0.5658773183822632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,1,balanced,0.8309546311696371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,1,balanced,1.0935413042704265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.01,0.244268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.01,0.3057663917541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,128,power_law_1.01,0.031276801228523256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,128,power_law_1.01,0.030899199843406677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,128,power_law_1.01,0.030937600135803222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,128,power_law_1.01,0.033062401413917544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,128,power_law_1.01,0.03315840065479279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,128,power_law_1.01,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,128,power_law_1.01,0.0342848002910614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,128,power_law_1.01,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,128,power_law_1.01,0.03470720052719116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,128,power_law_1.01,0.03464959859848023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,128,power_law_1.01,0.03501439988613129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,128,power_law_1.01,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,128,power_law_1.01,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,128,power_law_1.01,0.03708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,128,power_law_1.01,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,128,power_law_1.01,0.04007680118083954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,128,power_law_1.01,0.03996799886226654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,128,power_law_1.01,0.042387199401855466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,128,power_law_1.01,0.0440064013004303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,128,power_law_1.01,0.049747198820114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,128,power_law_1.01,0.05445759892463684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,128,power_law_1.01,0.06579840183258057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,128,power_law_1.01,0.07205119729042053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,128,power_law_1.01,0.09601920247077941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,128,power_law_1.01,0.11062400341033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,128,power_law_1.01,0.15386879444122314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,128,power_law_1.01,0.20112640857696534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,128,balanced,0.08329066634178162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,128,balanced,0.08116800089677174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,128,balanced,0.08192533254623413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,128,balanced,0.08553600311279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,128,balanced,0.08473599950472514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,128,balanced,0.09904000163078308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,128,balanced,0.11288533608118693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,128,balanced,0.11725866794586182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,128,balanced,0.10839999715487163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,128,balanced,0.11521066228548686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,128,balanced,0.11937066912651062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,128,balanced,0.12179199854532878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,128,balanced,0.12505599856376648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,128,balanced,0.13486933708190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,128,balanced,0.1343786617120107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,128,balanced,0.11986666917800903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,128,balanced,0.1274720033009847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,128,balanced,0.14592533310254416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,128,balanced,0.14412800470987955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,128,balanced,0.15346133708953857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,128,balanced,0.16522666811943054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,128,balanced,0.19132266441980997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,128,balanced,0.21790399154027304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,128,balanced,0.27447466055552167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,128,balanced,0.35785067081451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,128,balanced,0.4564053217569987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,128,balanced,0.5734879970550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,4,balanced,0.05030933519204458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,4,balanced,0.07288533449172974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,4,balanced,0.10358933607737224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,4,balanced,0.16037333011627197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,4,balanced,0.2571786642074585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,4,balanced,0.3121386567751567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,4,balanced,0.31305599212646484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,4,balanced,0.31780799229939777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,4,balanced,0.31406933069229126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,4,balanced,0.3174933393796285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,4,balanced,0.31701866785685223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,4,balanced,0.31850665807724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,4,balanced,0.3206719954808553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,4,balanced,0.3261760075887044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,4,balanced,0.3303626577059428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,4,balanced,0.3338720003763835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,4,balanced,0.3386826515197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,4,balanced,0.3859573205312093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,4,balanced,0.363045334815979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,4,balanced,0.4455466667811076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,4,balanced,0.4117813507715861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,4,balanced,0.6369440158208212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,4,balanced,0.5271413326263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,4,balanced,0.7547093232472738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,4,balanced,0.7842880090077718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,4,balanced,1.2734239896138508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,4,balanced,1.42412265141805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,32,power_law_1.01,0.049772799015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,32,power_law_1.01,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,32,power_law_1.01,0.04924159944057464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,32,power_law_1.01,0.04927360117435455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,32,power_law_1.01,0.04999040067195892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,32,power_law_1.01,0.04738560020923614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,32,power_law_1.01,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,32,power_law_1.01,0.05630080103874206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,32,power_law_1.01,0.05674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,32,power_law_1.01,0.059564799070358276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,32,power_law_1.01,0.0643775999546051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,32,power_law_1.01,0.06360960006713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,32,power_law_1.01,0.06803200244903565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,32,power_law_1.01,0.09659519791603088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,32,power_law_1.01,0.10409599542617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,32,power_law_1.01,0.11539839506149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,32,power_law_1.01,0.11722240447998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,32,power_law_1.01,0.1561087965965271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.01,0.19300479888916017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.01,0.23089919090270997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.01,0.14226560592651366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.01,0.18152960538864135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.01,0.23160319328308104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.01,0.34253439903259275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.01,0.389356803894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.01,0.580083179473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.01,0.7715519905090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,16,balanced,0.04385066529115041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,16,balanced,0.043663998444875084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,16,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,16,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,16,balanced,0.04986133178075155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,16,balanced,0.054773335655530296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,16,balanced,0.05571199953556061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,16,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,16,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,16,balanced,0.054458667834599815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,16,balanced,0.0562666654586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,16,balanced,0.06002133091290792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,16,balanced,0.05834666887919108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,16,balanced,0.060309335589408875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,16,balanced,0.06840000053246816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,16,balanced,0.06849066913127899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,16,balanced,0.07221333185831706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,16,balanced,0.07872533301512401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,16,balanced,0.08690667152404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,16,balanced,0.09923733274141948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,16,balanced,0.1114453375339508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,16,balanced,0.13740266362826029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,16,balanced,0.16305599610010782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,16,balanced,0.21615999937057495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,16,balanced,0.26130666335423786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,16,balanced,0.36925331751505536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,16,balanced,0.47208531697591144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,1,power_law_1.2,0.06465280055999756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,1,power_law_1.2,0.08156800270080566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,1,power_law_1.2,0.08414080142974853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,1,power_law_1.2,0.09393919706344604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,1,power_law_1.2,0.09871360063552856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,1,power_law_1.2,0.1233024001121521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,1,power_law_1.2,0.12923519611358641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,1,power_law_1.2,0.12957439422607422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,1,power_law_1.2,0.13578879833221436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,1,power_law_1.2,0.13735040426254272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,1,power_law_1.2,0.13790080547332764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,1,power_law_1.2,0.14428800344467163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,1,power_law_1.2,0.14553600549697876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,1,power_law_1.2,0.150764799118042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,1,power_law_1.2,0.14936959743499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,1,power_law_1.2,0.15279359817504884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,1,power_law_1.2,0.16094720363616943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,1,power_law_1.2,0.18098560571670533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,1,power_law_1.2,0.20074241161346434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,1,power_law_1.2,0.2482304096221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,1,power_law_1.2,0.2981760025024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,1,power_law_1.2,0.39880321025848386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,1,power_law_1.2,0.49694080352783204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,1,power_law_1.2,0.7029056072235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,1,power_law_1.2,0.9103039741516114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,1,power_law_1.2,1.307551956176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,1,power_law_1.2,1.739936065673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,4,power_law_1.01,0.08978559970855712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,4,power_law_1.01,0.13096319437026976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,4,power_law_1.01,0.1672063946723938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,4,power_law_1.01,0.20069758892059325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,4,power_law_1.01,0.23826560974121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,4,power_law_1.01,0.2695168018341064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,4,power_law_1.01,0.32381439208984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,4,power_law_1.01,0.3209023952484131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,4,power_law_1.01,0.33191039562225344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,4,power_law_1.01,0.3177727937698364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,4,power_law_1.01,0.3381184101104736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,4,power_law_1.01,0.3692863941192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,4,power_law_1.01,0.3500159978866577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,4,power_law_1.01,0.35445120334625246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,4,power_law_1.01,0.373638391494751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,4,power_law_1.01,0.4011839866638184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,4,power_law_1.01,0.3968384027481079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,4,power_law_1.01,0.41902718544006345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,4,power_law_1.01,0.44381442070007326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,4,power_law_1.01,0.5186240196228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,4,power_law_1.01,0.511187219619751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,4,power_law_1.01,0.6092031955718994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,4,power_law_1.01,0.7071104049682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,4,power_law_1.01,0.8775424003601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,4,power_law_1.01,1.0205696105957032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,4,power_law_1.01,1.3840959548950196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,4,power_law_1.01,1.664179229736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,8,balanced,0.04394133388996124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,8,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,8,balanced,0.047770669062932335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,8,balanced,0.05425066749254862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,8,balanced,0.08039466540018718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,8,balanced,0.08654933174451192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,8,balanced,0.08721066514650981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,8,balanced,0.08832533160845439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,8,balanced,0.08925333619117737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,8,balanced,0.09001066287358601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,8,balanced,0.08942400415738423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,8,balanced,0.09310400485992432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,8,balanced,0.09269866347312927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,8,balanced,0.09682133793830872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,8,balanced,0.10504000385602315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,8,balanced,0.10565333565076192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,8,balanced,0.10915733377138774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,8,balanced,0.11967466274897258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,8,balanced,0.12867732842763266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,8,balanced,0.16105066736539206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,8,balanced,0.17312532663345337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,8,balanced,0.23031999667485556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,8,balanced,0.26548266410827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,8,balanced,0.3850773175557454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,8,balanced,0.4446400006612142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,8,balanced,0.6210506757100424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,8,balanced,0.7966453234354655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,4,power_law_1.2,0.03214719891548157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,4,power_law_1.2,0.03452160060405731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,4,power_law_1.2,0.04674560129642487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,4,power_law_1.2,0.05571200251579285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,4,power_law_1.2,0.05630080103874206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,4,power_law_1.2,0.07007359862327575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,4,power_law_1.2,0.07265920042991639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,4,power_law_1.2,0.07176960110664368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,4,power_law_1.2,0.07534719705581665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,4,power_law_1.2,0.07260159850120544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,4,power_law_1.2,0.07489920258522034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,4,power_law_1.2,0.07634559869766236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,4,power_law_1.2,0.07910400032997131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,4,power_law_1.2,0.08947200179100037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,4,power_law_1.2,0.09498239755630493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,4,power_law_1.2,0.09965440034866332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,4,power_law_1.2,0.0986240029335022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,4,power_law_1.2,0.11139839887619019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,4,power_law_1.2,0.12876800298690796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,4,power_law_1.2,0.13691519498825072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,4,power_law_1.2,0.13808000087738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,4,power_law_1.2,0.16183680295944214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,4,power_law_1.2,0.18639999628067017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,4,power_law_1.2,0.25260798931121825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,4,power_law_1.2,0.3123071908950806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,4,power_law_1.2,0.42019200325012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,4,power_law_1.2,0.537062406539917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,1,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,1,balanced,0.03173333406448364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,1,balanced,0.035605333745479584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,1,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,1,balanced,0.0394400010506312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,1,balanced,0.03977599988381068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,1,balanced,0.03956799954175949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,1,balanced,0.04146666576464971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,1,balanced,0.04182399809360504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,1,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,1,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,1,balanced,0.044821331898371376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,1,balanced,0.048954665660858154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,1,balanced,0.05039466420809428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,1,balanced,0.06614399949709575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,1,balanced,0.06972800195217133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,1,balanced,0.07620266576608022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,1,balanced,0.09316800038019817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,1,balanced,0.10356799761454265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,1,balanced,0.13370666901270548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,1,balanced,0.1529973347981771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,1,balanced,0.21598400672276816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,1,balanced,0.2544426719347636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,1,balanced,0.36418131987253827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,1,balanced,0.4690026839574178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,1,balanced,0.6778293450673422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,1,balanced,0.874720017115275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,2,power_law_1.01,0.09049599766731262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,2,power_law_1.01,0.10979199409484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,2,power_law_1.01,0.12899199724197388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,2,power_law_1.01,0.1589568018913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,2,power_law_1.01,0.15370880365371703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,2,power_law_1.01,0.15975040197372437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,2,power_law_1.01,0.15984640121459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,2,power_law_1.01,0.16045440435409547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,2,power_law_1.01,0.16036479473114013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,2,power_law_1.01,0.1553536057472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,2,power_law_1.01,0.16242560148239135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,2,power_law_1.01,0.17571200132369996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,2,power_law_1.01,0.1906496047973633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,2,power_law_1.01,0.1934656023979187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,2,power_law_1.01,0.2172032117843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,2,power_law_1.01,0.220467209815979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,2,power_law_1.01,0.22985599040985108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,2,power_law_1.01,0.3065664052963257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,2,power_law_1.01,0.3326528072357178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,2,power_law_1.01,0.45090560913085936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,2,power_law_1.01,0.6332608222961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,2,power_law_1.01,0.8276736259460449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,2,power_law_1.01,1.0333951950073241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,2,power_law_1.01,1.5772224426269532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,2,power_law_1.01,1.7722496032714843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,2,power_law_1.01,3.0045312881469726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,16,power_law_1.01,0.12250239849090576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,16,power_law_1.01,0.15494400262832642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,16,power_law_1.01,0.14999040365219116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,16,power_law_1.01,0.15689599514007568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,16,power_law_1.01,0.15614080429077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,16,power_law_1.01,0.14742399454116822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,16,power_law_1.01,0.14473600387573243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,16,power_law_1.01,0.1459328055381775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,16,power_law_1.01,0.14581120014190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,16,power_law_1.01,0.14413440227508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,16,power_law_1.01,0.14844800233840943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,16,power_law_1.01,0.15527039766311646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,16,power_law_1.01,0.15695359706878662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,16,power_law_1.01,0.16053119897842408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,16,power_law_1.01,0.16930559873580933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,16,power_law_1.01,0.17448960542678832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,16,power_law_1.01,0.18957439661026002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,16,power_law_1.01,0.21439359188079835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,16,power_law_1.01,0.2224128007888794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,16,power_law_1.01,0.28146560192108155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,16,power_law_1.01,0.3077631950378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,16,power_law_1.01,0.38306560516357424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,16,power_law_1.01,0.41514878273010253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,16,power_law_1.01,0.5063039779663085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,16,power_law_1.01,0.6834879875183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,16,power_law_1.01,0.9530048370361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,16,power_law_1.2,0.05399680137634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,16,power_law_1.2,0.05148800015449524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,16,power_law_1.01,1.0796031951904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,16,power_law_1.2,0.047014400362968445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,16,power_law_1.2,0.04756479859352112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,16,power_law_1.2,0.04899199903011322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,16,power_law_1.2,0.04772480130195618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,16,power_law_1.2,0.04792959988117218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,16,power_law_1.2,0.0496832013130188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,16,power_law_1.2,0.050329601764678954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,16,power_law_1.2,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,16,power_law_1.2,0.05316479802131653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,16,power_law_1.2,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,16,power_law_1.2,0.05443199872970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,16,power_law_1.2,0.056409597396850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,16,power_law_1.2,0.06462720036506653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,16,power_law_1.2,0.06377599835395813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,16,power_law_1.2,0.06815999746322632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,16,power_law_1.2,0.07704319953918456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,16,power_law_1.2,0.0877888023853302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,16,power_law_1.2,0.10666240453720092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,16,power_law_1.2,0.12525440454483033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,16,power_law_1.2,0.1600767970085144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,16,power_law_1.2,0.20327041149139405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,16,power_law_1.2,0.27550721168518066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,16,power_law_1.2,0.33564159870147703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,16,power_law_1.2,0.4755392074584961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,2,power_law_1.01,3.3496448516845705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,16,power_law_1.2,0.6222784042358398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,128,balanced,0.058389330903689064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,128,balanced,0.06006933252016703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,128,balanced,0.058362667759259544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,128,balanced,0.062368000547091164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,128,balanced,0.06202666461467743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,128,balanced,0.0744053324063619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,128,balanced,0.08475200335184734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,128,balanced,0.08662933111190796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,128,balanced,0.0846506655216217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,128,balanced,0.0867786705493927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,128,balanced,0.09046933054924011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,128,balanced,0.08942932883898418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,128,balanced,0.08736000458399455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,128,balanced,0.08870933453241985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,128,balanced,0.09899200002352397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,128,balanced,0.0915786623954773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,128,balanced,0.0997173289457957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,128,balanced,0.11054399609565735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,128,balanced,0.12050132950146993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,128,balanced,0.1346453328927358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,128,balanced,0.1498026649157206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,128,balanced,0.17864533265431723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,128,balanced,0.2077066699663798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,128,balanced,0.2688586711883545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,128,balanced,0.35496532917022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,128,balanced,0.46371201674143475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,128,balanced,0.6036266485850016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,8,balanced,0.037861332297325134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,8,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,8,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,8,balanced,0.040287998815377556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,8,balanced,0.04197866717974345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,8,balanced,0.043290664752324425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,8,balanced,0.04229333500067393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,8,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,8,balanced,0.04669333497683207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,8,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,8,balanced,0.04645866652329763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,8,balanced,0.05519466598828634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,8,balanced,0.05417599777380625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,8,balanced,0.0620959997177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,8,balanced,0.06276266773541768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,8,balanced,0.06442666550477345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,8,balanced,0.07426666716734569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,8,balanced,0.07973333199818929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,8,balanced,0.08522133032480876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,8,balanced,0.09744000434875488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,8,balanced,0.10528000195821126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,8,balanced,0.15054933230082193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,8,balanced,0.1672746737798055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,8,balanced,0.23252799113591513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,8,balanced,0.2977653344472249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,8,balanced,0.4409866730372111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,8,balanced,0.5967413187026978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,4,power_law_1.01,0.04479359984397888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,4,power_law_1.01,0.044889599084854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,4,power_law_1.01,0.048256000876426695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,4,power_law_1.01,0.053862398862838744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,4,power_law_1.01,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,4,power_law_1.01,0.06046720147132874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,4,power_law_1.01,0.05943040251731872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,4,power_law_1.01,0.061875200271606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,4,power_law_1.01,0.06050559878349304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,4,power_law_1.01,0.06038399934768677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,4,power_law_1.01,0.06159359812736511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,4,power_law_1.01,0.06280320286750793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,4,power_law_1.01,0.06543359756469727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,4,power_law_1.01,0.06656000018119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,4,power_law_1.01,0.0704255998134613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,4,power_law_1.01,0.07166720032691956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,4,power_law_1.01,0.07434880137443542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,4,power_law_1.01,0.08938239812850952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,4,power_law_1.01,0.09503999948501587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,4,power_law_1.01,0.11711360216140747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,4,power_law_1.01,0.13725440502166747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,4,power_law_1.01,0.1698240041732788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,4,power_law_1.01,0.2083967924118042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,4,power_law_1.01,0.2929984092712402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,4,power_law_1.01,0.3768064022064209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,4,power_law_1.01,0.48638081550598145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,4,power_law_1.01,0.7284287929534912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,2,power_law_1.01,0.030086401104927062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,2,power_law_1.01,0.030022400617599487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,2,power_law_1.01,0.030054399371147157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,2,power_law_1.01,0.03020159900188446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,2,power_law_1.01,0.03052160143852234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,2,power_law_1.01,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,2,power_law_1.01,0.03309440016746521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,2,power_law_1.01,0.03445119857788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,2,power_law_1.01,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,2,power_law_1.01,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,2,power_law_1.01,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,2,power_law_1.01,0.03589119911193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,2,power_law_1.01,0.037747201323509214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,2,power_law_1.01,0.04015359878540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,2,power_law_1.01,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,2,power_law_1.01,0.0462336003780365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,2,power_law_1.01,0.04960640072822571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,2,power_law_1.01,0.05882880091667175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,2,power_law_1.01,0.06973440051078797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,2,power_law_1.01,0.08154879808425904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,2,power_law_1.01,0.1102336049079895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,2,power_law_1.01,0.14469759464263915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,2,power_law_1.01,0.16159360408782958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,2,power_law_1.01,0.24142079353332518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,2,power_law_1.01,0.3064512014389038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,2,power_law_1.01,0.42931838035583497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,2,power_law_1.01,0.5087935924530029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,4,balanced,0.07026666899522145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,4,balanced,0.0927946666876475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,4,balanced,0.13362666964530945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,4,balanced,0.2077066699663798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,4,balanced,0.33908267815907794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,4,balanced,0.4599626859029134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,4,balanced,0.45951998233795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,4,balanced,0.4628159999847412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,4,balanced,0.46491201718648273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,4,balanced,0.4678080081939697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,4,balanced,0.46528534094492596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,4,balanced,0.4717173178990682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,4,balanced,0.47410134474436444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,16,power_law_1.2,0.07694720029830933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,4,balanced,0.47600531578063965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,4,balanced,0.48817598819732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,16,power_law_1.2,0.09156479835510253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,4,balanced,0.49329066276550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,4,balanced,0.49834132194519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,16,power_law_1.2,0.08149120211601257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,4,balanced,0.6077173153559366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,4,balanced,0.5418560107549032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,16,power_law_1.2,0.09215999841690063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,4,balanced,0.6882719993591309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,4,balanced,0.6404373248418173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,4,balanced,1.012720028559367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,4,balanced,0.8053812980651855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,4,balanced,1.237061341603597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,4,balanced,1.2361013094584148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,16,power_law_1.2,0.08918399810791015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,4,balanced,1.8658453623453777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,4,balanced,2.0114293098449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,16,power_law_1.2,0.08613119721412658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,16,power_law_1.2,0.09113600254058837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,16,power_law_1.2,0.09321600198745728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,64,balanced,0.0440586656332016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,16,power_law_1.2,0.09146239757537841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,16,power_law_1.2,0.09030399918556213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,16,power_law_1.2,0.09607679843902588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,16,power_law_1.2,0.09442560076713562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,16,power_law_1.2,0.09568639993667602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,16,power_law_1.2,0.10111360549926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,64,balanced,0.043951998154322304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,64,balanced,0.043621331453323364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,64,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,64,balanced,0.04796266555786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,64,balanced,0.051882664362589516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,64,balanced,0.05193600058555603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,64,balanced,0.05186666548252106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,16,power_law_1.2,0.1080064058303833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,16,power_law_1.2,0.11122560501098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,16,power_law_1.2,0.11697920560836791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,64,balanced,0.05580266813437144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,64,balanced,0.05378133555253347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,64,balanced,0.0553706685702006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,2,power_law_1.01,0.041555199027061465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,2,power_law_1.01,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,2,power_law_1.01,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,2,power_law_1.01,0.05427839756011963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,2,power_law_1.01,0.055264002084732054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,2,power_law_1.01,0.07149440050125122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,2,power_law_1.01,0.06461439728736877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,64,balanced,0.0580320010582606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,2,power_law_1.01,0.08085759878158569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,64,balanced,0.05797333518664042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,2,power_law_1.01,0.07678080201148987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,16,power_law_1.2,0.12943999767303466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,2,power_law_1.01,0.09968640208244324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,16,power_law_1.2,0.14229120016098024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,2,power_law_1.01,0.11691520214080811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,16,power_law_1.2,0.17683839797973633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,2,power_law_1.01,0.12917120456695558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,16,power_law_1.2,0.2023616075515747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,2,power_law_1.01,0.10254080295562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,2,power_law_1.01,0.14678399562835692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,16,power_law_1.2,0.2704639911651611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,2,power_law_1.01,0.10436480045318604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,2,power_law_1.01,0.1549056053161621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,2,power_law_1.01,0.1090432047843933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,16,power_law_1.2,0.3400896072387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,2,power_law_1.01,0.10806399583816528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,2,power_law_1.01,0.15650559663772584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,2,power_law_1.01,0.11202559471130372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,16,power_law_1.2,0.4986368179321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,2,power_law_1.01,0.11363840103149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,2,power_law_1.01,0.16393599510192872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,2,power_law_1.01,0.11608959436416626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,2,power_law_1.01,0.16726399660110475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,2,power_law_1.01,0.11893119812011718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,2,power_law_1.01,0.12956160306930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,2,power_law_1.01,0.17497600317001344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,64,balanced,0.06232533355553945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,2,power_law_1.01,0.13962240219116212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,2,power_law_1.01,0.18012160062789917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,2,power_law_1.01,0.14736640453338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,2,power_law_1.01,0.14882559776306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,2,power_law_1.01,0.17365119457244874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.01,0.1886847972869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,16,power_law_1.2,0.6481599807739258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.01,0.22876160144805907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,64,balanced,0.07040533423423767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,16,power_law_1.2,0.9321536064147949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,2,power_law_1.01,0.19159679412841796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,2,power_law_1.01,0.19633280038833617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,16,power_law_1.2,1.2288191795349122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,2,power_law_1.01,0.21561601161956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,2,power_law_1.01,0.24500479698181152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.01,0.26584320068359374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,2,power_law_1.01,0.27463040351867674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,64,balanced,0.07211199899514516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.01,0.3523008108139038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,64,balanced,0.07261333366235097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,2,power_law_1.01,0.3350591897964478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.01,0.4361408233642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,2,power_law_1.01,0.39431679248809814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.01,0.5910399913787842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,2,power_law_1.01,0.5211584091186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.01,0.7745535850524903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,2,power_law_1.01,0.6431039810180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.01,1.047603225708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,64,balanced,0.0848479966322581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,64,balanced,0.09001066287358601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.01,1.4370880126953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,2,power_law_1.01,0.8776448249816895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,64,balanced,0.10103999574979146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,64,balanced,0.11683199803034465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,64,balanced,0.15013866623242697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,64,balanced,0.17696533600489298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,2,power_law_1.01,1.0791616439819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,64,balanced,0.22792534033457437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,64,balanced,0.2741546630859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,64,balanced,0.37489600976308185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,64,balanced,0.47091734409332275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,2,power_law_1.01,1.5612607955932618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,2,power_law_1.01,1.9982208251953124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,2,balanced,0.0644160012404124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,2,balanced,0.08685333530108134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,2,balanced,0.12027733524640401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,2,balanced,0.2002240022023519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,2,balanced,0.33510398864746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,2,balanced,0.46091198921203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,2,balanced,0.464682658513387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,2,balanced,0.46701331933339435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,2,balanced,0.46942933400472003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,2,balanced,0.47111467520395917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,2,balanced,0.47592000166575116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,2,balanced,0.47969599564870197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,2,balanced,0.4835573434829712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,2,balanced,0.4902506669362386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,2,balanced,0.5004479885101318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,2,balanced,0.5072906812032064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,2,balanced,0.5206666787465414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,2,balanced,0.6103306611378988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,2,balanced,0.5820480187733968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,2,balanced,0.7714666525522867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,2,balanced,0.7234933376312256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,2,balanced,0.9695146878560384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,2,balanced,0.9922986825307211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,2,balanced,1.51638396581014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,2,balanced,1.6405547459920247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,2,balanced,2.501135985056559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,2,balanced,2.9025920232137046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,2,power_law_1.01,0.043884798884391785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,2,power_law_1.01,0.046438398957252505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,2,power_law_1.01,0.04915199875831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,2,power_law_1.01,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,2,power_law_1.01,0.05367680191993714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,2,power_law_1.01,0.053388798236846925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,2,power_law_1.01,0.052767997980117796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,2,power_law_1.01,0.052851200103759766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,2,power_law_1.01,0.05284479856491089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,2,power_law_1.01,0.052223998308181765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,2,power_law_1.01,0.052537602186203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,2,power_law_1.01,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,2,power_law_1.01,0.057715201377868654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,2,power_law_1.01,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,16,balanced,0.04859200119972229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,16,balanced,0.050026665131251015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,16,balanced,0.051856001218159996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,16,balanced,0.05612266560395559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,16,balanced,0.05612266560395559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,16,balanced,0.06413866579532623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,16,balanced,0.06447466711203258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,16,balanced,0.06572266419728597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,16,balanced,0.06610666712125142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,16,balanced,0.0664160003264745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,16,balanced,0.06613866488138835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,16,balanced,0.06845866640408833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,16,balanced,0.07054399947325389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,2,power_law_1.01,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,16,balanced,0.06924800078074138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,16,balanced,0.07641066610813141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,16,balanced,0.07647466659545898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,16,balanced,0.07849599917729695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,16,balanced,0.08844266335169475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,16,balanced,0.09563733140627544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,16,balanced,0.11541866262753804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,16,balanced,0.13823999961217245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,16,balanced,0.1666933298110962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,16,balanced,0.1835199991861979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,16,balanced,0.24935466051101685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,16,balanced,0.3110453287760417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,2,power_law_1.01,0.06756479740142822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,2,power_law_1.01,0.07203199863433837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,16,balanced,0.42231468359629315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,16,balanced,0.5383253494898478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,2,power_law_1.01,0.08349440097808838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.01,0.09824000000953674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.01,0.11717120409011841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.01,0.126310396194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.01,0.17107199430465697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,4,balanced,0.04584000011285146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,4,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,4,balanced,0.0536053329706192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,4,balanced,0.07734933495521545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,4,balanced,0.10415466626485188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,4,balanced,0.1581013302008311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,4,balanced,0.2096853256225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,4,balanced,0.20943466822306314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,4,balanced,0.21345067024230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,4,balanced,0.21307732661565146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,4,balanced,0.214410662651062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,4,balanced,0.2227733333905538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,4,balanced,0.22881599267323813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,4,balanced,0.2249493400255839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,4,balanced,0.23797333240509033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,4,balanced,0.24011200666427612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,4,balanced,0.24953067302703857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,4,balanced,0.26904000838597614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,4,balanced,0.2895359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,4,balanced,0.3245013356208801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,4,balanced,0.3564213514328003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,4,balanced,0.43344000975290936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,4,balanced,0.5177653233210245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,4,balanced,0.6642986536026001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,4,balanced,0.8511306444803873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,4,balanced,1.1541226704915364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.01,0.21858561038970947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,4,balanced,1.4800693194071453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.01,0.32051839828491213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,16,power_law_1.2,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,16,power_law_1.2,0.046035200357437134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,16,power_law_1.2,0.044275200366973876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,16,power_law_1.2,0.0434688001871109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,16,power_law_1.2,0.04203520119190216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,16,power_law_1.2,0.04487679898738861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,16,power_law_1.2,0.045542401075363156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,16,power_law_1.2,0.04604800045490265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,16,power_law_1.2,0.04674560129642487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,16,power_law_1.2,0.04652160108089447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,16,power_law_1.2,0.04843519926071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,16,power_law_1.2,0.052262401580810545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,16,power_law_1.2,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,16,power_law_1.2,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,16,power_law_1.2,0.0599295973777771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,16,power_law_1.2,0.06447359919548035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,16,power_law_1.2,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,16,power_law_1.2,0.08173440098762512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.2,0.09786239862442017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.2,0.11368319988250733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.2,0.12483839988708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.2,0.17030400037765503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.01,0.38048639297485354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.2,0.19186559915542603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.2,0.2709376096725464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.2,0.36888320446014405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.2,0.4834752082824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.01,0.5118720054626464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.01,0.626803207397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,8,balanced,0.07361599802970886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,8,balanced,0.07269333302974701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,8,balanced,0.07445866862932841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,8,balanced,0.08583466211954753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,8,balanced,0.09847467144330342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,8,balanced,0.11752532919247945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,8,balanced,0.1344053347905477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,8,balanced,0.13365866740544638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,8,balanced,0.13673599561055502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,8,balanced,0.13363200426101685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,8,balanced,0.13563199838002524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,8,balanced,0.13914133111635843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,8,balanced,0.14031466841697693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,8,balanced,0.1430506706237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,8,balanced,0.15265066425005594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,8,balanced,0.15341867009798685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,8,balanced,0.15890133380889893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,8,balanced,0.16961600383122763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,8,balanced,0.17705066998799643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,8,balanced,0.19850132862726846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,8,balanced,0.22178133328755698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,8,balanced,0.27160000801086426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,8,balanced,0.3250346581141154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.2,0.6600831985473633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,8,balanced,0.4341919819513957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,8,balanced,0.5742026567459106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,8,balanced,0.7862239678700765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,8,balanced,1.0404640038808186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,32,power_law_1.2,0.060710400342941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,32,power_law_1.2,0.06063359975814819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,32,power_law_1.2,0.060127997398376466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,32,power_law_1.2,0.057094401121139525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,32,power_law_1.2,0.055219197273254396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,32,power_law_1.2,0.056620800495147706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,32,power_law_1.2,0.05927039980888367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,32,power_law_1.2,0.060812801122665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,32,power_law_1.2,0.06044800281524658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,32,power_law_1.2,0.061452800035476686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,32,power_law_1.2,0.06508799791336059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,32,power_law_1.2,0.06673280000686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,32,power_law_1.2,0.06763520240783691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,32,power_law_1.2,0.07077119946479797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,32,power_law_1.2,0.07701119780540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,8,balanced,0.03173333406448364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,8,balanced,0.03297066688537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,8,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,8,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,8,balanced,0.08171199758847554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,8,balanced,0.08260799944400787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,8,balanced,0.08197866876920064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,8,balanced,0.08312533299128215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,8,balanced,0.08556266625722249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,8,balanced,0.08407466610272725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,8,balanced,0.08502399921417236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,8,balanced,0.0846720039844513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,8,balanced,0.08642666538556416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,8,balanced,0.087909330924352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,8,balanced,0.09297600388526917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,8,balanced,0.09515200058619182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,8,balanced,0.09635200103123982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,8,balanced,0.1053013304869334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,8,balanced,0.11013866464296977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,8,balanced,0.12845333417256674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,8,balanced,0.13784533739089966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,8,balanced,0.17402132352193198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,32,power_law_1.2,0.0795199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,8,balanced,0.19885333379109701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,32,power_law_1.2,0.07950720191001892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,32,power_law_1.2,0.08447999954223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,32,power_law_1.2,0.09214079976081849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,32,power_law_1.2,0.10329600572586059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,32,power_law_1.2,0.115283203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,32,power_law_1.2,0.1508095979690552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,32,power_law_1.2,0.1676095962524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,32,power_law_1.2,0.2277695894241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,32,power_law_1.2,0.2898560047149658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,32,power_law_1.2,0.4390848159790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,32,power_law_1.2,0.5017536163330079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,8,balanced,0.2688053250312805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,8,balanced,0.32174400488535565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,8,balanced,0.4440693457921346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,8,balanced,0.5641546646753947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,2,power_law_1.01,0.01895039975643158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,2,power_law_1.01,0.020422400534152986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,2,power_law_1.01,0.026732799410820008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,2,power_law_1.01,0.031302401423454286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,2,power_law_1.01,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,2,power_law_1.01,0.043609601259231565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,2,power_law_1.01,0.044588801264762876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,2,power_law_1.01,0.04717440009117126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,2,power_law_1.01,0.04856959879398346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,2,power_law_1.01,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,2,power_law_1.01,0.05148800015449524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,2,power_law_1.01,0.052134400606155394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,2,power_law_1.01,0.05299199819564819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,2,power_law_1.01,0.05446400046348572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,2,power_law_1.01,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,2,power_law_1.01,0.05999360084533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,2,power_law_1.01,0.06229760050773621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,2,power_law_1.01,0.10911999940872193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.01,0.1204800009727478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.01,0.11850880384445191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.01,0.1229632019996643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.01,0.14700160026550294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.01,0.1702015995979309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.01,0.22194559574127198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.01,0.2685312032699585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.01,0.38806400299072263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.01,0.471014404296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,1,power_law_1.01,0.03139840066432953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,1,power_law_1.01,0.030476799607276915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,1,power_law_1.01,0.038489601016044615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,1,power_law_1.01,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,1,power_law_1.01,0.05533440113067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,1,power_law_1.01,0.07725440263748169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,1,power_law_1.01,0.07987200021743775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,1,power_law_1.01,0.08499199748039246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,1,power_law_1.01,0.08770560026168824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,1,power_law_1.01,0.08895360231399536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,1,power_law_1.01,0.09297279715538025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,1,power_law_1.01,0.09651839733123779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,1,power_law_1.01,0.09781119823455811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,1,power_law_1.01,0.10119040012359619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,1,power_law_1.01,0.10695680379867553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,1,power_law_1.01,0.11000959873199463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,1,power_law_1.01,0.12255359888076782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,1,power_law_1.01,0.13889919519424437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,1,power_law_1.01,0.15070079565048217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,1,power_law_1.01,0.1896448016166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,1,power_law_1.01,0.2078336000442505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,1,power_law_1.01,0.275219202041626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,1,power_law_1.01,0.3441663980484009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,1,power_law_1.01,0.47258877754211426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,1,power_law_1.01,0.6016767978668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,1,power_law_1.01,0.8483519554138184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,1,power_law_1.01,1.1018688201904296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,1,power_law_1.2,0.02380799949169159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,1,power_law_1.2,0.03343999981880188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,1,power_law_1.2,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,1,power_law_1.2,0.06483200192451477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,1,power_law_1.2,0.08398720026016235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,1,power_law_1.2,0.12431360483169555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,1,power_law_1.2,0.13579519987106323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,1,power_law_1.2,0.14479360580444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,1,power_law_1.2,0.14824960231781006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2688,1856,6,128,1,1,power_law_1.2,0.15052160024642944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2688,1856,6,128,1,1,power_law_1.2,0.161900794506073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2688,1856,6,128,1,1,power_law_1.2,0.1741312026977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2688,1856,6,128,1,1,power_law_1.2,0.16860159635543823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2688,1856,6,128,1,1,power_law_1.2,0.17659519910812377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2688,1856,6,128,1,1,power_law_1.2,0.19150079488754274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2688,1856,6,128,1,1,power_law_1.2,0.1812991976737976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2688,1856,6,128,1,1,power_law_1.2,0.19271680116653442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2688,1856,6,128,1,1,power_law_1.2,0.2118783950805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.2,0.23532159328460694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,32,power_law_1.2,0.03487359881401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,32,power_law_1.2,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,32,power_law_1.2,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,32,power_law_1.2,0.036294400691986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,32,power_law_1.2,0.03648000061511993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,32,power_law_1.2,0.03787519931793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,32,power_law_1.2,0.03767040073871612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,32,power_law_1.2,0.038124799728393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,32,power_law_1.2,0.03839359879493713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,32,power_law_1.2,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,32,power_law_1.2,0.03896960020065308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,32,power_law_1.2,0.04137600064277649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,32,power_law_1.2,0.04120959937572479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,32,power_law_1.2,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,32,power_law_1.2,0.04758400022983551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,32,power_law_1.2,0.048582398891448976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,32,power_law_1.2,0.05077120065689087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,32,power_law_1.2,0.05567359924316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,32,power_law_1.2,0.058713597059249875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,32,power_law_1.2,0.07031040191650391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,32,power_law_1.2,0.07531520128250122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,32,power_law_1.2,0.11357439756393432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,32,power_law_1.2,0.13916800022125245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,32,power_law_1.2,0.18217600584030152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,32,power_law_1.2,0.2283263921737671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,32,power_law_1.2,0.3512768030166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,32,power_law_1.01,0.03738879859447479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,32,power_law_1.01,0.03407999873161316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,32,power_law_1.01,0.033395200967788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,32,power_law_1.01,0.03255040049552917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,32,power_law_1.01,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,32,power_law_1.01,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,32,power_law_1.01,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,32,power_law_1.01,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,32,power_law_1.01,0.03537920117378235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,32,power_law_1.01,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,32,power_law_1.01,0.036582401394844054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,32,power_law_1.01,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,32,power_law_1.01,0.037273600697517395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,32,power_law_1.01,0.03894400000572205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,32,power_law_1.01,0.04227199852466583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,32,power_law_1.01,0.04316799938678741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,32,power_law_1.01,0.04343680143356323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,32,power_law_1.01,0.04707840085029602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,32,power_law_1.01,0.04924800097942352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,32,power_law_1.01,0.056383997201919556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,32,power_law_1.01,0.0643392026424408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,32,power_law_1.01,0.07393919825553893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,32,power_law_1.01,0.08772479891777038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,32,power_law_1.01,0.10412800312042236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,32,power_law_1.01,0.1369472026824951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,32,power_law_1.01,0.16906239986419677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,32,power_law_1.01,0.2351232051849365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,32,power_law_1.2,0.44815998077392577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,16,power_law_1.2,0.04166400134563446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,16,power_law_1.2,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,16,power_law_1.2,0.04239999949932098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,16,power_law_1.2,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,16,power_law_1.2,0.045151999592781066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,16,power_law_1.2,0.04565759897232056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,16,power_law_1.2,0.04555520117282867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,16,power_law_1.2,0.04769279956817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,16,power_law_1.2,0.04869759976863861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,16,power_law_1.2,0.05062400102615357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,16,power_law_1.2,0.05326719880104065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,16,power_law_1.2,0.0546176016330719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,16,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,16,power_law_1.2,0.056966400146484374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,16,power_law_1.2,0.0633791983127594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,16,power_law_1.2,0.06462720036506653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,16,power_law_1.2,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,16,power_law_1.2,0.08079360127449035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,16,power_law_1.2,0.0809216022491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,16,power_law_1.2,0.09889280200004577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,16,power_law_1.2,0.1059008002281189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,16,power_law_1.2,0.1341055989265442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,16,power_law_1.2,0.17203199863433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,16,power_law_1.2,0.21767039299011232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,16,power_law_1.2,0.2847743988037109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,16,power_law_1.2,0.39468159675598147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,16,power_law_1.2,0.5198847770690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,4,power_law_1.01,0.03777920007705689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,4,power_law_1.01,0.03834879994392395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,4,power_law_1.01,0.04477440118789673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,4,power_law_1.01,0.05118719935417175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,4,power_law_1.01,0.057004797458648684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,4,power_law_1.01,0.07071999907493591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,4,power_law_1.01,0.07217919826507568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,4,power_law_1.01,0.07138559818267823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,4,power_law_1.01,0.0738431990146637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,4,power_law_1.01,0.07332479953765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,4,power_law_1.01,0.07544320225715637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,4,power_law_1.01,0.07707520127296448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,4,power_law_1.01,0.0776639997959137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,4,power_law_1.01,0.07998080253601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,4,power_law_1.01,0.08663679957389832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,4,power_law_1.01,0.08652160167694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,4,power_law_1.01,0.08840960264205933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,4,power_law_1.01,0.10022399425506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,4,power_law_1.01,0.11802879571914673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,4,power_law_1.01,0.13557759523391724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,4,power_law_1.01,0.13434239625930786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,4,power_law_1.01,0.168121600151062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,4,power_law_1.01,0.19806079864501952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,4,power_law_1.01,0.2754175901412964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,4,power_law_1.01,0.3065792083740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,4,power_law_1.01,0.4097280025482178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,4,power_law_1.01,0.550713586807251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,4,balanced,0.043951998154322304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,4,balanced,0.045498669147491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,4,balanced,0.05054933329423269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,4,balanced,0.07926400005817413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,4,balanced,0.1125333309173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,4,balanced,0.12281066179275513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,4,balanced,0.12493866682052612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,4,balanced,0.12277332941691081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,4,balanced,0.12098133563995361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,4,balanced,0.12325333555539449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,4,balanced,0.12449600299199422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,4,balanced,0.12503467003504434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,4,balanced,0.1262506643931071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,4,balanced,0.131632000207901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,4,balanced,0.13506133357683817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,4,balanced,0.13917332887649536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,4,balanced,0.1432213286558787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,4,balanced,0.16106667121251425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,4,balanced,0.16565866271654764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,4,balanced,0.2037386695543925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,4,balanced,0.22055466969807944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,4,balanced,0.2946239908536275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,4,balanced,0.31274133920669556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,4,balanced,0.4592906634012858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,4,balanced,0.5327893495559692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,4,balanced,0.7154560089111328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,4,balanced,0.9343626499176025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,32,power_law_1.2,0.03686400055885315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,32,power_law_1.2,0.03285120129585266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,32,power_law_1.2,0.03249920010566711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,32,power_law_1.2,0.03086079955101013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,32,power_law_1.2,0.03170560002326965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,32,power_law_1.2,0.0336896002292633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,32,power_law_1.2,0.03372159898281098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,32,power_law_1.2,0.03372800052165985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,32,power_law_1.2,0.03357439935207367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,32,power_law_1.2,0.034067198634147644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,32,power_law_1.2,0.03511680066585541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,32,power_law_1.2,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,32,power_law_1.2,0.036473599076271054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,32,power_law_1.2,0.038438400626182555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,32,power_law_1.2,0.04099839925765991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,32,power_law_1.2,0.04425599873065948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,32,power_law_1.2,0.04645760059356689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,32,power_law_1.2,0.05623679757118225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,32,power_law_1.2,0.056985598802566526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,32,power_law_1.2,0.07011839747428894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,32,power_law_1.2,0.09180160164833069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,32,power_law_1.2,0.1053056001663208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,32,power_law_1.2,0.12174719572067261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,32,power_law_1.2,0.17711360454559327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,32,power_law_1.2,0.20766079425811768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,2,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,2,balanced,0.033301333586374916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,2,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,2,balanced,0.042581334710121155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,2,balanced,0.05580266813437144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,2,balanced,0.05589333176612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,2,balanced,0.05421333511670431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,2,balanced,0.059989333152770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,2,balanced,0.060032000144322716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,2,balanced,0.0610346645116806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,2,balanced,0.06047999858856201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,2,balanced,0.06232533355553945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,2,balanced,0.06504000226656596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,2,balanced,0.06845866640408833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,2,balanced,0.07523733377456665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,2,balanced,0.08041599889596303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,2,balanced,0.08662933111190796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,2,balanced,0.10283733407656352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,2,balanced,0.11896000305811565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,2,balanced,0.1439893345038096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,2,balanced,0.16907199223836264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,2,balanced,0.23836799462636313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,2,balanced,0.29657065868377686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,2,balanced,0.4337759812672933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,2,balanced,0.5486773252487183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,2,balanced,0.8101653258005778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,2,balanced,1.0531360308329265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,32,power_law_1.2,0.29390079975128175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,32,power_law_1.2,0.35697920322418214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,4,power_law_1.01,0.03711999952793121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,4,power_law_1.01,0.04749439954757691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,4,power_law_1.01,0.051622402667999265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,4,power_law_1.01,0.0577023983001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,4,power_law_1.01,0.06355199813842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,4,power_law_1.01,0.07871999740600585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,4,power_law_1.01,0.0749504029750824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,4,power_law_1.01,0.08155519962310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,4,power_law_1.01,0.08140159845352173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,4,power_law_1.01,0.08144000172615051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,4,power_law_1.01,0.08626559972763062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,4,power_law_1.01,0.08881919980049133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,4,power_law_1.01,0.09223039746284485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,4,power_law_1.01,0.09470080137252808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,4,power_law_1.01,0.10056960582733154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,4,power_law_1.01,0.10423040390014648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,4,power_law_1.01,0.11061760187149047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,4,power_law_1.01,0.12120959758758545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,32,power_law_1.2,0.07574399709701538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,32,power_law_1.2,0.06946560144424438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,32,power_law_1.2,0.06902400255203248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,32,power_law_1.2,0.0701632022857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,4,power_law_1.01,0.14161280393600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,32,power_law_1.2,0.071424001455307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,4,power_law_1.01,0.1650879979133606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,32,power_law_1.2,0.0709119975566864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,32,power_law_1.2,0.07311999797821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,32,power_law_1.2,0.07268480062484742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,32,power_law_1.2,0.07467520236968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,32,power_law_1.2,0.07392640113830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,32,power_law_1.2,0.07631999850273133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,4,power_law_1.01,0.1949504017829895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,4,power_law_1.01,0.24618880748748778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,4,power_law_1.01,0.3051520109176636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,4,power_law_1.01,0.4078207969665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,4,power_law_1.01,0.5287231922149658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,4,power_law_1.01,0.7342463970184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,32,power_law_1.2,0.07967360019683838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,32,power_law_1.2,0.08203520178794861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,4,power_law_1.01,0.9646271705627442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,32,power_law_1.2,0.08452479839324951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,32,power_law_1.2,0.0933184027671814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,32,power_law_1.2,0.09512959718704224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,32,power_law_1.2,0.10268160104751586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,32,power_law_1.2,0.12339199781417846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,32,power_law_1.01,0.0514240026473999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,32,power_law_1.2,0.13729920387268066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,32,power_law_1.01,0.03344640135765076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,32,power_law_1.01,0.032025599479675294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,32,power_law_1.01,0.02993920147418976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,32,power_law_1.01,0.030368000268936157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,32,power_law_1.01,0.03209599852561951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,32,power_law_1.01,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,32,power_law_1.01,0.03209599852561951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,32,power_law_1.01,0.03282560110092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,32,power_law_1.01,0.03284479975700379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,32,power_law_1.01,0.033267199993133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,32,power_law_1.01,0.03335680067539215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,32,power_law_1.01,0.03439359962940216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,32,power_law_1.01,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,32,power_law_1.01,0.039750400185585025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,32,power_law_1.01,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,32,power_law_1.01,0.043161600828170776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,32,power_law_1.01,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,32,power_law_1.01,0.05030400156974792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,32,power_law_1.2,0.1684671998023987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,32,power_law_1.01,0.06085759997367859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,32,power_law_1.2,0.21434240341186522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,32,power_law_1.01,0.058143997192382814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,32,power_law_1.2,0.27245440483093264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,32,power_law_1.2,0.2857280015945435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,32,power_law_1.2,0.4067647933959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,32,power_law_1.2,0.4974720001220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,32,power_law_1.01,0.06936320066452026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,32,power_law_1.01,0.08948479890823365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,32,power_law_1.2,0.6731647968292236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,32,power_law_1.01,0.10651520490646363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,32,power_law_1.01,0.12136319875717164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,32,power_law_1.2,0.8652480125427247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,32,power_law_1.01,0.19288959503173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,32,power_law_1.01,0.181222403049469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,2,balanced,0.06659733255704244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,2,balanced,0.06922666728496552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,2,balanced,0.07494399944941203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,2,balanced,0.08667199810345967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,2,balanced,0.11389333009719849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,2,balanced,0.17092265685399374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,2,balanced,0.17459199825922647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,2,balanced,0.17321600516637167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,2,balanced,0.17134400208791098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,2,balanced,0.17473600308100382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,2,balanced,0.17561600605646768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,2,balanced,0.1788960099220276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,2,balanced,0.18158932526906332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,2,balanced,0.186298668384552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,2,balanced,0.1904319922129313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,2,balanced,0.19555199146270752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,2,balanced,0.20440532763799033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,2,balanced,0.22696000337600708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,2,balanced,0.24388267596562704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,2,balanced,0.28835733731587726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,2,balanced,0.33005332946777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,2,balanced,0.4213493267695109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,2,balanced,0.5189280112584432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,2,balanced,0.7549920082092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,2,balanced,0.9551040331522623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,2,balanced,1.3943732579549153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,2,balanced,1.7963892618815105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,1,power_law_1.01,0.056704002618789676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,1,power_law_1.01,0.06092159748077393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,1,power_law_1.01,0.08168960213661194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,1,power_law_1.01,0.1158336043357849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,1,power_law_1.01,0.1473471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,1,power_law_1.01,0.17831679582595825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,1,power_law_1.01,0.23066239356994628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,1,power_law_1.01,0.23911681175231933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,1,power_law_1.01,0.25166079998016355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,1,power_law_1.01,0.2568896055221558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,1,power_law_1.01,0.2647871971130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,1,power_law_1.01,0.2815104007720947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,1,power_law_1.01,0.29210240840911866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,1,power_law_1.01,0.312825608253479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,1,power_law_1.01,0.32905600070953367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,1,power_law_1.01,0.35031681060791015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,1,power_law_1.01,0.3548991918563843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,1,power_law_1.01,0.41831040382385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,1,power_law_1.01,0.46334080696105956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,1,power_law_1.01,0.5647808074951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,1,power_law_1.01,0.6461952209472657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,1,power_law_1.01,0.8364352226257324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,1,power_law_1.01,1.0395903587341309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,1,power_law_1.01,1.4014528274536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,1,power_law_1.01,1.7639999389648438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,1,power_law_1.01,2.4812671661376955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,1,power_law_1.01,3.2134334564208986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,4,balanced,0.03165333221356074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,4,balanced,0.03253866732120514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,4,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,4,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,4,balanced,0.060405333836873375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,4,balanced,0.062261333068211876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,4,balanced,0.061936000982920326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,4,balanced,0.06281066437562306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,4,balanced,0.0621973325808843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,4,balanced,0.06249066690603892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,4,balanced,0.0624533345301946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,4,balanced,0.06432533264160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,4,balanced,0.06507200002670288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,4,balanced,0.0676693320274353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,4,balanced,0.07250133156776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,4,balanced,0.07461866736412048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,4,balanced,0.07603199779987335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,4,balanced,0.08134399851163228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,4,balanced,0.08848533034324646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,4,balanced,0.1065066655476888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,4,balanced,0.11476266384124756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,4,balanced,0.1450453301270803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,4,balanced,0.1693920095761617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,4,balanced,0.22604266802469888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,4,balanced,0.29445334275563556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,4,balanced,0.411077340443929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,4,balanced,0.5230933427810669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,4,power_law_1.01,0.07256320118904114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,4,power_law_1.01,0.0906175971031189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,4,power_law_1.01,0.12335360050201416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,4,power_law_1.01,0.12665599584579468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,4,power_law_1.01,0.15889919996261598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,4,power_law_1.01,0.16958719491958618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,4,power_law_1.01,0.18889600038528442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,4,power_law_1.01,0.22119040489196778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,4,power_law_1.01,0.20943360328674315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,4,power_law_1.01,0.21230080127716064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,4,power_law_1.01,0.2256256103515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,4,power_law_1.01,0.21658880710601808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,4,power_law_1.01,0.22344961166381835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,4,power_law_1.01,0.23493120670318604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,4,power_law_1.01,0.25173120498657225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,4,power_law_1.01,0.25674240589141845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,4,power_law_1.01,0.2671231985092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,4,power_law_1.01,0.2932415962219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,4,power_law_1.01,0.32125439643859866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,4,power_law_1.01,0.3707583904266357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,4,power_law_1.01,0.42665600776672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,4,power_law_1.01,0.5612736225128174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,4,power_law_1.01,0.6013311862945556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,4,power_law_1.01,0.7818367958068848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,4,power_law_1.01,0.9741120338439941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,4,power_law_1.01,1.3299519538879394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,4,power_law_1.01,1.6472448348999023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,8,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,8,balanced,0.033733333150545754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,8,balanced,0.035317334036032356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,8,balanced,0.03978666663169861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,8,balanced,0.04555733501911163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,8,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,8,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,8,balanced,0.04671466847260793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,8,balanced,0.046367997924486794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,8,balanced,0.04654933512210846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,8,balanced,0.0476693312327067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,8,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,8,balanced,0.04996266464392344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,8,balanced,0.049626668294270836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,8,balanced,0.05169066786766052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,8,balanced,0.05352533360322317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,8,balanced,0.05416533350944519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,8,balanced,0.05829333265622457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,8,balanced,0.06002133091290792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,8,balanced,0.067071999112765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,8,balanced,0.07051200171311696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,8,balanced,0.09242666761080424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,8,balanced,0.10605333248774211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,8,balanced,0.1400373379389445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,8,balanced,0.17325333754221597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,8,balanced,0.23412799835205078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,8,balanced,0.29811733961105347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,16,balanced,0.048698668678601585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,16,balanced,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,16,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,16,balanced,0.05418666700522105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,16,balanced,0.05589866638183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,16,balanced,0.06550399959087372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,16,balanced,0.06784533460934956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,16,balanced,0.06764799853165944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,16,balanced,0.06854400038719177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,16,balanced,0.06951466699441274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,16,balanced,0.06860800087451935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,16,balanced,0.07190933326880138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,16,balanced,0.07050666709740956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,16,balanced,0.073594664533933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,16,balanced,0.08682666222254436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,16,balanced,0.08514666557312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,16,balanced,0.09126399954160054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,16,balanced,0.10317333539326985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,16,balanced,0.11101866761843364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,16,balanced,0.12150933345158894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,16,balanced,0.13903466860453287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,16,balanced,0.17112533251444498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,16,balanced,0.20100265741348267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,16,balanced,0.27533332506815594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,16,balanced,0.34165334701538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,16,balanced,0.4950079917907715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,16,balanced,0.6385066509246826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,1,balanced,0.01942933350801468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,1,balanced,0.019071999937295914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,1,balanced,0.018826667219400406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,1,balanced,0.019120000302791595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,1,balanced,0.019253333409627277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,1,balanced,0.019215999792019527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,1,balanced,0.02110933264096578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,1,balanced,0.02103466788927714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,1,balanced,0.021327999730904896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,1,balanced,0.021733333667119343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,1,balanced,0.023002666731675465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,1,balanced,0.02500266581773758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,1,balanced,0.029178666571776073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,1,balanced,0.025792000194390614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,1,balanced,0.02740799884001414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,1,balanced,0.02940800040960312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,1,balanced,0.032474666833877563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,1,balanced,0.038405333956082664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,1,balanced,0.042117332418759666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,1,balanced,0.05615466833114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,1,balanced,0.06158933540185293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,1,balanced,0.08265600105126698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,1,balanced,0.09830400347709656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,1,balanced,0.12786666552225748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,1,balanced,0.16089600324630737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,1,balanced,0.22437334060668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,1,balanced,0.2884426712989807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,4,power_law_1.2,0.03204480111598969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,4,power_law_1.2,0.029657599329948426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,4,power_law_1.2,0.03160319924354553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,4,power_law_1.2,0.03480319976806641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,4,power_law_1.2,0.035436800122261046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,4,power_law_1.2,0.03973119854927063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,4,power_law_1.2,0.04004479944705963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,4,power_law_1.2,0.040908798575401306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,4,power_law_1.2,0.04232319891452789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,4,power_law_1.2,0.04270719885826111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,4,power_law_1.2,0.04454399943351746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,4,power_law_1.2,0.04744960069656372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,4,power_law_1.2,0.049958398938179015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,4,power_law_1.2,0.04997119903564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,4,power_law_1.2,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,4,power_law_1.2,0.060083198547363284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,4,power_law_1.2,0.06398079991340637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,4,power_law_1.2,0.07406079769134521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,4,power_law_1.2,0.0917631983757019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,4,power_law_1.2,0.10751359462738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,4,power_law_1.2,0.12116479873657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,4,power_law_1.2,0.15408639907836913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,4,power_law_1.2,0.20076799392700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,4,power_law_1.2,0.24766080379486083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,4,power_law_1.2,0.3718656063079834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,4,power_law_1.2,0.5338687896728516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,32,power_law_1.01,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,32,power_law_1.01,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,32,power_law_1.01,0.029260799288749695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,32,power_law_1.01,0.030099201202392577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,32,power_law_1.01,0.031692799925804135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,32,power_law_1.01,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,32,power_law_1.01,0.03249920010566711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,32,power_law_1.01,0.033267199993133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,32,power_law_1.01,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,32,power_law_1.01,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,32,power_law_1.01,0.034195199608802795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,32,power_law_1.01,0.034508800506591795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,32,power_law_1.01,0.03650560081005096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,32,power_law_1.01,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,32,power_law_1.01,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,32,power_law_1.01,0.04237439930438995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,32,power_law_1.01,0.04615040123462677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,32,power_law_1.01,0.049395200610160825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,32,power_law_1.01,0.05384960174560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,32,power_law_1.01,0.06372479796409607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,32,power_law_1.01,0.07798399925231933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,32,power_law_1.01,0.0925055980682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,32,power_law_1.01,0.09963520169258118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,32,power_law_1.01,0.1578879952430725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,32,power_law_1.01,0.1709247946739197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,32,power_law_1.01,0.22149760723114015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,32,power_law_1.01,0.30178558826446533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,8,power_law_1.2,0.05065600275993347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,8,power_law_1.2,0.04689919948577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,8,power_law_1.2,0.055302399396896365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,8,power_law_1.2,0.0586624026298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,8,power_law_1.2,0.06050559878349304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,8,power_law_1.2,0.06700800061225891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,8,power_law_1.2,0.06744959950447083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,8,power_law_1.2,0.06940159797668458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,8,power_law_1.2,0.07037439942359924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,8,power_law_1.2,0.07010560035705567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,8,power_law_1.2,0.07134720087051391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,8,power_law_1.2,0.07412480115890503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,8,power_law_1.2,0.07386239767074584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,8,power_law_1.2,0.07544959783554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,8,power_law_1.2,0.08292480111122132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,8,power_law_1.2,0.08373759984970093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,8,power_law_1.2,0.09385600090026855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,8,power_law_1.2,0.10485759973526002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,8,power_law_1.2,0.11995520591735839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,8,power_law_1.2,0.14583040475845338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,8,power_law_1.2,0.19227520227432252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,8,power_law_1.2,0.2629695892333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,8,power_law_1.2,0.29462399482727053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,8,power_law_1.2,0.4935616016387939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,8,power_law_1.2,0.5860352039337158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,4,power_law_1.2,0.727507209777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,8,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,8,balanced,0.041573333243529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,8,balanced,0.041946664452552795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,8,balanced,0.04305600126584371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,8,balanced,0.047744000951449074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,8,balanced,0.052144000927607216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,8,balanced,0.052501335740089417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,8,balanced,0.05394133428732554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,8,balanced,0.054474666714668274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,8,balanced,0.05413866539796194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,8,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,8,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,8,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,8,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,8,balanced,0.06404800216356914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,8,balanced,0.06410666803518932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,8,balanced,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,8,balanced,0.07602133353551228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,8,balanced,0.07886399825414021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,8,balanced,0.09386133154233296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,8,balanced,0.10302399595578511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,8,balanced,0.1448319951693217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,8,balanced,0.17913599809010824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,8,balanced,0.24876266717910767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,8,balanced,0.31013866265614826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,8,power_law_1.2,0.8842111587524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,8,balanced,0.434986670811971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,8,balanced,0.5581066608428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,8,power_law_1.2,1.2582271575927735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,32,balanced,0.050069332122802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,32,balanced,0.04901866614818573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,32,balanced,0.04044266790151596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,32,balanced,0.04555733501911163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,32,balanced,0.0462666650613149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,32,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,32,balanced,0.05178666611512502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,32,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,32,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,32,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,32,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,32,balanced,0.05402133365472158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,32,balanced,0.05580799778302511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,32,balanced,0.05622933308283488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,32,balanced,0.06029866635799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,32,balanced,0.06028266747792562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,32,balanced,0.06443733473618825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,32,balanced,0.06938666601975758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,32,balanced,0.07453866799672444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,32,balanced,0.0885599950949351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,32,balanced,0.09889066219329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,32,balanced,0.11965866883595784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,32,balanced,0.13729600111643472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,32,balanced,0.17893334229787192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,32,balanced,0.21394133567810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,32,balanced,0.2979466716448466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,32,balanced,0.3651680151621501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,128,power_law_1.2,0.03187839984893799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,128,power_law_1.2,0.03144319951534271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,128,power_law_1.2,0.031948798894882204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,128,power_law_1.2,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,128,power_law_1.2,0.034329599142074584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,128,power_law_1.2,0.034508800506591795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,128,power_law_1.2,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,128,power_law_1.2,0.035155200958251955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,128,power_law_1.2,0.03521920144557953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,128,power_law_1.2,0.0353983998298645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,128,power_law_1.2,0.036422398686408994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,128,power_law_1.2,0.03632639944553375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,128,power_law_1.2,0.036908799409866334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,128,power_law_1.2,0.03726080060005188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,128,power_law_1.2,0.03997440040111542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,128,power_law_1.2,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,128,power_law_1.2,0.040575999021530154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,128,power_law_1.2,0.04255360066890716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,128,power_law_1.2,0.04471679925918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,128,power_law_1.2,0.05132799744606018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,128,power_law_1.2,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,128,power_law_1.2,0.07404159903526306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,128,power_law_1.2,0.08668799996376038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,16,power_law_1.2,0.04100480079650879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,16,power_law_1.2,0.0486847996711731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,16,power_law_1.2,0.043532800674438474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,16,power_law_1.2,0.04442879855632782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,16,power_law_1.2,0.045542401075363156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,16,power_law_1.2,0.04567039906978607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,128,power_law_1.2,0.1223423957824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,16,power_law_1.2,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,16,power_law_1.2,0.04732159972190857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,16,power_law_1.2,0.04844799935817719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,16,power_law_1.2,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,16,power_law_1.2,0.05009920001029968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,16,power_law_1.2,0.050860798358917235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,16,power_law_1.2,0.052198398113250735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,16,power_law_1.2,0.05382400155067444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,16,power_law_1.2,0.05994240045547485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,16,power_law_1.2,0.061484801769256595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,16,power_law_1.2,0.06444159746170045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,16,power_law_1.2,0.07194240093231201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,128,power_law_1.2,0.14673919677734376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,16,power_law_1.2,0.07997440099716187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,16,power_law_1.2,0.09279360175132752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,16,power_law_1.2,0.1061568021774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,16,power_law_1.2,0.14175360202789306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,16,power_law_1.2,0.181222403049469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,16,power_law_1.2,0.254150390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,16,power_law_1.2,0.3126336097717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,16,power_law_1.2,0.454099178314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,16,power_law_1.2,0.5819712162017823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,128,power_law_1.2,0.1985983967781067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,128,power_law_1.2,0.27153921127319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,32,balanced,0.07162666817506154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,32,balanced,0.051967998345692955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,32,balanced,0.051781331499417625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,32,balanced,0.07231999933719635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,32,balanced,0.09847999612490337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,32,balanced,0.15158933401107788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,32,balanced,0.15204800168673197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,32,balanced,0.15266666809717813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,32,balanced,0.1523306667804718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,32,balanced,0.15449600418408713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,32,balanced,0.15237866838773093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,32,balanced,0.15546666582425436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,32,balanced,0.15571733315785727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,32,balanced,0.15555733442306519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,32,balanced,0.1609760026137034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,32,balanced,0.1625493367513021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,32,balanced,0.16266133387883505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,32,balanced,0.16991466283798218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,32,balanced,0.1753173271814982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,32,balanced,0.1858560045560201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,32,balanced,0.2015519936879476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,32,balanced,0.22340800364812216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,32,balanced,0.24837867418924967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,32,balanced,0.30505067110061646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,32,balanced,0.34404265880584717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,32,balanced,0.4556693236033122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,32,balanced,0.5410399834314982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,16,power_law_1.01,0.10412800312042236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,16,power_law_1.01,0.15511679649353027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,16,power_law_1.01,0.19705599546432495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,16,power_law_1.01,0.19207040071487427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,16,power_law_1.01,0.196396803855896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,16,power_law_1.01,0.18556159734725952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,16,power_law_1.01,0.19498239755630492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,16,power_law_1.01,0.16355199813842775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,16,power_law_1.01,0.18348159790039062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,16,power_law_1.01,0.16998399496078492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,16,power_law_1.01,0.1859455943107605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,16,power_law_1.01,0.1866688013076782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,16,power_law_1.01,0.1828287959098816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,16,power_law_1.01,0.18103040456771852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,16,power_law_1.01,0.18668160438537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,16,power_law_1.01,0.2026304006576538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,16,power_law_1.01,0.20000638961791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,16,power_law_1.01,0.21264638900756835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,16,power_law_1.01,0.24080638885498046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,16,power_law_1.01,0.2863039970397949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,16,power_law_1.01,0.2986624002456665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,16,power_law_1.01,0.3545536041259766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,16,power_law_1.01,0.38997759819030764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,16,power_law_1.01,0.49355521202087405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,16,power_law_1.01,0.5941184043884278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,16,power_law_1.01,0.7761600017547607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,16,power_law_1.01,0.9793279647827149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,1,power_law_1.2,0.08699520230293274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,1,power_law_1.2,0.09262080192565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,1,power_law_1.2,0.10369280576705933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,1,power_law_1.2,0.12855679988861085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,1,power_law_1.2,0.14792319536209106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,1,power_law_1.2,0.1714303970336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,1,power_law_1.2,0.21521921157836915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,1,power_law_1.2,0.21920640468597413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,1,power_law_1.2,0.2282111883163452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,1,power_law_1.2,0.23379840850830078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,1,power_law_1.2,0.23836159706115723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,1,power_law_1.2,0.24785280227661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,1,power_law_1.2,0.2488255977630615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,1,power_law_1.2,0.25966720581054686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,1,power_law_1.2,0.2708479881286621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,1,power_law_1.2,0.27657599449157716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,1,power_law_1.2,0.2923327922821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,1,power_law_1.2,0.33254399299621584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,1,power_law_1.2,0.3736576080322266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,1,power_law_1.2,0.4463871955871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,1,power_law_1.2,0.5184639930725098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,1,power_law_1.2,0.6597887992858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,1,power_law_1.2,0.8097727775573731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,1,power_law_1.2,1.0944064140319825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,1,power_law_1.2,1.3916352272033692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,1,power_law_1.2,1.9723455429077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,1,power_law_1.2,2.53690242767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,16,power_law_1.2,0.10995199680328369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,16,power_law_1.2,0.15632640123367308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,16,power_law_1.2,0.15155839920043945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,16,power_law_1.2,0.15112320184707642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,16,power_law_1.2,0.1592512011528015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,16,power_law_1.2,0.1320639967918396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,16,power_law_1.2,0.15701760053634645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,16,power_law_1.2,0.152838397026062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,16,power_law_1.2,0.14758399724960328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,16,power_law_1.2,0.15052160024642944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,16,power_law_1.2,0.1587391972541809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,16,power_law_1.2,0.15866880416870116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,16,power_law_1.2,0.15816320180892945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,16,power_law_1.2,0.16293120384216309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,16,power_law_1.2,0.1770815968513489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,16,power_law_1.2,0.18097920417785646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,16,power_law_1.2,0.19056639671325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,16,power_law_1.2,0.2032320022583008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,16,power_law_1.2,0.21589760780334472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,16,power_law_1.2,0.2543231964111328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,16,power_law_1.2,0.28181118965148927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,16,power_law_1.2,0.3762815952301025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,16,power_law_1.2,0.46229119300842286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,16,power_law_1.2,0.6605504035949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,16,power_law_1.2,0.8004159927368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,16,power_law_1.2,1.1029120445251466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,16,power_law_1.2,1.7473535537719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,64,power_law_1.01,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,64,power_law_1.01,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,64,power_law_1.01,0.0335999995470047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,64,power_law_1.01,0.035155200958251955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,64,power_law_1.01,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,64,power_law_1.01,0.03611519932746887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,64,power_law_1.01,0.03638400137424469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,64,power_law_1.01,0.03638400137424469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,64,power_law_1.01,0.036883199214935304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,64,power_law_1.01,0.03682560026645661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,64,power_law_1.01,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,64,power_law_1.01,0.03895680010318756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,64,power_law_1.01,0.03937920033931732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,64,power_law_1.01,0.039654400944709775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,64,power_law_1.01,0.043731200695037845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,64,power_law_1.01,0.04418559968471527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,64,power_law_1.01,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,64,power_law_1.01,0.04919039905071258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.01,0.05137280225753784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.01,0.062041598558425906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.01,0.06686720252037048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.01,0.08689919710159302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.01,0.10226559638977051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.01,0.13803520202636718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.01,0.16291199922561644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.01,0.23055360317230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.01,0.2793152093887329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,32,balanced,0.029546665648619335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,32,balanced,0.028431999186674755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,32,balanced,0.02755733331044515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,32,balanced,0.02923733244339625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,32,balanced,0.03327466547489166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,32,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,32,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,32,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,32,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,32,balanced,0.03369066615899404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,32,balanced,0.03521066655715307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,32,balanced,0.03393599887688955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,32,balanced,0.03654933224121729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,32,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,32,balanced,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,32,balanced,0.03979733337958654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,32,balanced,0.04010133445262909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,32,balanced,0.04153066625197729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,32,balanced,0.041738669077555336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,32,balanced,0.04785599807898203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,32,balanced,0.04821866750717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,32,balanced,0.056133334835370384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,32,balanced,0.06238399942715963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,32,balanced,0.09128000338872273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,32,balanced,0.11545600493748982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,32,balanced,0.1553759972254435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,32,balanced,0.18408000469207764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,8,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,8,balanced,0.037802666425704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,8,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,8,balanced,0.043680002291997276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,8,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,8,balanced,0.06217066446940104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,8,balanced,0.06462400158246358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,8,balanced,0.06417599817117055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,8,balanced,0.06585066517194112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,8,balanced,0.06413333117961884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,8,balanced,0.06598933537801106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,8,balanced,0.06694399813810985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,8,balanced,0.06726400057474773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,8,balanced,0.06898666421572368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,8,balanced,0.07381866872310638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,8,balanced,0.07453866799672444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,8,balanced,0.07781333227952321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,8,balanced,0.08815466364224751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,8,balanced,0.08877333005269368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,8,balanced,0.10360532999038696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,8,balanced,0.11319999893506368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,8,balanced,0.14402666687965393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,8,balanced,0.1662773291269938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,8,balanced,0.2227840026219686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,8,balanced,0.26576000452041626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,8,balanced,0.35734931627909344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,8,balanced,0.4570133288701375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,8,power_law_1.2,0.0723136007785797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,8,power_law_1.2,0.10001280307769775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,8,power_law_1.2,0.11668479442596436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,8,power_law_1.2,0.1336192011833191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,8,power_law_1.2,0.14984960556030275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,8,power_law_1.2,0.16918400526046753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,8,power_law_1.2,0.1511615991592407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,8,power_law_1.2,0.18302719593048095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,8,power_law_1.2,0.17084800004959105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,8,power_law_1.2,0.16810879707336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,8,power_law_1.2,0.1613759994506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,8,power_law_1.2,0.17701120376586915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,8,power_law_1.2,0.18736000061035157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,8,power_law_1.2,0.18529920578002929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,8,power_law_1.2,0.19481600522994996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,8,power_law_1.2,0.19701759815216063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,8,power_law_1.2,0.20952320098876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,8,power_law_1.2,0.2347327947616577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,8,power_law_1.2,0.25809919834136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,8,power_law_1.2,0.31146879196166993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,8,power_law_1.2,0.3175296068191528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,8,power_law_1.2,0.3935679912567139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,8,power_law_1.2,0.4402495861053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,8,power_law_1.2,0.5619775772094726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,8,power_law_1.2,0.7134143829345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,8,power_law_1.2,1.003763198852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,8,power_law_1.2,1.1593279838562012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,128,balanced,0.017423999806245167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,128,balanced,0.018944000204404194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,128,balanced,0.018895999838908512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,128,balanced,0.021040000021457672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,128,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,128,balanced,0.027274665733178455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,128,balanced,0.02743999908367793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,128,balanced,0.027098665634791057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,128,balanced,0.027552001178264618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,128,balanced,0.027952000498771667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,128,balanced,0.028789333999156952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,128,balanced,0.027461332579453785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,128,balanced,0.029264000554879505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,128,balanced,0.029215998947620392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,128,balanced,0.02978666623433431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,128,balanced,0.03120533376932144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,128,balanced,0.031354665756225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,128,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,128,balanced,0.03530666728814443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,128,balanced,0.03922666609287262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,128,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,128,balanced,0.049029335379600525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,128,balanced,0.05880000193913778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,128,balanced,0.07623466849327087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,128,balanced,0.09403733412424724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,128,balanced,0.1299253304799398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,128,balanced,0.16462399562199911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,4,power_law_1.2,0.03945600092411041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,4,power_law_1.2,0.03828479945659637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,4,power_law_1.2,0.045535999536514285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,4,power_law_1.2,0.051635199785232545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,4,power_law_1.2,0.06759679913520814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,4,power_law_1.2,0.07466239929199218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,4,power_law_1.2,0.07616000175476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,4,power_law_1.2,0.07779840230941773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,4,power_law_1.2,0.078847998380661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,4,power_law_1.2,0.0809664011001587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,4,power_law_1.2,0.0804095983505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,4,power_law_1.2,0.08254079818725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,4,power_law_1.2,0.08754559755325317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,4,power_law_1.2,0.09347839951515198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,4,power_law_1.2,0.10051840543746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,4,power_law_1.2,0.10651520490646363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,4,power_law_1.2,0.10165120363235473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,4,power_law_1.2,0.1137279987335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,4,power_law_1.2,0.13756799697875977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,4,power_law_1.2,0.15325440168380738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,4,power_law_1.2,0.1745151996612549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,4,power_law_1.2,0.2311039924621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,4,power_law_1.2,0.2546880006790161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,4,power_law_1.2,0.3901504039764404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,4,power_law_1.2,0.49955201148986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,8,balanced,0.024826665719350178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,8,balanced,0.025461333493391674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,8,balanced,0.030181333422660828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,8,balanced,0.043525333205858864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,8,balanced,0.056474665800730385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,8,balanced,0.08346666892369588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,8,balanced,0.08736000458399455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,8,balanced,0.08874133229255676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,8,balanced,0.09867200255393982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,8,balanced,0.09516800443331401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,8,balanced,0.09770133097966512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,8,balanced,0.1312320033709208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,8,balanced,0.13167466719945273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,8,balanced,0.13199466466903687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,8,balanced,0.19156799713770548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,8,balanced,0.19545066356658936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,8,balanced,0.20388267437616983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,8,balanced,0.19906665881474814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,8,balanced,0.20665599902470908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,8,balanced,0.22884267568588257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,8,balanced,0.25707733631134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,4,power_law_1.2,0.6474368095397949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,8,balanced,0.3028159936269124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,8,balanced,0.3531680107116699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,8,balanced,0.4392746686935425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,8,balanced,0.5363039970397949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,8,balanced,0.8432426452636719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,8,balanced,1.0404640038808186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,4,power_law_1.2,0.7950463771820069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,8,power_law_1.01,0.07182080149650574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,8,power_law_1.01,0.0843775987625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,8,power_law_1.01,0.0884223997592926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,8,power_law_1.01,0.10080000162124633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,8,power_law_1.01,0.105075204372406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,8,power_law_1.01,0.10528639554977418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,8,power_law_1.01,0.11063679456710815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,8,power_law_1.01,0.10698879957199096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,8,power_law_1.01,0.1133504033088684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,8,power_law_1.01,0.10924160480499268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,8,power_law_1.01,0.10853760242462158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,8,power_law_1.01,0.11350400447845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,8,power_law_1.01,0.11427839994430541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,8,power_law_1.01,0.11592320203781128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,8,power_law_1.01,0.12449920177459717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,8,power_law_1.01,0.12012799978256225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,8,power_law_1.01,0.12762880325317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,8,power_law_1.01,0.13984639644622804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.01,0.15255039930343628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.01,0.1769279956817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.01,0.20171520709991456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.01,0.26606719493865966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.01,0.31347839832305907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.01,0.4169600009918213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.01,0.54519681930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.01,0.7241536140441894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.01,0.939891242980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,16,power_law_1.2,0.09995520114898682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,16,power_law_1.2,0.15051519870758057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,16,power_law_1.2,0.189792001247406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,16,power_law_1.2,0.19120639562606812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,16,power_law_1.2,0.19364479780197144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,16,power_law_1.2,0.19070080518722535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,16,power_law_1.2,0.19009920358657836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,16,power_law_1.2,0.15665279626846312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,16,power_law_1.2,0.17318400144577026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,16,power_law_1.2,0.16839040517807008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,16,power_law_1.2,0.1681663990020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,16,power_law_1.2,0.17407360076904296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,16,power_law_1.2,0.1815168023109436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,16,power_law_1.2,0.19498879909515382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,16,power_law_1.2,0.1942911982536316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,16,power_law_1.2,0.18894720077514648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,16,power_law_1.2,0.2124351978302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,16,power_law_1.2,0.23710079193115235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,16,power_law_1.2,0.26049280166625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,16,power_law_1.2,0.3029887914657593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,16,power_law_1.2,0.30680959224700927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,16,power_law_1.2,0.3830143928527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,16,power_law_1.2,0.4291520118713379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,16,power_law_1.2,0.5523327827453614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,16,power_law_1.2,0.6697408199310303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,16,power_law_1.2,0.9904640197753907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,16,power_law_1.2,1.252787208557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,32,power_law_1.2,0.05025280117988586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,32,power_law_1.2,0.03123840093612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,32,power_law_1.2,0.030713599920272828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,32,power_law_1.2,0.029395198822021483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,32,power_law_1.2,0.030668801069259642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,32,power_law_1.2,0.031481599807739256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,32,power_law_1.2,0.032262399792671204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,32,power_law_1.2,0.03238399922847748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,32,power_law_1.2,0.032307198643684386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,32,power_law_1.2,0.03351680040359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,32,power_law_1.2,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,32,power_law_1.2,0.03406080007553101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,32,power_law_1.2,0.035123199224472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,32,power_law_1.2,0.03770880103111267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,32,power_law_1.2,0.041382399201393125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,32,power_law_1.2,0.04344319999217987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,32,power_law_1.2,0.045561599731445315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,32,power_law_1.2,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,32,power_law_1.2,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,32,power_law_1.2,0.06174719929695129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,32,power_law_1.2,0.06198400259017944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,32,power_law_1.2,0.07654399871826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,32,power_law_1.2,0.09983999729156494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,32,power_law_1.2,0.12428799867630005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,32,power_law_1.2,0.15172480344772338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,32,power_law_1.2,0.2719167947769165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,32,power_law_1.2,0.2522304058074951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,4,power_law_1.2,0.06361600160598754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,4,power_law_1.2,0.06977919936180114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,4,power_law_1.2,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,4,power_law_1.2,0.08563200235366822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,4,power_law_1.2,0.08647040128707886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,4,power_law_1.2,0.08785279989242553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,4,power_law_1.2,0.08814079761505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,4,power_law_1.2,0.08917760252952575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,4,power_law_1.2,0.09033600091934205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,4,power_law_1.2,0.09102720022201538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,4,power_law_1.2,0.09168639779090881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,4,power_law_1.2,0.11319040060043335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,4,power_law_1.2,0.11475199460983276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,4,power_law_1.2,0.11750400066375732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,4,power_law_1.2,0.129203200340271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,4,power_law_1.2,0.13409279584884642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,4,power_law_1.2,0.1848512053489685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,4,power_law_1.2,0.21480960845947267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,4,power_law_1.2,0.23036799430847169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,4,power_law_1.2,0.3495680093765259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,4,power_law_1.2,0.45255680084228517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,64,balanced,0.04956800242265066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,64,balanced,0.037647999823093414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,64,balanced,0.033285332222779594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,64,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,64,balanced,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,64,balanced,0.033999999364217125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,64,balanced,0.033589333295822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,64,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,64,balanced,0.03339733431736628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,64,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,64,balanced,0.03364266703526179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,64,balanced,0.0347680002450943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,64,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,64,balanced,0.03369600077470144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,64,balanced,0.03745600084463755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,64,balanced,0.03749333322048187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,64,balanced,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,64,balanced,0.039674667020638786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,64,balanced,0.037733333806196846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,64,balanced,0.044997334480285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,64,balanced,0.041946664452552795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,64,balanced,0.04971200227737427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,64,balanced,0.04997866849104563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,64,balanced,0.06534400085608165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,64,balanced,0.06615466872851054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,64,balanced,0.0841493308544159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,64,balanced,0.0937600036462148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,4,power_law_1.2,0.6229055881500244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,4,power_law_1.2,0.8105728149414062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,128,balanced,0.05586666862169901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,128,balanced,0.05804799993832906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,128,balanced,0.050144001841545105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,128,balanced,0.05589866638183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,128,balanced,0.05657066901524862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,128,balanced,0.05589866638183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,128,balanced,0.06102933486302694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,128,balanced,0.0562666654586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,128,balanced,0.0581279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,128,balanced,0.0584799995024999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,128,balanced,0.05866666634877523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,128,balanced,0.062405332922935486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,128,balanced,0.06205333272616068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,128,balanced,0.06246933341026306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,128,balanced,0.0681386689345042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,128,balanced,0.06914666791756947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,128,balanced,0.07248533268769582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,128,balanced,0.0928000013033549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,128,balanced,0.09059199690818787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,128,balanced,0.11944533387819926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,128,balanced,0.12731732924779257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,128,balanced,0.16704533497492471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,128,balanced,0.1551199952761332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,128,balanced,0.19588265816370645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,128,balanced,0.2128480076789856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,128,balanced,0.27907200654347736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,128,balanced,0.3315626581509908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,4,power_law_1.2,1.377683162689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,4,balanced,0.052005335688591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,4,balanced,0.07338133454322815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,4,balanced,0.0990559955437978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,4,balanced,0.1486133337020874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,4,balanced,0.24989867210388184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,4,balanced,0.4464053312937419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,4,balanced,0.44812798500061035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,4,balanced,0.4508853356043498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,4,balanced,0.45154666900634766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,4,balanced,0.4535893201828003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,4,balanced,0.45456000169118244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,4,balanced,0.4609280029932658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,4,balanced,0.4640959898630778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,4,balanced,0.46716801325480145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,4,balanced,0.4734826485315959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,4,balanced,0.47596267859141034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,4,balanced,0.4857226610183716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,4,balanced,0.5064640045166016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,4,balanced,0.5268213351567587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,4,balanced,0.5613173246383667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,4,balanced,0.6001813411712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,4,balanced,0.6677866776784261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,4,balanced,0.7300533453623453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,4,balanced,0.9523146947224935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,4,balanced,1.0395893255869548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,4,balanced,1.5607679684956868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,4,balanced,1.7404319445292156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,4,power_law_1.2,1.4560704231262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,4,power_law_1.2,2.0256832122802733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,4,power_law_1.2,3.0709184646606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,16,power_law_1.01,0.16368000507354735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,16,power_law_1.01,0.19056639671325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,16,power_law_1.01,0.176256000995636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,16,power_law_1.01,0.18033280372619628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,16,power_law_1.01,0.17133439779281617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,16,power_law_1.01,0.17772159576416016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,16,power_law_1.01,0.16260479688644408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,16,power_law_1.01,0.17368320226669312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,16,power_law_1.01,0.17920000553131105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,16,power_law_1.01,0.17609599828720093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,16,power_law_1.01,0.1757823944091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,16,power_law_1.01,0.17826559543609619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,16,power_law_1.01,0.1841920018196106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,16,power_law_1.01,0.18922239542007446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,16,power_law_1.01,0.2007551908493042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,16,power_law_1.01,0.21106560230255128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,16,power_law_1.01,0.22945919036865234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,16,power_law_1.01,0.27171199321746825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,16,power_law_1.01,0.28015999794006347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,16,power_law_1.01,0.3272128105163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,16,power_law_1.01,0.35560319423675535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,16,power_law_1.01,0.409657621383667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,16,power_law_1.01,0.4854720115661621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,16,power_law_1.01,0.6573887825012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,16,power_law_1.01,0.8590975761413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,16,power_law_1.01,1.1022591590881348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,16,power_law_1.01,1.4261759757995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,2,balanced,0.061861331264177956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,2,balanced,0.08718400200208028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,2,balanced,0.13634666800498962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,2,balanced,0.13923199971516928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,2,balanced,0.14059199889500937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,2,balanced,0.1414293348789215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,2,balanced,0.14114666978518167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,2,balanced,0.14416533708572388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,2,balanced,0.14588266611099243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,2,balanced,0.1462453305721283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,2,balanced,0.15036267042160034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,2,balanced,0.15033066272735596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,2,balanced,0.15396266182263693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,2,balanced,0.1600160002708435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,2,balanced,0.16988267501195273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,2,balanced,0.17333332697550455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,2,balanced,0.18726933002471924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,2,balanced,0.2182719906171163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,2,balanced,0.2305813431739807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,2,balanced,0.2959466576576233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,2,balanced,0.3453226486841838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,2,balanced,0.4783253272374471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,2,balanced,0.5932426850001017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,2,balanced,0.8596853415171305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,2,balanced,1.0944533348083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,2,balanced,1.626047929128011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,2,balanced,2.131728013356527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,64,power_law_1.2,0.07330560088157653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,64,power_law_1.2,0.07092480063438415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,64,power_law_1.2,0.06975359916687011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,64,power_law_1.2,0.0737280011177063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,64,power_law_1.2,0.07299200296401978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,64,power_law_1.2,0.07604479789733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,64,power_law_1.2,0.07155839800834655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,64,power_law_1.2,0.0769536018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,64,power_law_1.2,0.07656319737434387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,64,power_law_1.2,0.07626879811286927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,64,power_law_1.2,0.08672000169754028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,64,power_law_1.2,0.09086080193519593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,64,power_law_1.2,0.09186559915542603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,64,power_law_1.2,0.08643199801445008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,64,power_law_1.2,0.10160000324249267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,1,balanced,0.1260640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,64,power_law_1.2,0.10167039632797241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,1,balanced,0.19502399365107217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,1,balanced,0.3302133282025655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,64,power_law_1.2,0.10973440408706665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,1,balanced,0.6058346827824911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,64,power_law_1.2,0.1202623963356018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,1,balanced,1.1278026898701985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,64,power_law_1.2,0.12810239791870118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,1,balanced,1.6448747316996257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,64,power_law_1.2,0.15151360034942626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,1,balanced,1.654618740081787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,64,power_law_1.2,0.17182719707489014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,1,balanced,1.6628533999125164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,1,balanced,1.667850653330485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,1,balanced,1.6707359949747722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,1,balanced,1.6823946634928386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,64,power_law_1.2,0.23914880752563478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,1,balanced,1.688421408335368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,64,power_law_1.2,0.28825600147247316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,64,power_law_1.2,0.42741122245788576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,64,power_law_1.2,0.5196864128112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,64,power_law_1.2,0.8186495780944825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,1,balanced,1.7020799318949382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,1,balanced,1.7153973579406738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,64,power_law_1.2,1.1084480285644531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,1,balanced,1.7375466028849285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,1,balanced,1.7502880096435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,1,balanced,1.774602731068929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,1,balanced,2.1253867149353027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,1,balanced,1.905717372894287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,1,balanced,2.4324960708618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,1,balanced,2.1522773106892905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,1,balanced,2.7606932322184243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,1,balanced,2.646378676096598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,1,balanced,4.146634737650554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,1,balanced,4.170901298522949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,1,balanced,6.625338872273763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,1,balanced,6.992064158121745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,4,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,4,balanced,0.03462400039037069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,4,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,4,balanced,0.07025599976380666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,4,balanced,0.10498133301734924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,4,balanced,0.12443199753761292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,4,balanced,0.12568533420562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,4,balanced,0.12568533420562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,4,balanced,0.12706133723258972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,4,balanced,0.12706666191418967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,4,balanced,0.12571733196576437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,4,balanced,0.12615999579429626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,4,balanced,0.12888532876968384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,4,balanced,0.12843733032544455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,4,balanced,0.132341335217158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,4,balanced,0.13429866234461466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,4,balanced,0.1362879971663157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,4,balanced,0.14851733048756918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,4,balanced,0.1432319978872935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,4,balanced,0.16982932885487875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,4,balanced,0.16209600369135538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,4,balanced,0.23080533742904663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,4,balanced,0.2068693240483602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,4,balanced,0.30134934186935425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,4,balanced,0.308405339717865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,4,balanced,0.46291200319925946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,4,balanced,0.45267200469970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,8,power_law_1.01,0.03456639945507049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,8,power_law_1.01,0.04821119904518127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,8,power_law_1.01,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,8,power_law_1.01,0.05414400100708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,8,power_law_1.01,0.052076798677444455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,8,power_law_1.01,0.05859839916229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,8,power_law_1.01,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,8,power_law_1.01,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,8,power_law_1.01,0.06256639957427979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,8,power_law_1.01,0.08810240030288696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,8,power_law_1.01,0.08503680229187012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,8,power_law_1.01,0.07919999957084656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,8,power_law_1.01,0.1147007942199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,8,power_law_1.01,0.11634559631347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,8,power_law_1.01,0.12023040056228637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,8,power_law_1.01,0.13207039833068848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,8,power_law_1.01,0.1259775996208191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,8,power_law_1.01,0.12002559900283813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.01,0.11786880493164062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.01,0.13010560274124144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.01,0.1462847948074341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.01,0.18479360342025758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.01,0.2116544008255005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.01,0.28703999519348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.01,0.3794111967086792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.01,0.5408448219299317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.01,0.666476821899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,1,balanced,0.04790399968624115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,1,balanced,0.04799999793370565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,1,balanced,0.047584002216657005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,1,balanced,0.04960533479849497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,1,balanced,0.05082133412361145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,1,balanced,0.06284800171852112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,1,balanced,0.07832000156243642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,1,balanced,0.06982933481534322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,1,balanced,0.09116266171137492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,1,balanced,0.08197333415349324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,1,balanced,0.10592533151308696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,1,balanced,0.12268799543380737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,1,balanced,0.1369653344154358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,1,balanced,0.10242666800816853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,1,balanced,0.14677866299947104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,1,balanced,0.1269760032494863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,1,balanced,0.16360533237457275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,1,balanced,0.21308799584706625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,1,balanced,0.27686933676401776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,1,balanced,0.3911840120951335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,1,balanced,0.520522673924764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,1,balanced,0.7449920177459717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,1,balanced,0.9879360198974609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,1,balanced,1.4558240572611492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,1,balanced,1.9181067148844402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,1,balanced,2.868197441101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,1,balanced,3.799877484639486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,128,balanced,0.033733333150545754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,128,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,128,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,128,balanced,0.03956266740957896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,128,balanced,0.03764266769091288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,128,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,128,balanced,0.035546667873859406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,128,balanced,0.0349386657277743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,128,balanced,0.03540800015131632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,128,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,128,balanced,0.035349334279696144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,128,balanced,0.03572800010442734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,128,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,128,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,128,balanced,0.04001600046952566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,128,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,128,balanced,0.037802666425704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,128,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,128,balanced,0.041573333243529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,128,balanced,0.043925335009892784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,128,balanced,0.043578664461771645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,128,balanced,0.05318933228651682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,128,balanced,0.052933335304260254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,128,balanced,0.06533866624037425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,128,balanced,0.07227199772993724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,128,balanced,0.0855519970258077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,128,balanced,0.10378133257230122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,2,balanced,0.09854933619499207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,2,balanced,0.10306666294733684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,2,balanced,0.11117866635322571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,2,balanced,0.13165332873662314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,2,balanced,0.15453867117563883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,2,balanced,0.20545599857966104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,2,balanced,0.20639467239379883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,2,balanced,0.20691200097401938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,2,balanced,0.21050133307774863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,2,balanced,0.21006399393081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,2,balanced,0.21411732832590738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,2,balanced,0.21973333756128946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,2,balanced,0.2217973272005717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,2,balanced,0.22847465674082437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,2,balanced,0.23106666405995688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,2,balanced,0.23548799753189087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,2,balanced,0.24633600314458212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,2,balanced,0.27878399689992267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,2,balanced,0.3042293389638265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,2,balanced,0.3678133487701416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,2,balanced,0.42738668123881024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,2,balanced,0.5361013412475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,2,balanced,0.6484213272730509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,2,balanced,0.9588747024536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,2,balanced,1.1935199896494548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,2,balanced,1.740517298380534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,2,balanced,2.273263931274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,1,power_law_1.01,0.07601919770240784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,1,power_law_1.01,0.09909120202064514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,1,power_law_1.01,0.12547839879989625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,1,power_law_1.01,0.1804352045059204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,1,power_law_1.01,0.1918463945388794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,1,power_law_1.01,0.1758784055709839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,1,power_law_1.01,0.18480639457702636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,1,power_law_1.01,0.18058240413665771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,1,power_law_1.01,0.1723199963569641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,1,power_law_1.01,0.17701760530471802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,1,power_law_1.01,0.18107520341873168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,1,power_law_1.01,0.1709439992904663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,1,power_law_1.01,0.19129600524902343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,1,power_law_1.01,0.189683198928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,1,power_law_1.01,0.21048319339752197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,1,power_law_1.01,0.22375040054321288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,1,power_law_1.01,0.21911039352416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,1,power_law_1.01,0.267795205116272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.01,0.31554560661315917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.01,0.38198399543762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.01,0.46106882095336915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.01,0.6048831939697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.01,0.7555136203765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.01,1.0666751861572266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,32,power_law_1.01,0.04631040096282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,32,power_law_1.01,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,32,power_law_1.01,0.0442111998796463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,32,power_law_1.01,0.0430400013923645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,32,power_law_1.01,0.04431360065937042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,32,power_law_1.01,0.04508799910545349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,32,power_law_1.01,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,32,power_law_1.01,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,32,power_law_1.01,0.047225600481033324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,32,power_law_1.01,0.047513601183891294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,32,power_law_1.01,0.05006080269813538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,32,power_law_1.01,0.05324159860610962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,32,power_law_1.01,0.05459840297698974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,32,power_law_1.01,0.06072319746017456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,32,power_law_1.01,0.06297600269317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,32,power_law_1.01,0.06485120058059693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,32,power_law_1.01,0.06234239935874939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,32,power_law_1.01,0.06933119893074036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,32,power_law_1.01,0.07670400142669678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,32,power_law_1.01,0.08896639943122864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,32,power_law_1.01,0.09756799936294555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,32,power_law_1.01,0.1201024055480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,32,power_law_1.01,0.14264320135116576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,32,power_law_1.01,0.18592640161514282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,32,power_law_1.01,0.22335360050201417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,32,power_law_1.01,0.31882240772247317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,32,power_law_1.01,0.4021440029144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.01,1.3782464027404786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.01,1.9942592620849608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.01,2.6154239654541014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,4,power_law_1.2,0.07049599885940552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,4,power_law_1.2,0.09125120043754578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,4,power_law_1.2,0.09347839951515198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,4,power_law_1.2,0.13186559677124024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,4,power_law_1.2,0.15315840244293213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,4,power_law_1.2,0.15601919889450072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,4,power_law_1.2,0.19959039688110353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,4,power_law_1.2,0.21091198921203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,4,power_law_1.2,0.22385919094085693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,4,power_law_1.2,0.21262080669403077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,4,power_law_1.2,0.22135679721832274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,4,power_law_1.2,0.22741119861602782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,4,power_law_1.2,0.2324671983718872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,4,power_law_1.2,0.23306879997253419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,4,power_law_1.2,0.2484544038772583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,4,power_law_1.2,0.25743999481201174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,4,power_law_1.2,0.26917119026184083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,4,power_law_1.2,0.3040704011917114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,4,power_law_1.2,0.31290240287780763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,4,power_law_1.2,0.3851072072982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,4,power_law_1.2,0.432422399520874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,4,power_law_1.2,0.5267712116241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,4,power_law_1.2,0.5841087818145752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,4,power_law_1.2,0.8130368232727051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,4,power_law_1.2,0.8964287757873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,4,power_law_1.2,1.2620736122131349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,4,power_law_1.2,1.7234304428100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,4,power_law_1.2,0.060575997829437254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,4,power_law_1.2,0.04652160108089447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,4,power_law_1.2,0.05568000078201294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,4,power_law_1.2,0.057574397325515746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,4,power_law_1.2,0.05725439786911011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,4,power_law_1.2,0.061849600076675414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,4,power_law_1.2,0.06471040248870849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,4,power_law_1.2,0.0650111973285675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,4,power_law_1.2,0.06578559875488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,4,power_law_1.2,0.06581760048866273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,4,power_law_1.2,0.07032960057258605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,4,power_law_1.2,0.06869760155677795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,4,power_law_1.2,0.07090560197830201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,4,power_law_1.2,0.07808640003204345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,4,power_law_1.2,0.08093439936637878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,4,power_law_1.2,0.08398079872131348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,1,power_law_1.2,0.13265279531478882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,4,power_law_1.2,0.09621760249137878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,1,power_law_1.2,0.18922239542007446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,1,power_law_1.2,0.3237567901611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,4,power_law_1.2,0.12223999500274658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,1,power_law_1.2,0.5083712100982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,4,power_law_1.2,0.14354560375213624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,1,power_law_1.2,0.6666687965393067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,4,power_law_1.2,0.15630719661712647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,1,power_law_1.2,1.2064127922058105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,1,power_law_1.2,1.3147199630737305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,1,power_law_1.2,1.3672703742980956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,1,power_law_1.2,1.4088895797729493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,1,power_law_1.2,1.4430015563964844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,1,power_law_1.2,1.483891201019287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,1,power_law_1.2,1.5303615570068358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,4,power_law_1.2,0.1813696026802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,1,power_law_1.2,1.584992027282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,1,power_law_1.2,1.6519807815551757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,1,power_law_1.2,1.729689598083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,1,power_law_1.2,1.7570240020751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,4,power_law_1.2,0.23816320896148682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,1,power_law_1.2,1.8595327377319335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,1,power_law_1.2,2.081395149230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,1,power_law_1.2,2.034227180480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,1,power_law_1.2,2.316691207885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,1,power_law_1.2,2.3535423278808594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,1,power_law_1.2,2.847782325744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,1,power_law_1.2,3.188268852233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,1,power_law_1.2,3.897235107421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,4,power_law_1.2,0.3026560068130493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,1,power_law_1.2,4.772441482543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,1,power_law_1.2,6.457465362548828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,8,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,8,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,8,balanced,0.04774933556715647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,8,balanced,0.053786665201187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,8,balanced,0.07196799914042155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,8,balanced,0.10251733660697937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,8,balanced,0.1334773302078247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,8,balanced,0.13020267089207968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,8,balanced,0.12999467055002847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,8,balanced,0.13297067085901895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,8,balanced,0.13356266419092813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,8,balanced,0.1363200048605601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,8,balanced,0.1367093324661255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,8,balanced,0.13945600390434265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,8,balanced,0.14385066429773966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,8,balanced,0.14594133694966635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,8,balanced,0.15624533096949259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,8,balanced,0.16400532921155295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,8,balanced,0.17416000366210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,8,balanced,0.19405867656071982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,8,balanced,0.21462933222452799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,8,balanced,0.2537920077641805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,4,power_law_1.2,0.4331007957458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,8,balanced,0.30400532484054565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,8,balanced,0.38355199495951336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,8,balanced,0.47623999913533527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,8,balanced,0.63700799147288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,1,power_law_1.2,7.723244476318359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,8,balanced,0.8043519655863444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,4,power_law_1.2,0.5431104183197022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,32,power_law_1.01,0.06367999911308289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,32,power_law_1.01,0.06157439947128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,32,power_law_1.01,0.06017919778823853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,32,power_law_1.01,0.06458240151405334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,32,power_law_1.01,0.06386560201644897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,32,power_law_1.01,0.06356480121612548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,32,power_law_1.01,0.06652799844741822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,32,power_law_1.01,0.06851199865341187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,32,power_law_1.01,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,32,power_law_1.01,0.0671999990940094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,32,power_law_1.01,0.0698751986026764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,32,power_law_1.01,0.06969599723815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,32,power_law_1.01,0.07182080149650574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,32,power_law_1.01,0.07271040081977845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,32,power_law_1.01,0.07907840013504028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,32,power_law_1.01,0.07922559976577759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,32,power_law_1.01,0.08193920254707336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,32,power_law_1.01,0.0918720006942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,32,power_law_1.01,0.1024448037147522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,32,power_law_1.01,0.12285439968109131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,32,power_law_1.01,0.13801599740982057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,32,power_law_1.01,0.18268799781799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,32,power_law_1.01,0.2065216064453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,32,power_law_1.01,0.278873610496521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,32,power_law_1.01,0.3638079881668091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,32,power_law_1.01,0.542137622833252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,32,power_law_1.01,0.6770815849304199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,4,power_law_1.2,0.6900224208831787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,4,power_law_1.2,0.9138624191284179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,32,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,32,balanced,0.050026665131251015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,32,balanced,0.05022400120894114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,32,balanced,0.05641599992911021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,32,balanced,0.05813866853713989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,32,balanced,0.06237866481145223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,32,balanced,0.062080000837643944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,32,balanced,0.06428266565004985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,32,balanced,0.06460266808668773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,32,balanced,0.0662720004717509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,32,balanced,0.06252799928188324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,32,balanced,0.06603200236956279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,32,balanced,0.06851199766000111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,32,balanced,0.0654666672150294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,32,balanced,0.07043733199437459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,32,balanced,0.07044266661008199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,32,balanced,0.07241066793600719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,32,balanced,0.07853866616884868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,32,balanced,0.08342933654785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,32,balanced,0.0942186713218689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,32,balanced,0.10366400082906087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,32,balanced,0.13781332969665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,32,balanced,0.157151997089386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,32,balanced,0.20384534200032553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,32,balanced,0.2527573307355245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,32,balanced,0.33697601159413654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,32,balanced,0.43148799737294513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,1,power_law_1.01,0.16060160398483275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,1,power_law_1.01,0.26347520351409914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,1,power_law_1.01,0.4713280200958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,1,power_law_1.01,0.7454207897186279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,1,power_law_1.01,1.0028800010681151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,1,power_law_1.01,1.652716827392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,1,power_law_1.01,1.740729522705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,1,power_law_1.01,1.8475839614868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,1,power_law_1.01,1.8586111068725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,1,power_law_1.01,1.94781436920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,1,power_law_1.01,1.9763839721679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,1,power_law_1.01,2.0286592483520507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,1,power_law_1.01,2.0952127456665037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,1,power_law_1.01,2.0706623077392576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,1,power_law_1.01,2.1692928314208983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,1,power_law_1.01,2.2196735382080077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,1,power_law_1.01,2.292441558837891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,1,power_law_1.01,2.4521728515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,1,power_law_1.01,2.663929557800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,1,power_law_1.01,3.134342384338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,1,power_law_1.01,2.9779584884643553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,1,power_law_1.01,3.4007423400878904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,1,power_law_1.01,3.955219268798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,1,power_law_1.01,4.859545516967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,1,power_law_1.01,6.3471935272216795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,1,power_law_1.01,8.090188598632812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,1,power_law_1.01,10.210316467285157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,16,power_law_1.2,0.07082239985466003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,16,power_law_1.2,0.10020480155944825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,16,power_law_1.2,0.1176576018333435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,16,power_law_1.2,0.12428799867630005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,16,power_law_1.2,0.12472319602966309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,16,power_law_1.2,0.12027519941329956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,16,power_law_1.2,0.11117440462112427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,16,power_law_1.2,0.11855360269546508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,16,power_law_1.2,0.12353919744491577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,16,power_law_1.2,0.11919360160827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,16,power_law_1.2,0.11867519617080688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,16,power_law_1.2,0.12074879407882691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,16,power_law_1.2,0.12353919744491577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,16,power_law_1.2,0.12762880325317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,16,power_law_1.2,0.13816319704055785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,16,power_law_1.2,0.13702399730682374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,16,power_law_1.2,0.14313600063323975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,16,power_law_1.2,0.17467520236968995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,16,power_law_1.2,0.1814911961555481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,16,power_law_1.2,0.22603518962860109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,16,power_law_1.2,0.2577791929244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,16,power_law_1.2,0.3173504114151001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,16,power_law_1.2,0.3795135974884033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,16,power_law_1.2,0.5530240058898925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,16,power_law_1.2,0.6573056221008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,16,power_law_1.2,0.9464447975158692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,16,power_law_1.2,1.216038417816162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,8,balanced,0.0353973334034284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,8,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,8,balanced,0.0331839993596077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,8,balanced,0.03358400116364161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,8,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,8,balanced,0.0397119993964831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,8,balanced,0.06986666719118755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,8,balanced,0.07021866738796234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,8,balanced,0.0729013333717982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,8,balanced,0.06840533514817555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,8,balanced,0.060133333007494606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,8,balanced,0.09644800424575806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,8,balanced,0.0918880005677541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,8,balanced,0.07558933397134145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,8,balanced,0.10829866925875346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,8,balanced,0.08540800213813782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,8,balanced,0.09987200299898784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,8,balanced,0.14074666301409403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,8,balanced,0.16538666685422262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,8,balanced,0.23161600033442178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,8,balanced,0.2961120009422302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,8,balanced,0.42213865121205646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,8,balanced,0.5318719943364462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,8,balanced,0.7866453329722086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,8,balanced,1.0228479703267415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,8,balanced,1.5133066177368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,8,balanced,2.004997412363688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,4,balanced,0.04390933116277059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,4,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,4,balanced,0.07201066613197327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,4,balanced,0.11317867040634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,4,balanced,0.1750826636950175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,4,balanced,0.30002667506535846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,4,balanced,0.30266666412353516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,4,balanced,0.3033439914385478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,4,balanced,0.30390934149424237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,4,balanced,0.30505067110061646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,4,balanced,0.3043786684672038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,4,balanced,0.30644800265630084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,4,balanced,0.30931733051935834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,4,balanced,0.31252266963322956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,4,balanced,0.31648000081380206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,4,balanced,0.3200053373972575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,4,balanced,0.3233013351758321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,4,balanced,0.33269333839416504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,4,balanced,0.34355731805165607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,4,balanced,0.364303986231486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,4,balanced,0.3924266497294108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,4,balanced,0.42875198523203534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,4,balanced,0.4594186544418335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,4,balanced,0.5748853286107382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,4,balanced,0.6195146640141805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,4,balanced,0.8758506774902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,4,balanced,0.9898560047149658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,2,power_law_1.2,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,2,power_law_1.2,0.051801598072052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,2,power_law_1.2,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,2,power_law_1.2,0.06774399876594543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,2,power_law_1.2,0.08140159845352173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,2,power_law_1.2,0.09066240191459655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,2,power_law_1.2,0.11310080289840699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,2,power_law_1.2,0.11167999505996704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,2,power_law_1.2,0.11418240070343018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,2,balanced,0.01907733331123988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,2,balanced,0.019296000401178997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,2,balanced,0.019018666197856266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,2,balanced,0.020026666422684986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,2,balanced,0.02201066662867864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,2,balanced,0.023189333577950794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,2,balanced,0.02311466634273529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,2,balanced,0.023546665906906128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,2,balanced,0.025008000433444977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,2,balanced,0.02515733242034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,2,balanced,0.025392000873883564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,2,balanced,0.0436106671889623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,2,balanced,0.043578664461771645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,2,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,2,balanced,0.03700266778469086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,2,balanced,0.035002666215101876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,2,balanced,0.0359199990828832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,2,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,2,balanced,0.04916800061861674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,2,balanced,0.06464000046253204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,2,balanced,0.07771733403205872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,2,balanced,0.09912533561388652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,2,balanced,0.12302933136622111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,2,balanced,0.1662773291269938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,2,balanced,0.21055465936660767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,2,balanced,0.2964266737302144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,2,balanced,0.38308266798655194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,2,power_law_1.2,0.1136896014213562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,2,power_law_1.2,0.11779199838638306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,2,power_law_1.2,0.4056704044342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,2,power_law_1.2,0.12919679880142212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,2,power_law_1.2,0.1366719961166382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,2,power_law_1.2,0.14645760059356688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,2,power_law_1.2,0.14847359657287598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,2,power_law_1.2,0.16291840076446534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,2,power_law_1.2,0.1832576036453247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,2,power_law_1.2,0.20396161079406738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,2,power_law_1.2,0.2535936117172241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,2,power_law_1.2,0.27940480709075927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,2,power_law_1.2,0.3564543962478638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,2,power_law_1.2,0.4412735939025879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,2,power_law_1.2,0.5655807971954345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,2,power_law_1.2,0.7850687980651856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,2,power_law_1.2,1.0674304008483886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,2,power_law_1.2,1.4857024192810058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,2,balanced,0.0273333340883255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,2,balanced,0.027221334477265675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,2,balanced,0.02518400053183238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,2,balanced,0.02749866743882497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,2,balanced,0.02757866680622101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,2,balanced,0.03178133318821589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,2,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,2,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,2,balanced,0.05264533559481303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,2,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,2,balanced,0.04785599807898203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,2,balanced,0.08482666810353597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,2,balanced,0.07893333335717519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,2,balanced,0.06222933530807495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,2,balanced,0.09452799956003825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,2,balanced,0.07427733143170674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,2,balanced,0.09701866904894511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,2,balanced,0.11969600121180217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,2,balanced,0.14520532886187235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,2,balanced,0.20295999447504678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,2,balanced,0.2604479988416036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,2,balanced,0.3619413375854492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,2,balanced,0.4706559975941976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,2,balanced,0.6789759794871012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,2,balanced,0.8856426874796549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,2,balanced,1.3070826530456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,2,balanced,1.7262239456176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,1,balanced,0.056741332014401756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,1,balanced,0.05997333427270254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,1,balanced,0.06826133529345195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,1,balanced,0.09808533390363057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,1,balanced,0.14549332857131958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,1,balanced,0.22124266624450684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,1,balanced,0.22486400604248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,1,balanced,0.22750933965047201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,1,balanced,0.23019200563430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,1,balanced,0.2304533322652181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,1,balanced,0.2374080022176107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,1,balanced,0.24106667439142862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,1,balanced,0.24705066283543906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,1,balanced,0.25566933552424115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,1,balanced,0.26529600222905475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,1,balanced,0.2730453411738078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,1,balanced,0.2882240017255147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,1,balanced,0.33242666721343994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,1,balanced,0.3700000047683716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,1,balanced,0.4461119969685872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,1,balanced,0.5149973233540853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,1,balanced,0.7299520174662272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,1,balanced,0.8465119997660319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,1,balanced,1.2558293342590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,1,balanced,1.559765338897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,1,balanced,2.2749706904093423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,1,balanced,2.976149241129557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,32,balanced,0.027029333015282948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,32,balanced,0.02314666658639908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,32,balanced,0.023007998863856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,32,balanced,0.02573866645495097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,32,balanced,0.02513599892457326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,32,balanced,0.026378666361172993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,32,balanced,0.02604266752799352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,32,balanced,0.02672533442576726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,32,balanced,0.02712533374627431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,32,balanced,0.025407999753952026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,32,balanced,0.028560000161329906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,32,balanced,0.02914133419593175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,32,balanced,0.029045333464940388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,32,balanced,0.03124266614516576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,32,balanced,0.033999999364217125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,32,balanced,0.0348693331082662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,32,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,32,balanced,0.03737599899371465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,32,balanced,0.03751466671625773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,32,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,32,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,32,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,32,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,32,balanced,0.05305066704750061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,32,balanced,0.06017066538333893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,32,balanced,0.07423466444015503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,32,balanced,0.09319466352462769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,2,balanced,0.062074666221936546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,2,balanced,0.08064533273379008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,2,balanced,0.1367093324661255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,2,balanced,0.13759467005729675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,2,balanced,0.13704533378283182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,2,balanced,0.13800000150998434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,2,balanced,0.13860799868901572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,2,balanced,0.1388746698697408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,2,balanced,0.13959999879201254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,2,balanced,0.14019200205802917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,2,balanced,0.14375999569892883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,2,balanced,0.14191466569900513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,2,balanced,0.14350400368372598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,2,balanced,0.15212266643842062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,2,balanced,0.16739734013875326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,2,balanced,0.16741333405176798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,2,balanced,0.1856106718381246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,2,balanced,0.20661866664886475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,2,balanced,0.21348265806833902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,2,balanced,0.2621333400408427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,2,balanced,0.3050453265508016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,2,balanced,0.40587735176086426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,2,balanced,0.49025599161783856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,2,balanced,0.692624012629191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,2,balanced,0.8713493347167969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,2,balanced,1.2832427024841309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,2,balanced,1.6722559928894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,2,power_law_1.01,0.03308799862861633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,2,power_law_1.01,0.03364480137825012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,2,power_law_1.01,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,2,power_law_1.01,0.04336639940738678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,2,power_law_1.01,0.04798080027103424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,2,power_law_1.01,0.06843519806861878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,2,power_law_1.01,0.06822400093078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,2,power_law_1.01,0.070169597864151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,2,power_law_1.01,0.07017599940299987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,2,power_law_1.01,0.07163519859313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,2,power_law_1.01,0.07262719869613647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,2,power_law_1.01,0.07505919933319091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,2,power_law_1.01,0.07782400250434876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,2,power_law_1.01,0.08675199747085571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,2,power_law_1.01,0.09463040232658386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,2,power_law_1.01,0.09957759976387023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,2,power_law_1.01,0.09531520009040832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,2,power_law_1.01,0.10892800092697144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,2,power_law_1.01,0.12654080390930175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,2,power_law_1.01,0.1476415991783142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,2,power_law_1.01,0.1870527982711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,2,power_law_1.01,0.25351040363311766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,2,power_law_1.01,0.3113728046417236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,2,power_law_1.01,0.4269887924194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,2,power_law_1.01,0.5304384231567383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,2,power_law_1.01,0.7604415893554688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,2,power_law_1.01,0.9276351928710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,16,power_law_1.01,0.07301759719848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,16,power_law_1.01,0.11292159557342529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,16,power_law_1.01,0.10863360166549682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,16,power_law_1.01,0.11226240396499634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,16,power_law_1.01,0.11301120519638061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,16,power_law_1.01,0.10369919538497925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,16,power_law_1.01,0.10561280250549317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,16,power_law_1.01,0.10743680000305175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,16,power_law_1.01,0.10590720176696777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,16,power_law_1.01,0.10721919536590577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,16,power_law_1.01,0.10658559799194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,16,power_law_1.01,0.11228799819946289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,16,power_law_1.01,0.10718079805374145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,16,power_law_1.01,0.10909440517425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,16,power_law_1.01,0.11064319610595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,16,power_law_1.01,0.1184831976890564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,16,power_law_1.01,0.11910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,16,power_law_1.01,0.1411072015762329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,16,power_law_1.01,0.14139519929885863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,16,power_law_1.01,0.1581120014190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,16,power_law_1.01,0.16889599561691285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,16,power_law_1.01,0.2148224115371704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,16,power_law_1.01,0.2257472038269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,16,power_law_1.01,0.2969599962234497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,16,power_law_1.01,0.36932480335235596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,16,power_law_1.01,0.4923391819000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,16,power_law_1.01,0.6205952167510986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,2,power_law_1.2,0.05316479802131653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,2,power_law_1.2,0.07211520075798035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,2,power_law_1.2,0.08165119886398316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,2,power_law_1.2,0.10121599435806275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,2,power_law_1.2,0.12378239631652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,2,power_law_1.2,0.1417407989501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,2,power_law_1.2,0.1889407992362976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,2,power_law_1.2,0.20172159671783446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,2,power_law_1.2,0.2015552043914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,2,power_law_1.2,0.20582399368286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,2,power_law_1.2,0.2139456033706665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,2,power_law_1.2,0.21857280731201173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,2,power_law_1.2,0.23095040321350097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,2,power_law_1.2,0.23155839443206788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,2,power_law_1.2,0.24671359062194825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,2,power_law_1.2,0.25822720527648924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,2,power_law_1.2,0.2785856008529663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,2,power_law_1.2,0.31184000968933107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,2,power_law_1.2,0.34259839057922364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,2,power_law_1.2,0.39616639614105226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,2,power_law_1.2,0.45024638175964354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,2,power_law_1.2,0.5705088138580322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,2,power_law_1.2,0.7064832210540771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,2,power_law_1.2,0.9398336410522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,2,power_law_1.2,1.1296704292297364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,2,power_law_1.2,1.6536640167236327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,2,power_law_1.2,2.0664384841918944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,1,power_law_1.01,0.04101119935512543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,1,power_law_1.01,0.04480000138282776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,1,power_law_1.01,0.045875200629234315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,1,power_law_1.01,0.056364798545837404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,1,power_law_1.01,0.05680000185966492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,1,power_law_1.01,0.05390080213546753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,1,power_law_1.01,0.05756160020828247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,1,power_law_1.01,0.05546879768371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,1,power_law_1.01,0.05634559988975525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,1,power_law_1.01,0.05500800013542175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,1,power_law_1.01,0.055827200412750244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,1,power_law_1.01,0.05795199871063232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,1,power_law_1.01,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,1,power_law_1.01,0.06043519973754883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,1,power_law_1.01,0.06899840235710145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,1,power_law_1.01,0.07055360078811646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,1,power_law_1.01,0.07546240091323853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,1,power_law_1.01,0.08781440258026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.01,0.10045440196990967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.01,0.12228480577468873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.01,0.14264320135116576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.01,0.18735359907150267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,1,power_law_1.01,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,1,power_law_1.01,0.05602560043334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,1,power_law_1.01,0.06669440269470214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,1,power_law_1.01,0.07760639786720276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,1,power_law_1.01,0.08854399919509888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,1,power_law_1.01,0.11679999828338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,1,power_law_1.01,0.12201600074768067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,1,power_law_1.01,0.12780159711837769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,1,power_law_1.01,0.12943359613418579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,1,power_law_1.01,0.12837120294570922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,1,power_law_1.01,0.13568639755249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.01,0.23278720378875734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,1,power_law_1.01,0.136953604221344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,1,power_law_1.01,0.1403264045715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,1,power_law_1.01,0.14678399562835692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,1,power_law_1.01,0.154310405254364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,1,power_law_1.01,0.16085120439529418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,1,power_law_1.01,0.17703039646148683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,1,power_law_1.01,0.20821759700775147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.01,0.23791999816894532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.01,0.30601599216461184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.01,0.3696192026138306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.01,0.5059904098510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.01,0.6394944190979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.01,0.9094143867492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.01,0.32235519886016845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.01,1.1758655548095702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.01,1.698643112182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.01,2.2155071258544923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.01,0.41677441596984866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.01,0.5887423992156983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.01,0.7786752223968506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,1,power_law_1.2,0.11070719957351685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,1,power_law_1.2,0.16056959629058837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,1,power_law_1.2,0.25032958984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,1,power_law_1.2,0.44916481971740724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,1,power_law_1.2,0.681056022644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,1,power_law_1.2,0.022201600670814513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,1,power_law_1.2,0.02172800004482269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,1,power_law_1.2,0.9849344253540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,1,power_law_1.2,0.021753600239753722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,1,power_law_1.2,0.021881599724292756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,1,power_law_1.2,0.023263999819755556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,1,power_law_1.2,1.1441984176635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,1,power_law_1.2,1.5504768371582032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,1,power_law_1.2,1.6667264938354491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,1,power_law_1.2,0.025651198625564576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,1,power_law_1.2,0.02892799973487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,1,power_law_1.2,0.030956798791885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,1,power_law_1.2,1.6793535232543946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,1,power_law_1.2,0.04007680118083954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,1,power_law_1.2,0.034297600388526917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,1,power_law_1.2,0.03660799860954285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,1,power_law_1.2,1.8439104080200195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,1,power_law_1.2,0.04200319945812225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,1,power_law_1.2,0.045151999592781066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,1,power_law_1.2,0.047628799080848695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,1,power_law_1.2,1.8446144104003905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,1,power_law_1.2,0.06444159746170045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,1,power_law_1.2,0.05704960227012634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,1,power_law_1.2,1.9062271118164062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,1,power_law_1.2,0.06286720037460328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,1,power_law_1.2,2.0129983901977537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,1,power_law_1.2,0.08312960267066956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,1,power_law_1.2,2.032115173339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,1,power_law_1.2,2.1029888153076173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,1,power_law_1.2,2.150457572937012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.2,0.09896320104598999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,1,power_law_1.2,2.3617536544799806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,1,power_law_1.2,2.562803268432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,1,power_law_1.2,2.9259647369384765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,1,power_law_1.2,2.955788803100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.2,0.13829120397567748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,1,power_law_1.2,3.5302078247070314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,1,power_law_1.2,3.5088321685791017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,1,power_law_1.2,4.1846977233886715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,1,power_law_1.2,4.948998260498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.2,0.17214080095291137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,1,power_law_1.2,6.189094543457031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,1,power_law_1.2,8.410707092285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.2,0.2389631986618042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.2,0.29991679191589354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.2,0.4366015911102295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.2,0.5644544124603271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,32,power_law_1.01,0.017875200510025023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,32,power_law_1.01,0.017548799514770508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,32,power_law_1.01,0.017375999689102174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,32,power_law_1.01,0.01744000017642975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,32,power_law_1.01,0.0177279993891716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,32,power_law_1.01,0.018694399297237395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,32,power_law_1.01,0.022950400412082673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,32,power_law_1.01,0.02287999987602234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,32,power_law_1.01,0.03646079897880554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,32,power_law_1.01,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,32,power_law_1.01,0.0365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,32,power_law_1.01,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,32,power_law_1.01,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,32,power_law_1.01,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,32,power_law_1.01,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,32,power_law_1.01,0.03402239978313446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,32,power_law_1.01,0.03521279990673065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,32,power_law_1.01,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,32,power_law_1.01,0.03924480080604553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,1,balanced,0.022346665461858112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,32,power_law_1.01,0.04385280013084412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,1,balanced,0.021685334543387096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,1,balanced,0.023872000475724537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,1,balanced,0.029296000798543293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,1,balanced,0.040037333965301514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,1,balanced,0.06942933301130931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,1,balanced,0.07126933336257935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,1,balanced,0.0730453332265218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,1,balanced,0.0745119998852412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,1,balanced,0.07613866527875264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,1,balanced,0.08065600196520488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,1,balanced,0.08346133430798848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,1,balanced,0.08508267005284627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,1,balanced,0.08881066242853801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,1,balanced,0.09127466877301534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,1,balanced,0.09475732843081157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,1,balanced,0.09947733084360759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,1,balanced,0.11687999963760376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,1,balanced,0.12732799847920737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,1,balanced,0.16386133432388306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,32,power_law_1.01,0.04747520089149475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,1,balanced,0.18689066171646118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,1,balanced,0.25122666358947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,32,power_law_1.01,0.0580672025680542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,1,balanced,0.28751466671625775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,32,power_law_1.01,0.06793599724769592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,1,balanced,0.368010679880778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,32,power_law_1.01,0.09190400242805481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,32,power_law_1.01,0.11060479879379273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,32,power_law_1.01,0.15748480558395386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,32,power_law_1.01,0.189683198928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,1,balanced,0.4695306619008382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,1,balanced,0.7441226641337076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,1,balanced,0.8831146558125814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,1,balanced,0.03405333310365677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,1,balanced,0.04595200220743815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,1,balanced,0.060229331254959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,1,balanced,0.08846933643023173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,1,balanced,0.14838932951291403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,1,balanced,0.1848479906717936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,1,balanced,0.19087467590967813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,1,balanced,0.2002346714337667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,1,balanced,0.20749332507451376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,1,balanced,0.21618133783340454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.2,0.829139232635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,1,balanced,0.23354132970174155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,1,balanced,0.24978133042653403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,1,balanced,0.22208533684412637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,1,balanced,0.23869333664576212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,1,balanced,0.25439467032750446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,1,balanced,0.2360480030377706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,1,balanced,0.2531946698824565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,1,balanced,0.28171734015146893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,1,balanced,0.28249067068099976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,1,balanced,0.58787735303243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,1,balanced,0.6567519903182983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,1,balanced,0.48413864771525067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,1,balanced,0.48901331424713135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,1,balanced,0.8215200106302897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,1,balanced,0.8833066622416178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,1,balanced,1.289141337076823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,1,balanced,1.6747199694315593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.2,1.0908608436584473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,1,power_law_1.01,0.14985599517822265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,1,power_law_1.01,0.246726393699646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,1,power_law_1.01,0.43539838790893554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,1,power_law_1.01,0.8342720031738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,1,power_law_1.01,1.2713855743408202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,1,power_law_1.01,1.5892224311828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,1,power_law_1.01,2.3261632919311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,1,power_law_1.01,2.4126720428466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,1,power_law_1.01,2.497433662414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,1,power_law_1.01,2.5705408096313476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,1,power_law_1.01,2.634854316711426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,1,power_law_1.01,2.762188720703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,1,balanced,0.0562720000743866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,1,power_law_1.01,2.7822336196899413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,1,power_law_1.01,2.850399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,1,balanced,0.06039466460545858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,1,balanced,0.07779733339945476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,1,balanced,0.11448533336321513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,1,balanced,0.17557867368062338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,1,balanced,0.2948906620343526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,1,balanced,0.30426132678985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,1,balanced,0.3013813296953837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,1,balanced,0.3029866615931193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,1,balanced,0.3070240020751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,1,balanced,0.30713599920272827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,1,balanced,0.3173706730206807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,1,balanced,0.3172960082689921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,1,balanced,0.32766934235890705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,1,balanced,0.33371198177337646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,1,balanced,0.3429653247197469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,1,balanced,0.35865068435668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,1,balanced,0.40059200922648114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,1,balanced,0.43293333053588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,1,balanced,0.5100746552149454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,1,balanced,0.5716853141784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,1,balanced,0.7796479860941569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,1,balanced,0.8922399679819742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,1,power_law_1.01,2.899411201477051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,1,balanced,1.281440019607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,1,balanced,1.5679252942403157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,1,power_law_1.01,2.971308708190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,1,balanced,2.2633546193440757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,1,power_law_1.01,3.178771209716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,1,balanced,2.921999931335449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,1,power_law_1.01,3.3648193359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,1,power_law_1.01,3.491775894165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,1,power_law_1.01,3.937433624267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,1,power_law_1.01,3.6881664276123045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,1,power_law_1.01,4.189030456542969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,1,power_law_1.01,4.80126724243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,1,power_law_1.01,5.698259353637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,1,power_law_1.01,6.608633422851563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,1,power_law_1.01,8.88954849243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,1,power_law_1.01,10.602438354492188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,2,balanced,0.05641066531340281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,2,balanced,0.06656533479690552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,2,balanced,0.08412266770998637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,2,balanced,0.08886933326721191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,2,balanced,0.09054932991663615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,2,balanced,0.08898666501045227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,2,balanced,0.09066133697827657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,2,balanced,0.09180266658465068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,2,balanced,0.09066667159398396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,2,balanced,0.0909546713034312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,2,balanced,0.09102400143941243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,2,balanced,0.09325866897900899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,2,balanced,0.09341866771380107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,2,balanced,0.09486933549245198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,2,balanced,0.10133866469065349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,2,balanced,0.1032480001449585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,2,balanced,0.1076853374640147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,2,balanced,0.1351626714070638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,2,balanced,0.14079999923706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,2,balanced,0.18080000082651773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,2,balanced,0.22035733858744302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,2,balanced,0.2955626646677653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,2,balanced,0.37841065724690753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,2,balanced,0.5449066559473673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,2,balanced,0.7072479724884033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,2,balanced,1.0341653029123943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,2,balanced,1.3527679443359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,1,power_law_1.2,0.08357759714126586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,1,power_law_1.2,0.09639040231704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,1,power_law_1.2,0.12910720109939575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,1,power_law_1.2,0.170796799659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,1,power_law_1.2,0.22118399143218995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,1,power_law_1.2,0.3541248083114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,1,power_law_1.2,0.3840127944946289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,1,power_law_1.2,0.3992255926132202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,1,power_law_1.2,0.4167935848236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,1,power_law_1.2,0.42344961166381834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,1,power_law_1.2,0.43740158081054686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,1,power_law_1.2,0.4602816104888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,1,power_law_1.2,0.46886401176452636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,1,power_law_1.2,0.493503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,1,power_law_1.2,0.5206143856048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,1,power_law_1.2,0.5361087799072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,1,power_law_1.2,0.5758016109466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,1,power_law_1.2,0.6539391994476318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,1,power_law_1.2,0.7325376033782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,1,power_law_1.2,0.9057727813720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,1,power_law_1.2,1.1286463737487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,1,power_law_1.2,1.4383744239807128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,1,power_law_1.2,1.7470848083496093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,1,power_law_1.2,2.3595392227172853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,1,power_law_1.2,2.7948991775512697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,1,power_law_1.2,3.936569595336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,1,power_law_1.2,5.274867248535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,1,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,1,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,1,balanced,0.0414986660083135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,1,balanced,0.06151466568311056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,1,balanced,0.08982933561007182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,1,balanced,0.08776533603668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,1,balanced,0.09021332859992981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,1,balanced,0.0902880032857259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,1,balanced,0.09326933821042378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,1,balanced,0.09432533383369446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,1,balanced,0.09877866506576538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,1,balanced,0.09945600231488545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,1,balanced,0.10243733723958333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,1,balanced,0.11263466874758403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,1,balanced,0.12083199620246887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,1,balanced,0.125791996717453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,1,balanced,0.13157866398493448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,1,balanced,0.15344533324241638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,1,balanced,0.1729173262914022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,1,balanced,0.21107200781504312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,1,balanced,0.24621333678563437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,1,balanced,0.3529920180638631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,1,balanced,0.4291253487269084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,1,balanced,0.6256693204243978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,1,balanced,0.8039733568827311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,1,balanced,1.1701599756876628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,1,power_law_1.2,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,1,power_law_1.2,0.03456639945507049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,1,power_law_1.2,0.035641598701477054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,1,power_law_1.2,0.037171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,1,power_law_1.2,0.037862399220466615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,1,power_law_1.2,0.039078399538993835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,1,balanced,1.539199988047282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,1,power_law_1.2,0.03964160084724426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,1,power_law_1.2,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,1,power_law_1.2,0.04054400026798248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,1,power_law_1.2,0.041382399201393125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,1,power_law_1.2,0.0412416011095047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,1,power_law_1.2,0.043398401141166686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,1,power_law_1.2,0.045049598813056944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,1,power_law_1.2,0.04640640020370483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,1,power_law_1.2,0.05351679921150208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,1,power_law_1.2,0.0546176016330719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,1,power_law_1.2,0.05736960172653198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,1,power_law_1.2,0.06780160069465638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.2,0.07491840124130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.2,0.0953984022140503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,1,balanced,0.05121066669623057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,1,balanced,0.06362666686375935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,1,balanced,0.09257066249847412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,1,balanced,0.14422399799029031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,1,balanced,0.25340267022450763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,1,balanced,0.4811199903488159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,1,balanced,0.49402666091918945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,1,balanced,0.5085386832555135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,1,balanced,0.5202613274256388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,1,balanced,0.5336853265762329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,1,balanced,0.5317493279774984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,1,balanced,0.5496053298314413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,1,balanced,0.5664586623509725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,1,balanced,0.5590186516443888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,1,balanced,0.5736853281656901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,1,balanced,0.5930879910786947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,1,balanced,0.6236426830291748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,1,balanced,0.7016426722208658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,1,balanced,0.7053333123524984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.2,0.1102336049079895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,1,balanced,0.8529280026753744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,1,balanced,0.9647306601206461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,1,balanced,1.2800373236338298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,1,balanced,1.1209279696146648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,1,balanced,1.5289279619852703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,1,balanced,1.9394133885701497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,1,balanced,2.8046134312947593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,1,balanced,3.68668270111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.2,0.14968960285186766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.2,0.17434240579605104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.2,0.2412031888961792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.2,0.3111680030822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,64,balanced,0.05608533322811127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,64,balanced,0.056090667843818665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,64,balanced,0.05663466453552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,64,balanced,0.05929600199063619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,64,balanced,0.061664000153541565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,64,balanced,0.07436800003051758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,64,balanced,0.08413867155710857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,64,balanced,0.0827893316745758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,64,balanced,0.08271466692288716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,64,balanced,0.08311466872692108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,64,balanced,0.08793600400288899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,64,balanced,0.08487466971079509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,64,balanced,0.09537067015965779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,64,balanced,0.09101333220799764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,64,balanced,0.1009279986222585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,64,balanced,0.09816533327102661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,64,balanced,0.10644800464312236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,64,balanced,0.1143839955329895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,64,balanced,0.12475200494130452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,64,balanced,0.14670399824778238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,64,balanced,0.165583997964859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,64,balanced,0.21146667003631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,64,balanced,0.2524906595547994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,64,balanced,0.3377546469370524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,64,balanced,0.43751998742421466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,64,balanced,0.6011733214060465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,64,balanced,0.7710346380869547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.2,0.45024638175964354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.2,0.5838335990905762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,128,power_law_1.01,0.05829120278358459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,128,power_law_1.01,0.056934398412704465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,128,power_law_1.01,0.057145601511001586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,128,power_law_1.01,0.0592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,128,power_law_1.01,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,128,power_law_1.01,0.05928320288658142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,128,power_law_1.01,0.0621504008769989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,128,power_law_1.01,0.06102399826049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,8,balanced,0.04586133360862732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,8,balanced,0.04610133171081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,8,balanced,0.06118933359781901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,8,balanced,0.07659199833869934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,8,balanced,0.11269332965215047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,8,balanced,0.17350933949152628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,8,balanced,0.17684266964594522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,8,balanced,0.17784533898035684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,8,balanced,0.1763146718343099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,8,balanced,0.17857066790262857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,8,balanced,0.1795733372370402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,8,balanced,0.1800160010655721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,8,balanced,0.1816426714261373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,128,power_law_1.01,0.060108798742294314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,128,power_law_1.01,0.06286079883575439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,128,power_law_1.01,0.06204800009727478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,128,power_law_1.01,0.06228479743003845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,8,balanced,0.18383999665578207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,8,balanced,0.1881600022315979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,128,power_law_1.01,0.06361600160598754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,128,power_law_1.01,0.06462079882621766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,128,power_law_1.01,0.06766719818115234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,128,power_law_1.01,0.06826879978179931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,128,power_law_1.01,0.07088000178337098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,128,power_law_1.01,0.08005759716033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,128,power_law_1.01,0.08733440041542054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,8,balanced,0.18851200739542642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,8,balanced,0.19031999508539835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,8,balanced,0.19773866732915243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,8,balanced,0.20602132876714072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,8,balanced,0.2174826661745707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,8,balanced,0.23761065800984701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,128,power_law_1.01,0.10618239641189575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,8,balanced,0.2688373327255249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,128,power_law_1.01,0.12197760343551636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,8,balanced,0.28753066062927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,8,balanced,0.3552693525950114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,128,power_law_1.01,0.15397119522094727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,8,balanced,0.40788265069325763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,128,power_law_1.01,0.18396159410476684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,8,balanced,0.5546133518218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,128,power_law_1.01,0.24916479587554932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,8,balanced,0.6398133436838785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,128,power_law_1.01,0.31201279163360596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,128,power_law_1.01,0.43212161064147947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,128,power_law_1.01,0.5654208183288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,8,power_law_1.2,0.12956160306930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,8,power_law_1.2,0.1842560052871704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,8,power_law_1.2,0.18338559865951537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,8,power_law_1.2,0.22880640029907226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,8,power_law_1.2,0.27111680507659913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,8,power_law_1.2,0.21882879734039307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,8,power_law_1.2,0.2585472106933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,8,power_law_1.2,0.2604736089706421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,8,power_law_1.2,0.254419207572937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,8,power_law_1.2,0.26287360191345216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,8,power_law_1.2,0.26185600757598876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,8,power_law_1.2,0.2630592107772827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,8,power_law_1.2,0.2599040031433105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,8,power_law_1.2,0.26384639739990234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,8,power_law_1.2,0.27743360996246336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,8,power_law_1.2,0.2785856008529663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,8,power_law_1.2,0.28168959617614747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,8,power_law_1.2,0.3295743942260742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,8,power_law_1.2,0.33096320629119874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,8,power_law_1.2,0.39022719860076904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,8,power_law_1.2,0.42876157760620115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,8,power_law_1.2,0.5719679832458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,8,power_law_1.2,0.625651216506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,8,power_law_1.2,0.8172672271728516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,8,power_law_1.2,0.9506752014160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,8,power_law_1.2,1.3899904251098634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,8,power_law_1.2,1.8112640380859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,4,power_law_1.01,0.07465599775314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,4,power_law_1.01,0.09635199904441834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,4,power_law_1.01,0.08942720293998718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,4,power_law_1.01,0.11030399799346924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,4,power_law_1.01,0.12135679721832275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,4,power_law_1.01,0.15947519540786742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,4,power_law_1.01,0.15050239562988282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,4,power_law_1.01,0.15134079456329347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,4,power_law_1.01,0.16179200410842895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,4,power_law_1.01,0.16193920373916626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,4,power_law_1.01,0.16033920049667358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,4,power_law_1.01,0.16410239934921264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,4,power_law_1.01,0.16933759450912475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,4,power_law_1.01,0.17148799896240235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,4,power_law_1.01,0.1840831995010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,4,power_law_1.01,0.18903679847717286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,4,power_law_1.01,0.20048000812530517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,4,power_law_1.01,0.23208320140838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,4,power_law_1.01,0.2503551959991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,4,power_law_1.01,0.3203392028808594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,4,power_law_1.01,0.38484480381011965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,4,power_law_1.01,0.49667840003967284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,4,power_law_1.01,0.5430655956268311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,4,power_law_1.01,0.7422207832336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,4,power_law_1.01,0.9139264106750489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,4,power_law_1.01,1.289510440826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,4,power_law_1.01,1.6843135833740235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,8,power_law_1.2,0.04739840030670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,8,power_law_1.2,0.050329601764678954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,8,power_law_1.2,0.05138559937477112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,8,power_law_1.2,0.055801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,8,power_law_1.2,0.06312320232391358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,8,power_law_1.2,0.05829120278358459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,8,power_law_1.2,0.061817598342895505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,8,power_law_1.2,0.059411197900772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,8,power_law_1.2,0.059539198875427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,8,power_law_1.2,0.060659199953079224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,8,power_law_1.2,0.0651968002319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,8,power_law_1.2,0.06907520294189454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,8,power_law_1.2,0.0690559983253479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,8,power_law_1.2,0.07384960055351257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,8,power_law_1.2,0.08119680285453797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,8,power_law_1.2,0.08321279883384705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,8,power_law_1.2,0.08769919872283935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,8,power_law_1.2,0.10306559801101685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,8,power_law_1.2,0.10805120468139648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,8,power_law_1.2,0.13048319816589354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,8,power_law_1.2,0.1499392032623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,8,power_law_1.2,0.1919935941696167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,8,power_law_1.2,0.2381376028060913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,8,power_law_1.2,0.32895359992980955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,8,power_law_1.2,0.42281599044799806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,8,power_law_1.2,0.5907455921173096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,8,power_law_1.2,0.8224512100219726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,8,power_law_1.2,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,8,power_law_1.2,0.04989440143108368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,8,power_law_1.2,0.04894079864025116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,8,power_law_1.2,0.05735039710998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,8,power_law_1.2,0.060787200927734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,8,power_law_1.2,0.062118399143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,8,power_law_1.2,0.06355199813842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,8,power_law_1.2,0.06609280109405517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,8,power_law_1.2,0.06529279947280883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,8,power_law_1.2,0.06653439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,8,power_law_1.2,0.0698303997516632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,8,power_law_1.2,0.07263360023498536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,8,power_law_1.2,0.07561600208282471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,8,power_law_1.2,0.08177279829978942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,8,power_law_1.2,0.09120000004768372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,8,power_law_1.2,0.09717760086059571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,8,power_law_1.2,0.10116480588912964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,8,power_law_1.2,0.11893759965896607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,8,power_law_1.2,0.13288960456848145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,8,power_law_1.2,0.16368639469146729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,8,power_law_1.2,0.2000191926956177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,8,power_law_1.2,0.26518399715423585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,8,power_law_1.2,0.3365504026412964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,8,power_law_1.2,0.47343997955322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,8,power_law_1.2,0.5971968173980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,8,power_law_1.2,0.8136384010314941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,8,power_law_1.2,1.1355968475341798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,16,power_law_1.2,0.05739520192146301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,16,power_law_1.2,0.06694399714469909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,16,power_law_1.2,0.06147199869155884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,16,power_law_1.2,0.06313599944114685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,16,power_law_1.2,0.061452800035476686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,16,power_law_1.2,0.05829120278358459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,16,power_law_1.2,0.0644864022731781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,16,power_law_1.2,0.06610559821128845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,16,power_law_1.2,0.06539520025253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,16,power_law_1.2,0.06784639954566955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,16,power_law_1.2,0.06765440106391907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,16,power_law_1.2,0.06962559819221496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,16,power_law_1.2,0.07015039920806884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,16,power_law_1.2,0.07137280106544494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,16,power_law_1.2,0.08275840282440186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,16,power_law_1.2,0.08300160169601441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,16,power_law_1.2,0.09220479726791382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,16,power_law_1.2,0.10424319505691529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,16,power_law_1.2,0.11761280298233032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,16,power_law_1.2,0.1428544044494629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,16,power_law_1.2,0.15708800554275512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,16,power_law_1.2,0.20941441059112548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,16,power_law_1.2,0.2376703977584839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,16,power_law_1.2,0.343340802192688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,16,power_law_1.2,0.436627197265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,16,power_law_1.2,0.5890111923217773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,16,power_law_1.2,0.7667263984680176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,32,power_law_1.2,0.03438720107078552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,32,power_law_1.2,0.03392640054225922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,32,power_law_1.2,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,32,power_law_1.2,0.03485440015792847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,32,power_law_1.2,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,32,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,32,power_law_1.2,0.03763839900493622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,32,power_law_1.2,0.03782399892807007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,32,power_law_1.2,0.038099199533462524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,32,power_law_1.2,0.03823359906673431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,32,power_law_1.2,0.03900800049304962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,32,power_law_1.2,0.03940480053424835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,32,power_law_1.2,0.03948799967765808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,32,power_law_1.2,0.04041599929332733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,32,power_law_1.2,0.04412800073623657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,32,power_law_1.2,0.044614401459693906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,32,power_law_1.2,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,32,power_law_1.2,0.05015680193901062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,32,power_law_1.2,0.05304960012435913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,32,power_law_1.2,0.06577280163764954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,32,power_law_1.2,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,32,power_law_1.2,0.08572800159454345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,16,power_law_1.01,0.019596800208091736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,16,power_law_1.01,0.019897599518299103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,16,power_law_1.01,0.01825280040502548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,16,power_law_1.01,0.01988479942083359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,16,balanced,0.026159999271233875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,16,power_law_1.01,0.01930239945650101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,16,power_law_1.01,0.01976960003376007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,16,power_law_1.01,0.02129279971122742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,16,power_law_1.01,0.020735999941825865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,16,power_law_1.01,0.021798400580883025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,16,power_law_1.01,0.02194560021162033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,16,power_law_1.01,0.02240000069141388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,16,balanced,0.02714666724205017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,16,power_law_1.01,0.02353920042514801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,16,power_law_1.01,0.024223999679088594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,16,power_law_1.01,0.027187201380729675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,16,power_law_1.01,0.028089600801467895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,32,power_law_1.2,0.09961599707603455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,16,power_law_1.01,0.029241600632667543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,16,power_law_1.01,0.042124798893928526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,16,balanced,0.027295999228954315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,16,balanced,0.030181333422660828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,16,balanced,0.031311998764673867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,16,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,16,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,16,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,16,balanced,0.03365333378314972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,16,balanced,0.03331733246644338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,16,balanced,0.033759998778502144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,16,balanced,0.03377600014209747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,16,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,16,balanced,0.03532266616821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,16,balanced,0.03930133332808813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,16,balanced,0.03736000011364619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,16,balanced,0.04153066625197729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,16,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,16,balanced,0.04586133360862732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,16,power_law_1.01,0.04528000056743622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,16,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,16,balanced,0.05169600248336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,16,power_law_1.01,0.0440447986125946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,16,balanced,0.06117333471775055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,16,balanced,0.07063466807206471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,16,power_law_1.01,0.048902401328086854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,16,balanced,0.09327466289202373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,16,balanced,0.10995200276374817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,16,power_law_1.01,0.05255039930343628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,32,power_law_1.2,0.12200319766998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,16,power_law_1.01,0.06316159963607788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,16,power_law_1.01,0.07920640110969543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,16,power_law_1.01,0.09783679842948914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,16,power_law_1.01,0.11956479549407958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,16,power_law_1.01,0.16368000507354735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,16,power_law_1.01,0.20288000106811524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,16,balanced,0.14685866236686707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,16,balanced,0.17547732591629028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,32,power_law_1.2,0.17278079986572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,1,power_law_1.01,0.04517120122909546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,1,power_law_1.01,0.05235840082168579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,1,power_law_1.01,0.05741440057754517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,1,power_law_1.01,0.07633919715881347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,1,power_law_1.01,0.07038080096244811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,1,power_law_1.01,0.0721343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,1,power_law_1.01,0.07259520292282104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,1,power_law_1.01,0.07323520183563233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,1,power_law_1.01,0.072326397895813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,1,power_law_1.01,0.07230079770088196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,1,power_law_1.01,0.07382400035858154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,1,power_law_1.01,0.07568640112876893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,1,power_law_1.01,0.07936639785766601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,1,power_law_1.01,0.08019199967384338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,1,power_law_1.01,0.09008640050888062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,1,power_law_1.01,0.09249920248985291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,1,power_law_1.01,0.09247360229492188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,1,power_law_1.01,0.11312639713287354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.01,0.12531839609146117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.01,0.1525056004524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.01,0.18071680068969725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.01,0.236844801902771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,32,power_law_1.2,0.26489601135253904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.01,0.29044480323791505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.01,0.4113920211791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,2,power_law_1.2,0.018956799805164338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,2,power_law_1.2,0.020211200416088104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,2,power_law_1.2,0.024588799476623534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,2,power_law_1.2,0.029977598786354066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,2,power_law_1.2,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,2,power_law_1.2,0.04239999949932098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,2,power_law_1.2,0.045363199710845944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,2,power_law_1.2,0.047788798809051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,2,power_law_1.2,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,2,power_law_1.2,0.05006719827651977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,2,power_law_1.2,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,2,power_law_1.2,0.052537602186203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,2,power_law_1.2,0.05340800285339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,2,power_law_1.2,0.05611519813537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,2,power_law_1.2,0.06259840130805969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,2,power_law_1.2,0.06227840185165405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,2,power_law_1.2,0.06362239718437195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,2,power_law_1.2,0.114028799533844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.2,0.12392959594726563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.2,0.12065279483795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.2,0.12316160202026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.2,0.1498751997947693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.2,0.17381759881973266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.01,0.5177279949188233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,16,balanced,0.02550933261712392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,16,balanced,0.025061334172884624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,16,balanced,0.026149332523345947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,16,balanced,0.02703999976317088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.2,0.22850561141967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,16,balanced,0.042768001556396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,16,balanced,0.04756266872088114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,16,balanced,0.057445332407951355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.2,0.26947200298309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,16,balanced,0.055311997731526695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,16,balanced,0.06573866804440816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,16,balanced,0.0862506628036499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,16,balanced,0.08498666683832805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,16,balanced,0.08710933725039165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,16,balanced,0.08689066767692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,16,balanced,0.08457600076993306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,16,balanced,0.09673066933949788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,16,balanced,0.09519466757774353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,16,balanced,0.09617599844932556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,16,balanced,0.10019200046857198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,16,balanced,0.12261866529782613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,16,balanced,0.14550933241844177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,16,balanced,0.166512002547582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,16,balanced,0.21688000361124674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,16,balanced,0.25827733675638836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,16,balanced,0.3766293525695801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,16,balanced,0.4699999888737996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,16,balanced,0.6887733141581217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,16,balanced,0.9068000316619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.01,0.7359936237335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.2,0.41805438995361327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,32,power_law_1.2,0.30972158908843994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.01,0.9708800315856934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,2,power_law_1.2,0.054579198360443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,2,power_law_1.2,0.0754688024520874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,2,power_law_1.2,0.08722559809684753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,2,power_law_1.2,0.11503360271453858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,2,power_law_1.2,0.1351423978805542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,2,power_law_1.2,0.15548160076141357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,64,power_law_1.2,0.10005760192871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,2,power_law_1.2,0.21338880062103271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,64,power_law_1.2,0.09742720127105713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,2,power_law_1.2,0.21870720386505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,64,power_law_1.2,0.09655680060386658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,2,power_law_1.2,0.22293760776519775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,2,power_law_1.2,0.23049600124359132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,64,power_law_1.2,0.10367360115051269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,2,power_law_1.2,0.23172481060028077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,2,power_law_1.2,0.24872961044311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,2,power_law_1.2,0.24627840518951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,2,power_law_1.2,0.2605247974395752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,2,power_law_1.2,0.27523839473724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,2,power_law_1.2,0.2838592052459717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,2,power_law_1.2,0.3071295976638794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,2,power_law_1.2,0.34662399291992185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,2,power_law_1.2,0.3738176107406616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,2,power_law_1.2,0.4517375946044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,2,power_law_1.2,0.5001088142395019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,2,power_law_1.2,0.6607232093811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,64,power_law_1.2,0.1003775954246521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,2,power_law_1.2,0.8051199913024902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,2,power_law_1.2,1.057868766784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,2,power_law_1.2,1.3374208450317382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,2,power_law_1.2,1.867910385131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,64,power_law_1.2,0.10384639501571655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,2,power_law_1.2,2.418911933898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,64,power_law_1.2,0.11424640417099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,64,power_law_1.2,0.10641280412673951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,64,power_law_1.2,0.11400320529937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,64,power_law_1.2,0.11847679615020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,64,power_law_1.2,0.11919360160827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,64,power_law_1.2,0.12279679775238037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,64,power_law_1.2,0.12399359941482543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,64,power_law_1.2,0.12477439641952515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,64,power_law_1.2,0.13082239627838135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,64,power_law_1.2,0.1302016019821167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,64,power_law_1.2,0.14173439741134644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,64,power_law_1.2,0.16142079830169678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,64,power_law_1.2,0.19157760143280028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,64,power_law_1.2,0.2428287982940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,64,power_law_1.2,0.30176639556884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,64,power_law_1.2,0.39118080139160155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,64,power_law_1.2,0.45372800827026366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,64,power_law_1.2,0.6376639842987061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,64,power_law_1.2,0.9833151817321777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,64,power_law_1.2,1.3152000427246093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,64,power_law_1.2,1.5845376014709474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,1,balanced,0.0234400009115537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,1,balanced,0.02346666653951009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,1,balanced,0.02552533398071925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,1,balanced,0.02994133283694585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,1,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,1,balanced,0.07762133578459422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,1,balanced,0.08043733239173889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,1,balanced,0.08427733182907104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,1,balanced,0.08611733714739482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,1,balanced,0.08821866909662883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,1,balanced,0.08896000186602275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,1,balanced,0.09271466732025146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,1,balanced,0.09504000345865886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,1,balanced,0.09705066680908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,1,balanced,0.10162132978439331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,1,balanced,0.10519466797510783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,1,balanced,0.11222400267918904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,1,balanced,0.12772799531618753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,1,balanced,0.14222400387128195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,1,balanced,0.18317866325378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,1,balanced,0.20984532435735068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,1,balanced,0.28307199478149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,1,balanced,0.29928000768025714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,1,balanced,0.4177279869715373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,1,balanced,0.5331146717071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,1,balanced,0.768064022064209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,1,balanced,1.005776007970174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.2,0.4960768222808838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,64,power_law_1.01,0.017241600155830383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,64,power_law_1.01,0.017107200622558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,64,power_law_1.01,0.017715199291706084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,64,power_law_1.01,0.01817599982023239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,64,power_law_1.01,0.020665599405765532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,64,power_law_1.01,0.0381632000207901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,64,power_law_1.01,0.029446399211883544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,64,power_law_1.01,0.02993279993534088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,64,power_law_1.01,0.030451199412345885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,64,power_law_1.01,0.03017599880695343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,64,power_law_1.01,0.029919999837875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,64,power_law_1.01,0.030374398827552794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,64,power_law_1.01,0.03146879971027374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,64,power_law_1.01,0.031219199299812317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,64,power_law_1.01,0.03281919956207276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,64,power_law_1.01,0.03364480137825012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,64,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,64,power_law_1.01,0.03975679874420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.01,0.043110400438308716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.01,0.04927999973297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.01,0.05249919891357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.01,0.06312959790229797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.01,0.07781760096549988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.01,0.103603196144104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.01,0.12752000093460084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.01,0.18508800268173217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.01,0.23647360801696776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,64,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,64,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,64,balanced,0.031845333675543465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,64,balanced,0.03469333300987879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,64,balanced,0.03572266548871994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,64,balanced,0.03642133375008901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,64,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,64,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,64,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,64,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,64,balanced,0.03757333258787791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,64,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,64,balanced,0.03955200066169103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,64,balanced,0.03961066653331121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,64,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,64,balanced,0.04354666670163473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,64,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,64,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,64,balanced,0.04560000201066335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,64,balanced,0.04987200101216634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,64,balanced,0.05286933481693268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,64,balanced,0.05816000203291575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,64,balanced,0.06518933176994324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,64,balanced,0.085807998975118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,64,balanced,0.09918933113416036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,64,balanced,0.1288266678651174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,64,balanced,0.1613653302192688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,1,power_law_1.01,0.038099199533462524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,1,power_law_1.01,0.03611519932746887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,1,power_law_1.01,0.0412992000579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,1,power_law_1.01,0.05243520140647888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,1,power_law_1.01,0.06159359812736511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,1,power_law_1.01,0.07898240089416504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,1,power_law_1.01,0.08485119938850402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,1,power_law_1.01,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,1,power_law_1.01,0.08849920034408569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,1,power_law_1.01,0.08954880237579346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,1,power_law_1.01,0.09437440037727356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,1,power_law_1.01,0.09908480048179627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,1,power_law_1.01,0.10231679677963257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,1,power_law_1.01,0.10926719903945922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,1,power_law_1.01,0.11936639547348023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,1,power_law_1.01,0.12558720111846924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,1,power_law_1.01,0.1326143980026245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,1,power_law_1.01,0.1559999942779541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,1,power_law_1.01,0.17646080255508423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,1,power_law_1.01,0.22108800411224366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,1,power_law_1.01,0.2595839977264404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,1,power_law_1.01,0.35393919944763186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,1,power_law_1.01,0.44464640617370604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,1,power_law_1.01,0.6357823848724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,1,power_law_1.01,0.8148927688598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,1,power_law_1.01,1.176211166381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,1,power_law_1.01,1.5340543746948243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,128,power_law_1.01,0.02144639939069748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,128,power_law_1.01,0.021158400177955627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,128,power_law_1.01,0.020844799280166627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,128,power_law_1.01,0.02197760045528412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,128,power_law_1.01,0.026233598589897156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,128,power_law_1.01,0.05141760110855102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,128,power_law_1.01,0.041171199083328246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,128,power_law_1.01,0.041228801012039185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,128,power_law_1.01,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,128,power_law_1.01,0.042284798622131345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,128,power_law_1.01,0.04173440039157868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,128,power_law_1.01,0.04237439930438995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,128,power_law_1.01,0.04137600064277649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,128,power_law_1.01,0.040940800309181215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,128,power_law_1.01,0.04408960044384003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,128,power_law_1.01,0.042796799540519716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,128,power_law_1.01,0.048256000876426695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,128,power_law_1.01,0.048979198932647704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.01,0.05581439733505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.01,0.06564480066299438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.01,0.08700159788131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.01,0.10711679458618165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.01,0.13097599744796753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.01,0.17314560413360597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.01,0.22451839447021485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.01,0.3268480062484741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.01,0.45795202255249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,4,balanced,0.045978665351867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,4,balanced,0.04952000081539154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,4,balanced,0.07239999870459239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,4,balanced,0.10238933563232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,4,balanced,0.15576000014940897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,4,balanced,0.1839466691017151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,4,balanced,0.1850879987080892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,4,balanced,0.18545067310333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,4,balanced,0.187882661819458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,4,balanced,0.18784532944361368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,4,balanced,0.18926932414372763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,4,balanced,0.19226133823394775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,4,balanced,0.19570134083429971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,4,balanced,0.19671465953191122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,4,balanced,0.20497065782546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,4,balanced,0.20794665813446045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,4,balanced,0.21544533967971802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,4,balanced,0.2302346626917521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,4,balanced,0.2455093264579773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,4,balanced,0.3025493423144023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,4,balanced,0.30374934275945026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,4,balanced,0.4360160032908122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,4,balanced,0.4434293508529663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,4,balanced,0.6280053456624349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,4,balanced,0.7460319995880127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,4,balanced,1.1101866563161213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,4,balanced,1.3510079383850098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,32,power_law_1.01,0.060134398937225345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,32,power_law_1.01,0.05751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,32,power_law_1.01,0.05694079995155334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,32,power_law_1.01,0.05826560258865356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,32,power_law_1.01,0.05611519813537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,32,power_law_1.01,0.060710400342941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,32,power_law_1.01,0.058745598793029784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,32,power_law_1.01,0.0639680027961731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,32,power_law_1.01,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,32,power_law_1.01,0.06633599996566772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,32,power_law_1.01,0.06858239769935608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,32,power_law_1.01,0.07120640277862549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,32,power_law_1.01,0.07303680181503296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,32,power_law_1.01,0.07176960110664368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,32,power_law_1.01,0.08495360016822814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,32,power_law_1.01,0.08533759713172913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,32,power_law_1.01,0.09098880290985108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,32,power_law_1.01,0.10588159561157226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,32,power_law_1.01,0.10986239910125732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,32,power_law_1.01,0.1368191957473755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,32,power_law_1.01,0.15700479745864868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,32,power_law_1.01,0.19143680334091187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,32,power_law_1.01,0.24517760276794434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,32,power_law_1.01,0.31959679126739504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,32,power_law_1.01,0.4438271999359131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,32,power_law_1.01,0.6478528022766114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,32,power_law_1.01,0.7516736030578614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,4,power_law_1.2,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,4,power_law_1.2,0.0695039987564087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,4,power_law_1.2,0.07338240146636962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,4,power_law_1.2,0.09191679954528809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,4,power_law_1.2,0.09739519953727722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,4,power_law_1.2,0.11076480150222778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,4,power_law_1.2,0.13116159439086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,4,power_law_1.2,0.12458239793777466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,4,power_law_1.2,0.1213312029838562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,4,power_law_1.2,0.13473279476165773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,4,power_law_1.2,0.13771519660949708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,4,power_law_1.2,0.13447680473327636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,4,power_law_1.2,0.14102400541305543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,4,power_law_1.2,0.14410879611968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,4,power_law_1.2,0.1533951997756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,4,power_law_1.2,0.15701760053634645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,4,power_law_1.2,0.16790399551391602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,4,power_law_1.2,0.18502399921417237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,4,power_law_1.2,0.20382080078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,4,power_law_1.2,0.24392321109771728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,4,power_law_1.2,0.26224000453948976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,4,power_law_1.2,0.3389440059661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,4,power_law_1.2,0.41057281494140624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,4,power_law_1.2,0.5444608211517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,4,power_law_1.2,0.6416895866394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,4,power_law_1.2,0.9995327949523926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,4,power_law_1.2,1.157759952545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,32,2,power_law_1.2,0.02030719965696335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,32,2,power_law_1.2,0.021241599321365358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,32,2,power_law_1.2,0.023097600042819976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,32,2,power_law_1.2,0.027456000447273254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,32,2,power_law_1.2,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,32,2,power_law_1.2,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,32,2,power_law_1.2,0.04097920060157776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,32,2,power_law_1.2,0.04317440092563629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,32,2,power_law_1.2,0.045433598756790164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,32,2,power_law_1.2,0.04831359982490539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,32,2,power_law_1.2,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,32,2,power_law_1.2,0.05069440007209778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,32,2,power_law_1.2,0.05779839754104614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,32,2,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,32,2,power_law_1.2,0.057081598043441775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,32,2,power_law_1.2,0.06622080206871032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,32,2,power_law_1.2,0.07128319740295411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,32,2,power_law_1.2,0.08432639837265014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,32,2,power_law_1.2,0.0967743992805481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,32,2,power_law_1.2,0.12674560546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,32,2,power_law_1.2,0.14765440225601195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,32,2,power_law_1.2,0.19909759759902954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,32,2,power_law_1.2,0.22885119915008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,32,2,power_law_1.2,0.29063680171966555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,32,2,power_law_1.2,0.34355199337005615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,32,2,power_law_1.2,0.4622208118438721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,32,2,power_law_1.2,0.5757631778717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,8,power_law_1.2,0.03526400029659271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,8,power_law_1.2,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,8,power_law_1.2,0.03173120021820068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,8,power_law_1.2,0.031667199730873105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,8,power_law_1.2,0.0322816014289856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,8,power_law_1.2,0.0324864000082016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,8,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,8,power_law_1.2,0.03447679877281189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,8,power_law_1.2,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,8,power_law_1.2,0.03929600119590759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,8,power_law_1.2,0.037036800384521486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,8,power_law_1.2,0.0409855991601944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,8,power_law_1.2,0.04123519957065582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,8,power_law_1.2,0.041043201088905336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,8,power_law_1.2,0.050988799333572386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,8,power_law_1.2,0.05262719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,8,power_law_1.2,0.0574400007724762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,8,power_law_1.2,0.07233279943466187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,8,power_law_1.2,0.07881600260734559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,8,power_law_1.2,0.10179200172424316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,8,power_law_1.2,0.10586240291595458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,8,power_law_1.2,0.15022079944610595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,8,power_law_1.2,0.1771008014678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,8,power_law_1.2,0.2336127996444702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,8,power_law_1.2,0.27898240089416504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,8,power_law_1.2,0.40321922302246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,8,power_law_1.2,0.47879681587219236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,32,balanced,0.05197866757710775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,32,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,32,balanced,0.051882664362589516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,32,balanced,0.0603413333495458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,32,balanced,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,32,balanced,0.08678399523099263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,32,balanced,0.08616000413894653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,32,balanced,0.08650133013725281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,32,balanced,0.0846666693687439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,32,balanced,0.08450667063395183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,32,balanced,0.08963732918103536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,32,balanced,0.08690667152404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,32,balanced,0.08900800347328186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,32,balanced,0.09125866492589314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,32,balanced,0.09445333480834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,32,balanced,0.09769599636395772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,32,balanced,0.10050132870674133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,32,balanced,0.10950932900110881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,32,balanced,0.1164959967136383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,32,balanced,0.1360160013039907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,32,balanced,0.15053332845369974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,32,balanced,0.18758400281270346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,32,balanced,0.22164267301559448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,32,balanced,0.29950400193532306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,32,balanced,0.36602667967478436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,32,balanced,0.513157327969869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,32,balanced,0.6585439840952555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,1,power_law_1.01,0.018387199938297273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,1,power_law_1.01,0.022316800057888032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,1,power_law_1.01,0.027673599123954774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,1,power_law_1.01,0.03543039858341217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,1,power_law_1.01,0.04549759924411774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,1,power_law_1.01,0.056358402967453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,1,power_law_1.01,0.058271998167037965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,1,power_law_1.01,0.061536002159118655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,1,power_law_1.01,0.06346880197525025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,1,power_law_1.01,0.06494719982147217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,1,power_law_1.01,0.0700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,1,power_law_1.01,0.0681984007358551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,1,power_law_1.01,0.06815360188484192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,1,power_law_1.01,0.07241600155830383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,1,power_law_1.01,0.07263360023498536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,1,power_law_1.01,0.07268480062484742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,1,power_law_1.01,0.0768064022064209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,1,power_law_1.01,0.08323839902877808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,1,power_law_1.01,0.09514240026474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,1,power_law_1.01,0.1892032027244568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,1,power_law_1.01,0.21249918937683104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,1,power_law_1.01,0.16643199920654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,1,power_law_1.01,0.192684805393219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,1,power_law_1.01,0.2439296007156372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,1,power_law_1.01,0.30247039794921876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,1,power_law_1.01,0.408403205871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,1,power_law_1.01,0.5141056060791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,1,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,1,balanced,0.037733333806196846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,1,balanced,0.04695466657479604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,1,balanced,0.054485330979029335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,1,balanced,0.0828000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,1,balanced,0.0883146623770396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,1,balanced,0.08833600083986919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,1,balanced,0.08881066242853801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,1,balanced,0.09095999598503113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,1,balanced,0.0906880001227061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,1,balanced,0.09404800335566203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,1,balanced,0.09822932879130046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,1,balanced,0.09940800070762634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,1,balanced,0.10113599896430969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,1,balanced,0.107013334830602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,1,balanced,0.11049600442250569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,1,balanced,0.12518399953842163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,1,balanced,0.1401653289794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,1,balanced,0.15958399573961893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,1,balanced,0.21740800142288208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,1,balanced,0.24940800666809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,1,balanced,0.3752959966659546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,1,balanced,0.42711468537648517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,1,balanced,0.612117330233256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,1,balanced,0.7983893553415934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,1,balanced,1.162069320678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,1,balanced,1.4997493426005046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,2,power_law_1.01,0.03470079898834229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,2,power_law_1.01,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,2,power_law_1.01,0.04780800044536591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,2,power_law_1.01,0.05550079941749573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,2,power_law_1.01,0.06142079830169678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,2,power_law_1.01,0.07865599989891052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,2,power_law_1.01,0.08254719972610473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,2,power_law_1.01,0.0839743971824646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,2,power_law_1.01,0.08739200234413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,2,power_law_1.01,0.08944640159606934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,2,power_law_1.01,0.09178239703178406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,2,power_law_1.01,0.09594879746437072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,2,power_law_1.01,0.09779199957847595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,2,power_law_1.01,0.1026304006576538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,2,power_law_1.01,0.10964479446411132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,2,power_law_1.01,0.11031039953231811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,2,power_law_1.01,0.13315199613571166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,2,power_law_1.01,0.1447808027267456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,2,power_law_1.01,0.1493888020515442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,2,power_law_1.01,0.18054399490356446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,2,power_law_1.01,0.20578560829162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,2,power_law_1.01,0.26796159744262693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,2,power_law_1.01,0.35326719284057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,2,power_law_1.01,0.4467455863952637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,2,power_law_1.01,0.5672704219818115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,2,power_law_1.01,0.7582272052764892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,2,power_law_1.01,1.0865471839904786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,128,balanced,0.07346133391062419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,128,balanced,0.07266133526961009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,128,balanced,0.07268799841403961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,128,balanced,0.07602666815121968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,128,balanced,0.07597866654396057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,128,balanced,0.09085866808891296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,128,balanced,0.09285333752632141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,128,balanced,0.0960106650988261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,128,balanced,0.09706667065620422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,128,balanced,0.09292266766230266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,128,balanced,0.10478400190671285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,128,balanced,0.09328533212343852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,128,balanced,0.09465066591898601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,128,balanced,0.09346133470535278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,128,balanced,0.10297066966692607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,128,balanced,0.10053867101669312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,128,balanced,0.10107733805974324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,128,balanced,0.1055626670519511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,128,balanced,0.1081119974454244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,128,balanced,0.12569600343704224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,128,balanced,0.1253973344961802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,128,balanced,0.14646933476130167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,128,balanced,0.16761600971221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,128,balanced,0.21448000272115073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,128,balanced,0.2525706688563029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,128,balanced,0.3487360080083211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,128,balanced,0.42632532119750977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,64,power_law_1.2,0.030937600135803222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,64,power_law_1.2,0.031974399089813234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,64,power_law_1.2,0.028806400299072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,64,power_law_1.2,0.03049600124359131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,64,power_law_1.2,0.03128960132598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,64,power_law_1.2,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,64,power_law_1.2,0.03163520097732544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,64,power_law_1.2,0.03190400004386902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,64,power_law_1.2,0.03200640082359314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,64,power_law_1.2,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,64,power_law_1.2,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,64,power_law_1.2,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,64,power_law_1.2,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,64,power_law_1.2,0.03642880022525787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,64,power_law_1.2,0.04043520092964172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,64,power_law_1.2,0.04209280014038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,64,power_law_1.2,0.04424319863319397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,64,power_law_1.2,0.05067520141601563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,64,power_law_1.2,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,64,power_law_1.2,0.06394879817962647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,64,power_law_1.2,0.07256960272789001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,64,power_law_1.2,0.09139840006828308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,64,power_law_1.2,0.11591039896011353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,64,power_law_1.2,0.15066239833831788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,64,power_law_1.2,0.20819199085235596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,64,power_law_1.2,0.27319040298461916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,64,power_law_1.2,0.04152320027351379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,64,power_law_1.2,0.04120959937572479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,64,power_law_1.2,0.04200319945812225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,64,power_law_1.2,0.043059200048446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,64,power_law_1.2,0.044838398694992065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,64,power_law_1.2,0.04611839950084686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,64,power_law_1.2,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,64,power_law_1.2,0.04725759923458099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,64,power_law_1.2,0.04943360090255737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,64,power_law_1.2,0.04958719909191132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,64,power_law_1.2,0.05008640289306641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,64,power_law_1.2,0.05085440278053284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,64,power_law_1.2,0.05146239995956421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,64,power_law_1.2,0.05583999752998352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,64,power_law_1.2,0.06079360246658325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,64,power_law_1.2,0.05878400206565857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,64,power_law_1.2,0.06346240043640136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,64,power_law_1.2,0.07284479737281799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,64,power_law_1.2,0.08372480273246766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,64,power_law_1.2,0.10471680164337158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,64,power_law_1.2,0.12252160310745239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,64,power_law_1.2,0.15589760541915892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,64,power_law_1.2,0.2017535924911499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,64,power_law_1.2,0.2662528038024902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,64,power_law_1.2,0.33805439472198484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,64,power_law_1.2,0.5139328002929687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,64,power_law_1.2,0.8542336463928223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,64,power_law_1.2,0.3771647930145264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,16,power_law_1.2,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,16,power_law_1.2,0.07854719758033753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,16,power_law_1.2,0.07419520020484924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,16,power_law_1.2,0.07716479897499084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,16,power_law_1.2,0.0790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,16,power_law_1.2,0.07146239876747132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,16,power_law_1.2,0.07184000015258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,16,power_law_1.2,0.07502080202102661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,16,power_law_1.2,0.07281919717788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,16,power_law_1.2,0.07297279834747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,16,power_law_1.2,0.07472000122070313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,16,power_law_1.2,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,16,power_law_1.2,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,16,power_law_1.2,0.07937920093536377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,16,power_law_1.2,0.08373759984970093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,16,power_law_1.2,0.08199040293693542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,16,power_law_1.2,0.09072639942169189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,16,power_law_1.2,0.1014143943786621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,16,power_law_1.2,0.10970879793167114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,16,power_law_1.2,0.126636803150177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,16,power_law_1.2,0.1385856032371521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,16,power_law_1.2,0.1769919991493225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,16,power_law_1.2,0.21570560932159424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,16,power_law_1.2,0.2913727998733521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,16,power_law_1.2,0.37278079986572266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,16,power_law_1.2,0.5115903854370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,16,power_law_1.2,0.6636864185333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,8,balanced,0.036330667634805046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,8,balanced,0.03559466699759165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,8,balanced,0.03590933233499527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,8,balanced,0.03977066775163015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,8,balanced,0.04595733185609182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,8,balanced,0.050527999798456825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,8,balanced,0.048528000712394714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,8,balanced,0.05089599887530009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,8,balanced,0.05012266834576925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,8,balanced,0.050341332952181496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,8,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,8,balanced,0.051072001457214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,8,balanced,0.05409599840641022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,8,balanced,0.053504000107447304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,8,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,8,balanced,0.058005332946777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,8,balanced,0.06293866535027821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,8,balanced,0.06648000081380208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,8,balanced,0.06876799960931142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,8,balanced,0.08833066622416179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,8,balanced,0.08954133590062459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,8,balanced,0.11186666289965312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,8,balanced,0.11963733037312825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,8,balanced,0.16244799892107645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,8,balanced,0.1861813267072042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,8,balanced,0.24864532550175986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,8,balanced,0.3043786684672038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,1,power_law_1.01,0.020601600408554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,1,power_law_1.01,0.022771200537681578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,1,power_law_1.01,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,1,power_law_1.01,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,1,power_law_1.01,0.06568959951400757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,1,power_law_1.01,0.08974080085754395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,1,power_law_1.01,0.10920959711074829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,1,power_law_1.01,0.11610879898071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,1,power_law_1.01,0.1217919945716858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,1,power_law_1.01,0.12434560060501099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,1,power_law_1.01,0.12928639650344848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,1,power_law_1.01,0.13163520097732545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,1,power_law_1.01,0.13624320030212403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,1,power_law_1.01,0.13786239624023439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,1,power_law_1.01,0.14583679437637329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,1,power_law_1.01,0.14835200309753419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,1,power_law_1.01,0.15845119953155518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,1,power_law_1.01,0.17959680557250976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.01,0.19591679573059081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.01,0.23863680362701417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.01,0.27898879051208497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.01,0.3691776037216187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.01,0.3883455991744995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.01,0.5182528018951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.01,0.643616008758545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.01,0.9421119689941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.01,1.16178560256958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,32,power_law_1.2,0.16569600105285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,32,power_law_1.2,0.1639680027961731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,32,power_law_1.2,0.15859839916229249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,32,power_law_1.2,0.16920959949493408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,32,power_law_1.2,0.15904639959335326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,32,power_law_1.2,0.14673919677734376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,32,power_law_1.2,0.16100480556488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,32,power_law_1.2,0.1600000023841858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,32,power_law_1.2,0.14999680519104003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,32,power_law_1.2,0.16424959897994995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,32,power_law_1.2,0.1722048044204712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,32,power_law_1.2,0.16600320339202881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,32,power_law_1.2,0.17079039812088012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,32,power_law_1.2,0.15619200468063354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,32,power_law_1.2,0.17583999633789063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,32,power_law_1.2,0.17211519479751586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,32,power_law_1.2,0.18509440422058104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,32,power_law_1.2,0.2025536060333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,32,power_law_1.2,0.2192768096923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,2,power_law_1.2,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,32,power_law_1.2,0.2549247980117798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,2,power_law_1.2,0.036032000184059144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,32,power_law_1.2,0.29738879203796387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,2,power_law_1.2,0.039743998646736146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,2,power_law_1.2,0.053401601314544675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,2,power_law_1.2,0.05496320128440857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,2,power_law_1.2,0.05841919779777527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,2,power_law_1.2,0.059001600742340087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,2,power_law_1.2,0.06270719766616821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,2,power_law_1.2,0.06341760158538819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,32,power_law_1.2,0.42543997764587405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,2,power_law_1.2,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,32,power_law_1.2,0.4440000057220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,2,power_law_1.2,0.06395519971847534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,2,power_law_1.2,0.06615039706230164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,32,power_law_1.2,0.626585578918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,2,power_law_1.2,0.06744959950447083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,32,power_law_1.2,0.8447744369506835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,2,power_law_1.2,0.07580159902572632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,32,power_law_1.2,1.1233087539672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,2,power_law_1.2,0.08179200291633607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,2,power_law_1.2,0.08551679849624634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,2,power_law_1.2,0.10085760354995728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,2,power_law_1.2,0.10977920293807983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,32,power_law_1.2,1.5964991569519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,2,power_law_1.2,0.11953920125961304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,2,power_law_1.2,0.14967039823532105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,2,power_law_1.2,0.19022079706192016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,2,power_law_1.2,0.2350719928741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,2,power_law_1.2,0.3009664058685303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,2,power_law_1.2,0.4545599937438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,1,power_law_1.2,0.07623040080070495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,1,power_law_1.2,0.08946560025215149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,1,power_law_1.2,0.11715199947357177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,1,power_law_1.2,0.1502400040626526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,1,power_law_1.2,0.1788480043411255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,1,power_law_1.2,0.2659008026123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,1,power_law_1.2,0.27302401065826415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,1,power_law_1.2,0.27853438854217527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,1,power_law_1.2,0.28512001037597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,1,power_law_1.2,0.2859519958496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,1,power_law_1.2,0.28843519687652586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,1,power_law_1.2,0.28589439392089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,1,power_law_1.2,0.290393590927124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,1,power_law_1.2,0.29564800262451174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,1,power_law_1.2,0.3100672006607056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,1,power_law_1.2,0.3149503946304321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,1,power_law_1.2,0.33569281101226806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,1,power_law_1.2,0.3780287981033325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.2,0.421619176864624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.2,0.5062016010284424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.2,0.5778048038482666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.2,0.749286413192749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.2,0.9025664329528809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.2,1.2243007659912108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.2,1.5443008422851563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,2,power_law_1.2,0.4899712085723877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,32,4,power_law_1.01,0.020710399746894835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,2,power_law_1.2,0.8006912231445312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,32,4,power_law_1.01,0.021241599321365358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,32,4,power_law_1.01,0.023123200237751006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,32,4,power_law_1.01,0.026668798923492432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,32,4,power_law_1.01,0.02781440019607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,32,4,power_law_1.01,0.02943359911441803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,32,4,power_law_1.01,0.03224320113658905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,32,4,power_law_1.01,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,32,4,power_law_1.01,0.03511680066585541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,32,4,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,32,4,power_law_1.01,0.03637759983539581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,32,4,power_law_1.01,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,32,4,power_law_1.01,0.03874559998512268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,32,4,power_law_1.01,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,32,4,power_law_1.01,0.04389120042324066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,32,4,power_law_1.01,0.04619520008563995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,32,4,power_law_1.01,0.055264002084732054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,32,4,power_law_1.01,0.06578559875488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,32,4,power_law_1.01,0.09645439982414246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,32,4,power_law_1.01,0.11937919855117798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,32,4,power_law_1.01,0.1235584020614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,32,4,power_law_1.01,0.13982080221176146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,32,4,power_law_1.01,0.1631999969482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,32,4,power_law_1.01,0.2118783950805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,32,4,power_law_1.01,0.2663232088088989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,32,4,power_law_1.01,0.350163197517395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,32,4,power_law_1.01,0.4588223934173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,16,power_law_1.2,0.0521664023399353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,16,power_law_1.2,0.06389120221138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,16,power_law_1.2,0.056518399715423585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,16,power_law_1.2,0.060838401317596436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,16,power_law_1.2,0.06372479796409607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,16,power_law_1.2,0.055103999376296994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,16,power_law_1.2,0.059673601388931276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,16,power_law_1.2,0.05928320288658142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,16,power_law_1.2,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,16,power_law_1.2,0.06293119788169861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,16,power_law_1.2,0.06409599781036376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,16,power_law_1.2,0.06455039978027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,16,power_law_1.2,0.06465280055999756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,16,power_law_1.2,0.06641280055046081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,16,power_law_1.2,0.07575039863586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,16,power_law_1.2,0.07959679961204529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,16,power_law_1.2,0.08396160006523132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,16,power_law_1.2,0.10145280361175538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,16,power_law_1.2,0.10451840162277222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,16,power_law_1.2,0.12564480304718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,16,power_law_1.2,0.13791359663009645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,16,power_law_1.2,0.17855360507965087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,16,power_law_1.2,0.2235584020614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,16,power_law_1.2,0.29663360118865967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,16,power_law_1.2,0.3825727939605713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,16,power_law_1.2,0.5284863948822022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,16,power_law_1.2,0.7210048198699951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.2,2.19039363861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,1,power_law_1.2,0.03798399865627289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,1,power_law_1.2,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,1,power_law_1.2,0.07384960055351257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,1,power_law_1.2,0.09835519790649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,1,power_law_1.2,0.11739519834518433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,1,power_law_1.2,0.18258559703826904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,1,power_law_1.2,0.19949439764022828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,1,power_law_1.2,0.20103681087493896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,1,power_law_1.2,0.21142399311065674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,1,power_law_1.2,0.21071999073028563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,1,power_law_1.2,0.21989119052886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,1,power_law_1.2,0.22621440887451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.2,2.832057571411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,1,power_law_1.2,0.2331455945968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,1,power_law_1.2,0.2389888048171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,1,power_law_1.2,0.2515199899673462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,1,power_law_1.2,0.26733438968658446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,1,power_law_1.2,0.27589120864868166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,1,power_law_1.2,0.31816959381103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,1,power_law_1.2,0.3479871988296509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,1,power_law_1.2,0.43067522048950196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,1,power_law_1.2,0.4952832221984863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,1,power_law_1.2,0.6459263801574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,1,power_law_1.2,0.8006464004516601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,1,power_law_1.2,1.097817611694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,2,power_law_1.2,0.904377555847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,1,power_law_1.2,1.349734401702881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,1,power_law_1.2,1.9160255432128905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,1,power_law_1.2,2.4823999404907227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,32,power_law_1.01,0.058796799182891844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,32,power_law_1.01,0.05424000024795532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,32,power_law_1.01,0.052198398113250735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,32,power_law_1.01,0.053881597518920896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,32,power_law_1.01,0.057126402854919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,32,power_law_1.01,0.05692160129547119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,32,power_law_1.01,0.061299198865890504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,32,power_law_1.01,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,32,power_law_1.01,0.059462398290634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,32,power_law_1.01,0.06027519702911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,32,power_law_1.01,0.060710400342941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,32,power_law_1.01,0.06399999856948853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,32,power_law_1.01,0.06455680131912231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,32,power_law_1.01,0.06682239770889283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,32,power_law_1.01,0.07544959783554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,32,power_law_1.01,0.0776639997959137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,32,power_law_1.01,0.08297600150108338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,32,power_law_1.01,0.10149120092391968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,32,power_law_1.01,0.10573439598083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,32,power_law_1.01,0.12819199562072753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,32,power_law_1.01,0.14633599519729615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,32,power_law_1.01,0.18960000276565553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,32,power_law_1.01,0.2273344039916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,32,power_law_1.01,0.313919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,32,power_law_1.01,0.39635839462280276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,32,power_law_1.01,0.6167871952056885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,32,power_law_1.01,0.7781951904296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,8,power_law_1.01,0.11370879411697388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,8,power_law_1.01,0.11767679452896118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,8,power_law_1.01,0.11635839939117432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,8,power_law_1.01,0.11809920072555542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,8,power_law_1.01,0.11770880222320557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,8,power_law_1.01,0.11671040058135987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,8,power_law_1.01,0.11304320096969604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,8,power_law_1.01,0.11235840320587158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,8,power_law_1.01,0.10258560180664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,8,power_law_1.01,0.10687359571456909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,8,power_law_1.01,0.0948095977306366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,8,power_law_1.01,0.10769920349121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,8,power_law_1.01,0.1310464024543762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,8,power_law_1.01,0.1282431960105896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,8,power_law_1.01,0.16166399717330932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,8,power_law_1.01,0.14510719776153563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,8,power_law_1.01,0.17201919555664064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,8,power_law_1.01,0.2453632116317749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.01,0.3267199993133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.01,0.3864383935928345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.01,0.5616064071655273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,1,power_law_1.01,0.08634240031242371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,1,power_law_1.01,0.10839680433273316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,1,power_law_1.01,0.1347391963005066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,1,power_law_1.01,0.1884735941886902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,1,power_law_1.01,0.24245760440826417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,1,power_law_1.01,0.3736000061035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,1,power_law_1.01,0.38830718994140623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,1,power_law_1.01,0.411846399307251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.01,0.7996287822723389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,1,power_law_1.01,0.4127039909362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,1,power_law_1.01,0.4242559909820557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,1,power_law_1.01,0.44223361015319823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,1,power_law_1.01,0.44764161109924316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,1,power_law_1.01,0.46364798545837405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,1,power_law_1.01,0.4769023895263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,1,power_law_1.01,0.5024640083312988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,1,power_law_1.01,0.5171520233154296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,1,power_law_1.01,0.543507194519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,1,power_law_1.01,0.6250751972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,1,power_law_1.01,0.6954175949096679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,1,power_law_1.01,0.8591423988342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,1,power_law_1.01,1.0328127861022949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,1,power_law_1.01,1.3143487930297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.01,0.9496447563171386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,1,power_law_1.01,1.5244288444519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,1,power_law_1.01,2.0796735763549803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,1,power_law_1.01,2.5557632446289062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,1,power_law_1.01,3.5791553497314452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,1,power_law_1.01,4.728684616088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.01,1.483903980255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,64,power_law_1.01,0.03956480026245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,64,power_law_1.01,0.03834879994392395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,64,power_law_1.01,0.03898240029811859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,64,power_law_1.01,0.040115201473236085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,64,power_law_1.01,0.039340800046920775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.01,1.9280832290649415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,64,power_law_1.01,0.041254401206970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,64,power_law_1.01,0.04059520065784454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,64,power_law_1.01,0.040575999021530154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,64,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,64,power_law_1.01,0.041075199842453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,64,power_law_1.01,0.041094401478767396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,64,power_law_1.01,0.0414463996887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,64,power_law_1.01,0.041766399145126344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,64,power_law_1.01,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,64,power_law_1.01,0.044838398694992065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,64,power_law_1.01,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,64,power_law_1.01,0.04616959989070892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,64,power_law_1.01,0.0514240026473999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,64,power_law_1.01,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,64,power_law_1.01,0.06399359703063964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,64,power_law_1.01,0.0680895984172821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,64,power_law_1.01,0.08373119831085205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,64,power_law_1.01,0.09685760140419006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,64,power_law_1.01,0.12813440561294556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,64,power_law_1.01,0.15926400423049927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,64,power_law_1.01,0.20776960849761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,64,power_law_1.01,0.25677440166473386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,2,power_law_1.01,0.07582079768180847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,2,power_law_1.01,0.08233600258827209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,2,power_law_1.01,0.09364479780197144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,2,power_law_1.01,0.10784640312194824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,2,power_law_1.01,0.1299008011817932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,2,power_law_1.01,0.15033600330352784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,2,power_law_1.01,0.1641088008880615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,2,power_law_1.01,0.18240640163421631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,2,power_law_1.01,0.19363199472427367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,2,power_law_1.01,0.18958719968795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,2,power_law_1.01,0.19924479722976685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,2,power_law_1.01,0.20613760948181153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,2,power_law_1.01,0.2149183988571167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,2,power_law_1.01,0.21905920505523682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,2,power_law_1.01,0.22497279644012452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,2,power_law_1.01,0.23158400058746337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,2,power_law_1.01,0.2517695903778076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,2,power_law_1.01,0.26915199756622316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,2,power_law_1.01,0.30348160266876223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,2,power_law_1.01,0.36186239719390867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,2,power_law_1.01,0.410975980758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,2,power_law_1.01,0.5328383922576905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,2,power_law_1.01,0.6494783878326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,2,power_law_1.01,0.9239232063293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,2,power_law_1.01,1.125119972229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,2,power_law_1.01,1.5684096336364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,2,power_law_1.01,2.052243232727051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.01,2.9334463119506835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,2,power_law_1.2,0.04751999974250794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,2,power_law_1.2,0.05100799798965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,2,power_law_1.2,0.05790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,2,power_law_1.2,0.061536002159118655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,2,power_law_1.2,0.06648319959640503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,2,power_law_1.2,0.08568320274353028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,2,power_law_1.2,0.08575999736785889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,2,power_law_1.2,0.08648319840431214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,2,power_law_1.2,0.0864575982093811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,2,power_law_1.2,0.08689919710159302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,2,power_law_1.2,0.08699520230293274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,2,power_law_1.2,0.08920320272445678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,2,power_law_1.2,0.09184640049934387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,2,power_law_1.2,0.09523199796676636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,2,power_law_1.2,0.10258560180664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,2,power_law_1.2,0.10576640367507935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,2,power_law_1.2,0.11184639930725097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,2,power_law_1.2,0.1313088059425354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,2,power_law_1.2,0.14606080055236817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,2,power_law_1.2,0.17918720245361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,2,power_law_1.2,0.206060791015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,2,power_law_1.2,0.2699903964996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,2,power_law_1.2,0.32584319114685056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,2,power_law_1.2,0.5079296112060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.01,3.7180992126464845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,2,power_law_1.2,0.6228799819946289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,2,power_law_1.2,0.9252736091613769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,2,power_law_1.01,0.02311040014028549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,2,power_law_1.01,0.0301503986120224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,4,power_law_1.01,0.0693120002746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,2,power_law_1.01,0.04199680089950562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,2,power_law_1.01,0.055206400156021115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,2,power_law_1.01,0.06025599837303162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,2,power_law_1.01,0.08803200125694274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,2,power_law_1.01,0.08941439986228943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,2,power_law_1.01,0.09065600037574768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,2,power_law_1.01,0.09712640047073365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,2,power_law_1.01,0.09919999837875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,2,power_law_1.01,0.10083199739456176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,4,power_law_1.01,0.08375679850578308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,2,power_law_1.01,0.10611840486526489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,2,power_law_1.01,0.10892159938812256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,4,power_law_1.01,0.09323520064353943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,2,power_law_1.01,0.10725120306015015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,2,power_law_1.01,0.11563520431518555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,4,power_law_1.01,0.10154240131378174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,2,power_law_1.01,0.1187391996383667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,2,power_law_1.01,0.12636799812316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,4,power_law_1.01,0.11075839996337891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,2,power_law_1.01,0.1339840054512024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.01,0.15455360412597657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.01,0.18395520448684693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.01,0.22012801170349122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.01,0.2594048023223877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.01,0.297273588180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.01,0.38551039695739747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,4,power_law_1.01,0.12481919527053834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.01,0.4716671943664551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,4,power_law_1.01,0.13229440450668334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.01,0.6644351959228516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,4,power_law_1.01,0.13930879831314086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.01,0.8231040000915527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,4,power_law_1.01,0.14179840087890624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,4,power_law_1.01,0.1413375973701477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,4,power_law_1.01,0.14538240432739258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,4,power_law_1.01,0.1414528012275696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,4,power_law_1.01,0.1494271993637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,4,power_law_1.01,0.14905600547790526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,4,power_law_1.01,0.15432319641113282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,4,power_law_1.01,0.15710079669952393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,4,power_law_1.01,0.16669440269470215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,2,power_law_1.2,1.120582389831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,4,power_law_1.01,0.17448320388793945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,4,power_law_1.01,0.18575359582901002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,4,power_law_1.01,0.21967999935150145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,4,power_law_1.01,0.24138240814208983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,4,power_law_1.01,0.29363839626312255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,4,power_law_1.01,0.37235839366912843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,4,power_law_1.01,0.5053952217102051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,4,power_law_1.01,0.6369664192199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,4,power_law_1.01,0.9271360397338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,4,power_law_1.01,1.1728063583374024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,4,balanced,0.04486933350563049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,4,balanced,0.04529066880544027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,4,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,4,balanced,0.0787306676308314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,4,balanced,0.10699199636777242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,4,balanced,0.12451199690500896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,4,balanced,0.12379200259844463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,4,balanced,0.12692800164222717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,4,balanced,0.12772267063458762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,4,balanced,0.1285760005315145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,4,balanced,0.12779200077056885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,4,balanced,0.13090133666992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,4,balanced,0.13397333025932312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,4,balanced,0.13412800431251526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,4,balanced,0.14326933026313782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,4,balanced,0.146997332572937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,4,balanced,0.1519200007120768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,4,balanced,0.17319466670354208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,4,balanced,0.180293341477712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,4,balanced,0.22446932395299277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,4,balanced,0.24122132857640585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,4,balanced,0.31942933797836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,4,balanced,0.37137067317962646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,4,balanced,0.5220213333765665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,4,balanced,0.6446453332901001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,4,balanced,0.9007786909739176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,4,balanced,1.1678400039672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,2,power_law_1.01,0.06751999855041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,2,power_law_1.01,0.0948736011981964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,2,power_law_1.01,0.11877119541168213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,2,power_law_1.01,0.13240959644317626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,2,power_law_1.01,0.13616640567779542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,2,power_law_1.01,0.1394495964050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,2,power_law_1.01,0.14387199878692628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,2,power_law_1.01,0.14563839435577391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,2,power_law_1.01,0.14754559993743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,2,power_law_1.01,0.14870400428771974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,2,power_law_1.01,0.15953919887542725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,2,power_law_1.01,0.16061439514160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,2,power_law_1.01,0.1636415958404541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,2,power_law_1.01,0.17301759719848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,2,power_law_1.01,0.1892416000366211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,2,power_law_1.01,0.19106559753417968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,2,power_law_1.01,0.21468799114227294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,2,power_law_1.01,0.24547839164733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,2,power_law_1.01,0.2755520105361938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,2,power_law_1.01,0.3770047903060913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,2,power_law_1.01,0.3890496015548706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,2,power_law_1.01,0.6002175807952881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,2,power_law_1.01,0.7021952152252198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,2,power_law_1.01,1.000115203857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,2,power_law_1.01,1.0401663780212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,2,power_law_1.01,2.049875259399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,1,power_law_1.01,0.05618559718132019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,1,power_law_1.01,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,1,power_law_1.01,0.09237759709358215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,1,power_law_1.01,0.14056320190429689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,1,power_law_1.01,0.18904320001602173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,1,power_law_1.01,0.237555193901062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,1,power_law_1.01,0.305401611328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,1,power_law_1.01,0.3232959985733032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,1,power_law_1.01,0.3323071956634521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,1,power_law_1.01,0.3510080099105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,1,power_law_1.01,0.3545664072036743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,1,power_law_1.01,0.37204480171203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,1,power_law_1.01,0.37449600696563723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,1,power_law_1.01,0.39509119987487795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,1,power_law_1.01,0.40685439109802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,1,power_law_1.01,0.41730561256408694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,1,power_law_1.01,0.4445312023162842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,1,power_law_1.01,0.4952064037322998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,1,power_law_1.01,0.527564811706543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,1,power_law_1.01,0.6232895851135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,1,power_law_1.01,0.7098944187164307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,1,power_law_1.01,0.8894783973693847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,1,power_law_1.01,1.019007968902588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,1,balanced,0.06214933097362518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,1,balanced,0.07658133407433827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,1,power_law_1.01,1.3316287994384766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,1,power_law_1.01,1.644268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,1,power_law_1.01,2.234060859680176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,1,balanced,0.10194133718808492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,1,balanced,0.15633599956830344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,1,balanced,0.25567466020584106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,1,balanced,0.45685867468516034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,1,balanced,0.6554400126139323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,2,power_law_1.01,0.018572799861431122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,1,power_law_1.01,2.8230911254882813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,1,balanced,0.6584639946619669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,1,balanced,0.6763520240783691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,1,balanced,0.6673813660939535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,1,balanced,0.682474692662557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,1,balanced,0.6929226716359457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,8,balanced,0.04165333261092504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,1,balanced,0.6925386587778727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,8,balanced,0.04192533095677694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,8,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,1,balanced,0.6997066338857015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,8,balanced,0.04971733192602793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,8,balanced,0.05173333485921224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,8,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,2,power_law_1.01,0.018118399381637573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,2,power_law_1.01,0.018156799674034118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,2,power_law_1.01,0.018572799861431122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,2,power_law_1.01,0.01865600049495697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,2,power_law_1.01,0.019545599818229675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,2,power_law_1.01,0.019788800179958342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,2,power_law_1.01,0.020659199357032774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,2,power_law_1.01,0.021350400149822236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,2,power_law_1.01,0.020972800254821778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,1,balanced,0.7112853527069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,2,power_law_1.01,0.02341119945049286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,8,balanced,0.06203199923038483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,1,balanced,0.7232639789581299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,2,power_law_1.01,0.023161600530147552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,1,balanced,0.740943988164266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,2,power_law_1.01,0.024307200312614442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,1,balanced,0.7805546919504801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,2,power_law_1.01,0.02632319927215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,2,power_law_1.01,0.027430400252342224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,2,power_law_1.01,0.029164800047874452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,2,power_law_1.01,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,8,balanced,0.06265600025653839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,8,balanced,0.06332266827424367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,8,balanced,0.06534400085608165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,8,balanced,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,8,balanced,0.06605333089828491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,8,balanced,0.06924266616503398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,1,balanced,0.8229440053304037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,1,balanced,0.8922932942708334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,2,power_law_1.01,2.6325504302978517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,1,balanced,0.9773866335550944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,1,balanced,1.1226186752319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,1,balanced,1.3224533398946126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,2,power_law_1.01,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,8,balanced,0.06849066913127899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,8,balanced,0.07632533212502797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,8,balanced,0.07844266792138417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,8,balanced,0.08146133522192638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,8,balanced,0.09346133470535278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,8,balanced,0.10296000043551128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,8,balanced,0.13397866487503052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,8,balanced,0.15422399838765463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.01,0.046003198623657225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,8,balanced,0.21144533157348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,8,balanced,0.25524266560872394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,8,balanced,0.36161601543426514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,1,balanced,1.6690719922383626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,8,balanced,0.44969598452250165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.01,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,8,balanced,0.6482933362325033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,1,balanced,2.1179679234822593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,8,balanced,0.832805315653483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,1,balanced,2.7225974400838218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,1,balanced,3.6653760274251304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.01,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.01,0.08588160276412964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.01,0.09926400184631348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.01,0.1350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.01,0.17177599668502808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.01,0.24442241191864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.01,0.3058367967605591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,8,power_law_1.2,0.0901311993598938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,8,power_law_1.2,0.10383360385894776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,8,power_law_1.2,0.12268160581588745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,8,power_law_1.2,0.14190720319747924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,8,power_law_1.2,0.14971519708633424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,8,power_law_1.2,0.15397119522094727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,8,power_law_1.2,0.1673535943031311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,8,power_law_1.2,0.16161919832229615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,8,power_law_1.2,0.15720319747924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,8,power_law_1.2,0.16075520515441893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,8,power_law_1.2,0.15323519706726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,8,power_law_1.2,0.1657663941383362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,8,power_law_1.2,0.15854719877243043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,8,power_law_1.2,0.17056000232696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,8,power_law_1.2,0.1767040014266968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,8,power_law_1.2,0.1820479989051819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,8,power_law_1.2,0.18652160167694093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,8,power_law_1.2,0.19957120418548585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,4,power_law_1.01,0.0315200001001358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,4,power_law_1.01,0.03230080008506775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,4,power_law_1.01,0.034297600388526917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,32,power_law_1.01,0.07239680290222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,4,power_law_1.01,0.03790079951286316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,4,power_law_1.01,0.03916159868240356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,4,power_law_1.01,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,4,power_law_1.01,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,4,power_law_1.01,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,4,power_law_1.01,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,4,power_law_1.01,0.049369600415229795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,4,power_law_1.01,0.053324800729751584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,32,power_law_1.01,0.06698240041732788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,4,power_law_1.01,0.05550079941749573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,32,power_law_1.01,0.06613119840621948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,4,power_law_1.01,0.060159999132156375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,32,power_law_1.01,0.06641280055046081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,8,power_law_1.2,0.22124800682067872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,32,power_law_1.01,0.067603200674057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,8,power_law_1.2,0.2620863914489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,32,power_law_1.01,0.0640447974205017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,32,power_law_1.01,0.06762880086898804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,8,power_law_1.2,0.3043904066085815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,32,power_law_1.01,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,8,power_law_1.2,0.3743743896484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,32,power_law_1.01,0.06727039813995361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,32,power_law_1.01,0.0687936007976532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,4,power_law_1.01,0.06169599890708923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,8,power_law_1.2,0.46430082321166993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,32,power_law_1.01,0.07055360078811646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,32,power_law_1.01,0.07227519750595093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,8,power_law_1.2,0.6589056015014648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,32,power_law_1.01,0.07059839963912964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,64,power_law_1.01,0.054745602607727054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,32,power_law_1.01,0.07332479953765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,8,power_law_1.2,0.8807552337646485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,32,power_law_1.01,0.07986559867858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,32,power_law_1.01,0.07955200076103211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,32,power_law_1.01,0.08072959780693054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,8,power_law_1.2,1.2482111930847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,32,power_law_1.01,0.09260159730911255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,32,power_law_1.01,0.10298880338668823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,32,power_law_1.01,0.11909760236740112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,8,power_law_1.2,1.6470655441284179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,32,power_law_1.01,0.13848960399627686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,64,power_law_1.01,0.05146880149841308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,32,power_law_1.01,0.17333120107650757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,64,power_law_1.01,0.051475197076797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,4,power_law_1.01,0.0706496000289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,64,power_law_1.01,0.05414400100708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,64,power_law_1.01,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,64,power_law_1.01,0.05864959955215454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,64,power_law_1.01,0.05791360139846802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,32,power_law_1.01,0.2072000026702881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,64,power_law_1.01,0.05916159749031067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,4,power_law_1.01,0.07582719922065735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,4,power_law_1.01,0.07341439723968506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,4,power_law_1.01,0.08963199853897094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,4,power_law_1.01,0.10063359737396241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,4,power_law_1.01,0.12874879837036132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,32,power_law_1.01,0.27536640167236326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,4,power_law_1.01,0.14835200309753419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,32,power_law_1.01,0.34944639205932615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,4,power_law_1.01,0.19214719533920288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,64,power_law_1.01,0.06496639847755432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,4,power_law_1.01,0.22992639541625975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,64,power_law_1.01,0.06378239989280701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,4,power_law_1.01,0.33413760662078856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,64,power_law_1.01,0.06453760266304016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,4,power_law_1.01,0.43791999816894533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,4,power_law_1.01,0.6594944000244141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,4,power_law_1.01,0.9239808082580566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,64,power_law_1.01,0.07266560196876526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,64,power_law_1.01,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,64,power_law_1.01,0.06949759721755981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,32,power_law_1.01,0.5065855979919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,32,power_law_1.01,0.6389823913574219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,64,power_law_1.01,0.08232960104942322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,64,power_law_1.01,0.08592640161514283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,64,power_law_1.01,0.08303359746932984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,64,power_law_1.01,0.10170880556106568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,64,power_law_1.01,0.10702719688415527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,64,power_law_1.01,0.12672640085220338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,64,power_law_1.01,0.14283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,64,power_law_1.01,0.17841919660568237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,64,power_law_1.01,0.23032960891723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,64,power_law_1.01,0.30376319885253905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,64,power_law_1.01,0.3905600070953369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,64,power_law_1.01,0.6163712024688721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,64,power_law_1.01,0.7886528015136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,8,power_law_1.2,0.03252480030059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,8,power_law_1.2,0.03266560137271881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,8,power_law_1.2,0.03407999873161316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,8,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,8,power_law_1.2,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,8,power_law_1.2,0.03720960021018982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,8,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,8,power_law_1.2,0.0385919988155365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,8,power_law_1.2,0.03909119963645935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,8,power_law_1.2,0.03971840143203735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,8,power_law_1.2,0.04147199988365173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,8,power_law_1.2,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,8,power_law_1.2,0.04424319863319397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,8,power_law_1.2,0.04480000138282776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,8,power_law_1.2,0.049267199635505673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,8,power_law_1.2,0.05249279737472534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,8,power_law_1.2,0.05713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,8,power_law_1.2,0.06249600052833557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,8,power_law_1.2,0.07044479846954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,8,power_law_1.2,0.0856768012046814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,8,power_law_1.2,0.10449919700622559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,8,power_law_1.2,0.14216959476470947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,8,power_law_1.2,0.17837439775466918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,8,power_law_1.2,0.25564160346984866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,8,power_law_1.2,0.3650815963745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,8,power_law_1.2,0.5390912055969238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,32,balanced,0.019146667172511418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,32,balanced,0.019002666076024372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,32,balanced,0.01913600042462349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,32,balanced,0.01903466631968816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,32,balanced,0.018992000569899876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,32,balanced,0.019402666638294857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,32,balanced,0.023503998915354412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,32,balanced,0.023376000424226124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,32,balanced,0.03839999934037527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,32,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,32,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,32,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,32,balanced,0.03436266630887985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,32,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,32,balanced,0.03864533454179764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,32,balanced,0.0391146664818128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,32,balanced,0.03957333415746689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,32,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,32,balanced,0.043466667334238686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,32,balanced,0.04974400003751119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,32,balanced,0.058970664938290916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,32,balanced,0.07646933197975159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,32,balanced,0.08893332878748576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,32,balanced,0.11955199639002483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,32,balanced,0.1504533290863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,32,balanced,0.21316266059875488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,32,balanced,0.2735360066095988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,8,power_law_1.2,0.7351552009582519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,32,power_law_1.2,0.0491456001996994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,32,power_law_1.2,0.050310397148132326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,32,power_law_1.2,0.045977601408958436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,32,power_law_1.2,0.049491199851036075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,32,power_law_1.2,0.050496000051498416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,32,power_law_1.2,0.05198079943656921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,32,power_law_1.2,0.05453439950942993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,32,power_law_1.2,0.05331839919090271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,32,power_law_1.2,0.055238401889801024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,32,power_law_1.2,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,32,power_law_1.2,0.05756160020828247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,32,power_law_1.2,0.060915201902389526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,32,power_law_1.2,0.06106240153312683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,32,power_law_1.2,0.06850559711456299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,32,power_law_1.2,0.07686399817466735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,32,power_law_1.2,0.08291199803352356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,32,power_law_1.2,0.08794239759445191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,32,power_law_1.2,0.1068992018699646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,32,power_law_1.2,0.10902400016784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,32,power_law_1.2,0.1424512028694153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,32,power_law_1.2,0.16121599674224854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,32,power_law_1.2,0.204038405418396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,32,power_law_1.2,0.26317439079284666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,32,power_law_1.2,0.3676608085632324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,32,power_law_1.2,0.4819647789001465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,32,power_law_1.2,0.6953152179718017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,256,power_law_1.2,0.04702720046043396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,32,power_law_1.2,0.8759552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,256,power_law_1.2,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,256,power_law_1.2,0.04766719937324524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,256,power_law_1.2,0.049292799830436704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,256,power_law_1.2,0.04956159889698029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,256,power_law_1.2,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,256,power_law_1.2,0.05175679922103882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,256,power_law_1.2,0.05156480073928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,256,power_law_1.2,0.05259519815444946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,256,power_law_1.2,0.052851200103759766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,256,power_law_1.2,0.05244799852371216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,256,power_law_1.2,0.052723199129104614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,256,power_law_1.2,0.054636800289154054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,256,power_law_1.2,0.05466880202293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,256,power_law_1.2,0.05790079832077026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,256,power_law_1.2,0.05788159966468811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,256,power_law_1.2,0.06348800063133239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,256,power_law_1.2,0.06881279945373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,256,power_law_1.2,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,256,power_law_1.2,0.08814719915390015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,256,power_law_1.2,0.09926400184631348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,256,power_law_1.2,0.12620799541473388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,256,power_law_1.2,0.15935360193252562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,256,power_law_1.2,0.21340160369873046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,256,power_law_1.2,0.27612159252166746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,256,power_law_1.2,0.40485119819641113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,256,power_law_1.2,0.5032896041870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,16,balanced,0.043882668018341064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,16,balanced,0.043920000394185386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,16,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,16,balanced,0.055999999245007835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,16,balanced,0.07210666437943776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,16,balanced,0.09882666667302449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,16,balanced,0.12917332847913107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,16,balanced,0.12683199842770895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,16,balanced,0.1256480018297831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,16,balanced,0.12811199824015299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,16,balanced,0.12867200374603271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,16,balanced,0.13125866651535034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,16,balanced,0.1312373379866282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,16,balanced,0.13392000397046408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,16,balanced,0.13851199547449747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,16,balanced,0.14102933804194132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,16,balanced,0.1434506674607595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,16,balanced,0.1507253348827362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,16,balanced,0.15729600191116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,16,balanced,0.1748639941215515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,16,balanced,0.18918933471043906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,16,balanced,0.22223466634750366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,16,balanced,0.26543466250101727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,16,balanced,0.32716800769170123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,16,balanced,0.39343468348185223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,16,balanced,0.5194506645202637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,16,balanced,0.6564053297042847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,16,power_law_1.01,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,16,power_law_1.01,0.038841599225997926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,16,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,16,power_law_1.01,0.03992320001125336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,16,power_law_1.01,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,16,power_law_1.01,0.04046080112457275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,8,balanced,0.045791998505592346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,8,balanced,0.04602666695912679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,16,power_law_1.01,0.04069119989871979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,8,balanced,0.04986133178075155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,8,balanced,0.055829331278800964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,16,power_law_1.01,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,8,balanced,0.07495999832948048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,8,balanced,0.10619200269381206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,16,power_law_1.01,0.041766399145126344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,8,balanced,0.10729599992434184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,8,balanced,0.10863999525705974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,8,balanced,0.10975466171900432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,8,balanced,0.11109333237012227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,8,balanced,0.11152533690134685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,8,balanced,0.11522133151690166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,8,balanced,0.11565867066383362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,8,balanced,0.1181813379128774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,8,balanced,0.1267626682917277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,8,balanced,0.12689600388209024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,8,balanced,0.13200533390045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,8,balanced,0.14727999766667685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,8,balanced,0.15593066811561584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,8,balanced,0.1823093295097351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,8,balanced,0.20886399348576865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,8,balanced,0.2557599941889445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,8,balanced,0.3073066671689351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,8,balanced,0.4116693337758382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,8,balanced,0.5106879870096842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,8,balanced,0.7196746667226156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,8,balanced,0.9146347045898438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,16,power_law_1.01,0.04190720021724701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,16,power_law_1.01,0.04287999868392944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,16,power_law_1.01,0.043532800674438474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,16,power_law_1.01,0.04405759871006012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,16,power_law_1.01,0.04683519899845123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,16,power_law_1.01,0.050969600677490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,16,power_law_1.01,0.052883201837539674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,16,power_law_1.01,0.05802239775657654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,16,power_law_1.01,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.01,0.06913920044898987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.01,0.08394240140914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.01,0.09196799993515015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.01,0.12175359725952148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.01,0.15586559772491454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.01,0.20729598999023438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.01,0.2363840103149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.01,0.35012478828430177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.01,0.44389119148254397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,1,power_law_1.2,0.10762879848480225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,1,power_law_1.2,0.130950403213501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,1,power_law_1.2,0.15607680082321168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,1,power_law_1.2,0.21894400119781493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,1,power_law_1.2,0.2996992111206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,1,power_law_1.2,0.39514880180358886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,2,power_law_1.2,0.03797760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,1,power_law_1.2,0.4452479839324951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,2,power_law_1.2,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,2,power_law_1.2,0.04267520010471344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,1,power_law_1.2,0.5876160144805909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,1,power_law_1.2,0.6241151809692382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,1,power_law_1.2,0.6433023929595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,1,power_law_1.2,0.6690624237060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,2,power_law_1.2,0.04806399941444397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,2,power_law_1.2,0.0453247994184494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,1,power_law_1.2,0.6831615924835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,2,power_law_1.2,0.04767360091209412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,2,power_law_1.2,0.0455808013677597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,2,power_law_1.2,0.04629760086536407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,1,power_law_1.2,0.711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,2,power_law_1.2,0.04567039906978607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,2,power_law_1.2,0.04520959854125976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,1,power_law_1.2,0.7377920150756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,2,power_law_1.2,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,2,power_law_1.2,0.047539201378822324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,2,power_law_1.2,0.04831359982490539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,2,power_law_1.2,0.049292799830436704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,2,power_law_1.2,0.05533440113067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,2,power_law_1.2,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,2,power_law_1.2,0.06167680025100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,2,power_law_1.2,0.07141119837760926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,2,power_law_1.2,0.03612160086631775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,2,power_law_1.2,0.04083200097084046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,2,power_law_1.2,0.05008640289306641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,2,power_law_1.2,0.066048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,2,power_law_1.2,0.07758079767227173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,2,power_law_1.2,0.11112320423126221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,2,power_law_1.2,0.11155200004577637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,1,power_law_1.2,0.7673215866088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,2,power_law_1.2,0.11221760511398315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,2,power_law_1.2,0.1209663987159729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,1,power_law_1.2,0.7921152114868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,1,power_law_1.2,0.8325504302978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,1,power_law_1.2,0.9266304016113281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.2,0.0813759982585907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.2,1.031545639038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,2,power_law_1.2,0.12396160364151002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.2,1.1880000114440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,2,power_law_1.2,0.12407040596008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,2,power_law_1.2,0.1282431960105896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.2,1.259436798095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.2,1.506284809112549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.2,0.08600320219993592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.2,1.7244224548339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,2,power_law_1.2,0.12629120349884032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,2,power_law_1.2,0.1381824016571045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.2,2.202003288269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,2,power_law_1.2,0.14226560592651366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.2,0.10610560178756714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,1,power_law_1.2,2.8678720474243162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,2,power_law_1.2,0.15139199495315553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.2,0.13645440340042114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,2,power_law_1.2,0.15412479639053345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,2,power_law_1.2,0.17552000284194946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,1,power_law_1.2,3.7828609466552736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,2,power_law_1.2,0.18354560136795045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,2,power_law_1.2,0.20938239097595215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,2,power_law_1.2,0.21473920345306396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,1,power_law_1.2,4.592761611938476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,2,power_law_1.2,0.26270079612731934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,2,power_law_1.2,0.29869439601898196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,2,power_law_1.2,0.3984639883041382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,2,power_law_1.2,0.4571839809417725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.2,0.1565440058708191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,2,power_law_1.2,0.6698239803314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,2,power_law_1.2,0.8530303955078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.2,0.2666624069213867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.2,0.28843519687652586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.2,0.4195712089538574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,1,balanced,0.09278399745623271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,1,balanced,0.12640000383059183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,1,balanced,0.18941332896550497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,1,balanced,0.3227360049883525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,1,balanced,0.5797333319981893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,1,balanced,0.7143039703369141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,1,balanced,0.7225279808044434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,1,balanced,0.7221173445383707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,1,balanced,0.7276426951090494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,1,balanced,0.7289706865946451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,1,balanced,0.735765298207601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,1,balanced,0.7416586875915527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,1,balanced,0.7468159993489584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,1,balanced,0.7569706439971924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,1,balanced,0.7726720174153646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,1,balanced,0.7801973025004069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,1,balanced,0.8031307061513265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,1,balanced,0.8454240163167318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,1,balanced,0.8893653551737467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,1,balanced,1.1387840112050374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,1,balanced,1.0873226324717205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.2,0.5555327892303467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,1,balanced,1.4851786295572917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,1,balanced,1.5933653513590496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,1,balanced,2.3985865910847983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,1,balanced,2.8277759552001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,1,balanced,4.077247937520345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,1,balanced,5.172842661539714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,32,power_law_1.2,0.016710400581359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,32,power_law_1.2,0.016230399906635284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,32,power_law_1.2,0.016441600024700166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,32,power_law_1.2,0.016575999557971954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,32,power_law_1.2,0.01703680008649826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,32,power_law_1.2,0.016972799599170686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,32,power_law_1.2,0.01736319959163666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,32,power_law_1.2,0.02003840059041977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,32,power_law_1.2,0.020377600193023683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,32,power_law_1.2,0.021011200547218323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,32,power_law_1.2,0.027481600642204285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,32,power_law_1.2,0.027852800488471986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,32,power_law_1.2,0.028723201155662535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,32,power_law_1.2,0.02847360074520111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,32,power_law_1.2,0.028307199478149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,32,power_law_1.2,0.02876800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,32,power_law_1.2,0.0296640008687973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,32,power_law_1.2,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,32,power_law_1.2,0.034969601035118106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,32,power_law_1.2,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,32,power_law_1.2,0.042208001017570496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,32,power_law_1.2,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,32,power_law_1.2,0.06204800009727478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,32,power_law_1.2,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,4,power_law_1.2,0.05231999754905701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,32,power_law_1.2,0.10058879852294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,4,power_law_1.2,0.06785920262336731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,4,power_law_1.2,0.07219840288162231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,4,power_law_1.2,0.08980479836463928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,4,power_law_1.2,0.09957759976387023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,4,power_law_1.2,0.11733119487762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,4,power_law_1.2,0.13011840581893921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,4,power_law_1.2,0.1353983998298645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,4,power_law_1.2,0.14353280067443847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,4,power_law_1.2,0.14300800561904908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,4,power_law_1.2,0.1479423999786377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,4,power_law_1.2,0.14859520196914672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,4,power_law_1.2,0.15253119468688964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,4,power_law_1.2,0.16487679481506348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,4,power_law_1.2,0.16609920263290406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,4,power_law_1.2,0.17550079822540282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,4,power_law_1.2,0.18085119724273682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,4,power_law_1.2,0.20814719200134277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,4,power_law_1.2,0.22049920558929442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,4,power_law_1.2,0.2622848033905029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,4,power_law_1.2,0.3053440093994141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,4,power_law_1.2,0.36401278972625734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,4,power_law_1.2,0.4560192108154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,4,power_law_1.2,0.6084799766540527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,4,power_law_1.2,0.7541440010070801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,4,power_law_1.2,0.9905664443969726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,4,power_law_1.2,1.3092032432556153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,8,power_law_1.2,0.08392959833145142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,8,power_law_1.2,0.0936896026134491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,8,power_law_1.2,0.1027135968208313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,8,power_law_1.2,0.11765120029449463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,32,power_law_1.2,0.1363584041595459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,8,power_law_1.2,0.12342400550842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,8,power_law_1.2,0.13292800188064574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,8,power_law_1.2,0.12348159551620483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,8,power_law_1.2,0.12671999931335448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,8,power_law_1.2,0.12968319654464722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,8,power_law_1.2,0.13178240060806273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,8,power_law_1.2,0.13542399406433106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,8,power_law_1.2,0.1431615948677063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,8,power_law_1.2,0.14468480348587037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,8,power_law_1.2,0.15016319751739501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,8,power_law_1.2,0.16289279460906983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,8,power_law_1.2,0.16814080476760865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,8,power_law_1.2,0.18016639947891236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,8,power_law_1.2,0.2111232042312622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,8,power_law_1.2,0.2341439962387085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,8,power_law_1.2,0.3052736043930054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,8,power_law_1.2,0.36143999099731444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,8,power_law_1.2,0.44048638343811036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,8,power_law_1.2,0.551475191116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,8,power_law_1.2,0.7777728080749512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,8,power_law_1.2,0.8454015731811524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,8,power_law_1.2,1.3220928192138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,8,power_law_1.2,1.7069311141967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,32,power_law_1.2,0.17351679801940917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,64,power_law_1.2,0.07832319736480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,64,power_law_1.2,0.07640320062637329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,64,power_law_1.2,0.07526400089263915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,64,power_law_1.2,0.07734400033950806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,64,power_law_1.2,0.07631360292434693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,64,power_law_1.2,0.0813759982585907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,64,power_law_1.2,0.08176000118255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,64,power_law_1.2,0.08419839739799499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,64,power_law_1.2,0.08595200181007386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,64,power_law_1.2,0.08372480273246766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,64,power_law_1.2,0.08380159735679626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,64,power_law_1.2,0.0864192008972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,64,power_law_1.2,0.08926720023155213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,64,power_law_1.2,0.0833407998085022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,64,power_law_1.2,0.09473919868469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,64,power_law_1.2,0.09720960259437561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,64,power_law_1.2,0.10152319669723511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,8,power_law_1.2,0.11726080179214478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,64,power_law_1.2,0.12037119865417481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,8,power_law_1.2,0.14669439792633057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,8,power_law_1.2,0.16350719928741456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,8,power_law_1.2,0.20697600841522218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,8,power_law_1.2,0.22696959972381592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,64,power_law_1.2,0.13431040048599244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,8,power_law_1.2,0.23766400814056396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,64,power_law_1.2,0.17114239931106567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,64,power_law_1.2,0.19226880073547364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,64,power_law_1.2,0.24439680576324463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,64,power_law_1.2,0.29728639125823975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,64,power_law_1.2,0.4711296081542969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,64,power_law_1.2,0.5960832118988038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,8,power_law_1.2,0.22167680263519288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,2,balanced,0.11582400401433308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,2,balanced,0.1796320080757141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,2,balanced,0.28933332363764447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,8,power_law_1.2,0.21989119052886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,2,balanced,0.2877333362897237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,2,balanced,0.287498672803243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,8,power_law_1.2,0.231878399848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,2,balanced,0.27391467491785687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,2,balanced,0.29197333256403607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,8,power_law_1.2,0.23671040534973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,64,power_law_1.2,0.9102656364440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,8,power_law_1.2,0.24257280826568603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,2,balanced,0.30324800809224445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,8,power_law_1.2,0.246124792098999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,64,power_law_1.2,1.0979968070983888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,2,balanced,0.2980533242225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,2,balanced,0.2701759934425354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,2,balanced,0.2861280043919881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,2,balanced,0.2704426646232605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,2,balanced,0.2577333251635234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,2,balanced,0.25470399856567383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,2,balanced,0.21979733308156332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,2,balanced,0.23625600337982178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,2,balanced,0.20150399208068848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,2,balanced,0.4036959807078044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,8,power_law_1.2,0.24982399940490724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,8,power_law_1.2,0.2632319927215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,8,power_law_1.2,0.2688512086868286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,8,power_law_1.2,0.2829056024551392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,8,power_law_1.2,0.31910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,2,balanced,0.33593066533406574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,2,balanced,0.45685335000356037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,2,balanced,0.5856373310089111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,2,balanced,0.8318453629811605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,2,balanced,1.0663359959920247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,2,balanced,1.5512159665425618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,8,power_law_1.2,0.3625407934188843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,2,balanced,2.036442597707113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,2,balanced,2.9722347259521484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,2,balanced,3.942805290222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,8,power_law_1.2,0.36574718952178953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,8,power_law_1.2,0.4316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,8,power_law_1.2,0.503276777267456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,8,power_law_1.2,0.6389440059661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,8,power_law_1.2,0.6138304233551025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,8,power_law_1.2,0.8156352043151855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,8,power_law_1.2,1.0811391830444337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,8,power_law_1.2,1.476153564453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,8,power_law_1.2,1.7122880935668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,4,power_law_1.2,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,4,power_law_1.2,0.03273600041866302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,4,power_law_1.2,0.03128960132598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,4,power_law_1.2,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,4,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,4,power_law_1.2,0.03540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,4,power_law_1.2,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,4,power_law_1.2,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,4,power_law_1.2,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,4,power_law_1.2,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,4,power_law_1.2,0.03704319894313812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,4,power_law_1.2,0.038790398836135866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,4,power_law_1.2,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,4,power_law_1.2,0.03943679928779602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,4,power_law_1.2,0.045510399341583255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,4,power_law_1.2,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,4,power_law_1.2,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,4,power_law_1.2,0.0546239972114563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,4,power_law_1.2,0.05669119954109192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,4,power_law_1.2,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,4,power_law_1.2,0.07365760207176208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,4,power_law_1.2,0.09681280255317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,2,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,2,balanced,0.03585600107908249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,2,balanced,0.041738669077555336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,2,balanced,0.04420800010363261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,2,balanced,0.045968001087506614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,2,balanced,0.04584000011285146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,2,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,2,balanced,0.04774933556715647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,2,balanced,0.0472320020198822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,2,balanced,0.04796266555786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,2,balanced,0.04791999856630961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,2,balanced,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,2,balanced,0.04956800242265066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,2,balanced,0.052058666944503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,2,balanced,0.0620959997177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,2,balanced,0.06442133088906606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,2,balanced,0.0642986645301183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,2,balanced,0.08261866867542267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,2,balanced,0.08233066896597545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,2,balanced,0.10314133763313293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,2,balanced,0.11876799662907918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,2,balanced,0.15411733587582907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,2,balanced,0.19234132766723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,2,balanced,0.26930665969848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,2,balanced,0.3468000094095866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,2,balanced,0.5044800043106079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,4,power_law_1.01,0.06535680294036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,2,balanced,0.6521973212560018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,4,power_law_1.01,0.07419520020484924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,4,power_law_1.2,0.12541439533233642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,4,power_law_1.01,0.07859200239181519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,4,power_law_1.01,0.08564479947090149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,4,power_law_1.01,0.0871999979019165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,4,power_law_1.01,0.08803840279579163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,4,power_law_1.01,0.08760960102081299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,4,power_law_1.01,0.0900160014629364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,4,power_law_1.01,0.09030399918556213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,4,power_law_1.01,0.09056640267372132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,4,power_law_1.01,0.09282559752464295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,1,power_law_1.01,0.01973759979009628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,1,power_law_1.01,0.01990399956703186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,1,power_law_1.01,0.021164800226688384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,1,power_law_1.01,0.0211776003241539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,4,power_law_1.01,0.10287359952926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,4,power_law_1.01,0.11688319444656373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,4,power_law_1.01,0.12030719518661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,1,power_law_1.01,0.023238399624824525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,1,power_law_1.01,0.02844800055027008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,1,power_law_1.01,0.027699199318885804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,1,power_law_1.01,0.028863999247550964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,1,power_law_1.01,0.027987200021743774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,1,power_law_1.01,0.029292801022529603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,1,power_law_1.01,0.03081600069999695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,4,power_law_1.01,0.12973439693450928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,1,power_law_1.01,0.03131519854068756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,1,power_law_1.01,0.03224959969520569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,1,power_law_1.01,0.03501439988613129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,4,power_law_1.01,0.13546240329742432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,1,power_law_1.01,0.073471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,4,power_law_1.01,0.14359040260314943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,4,power_law_1.01,0.19466880559921265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,1,power_law_1.01,0.07557759881019592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,1,power_law_1.01,0.07868800163269044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,4,power_law_1.01,0.23779840469360353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,1,power_law_1.01,0.05504000186920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,4,power_law_1.01,0.3393791913986206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.01,0.06039040088653565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,4,power_law_1.01,0.4471104145050049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.01,0.08141440153121948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.01,0.10188800096511841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,4,power_law_1.2,0.1701248049736023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,4,power_law_1.01,0.574067211151123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.01,0.13720959424972534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,4,power_law_1.01,0.703219223022461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.01,0.1652672052383423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,4,power_law_1.01,1.123366355895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.01,0.23578879833221436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,4,power_law_1.01,1.8384960174560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,4,power_law_1.01,2.4626623153686524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,4,power_law_1.2,0.2054527997970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.01,0.30176639556884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,4,power_law_1.2,0.2922559976577759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,32,2,balanced,0.07127466797828674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,32,2,balanced,0.07460799813270569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,32,2,balanced,0.08316799998283386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,32,2,balanced,0.10502400000890096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,32,2,balanced,0.13452266653378805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,32,2,balanced,0.21068267027537027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,32,2,balanced,0.21855467557907104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,32,2,balanced,0.2339306672414144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,32,2,balanced,0.21823465824127197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,32,2,balanced,0.21585599581400552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,32,2,balanced,0.2214933236440023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,32,2,balanced,0.22528000672658285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,32,2,balanced,0.22039467096328735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,32,2,balanced,0.21545066436131796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,32,2,balanced,0.2239039937655131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,32,2,balanced,0.2250666618347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,32,2,balanced,0.23221333821614584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,32,2,balanced,0.25616000096003216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,32,2,balanced,0.2717439929644267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,32,2,balanced,0.30830933650334674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,32,2,balanced,0.34441065788269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,32,2,balanced,0.4296586513519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,32,2,balanced,0.5065493186314901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,32,2,balanced,0.7599733670552572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,32,2,balanced,0.9261546929677328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,32,2,balanced,1.322698672612508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,16,balanced,0.03757333258787791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,16,balanced,0.036144000788529716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,16,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,16,balanced,0.043605332573254905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,16,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,16,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,16,balanced,0.05115200082461039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,16,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,16,balanced,0.04996799925963084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,16,balanced,0.04981866478919983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,16,balanced,0.049653331438700356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,16,balanced,0.050053333242734276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,16,balanced,0.05169600248336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,16,balanced,0.050197333097457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,16,balanced,0.05379733443260193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,16,balanced,0.05590933561325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,16,balanced,0.056202664971351624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,16,balanced,0.05992533266544342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,16,balanced,0.06141333281993866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,16,balanced,0.06894933183987935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,16,balanced,0.07437333464622498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,16,balanced,0.09857066472371419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,16,balanced,0.11143466830253601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,16,balanced,0.14222932855288187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,32,power_law_1.01,0.08341119885444641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,32,power_law_1.01,0.08541439771652222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,32,power_law_1.01,0.08452479839324951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,32,power_law_1.01,0.08577280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,32,2,balanced,1.7109012603759766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,16,balanced,0.17293866475423178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,32,power_law_1.01,0.08561279773712158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,16,balanced,0.23248000939687094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,32,power_law_1.01,0.08354560136795045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,32,power_law_1.01,0.0795199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,32,power_law_1.01,0.08535680174827576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,32,power_law_1.01,0.08488320112228394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,32,power_law_1.01,0.08170239925384522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,32,power_law_1.01,0.08268160223960877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,32,power_law_1.01,0.08396160006523132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,16,balanced,0.29583466053009033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,32,power_law_1.01,0.08886399865150452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,32,power_law_1.01,0.08977280259132385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,32,power_law_1.01,0.09595519900321961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,32,power_law_1.01,0.09475839734077454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,32,power_law_1.01,0.09998080134391785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,32,power_law_1.01,0.11449600458145141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,32,power_law_1.01,0.12599040269851686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,32,power_law_1.01,0.14692480564117433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.01,0.4373568058013916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,32,power_law_1.01,0.16189440488815307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,32,power_law_1.01,0.19515520334243774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,4,power_law_1.01,2.774950408935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,32,power_law_1.01,0.24000000953674316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,32,power_law_1.01,0.3142528057098389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,32,power_law_1.01,0.39174399375915525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,32,power_law_1.01,0.5443456172943115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,32,power_law_1.01,0.6625088214874267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.01,0.5732672214508057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,4,power_law_1.2,0.3959295988082886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,1,power_law_1.2,0.05355520248413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,1,power_law_1.2,0.0581055998802185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,1,power_law_1.2,0.0639743983745575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,1,power_law_1.2,0.08310400247573853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,1,power_law_1.2,0.10435199737548828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,1,power_law_1.2,0.12149120569229126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,1,power_law_1.2,0.15282560586929322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,1,power_law_1.2,0.16057599782943727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,1,power_law_1.2,0.16643199920654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,1,power_law_1.2,0.1715775966644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,1,power_law_1.2,0.18260480165481568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,1,power_law_1.2,0.196288001537323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,1,power_law_1.2,0.2066431999206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,1,power_law_1.2,0.2212928056716919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,1,power_law_1.2,0.24057600498199463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,1,power_law_1.2,0.25603199005126953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,64,power_law_1.01,0.06220160126686096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,1,power_law_1.2,0.2551743984222412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,64,power_law_1.01,0.059961599111557004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,1,power_law_1.2,0.30979840755462645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,64,power_law_1.01,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,64,power_law_1.01,0.061952000856399535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,64,power_law_1.01,0.06156799793243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,64,power_law_1.01,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,64,power_law_1.01,0.06468480229377746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,1,power_law_1.2,0.35360639095306395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,64,power_law_1.01,0.06727679967880248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,1,power_law_1.2,0.4441023826599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,64,power_law_1.01,0.06578559875488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,1,power_law_1.2,0.5282559871673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,1,power_law_1.2,0.7015744209289551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,1,power_law_1.2,0.8922752380371094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,2,power_law_1.2,0.03596799969673157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,2,power_law_1.2,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,1,power_law_1.2,1.2162752151489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,2,power_law_1.2,0.06318079829216003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,64,power_law_1.01,0.06477439999580384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,2,power_law_1.2,0.0858240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,64,power_law_1.01,0.06763520240783691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,2,power_law_1.2,0.10213760137557984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,1,power_law_1.2,1.5531264305114747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,64,power_law_1.01,0.06933119893074036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,2,power_law_1.2,0.14426239728927612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,64,power_law_1.01,0.06982399821281433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,2,power_law_1.2,0.14714879989624025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,64,power_law_1.01,0.07131519913673401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,2,power_law_1.2,0.15726720094680785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,64,power_law_1.01,0.07553920149803162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,1,power_law_1.2,2.2221439361572264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,2,power_law_1.2,0.163372802734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,64,power_law_1.01,0.07541120052337646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,2,power_law_1.2,0.17045120000839234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,64,power_law_1.01,0.08206080198287964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,2,power_law_1.2,0.16995840072631835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,64,power_law_1.01,0.09843840003013611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,2,power_law_1.2,0.1725376009941101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,64,power_law_1.01,0.10670080184936523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,1,power_law_1.2,2.9107776641845704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,2,power_law_1.2,0.1810495972633362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,64,power_law_1.01,0.128057599067688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,2,power_law_1.2,0.18272639513015748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,64,power_law_1.01,0.15411200523376464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,2,power_law_1.2,0.1947648048400879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,64,power_law_1.01,0.19537919759750366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,2,power_law_1.2,0.20244479179382324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,16,balanced,0.04484800000985464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,2,power_law_1.2,0.21310079097747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,2,power_law_1.2,0.22918400764465333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.2,0.2507839918136597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.2,0.3094847917556763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.2,0.35196800231933595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.2,0.40543360710144044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,16,balanced,0.04568533102671305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.2,0.4730368137359619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.2,0.5787392139434815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,64,power_law_1.01,0.23313279151916505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.2,0.7057919979095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,64,power_law_1.01,0.33956480026245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,16,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,16,balanced,0.05239466826121012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,16,balanced,0.07789333164691925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,16,balanced,0.08473066488901775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,16,balanced,0.0853760043780009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,16,balanced,0.08634666601816814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,16,balanced,0.08661866188049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,16,balanced,0.08710400263468425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,16,balanced,0.0881226658821106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,16,balanced,0.08980799714724223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,16,balanced,0.09242133299509685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.2,0.9636799812316894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,16,balanced,0.09282666444778442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,16,balanced,0.09763200084368388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,16,balanced,0.09927999973297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,16,balanced,0.10233066479365031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,16,balanced,0.11875733733177185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,16,balanced,0.12113599975903828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,16,balanced,0.14402133226394653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,16,balanced,0.15637333194414774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,16,balanced,0.20094400644302368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,16,balanced,0.22238399585088095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,16,balanced,0.33431466420491535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,16,balanced,0.3725279966990153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,16,balanced,0.5213386615117391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,64,power_law_1.01,0.41137280464172366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,16,balanced,0.6588213443756104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,64,power_law_1.01,0.5808703899383545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,64,power_law_1.01,0.7736639976501465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,8,power_law_1.01,0.04968959987163544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,8,power_law_1.01,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,8,power_law_1.01,0.053260797262191774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,8,power_law_1.01,0.057657599449157715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,8,power_law_1.01,0.058796799182891844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,8,power_law_1.01,0.06175360083580017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,8,power_law_1.01,0.061324799060821535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,8,power_law_1.01,0.06338559985160827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,8,power_law_1.01,0.0633791983127594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,8,power_law_1.01,0.06423680186271667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,8,power_law_1.01,0.06352639794349671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,8,power_law_1.01,0.0655232012271881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,8,power_law_1.01,0.06703360080718994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,8,power_law_1.01,0.06826239824295044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,8,power_law_1.01,0.07409279942512512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,8,power_law_1.01,0.0762112021446228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,8,power_law_1.01,0.08223360180854797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,8,power_law_1.01,0.09255679845809936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,8,power_law_1.01,0.09996799826622009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,8,power_law_1.01,0.1232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,8,power_law_1.01,0.1417088031768799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,8,power_law_1.01,0.18630399703979492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,8,power_law_1.01,0.22646400928497315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,8,power_law_1.01,0.27964799404144286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,8,power_law_1.01,0.36376960277557374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,8,power_law_1.01,0.4980288028717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,8,power_law_1.01,0.6763455867767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.2,1.1927616119384765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,32,balanced,0.03977599988381068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,32,balanced,0.035599999129772186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,32,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,32,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,32,balanced,0.03600533306598663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,32,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,32,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,32,balanced,0.03804266701141993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,32,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,32,balanced,0.03764266769091288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,32,balanced,0.037647999823093414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,32,balanced,0.03938666731119156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,32,balanced,0.03858133405447006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,32,balanced,0.040821333726247154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,32,balanced,0.04351999859015147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,32,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,32,balanced,0.04378133515516917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,32,balanced,0.045850664377212524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,32,balanced,0.04700266818205515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,32,balanced,0.05381333331267039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,32,balanced,0.05407466491063436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,32,balanced,0.06493866443634033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,32,balanced,0.07043733199437459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,32,balanced,0.08967999617258708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,32,balanced,0.10108266274134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,32,balanced,0.12970667084058127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,32,balanced,0.15474133690198263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,8,power_law_1.01,0.02800639867782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,8,power_law_1.01,0.029158401489257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,8,power_law_1.01,0.029811200499534608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,8,power_law_1.01,0.028307199478149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,8,power_law_1.01,0.0321152001619339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,8,power_law_1.01,0.03216640055179596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,8,power_law_1.01,0.032518398761749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,8,power_law_1.01,0.030636799335479737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,8,power_law_1.01,0.03240320086479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,8,power_law_1.01,0.03383679986000061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,8,power_law_1.01,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,8,power_law_1.01,0.040115201473236085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,8,power_law_1.01,0.042668798565864564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,8,power_law_1.01,0.04239999949932098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,8,power_law_1.01,0.04721280038356781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,8,power_law_1.01,0.04898560047149658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,8,power_law_1.01,0.05172479748725891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,8,power_law_1.01,0.058483201265335086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,8,power_law_1.01,0.058143997192382814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,8,power_law_1.01,0.07214080095291138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,8,power_law_1.01,0.07978240251541138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,8,power_law_1.01,0.10364799499511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,8,power_law_1.01,0.11128319501876831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,8,power_law_1.01,0.14815360307693481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,8,power_law_1.01,0.18595199584960936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,8,power_law_1.01,0.2756671905517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,8,power_law_1.01,0.34955520629882814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,16,balanced,0.03655466685692469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,16,balanced,0.03580799947182337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,16,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,16,balanced,0.04182933270931244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,16,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,16,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,16,balanced,0.04562666515509287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,16,balanced,0.045824001232783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,16,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,16,balanced,0.046336000164349876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,16,balanced,0.04744000236193339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,16,balanced,0.04762666424115499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,16,balanced,0.04888000090916952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,16,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,16,balanced,0.05189866820971171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,16,balanced,0.05377600093682607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,16,balanced,0.0540533314148585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,16,balanced,0.058143998185793556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,16,balanced,0.05881066620349884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,16,balanced,0.06832000116507213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,16,balanced,0.07229333122571309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,16,balanced,0.09090133508046468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,16,balanced,0.10572266578674316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,16,balanced,0.13737066586812338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,16,balanced,0.16709866126378378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,16,balanced,0.2242506742477417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,16,balanced,0.2834773262341817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,32,power_law_1.2,0.025721600651741026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,32,power_law_1.2,0.02162559926509857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,32,power_law_1.2,0.021491199731826782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,32,power_law_1.2,0.024383999407291412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,32,power_law_1.2,0.02330880016088486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,32,power_law_1.2,0.026528000831604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,32,power_law_1.2,0.02632960081100464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,32,power_law_1.2,0.025440001487731935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,32,power_law_1.2,0.029824000597000123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,32,power_law_1.2,0.031251201033592226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,32,power_law_1.2,0.03107840120792389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,32,power_law_1.2,0.031673601269721983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,32,power_law_1.2,0.031788799166679385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,32,power_law_1.2,0.03200640082359314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,2,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,32,power_law_1.2,0.03452160060405731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,2,balanced,0.05184000233809153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,32,power_law_1.2,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,32,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,32,power_law_1.2,0.04066559970378876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,32,power_law_1.2,0.042771199345588685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,32,power_law_1.2,0.052025598287582395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,32,power_law_1.2,0.05141760110855102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,2,balanced,0.06541866560777028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,2,balanced,0.09475732843081157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,2,balanced,0.14007467031478882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,2,balanced,0.22935465971628824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,2,balanced,0.22721600532531738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,2,balanced,0.2306613326072693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,2,balanced,0.23101866245269775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,2,balanced,0.23405865828196207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,2,balanced,0.2360373338063558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,2,balanced,0.23748799165089926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,2,balanced,0.2422879934310913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,2,balanced,0.2472533384958903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,2,balanced,0.254805326461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,2,balanced,0.26131733258565265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,2,balanced,0.2683200041453044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,2,balanced,0.2900213400522868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,2,balanced,0.3071253299713135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,2,balanced,0.3460373481114705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,2,balanced,0.3818399906158447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,2,balanced,0.4792106548945109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,2,balanced,0.5389440059661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,2,balanced,0.7379626433054606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,32,power_law_1.2,0.0636672019958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,2,balanced,0.8727946281433105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,2,balanced,1.244213342666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,2,balanced,1.5810400644938152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,32,power_law_1.2,0.08362240195274354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,32,power_law_1.2,0.11041920185089112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,32,power_law_1.2,0.14422399997711183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,32,power_law_1.2,0.17292159795761108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,32,power_law_1.2,0.2032831907272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,4,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,4,balanced,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,4,balanced,0.0621013343334198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,4,balanced,0.06643733382225037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,4,balanced,0.07844799757003784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,4,balanced,0.10577600200970967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,4,balanced,0.10941867033640544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,4,balanced,0.10932266712188721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,4,balanced,0.11123733719189961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,4,balanced,0.1149013340473175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,4,balanced,0.113237331310908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,4,balanced,0.11481600006421407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,4,balanced,0.11703466375668843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,4,balanced,0.12166399757067363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,4,balanced,0.12778133153915405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,4,balanced,0.12943999965985617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,4,balanced,0.13396267096201578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,4,balanced,0.14866666992505392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,4,balanced,0.16064000129699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,4,balanced,0.18672533830006918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,4,balanced,0.20971200863520303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,4,balanced,0.2630186676979065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,4,balanced,0.3211733301480611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,4,balanced,0.4604959885279338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,4,balanced,0.5796906550725301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,4,balanced,0.8388160069783529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,4,balanced,1.088320016860962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,4,power_law_1.01,0.07260159850120544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,4,power_law_1.01,0.08328959941864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,4,power_law_1.01,0.09383040070533752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,4,power_law_1.01,0.10536320209503174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,4,power_law_1.01,0.1131775975227356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,4,power_law_1.01,0.11758719682693482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,4,power_law_1.01,0.12845439910888673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,4,power_law_1.01,0.1308351993560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,4,power_law_1.01,0.12973439693450928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,4,power_law_1.01,0.13288320302963258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,4,power_law_1.01,0.13196799755096436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,4,power_law_1.01,0.14030719995498658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,4,power_law_1.01,0.13781119585037233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,4,power_law_1.01,0.14347519874572753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,4,power_law_1.01,0.14888960123062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,4,power_law_1.01,0.1538624048233032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,4,power_law_1.01,0.15879679918289186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,4,power_law_1.01,0.17475199699401855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,4,power_law_1.01,0.19510400295257568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,4,power_law_1.01,0.2408128023147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,4,power_law_1.01,0.29057281017303466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,4,power_law_1.01,0.3782144069671631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,4,power_law_1.01,0.4809279918670654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,4,power_law_1.01,0.67259521484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,4,power_law_1.01,0.8339008331298828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,4,power_law_1.01,1.258835220336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,4,power_law_1.01,1.7066944122314454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,4,power_law_1.01,0.07283200025558471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,4,power_law_1.01,0.0836736023426056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,4,power_law_1.01,0.0907904028892517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,4,power_law_1.01,0.10913920402526855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,4,power_law_1.01,0.11299200057983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,4,power_law_1.01,0.13033599853515626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,4,power_law_1.01,0.14264320135116576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,4,power_law_1.01,0.14101120233535766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,4,power_law_1.01,0.1427199959754944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,4,power_law_1.01,0.14647040367126465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,4,power_law_1.01,0.1422144055366516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,4,power_law_1.01,0.15012480020523072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,4,power_law_1.01,0.15159039497375487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,4,power_law_1.01,0.15567359924316407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,4,power_law_1.01,0.15711359977722167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,4,power_law_1.01,0.16104320287704468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,4,power_law_1.01,0.16517759561538697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,4,power_law_1.01,0.18165760040283202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,4,power_law_1.01,0.19779200553894044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,4,power_law_1.01,0.2357248067855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,4,power_law_1.01,0.25985279083251955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,4,power_law_1.01,0.33829760551452637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,4,power_law_1.01,0.39095680713653563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,4,power_law_1.01,0.5457920074462891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,4,power_law_1.01,0.7012288093566894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,4,power_law_1.01,0.9727871894836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,4,power_law_1.01,1.2090815544128417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,16,balanced,0.019146667172511418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,16,balanced,0.018992000569899876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,16,balanced,0.019653332730134327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,16,balanced,0.019359999646743137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,16,balanced,0.021333334346612293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,16,balanced,0.02317333221435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,16,balanced,0.025472000241279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,16,balanced,0.025450666745503742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,16,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,16,balanced,0.03531199942032496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,16,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,16,balanced,0.03923733284076055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,16,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,16,balanced,0.03486400097608566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,16,balanced,0.042250668009122215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,16,balanced,0.04199466605981191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,16,balanced,0.039120001097520195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,16,balanced,0.046015997727712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,16,balanced,0.04897066454092661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,16,balanced,0.057914664347966514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,16,balanced,0.06781866649786632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,16,balanced,0.08295999964078267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,16,balanced,0.09891733527183533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,16,balanced,0.13616533080736795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,16,balanced,0.17203199863433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,16,balanced,0.2424373428026835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,16,balanced,0.3146666685740153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,2,power_law_1.2,0.017875200510025023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,2,power_law_1.2,0.01958400011062622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,2,power_law_1.2,0.022912000119686127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,2,power_law_1.2,0.028780800104141236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,2,power_law_1.2,0.03849599957466125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,2,power_law_1.2,0.043507200479507444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,2,power_law_1.2,0.05060480237007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,2,power_law_1.2,0.05120639801025391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,2,power_law_1.2,0.05381760001182556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,2,power_law_1.2,0.05466880202293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,2,power_law_1.2,0.05704960227012634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,2,power_law_1.2,0.0582144021987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,2,power_law_1.2,0.06103039979934692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,2,power_law_1.2,0.06125440001487732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,2,power_law_1.2,0.06350079774856568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,2,power_law_1.2,0.06643840074539184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,2,power_law_1.2,0.07235199809074402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,2,power_law_1.2,0.08201599717140198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,2,power_law_1.2,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,2,power_law_1.2,0.10890239477157593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,2,power_law_1.2,0.11061760187149047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,2,power_law_1.2,0.13945599794387817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,2,power_law_1.2,0.17475199699401855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,2,power_law_1.2,0.2162303924560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,2,power_law_1.2,0.2554559946060181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,2,power_law_1.2,0.3397952079772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,2,power_law_1.2,0.425600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,32,power_law_1.2,0.10257279872894287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,32,power_law_1.2,0.09924479722976684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,32,power_law_1.2,0.09826560020446777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,32,power_law_1.2,0.10540159940719604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,32,power_law_1.2,0.10169600248336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,32,power_law_1.2,0.09962239861488342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,32,power_law_1.2,0.10347520112991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,32,power_law_1.2,0.10409599542617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,32,power_law_1.2,0.10497280359268188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,32,power_law_1.2,0.10378880500793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,32,power_law_1.2,0.10655360221862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,32,power_law_1.2,0.1109760046005249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,32,power_law_1.2,0.11301120519638061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,32,power_law_1.2,0.11560319662094116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,32,power_law_1.2,0.12242560386657715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,32,power_law_1.2,0.12986880540847778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,32,power_law_1.2,0.1472640037536621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,32,power_law_1.2,0.16723840236663817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,32,power_law_1.2,0.17404160499572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,32,power_law_1.2,0.2213184118270874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,32,power_law_1.2,0.2525376081466675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,32,power_law_1.2,0.32101759910583494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,32,power_law_1.2,0.3199359893798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,32,power_law_1.2,0.426470422744751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,32,power_law_1.2,0.5726975917816162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,32,power_law_1.2,0.8216383934020997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,32,power_law_1.2,0.9982336044311524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,8,power_law_1.2,0.03185920119285583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,8,power_law_1.2,0.032339200377464294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,8,power_law_1.2,0.03172479867935181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,8,power_law_1.2,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,8,power_law_1.2,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,8,power_law_1.2,0.037599998712539676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,8,power_law_1.2,0.037676799297332766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,8,power_law_1.2,0.03925119936466217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,8,power_law_1.2,0.03982079923152924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,8,power_law_1.2,0.04071680009365082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,8,power_law_1.2,0.04232319891452789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,8,power_law_1.2,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,8,power_law_1.2,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,8,power_law_1.2,0.047014400362968445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,8,power_law_1.2,0.0545087993144989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,8,power_law_1.2,0.06027519702911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,8,power_law_1.2,0.058303999900817874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,8,power_law_1.2,0.0667519986629486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,8,power_law_1.2,0.08632959723472595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,8,power_law_1.2,0.10250879526138305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,8,power_law_1.2,0.1217087984085083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,8,power_law_1.2,0.17570559978485106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,8,power_law_1.2,0.20035839080810547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,8,power_law_1.2,0.2767616033554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,8,power_law_1.2,0.3951040029525757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,8,power_law_1.2,0.5835584163665771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,1,balanced,0.039919999738534294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,1,balanced,0.05602133274078369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,1,balanced,0.0775733341773351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,1,balanced,0.12243200341860454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,1,balanced,0.20948266983032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,1,balanced,0.21518399318059286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,1,balanced,0.21834667523701987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,1,balanced,0.2156320015589396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,1,balanced,0.21782932678858438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,1,balanced,0.21755200624465942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,1,balanced,0.2226453423500061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,1,balanced,0.22473599513371786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,1,balanced,0.2270080049832662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,1,balanced,0.23286932706832886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,1,balanced,0.24143999814987183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,1,balanced,0.24764267603556314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,1,balanced,0.2564213275909424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,1,balanced,0.278383990128835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,1,balanced,0.3039733370145162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,1,balanced,0.37652798493703205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,1,balanced,0.4151253302892049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,1,balanced,0.5505866607030233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,1,balanced,0.6171360015869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,1,balanced,0.8574026425679525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,1,balanced,1.0563626289367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,1,balanced,1.5163946151733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,1,balanced,1.9398399988810222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,64,balanced,0.042634665966033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,64,balanced,0.04177066683769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,64,balanced,0.04180799921353658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,64,balanced,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,64,balanced,0.045850664377212524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,64,balanced,0.049653331438700356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,64,balanced,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,64,balanced,0.05006400247414907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,64,balanced,0.05795733133951823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,64,balanced,0.05608533322811127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,64,balanced,0.05653866628805796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,64,balanced,0.05606399973233541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,64,balanced,0.054485330979029335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,64,balanced,0.06014933188756307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,64,balanced,0.0668639987707138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,64,balanced,0.0643093337615331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,64,balanced,0.06796266635258992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,64,balanced,0.07807466884454091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,64,balanced,0.078575998544693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,64,balanced,0.09282666444778442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,64,balanced,0.10243200262387593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,64,balanced,0.12648000319798788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,64,balanced,0.14828266700108847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,64,balanced,0.194650669892629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,64,balanced,0.23598933219909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,64,balanced,0.3309333324432373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,64,balanced,0.41439465681711835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,8,power_law_1.2,0.6900991916656494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,32,balanced,0.043141335248947144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,32,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,32,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,32,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,32,balanced,0.06010133524735769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,32,balanced,0.07685866455237071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,32,balanced,0.07671999931335449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,32,balanced,0.07819733520348866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,32,balanced,0.07955733438332875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,32,balanced,0.07920533418655396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,32,balanced,0.07858133316040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,32,balanced,0.08054399987061818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,32,balanced,0.08197333415349324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,32,balanced,0.08109866579373677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,32,balanced,0.08709866801897685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,32,balanced,0.08826133608818054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,32,balanced,0.09114666779836018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,32,balanced,0.09726400176684062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,32,balanced,0.09891200065612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,32,balanced,0.1071626643339793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,32,balanced,0.11711466312408447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,32,balanced,0.1339306632677714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,32,balanced,0.1523253321647644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,32,balanced,0.187717338403066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,32,balanced,0.22404799858729044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,32,balanced,0.2962453365325928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,32,balanced,0.3640480041503906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,16,power_law_1.2,0.08826239705085755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,16,power_law_1.2,0.12023680210113526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,16,power_law_1.2,0.11614079475402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,16,power_law_1.2,0.1210368037223816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,16,power_law_1.2,0.12143360376358033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,16,power_law_1.2,0.10103679895401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,16,power_law_1.2,0.11368319988250733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,16,power_law_1.2,0.12005759477615356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,16,power_law_1.2,0.11783679723739623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,16,power_law_1.2,0.11619199514389038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,16,power_law_1.2,0.11939840316772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,16,power_law_1.2,0.12017920017242431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,16,power_law_1.2,0.12446080446243286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,16,power_law_1.2,0.1271232008934021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,16,power_law_1.2,0.1414528012275696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,16,power_law_1.2,0.14305280447006224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,16,power_law_1.2,0.14545279741287231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,16,power_law_1.2,0.16615040302276612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,16,power_law_1.2,0.17508480548858643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,16,power_law_1.2,0.2134079933166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,16,power_law_1.2,0.24273920059204102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,16,power_law_1.2,0.32963199615478517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,16,power_law_1.2,0.39962880611419677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,16,power_law_1.2,0.5674304008483887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,16,power_law_1.2,0.7446591854095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,16,power_law_1.2,1.044268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,16,power_law_1.2,1.5595328330993652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,16,power_law_1.2,0.12761600017547609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,16,power_law_1.2,0.20890240669250487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,16,power_law_1.2,0.1905408024787903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,16,power_law_1.2,0.20227839946746826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,16,power_law_1.2,0.2137984037399292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,16,power_law_1.2,0.185971200466156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,16,power_law_1.2,0.19914239645004272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,16,power_law_1.2,0.17860480546951293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,16,power_law_1.2,0.22016000747680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,16,power_law_1.2,0.20222721099853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,16,power_law_1.2,0.1904255986213684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,16,power_law_1.2,0.20217599868774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,16,power_law_1.2,0.20765440464019774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,16,power_law_1.2,0.20855679512023925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,16,power_law_1.2,0.22772479057312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,16,power_law_1.2,0.23797121047973632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,16,power_law_1.2,0.244051194190979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,16,power_law_1.2,0.295033597946167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,16,power_law_1.2,0.2749056100845337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,16,power_law_1.2,0.3675008058547974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,16,power_law_1.2,0.34466559886932374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,16,power_law_1.2,0.43702402114868166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,16,power_law_1.2,0.5120448112487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,16,power_law_1.2,0.6693376064300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,16,power_law_1.2,0.7987520217895507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,16,power_law_1.2,1.052012825012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,16,power_law_1.2,1.5462592124938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,8,power_law_1.2,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,8,power_law_1.2,0.06231679916381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,8,power_law_1.2,0.05934720039367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,8,power_law_1.2,0.0652288019657135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,8,power_law_1.2,0.07246080040931702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,8,power_law_1.2,0.06357759833335877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,8,power_law_1.2,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,8,power_law_1.2,0.07523840069770812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,8,power_law_1.2,0.07790079712867737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,8,power_law_1.2,0.07999359965324401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,8,power_law_1.2,0.07802240252494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,8,power_law_1.2,0.08131840229034423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,8,power_law_1.2,0.08513919711112976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,8,power_law_1.2,0.08513280153274536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,8,power_law_1.2,0.09846400022506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,8,power_law_1.2,0.10367360115051269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,8,power_law_1.2,0.1089792013168335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,8,power_law_1.2,0.11974400281906128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,8,power_law_1.2,0.13114240169525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,8,power_law_1.2,0.15685759782791137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,8,power_law_1.2,0.1951807975769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,8,power_law_1.2,0.2383359909057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,8,power_law_1.2,0.29013121128082275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,8,power_law_1.2,0.41597437858581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,8,power_law_1.2,0.5064127922058106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,8,power_law_1.2,0.6649151802062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,8,power_law_1.2,0.9014719963073731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,64,power_law_1.01,0.02449280023574829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,64,power_law_1.01,0.021817600727081297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,64,power_law_1.01,0.022144000232219695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,64,power_law_1.01,0.02142080068588257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,64,power_law_1.01,0.021055999398231506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,64,power_law_1.01,0.02253440022468567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,64,power_law_1.01,0.022598400712013245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,64,power_law_1.01,0.025964799523353576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,64,power_law_1.01,0.026169601082801818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,64,power_law_1.01,0.02648960053920746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,64,power_law_1.01,0.04429439902305603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,64,power_law_1.01,0.04442239999771118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,64,power_law_1.01,0.04345600008964538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,64,power_law_1.01,0.04462080001831055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,64,power_law_1.01,0.04368639886379242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,64,power_law_1.01,0.04699519872665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,64,power_law_1.01,0.04880639910697937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,64,power_law_1.01,0.051551997661590576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.01,0.05557119846343994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.01,0.0650816023349762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.01,0.07740160226821899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.01,0.10792959928512573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.01,0.12741119861602784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.01,0.17175040245056153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.01,0.21917440891265869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.01,0.31020801067352294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.01,0.420300817489624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,32,power_law_1.2,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,32,power_law_1.2,0.02852480113506317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,32,power_law_1.2,0.02855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,32,power_law_1.2,0.029996800422668456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,32,power_law_1.2,0.03052160143852234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,32,power_law_1.2,0.031718400120735166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,32,power_law_1.2,0.03199360072612763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,32,power_law_1.2,0.03196159899234772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,32,power_law_1.2,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,32,power_law_1.2,0.032601600885391234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,32,power_law_1.2,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,32,power_law_1.2,0.034246399998664856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,32,power_law_1.2,0.0349375993013382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,32,power_law_1.2,0.036268800497055054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,32,power_law_1.2,0.04068480134010315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,32,power_law_1.2,0.040703999996185306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,32,power_law_1.2,0.04298880100250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,32,power_law_1.2,0.047891199588775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,32,power_law_1.2,0.05059840083122254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,32,power_law_1.2,0.0597055971622467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,32,power_law_1.2,0.06650879979133606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,32,power_law_1.2,0.08578559756278992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,32,power_law_1.2,0.10531200170516967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,32,power_law_1.2,0.15383679866790773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,1,power_law_1.2,0.05330560207366943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,1,power_law_1.2,0.05717759728431702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,1,power_law_1.2,0.06567680239677429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,1,power_law_1.2,0.09114239811897278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,256,power_law_1.01,0.018137599527835845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,1,power_law_1.2,0.11388800144195557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,256,power_law_1.01,0.018329599499702455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,1,power_law_1.2,0.13694080114364623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,256,power_law_1.01,0.01910399943590164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,1,power_law_1.2,0.17799680233001708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,256,power_law_1.01,0.021331200003623964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,1,power_law_1.2,0.18982399702072145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,256,power_law_1.01,0.03560320138931274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,1,power_law_1.2,0.19466240406036378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,256,power_law_1.01,0.029446399211883544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,1,power_law_1.2,0.19816319942474364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,1,balanced,0.052895997961362205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,256,power_law_1.01,0.03054719865322113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,1,power_law_1.2,0.21571838855743408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,256,power_law_1.01,0.030169600248336793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,1,power_law_1.2,0.226310396194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,256,power_law_1.01,0.030060800909996032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,1,power_law_1.2,0.23416318893432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,256,power_law_1.01,0.03059839904308319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,256,power_law_1.01,0.03048959970474243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,256,power_law_1.01,0.03108479976654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,256,power_law_1.01,0.030931198596954347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,256,power_law_1.01,0.03130879998207092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,256,power_law_1.01,0.03215999901294708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,256,power_law_1.01,0.033792001008987424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,1,power_law_1.2,0.25382399559020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,256,power_law_1.01,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,1,power_law_1.2,0.2730815887451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,256,power_law_1.01,0.037036800384521486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,1,power_law_1.2,0.28645761013031007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,256,power_law_1.01,0.037574398517608645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,1,balanced,0.0641653339068095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,256,power_law_1.01,0.04121600091457367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,1,balanced,0.085125337044398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,1,balanced,0.1185653309027354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,256,power_law_1.01,0.04552960097789764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,256,power_law_1.01,0.05966079831123352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,256,power_law_1.01,0.07267839908599853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,256,power_law_1.01,0.09781119823455811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,256,power_law_1.01,0.12488960027694702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,1,power_law_1.2,0.28331520557403567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,1,balanced,0.18570133050282797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,1,power_law_1.2,0.3383935928344727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,1,balanced,0.31405866146087646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,1,balanced,0.32266666491826373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,1,power_law_1.2,0.3799360036849976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,1,balanced,0.32108267148335773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,1,balanced,0.3245760003725688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,1,power_law_1.2,0.4771967887878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,1,power_law_1.2,0.5576640129089355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,256,power_law_1.01,0.16914559602737428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,1,power_law_1.2,0.7347839832305908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,256,power_law_1.01,0.2230015993118286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,1,power_law_1.2,0.9256319999694824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,1,balanced,0.32233599821726483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,1,balanced,0.32822932799657184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,1,power_law_1.2,1.2708160400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,1,balanced,0.33797868092854816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,1,balanced,0.3382026751836141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,1,balanced,0.34515198071797687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,1,balanced,0.35339732964833576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,1,power_law_1.2,1.5974783897399902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,1,balanced,0.3568906784057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,1,balanced,0.3703039884567261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,1,balanced,0.40088534355163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,1,balanced,0.42690134048461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,32,power_law_1.2,0.18740479946136473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,1,balanced,0.4885066747665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,1,power_law_1.2,2.269388771057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,1,balanced,0.5386399825414022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,1,balanced,0.6775573094685873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,1,balanced,0.7913653055826823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,1,power_law_1.2,2.970182418823242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,1,balanced,1.0976746877034504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,1,balanced,1.327386697133382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,1,balanced,1.9262879689534504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,1,balanced,2.4124372800191245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,32,power_law_1.2,0.23530879020690917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,32,power_law_1.2,0.3230655908584595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,64,balanced,0.03326933334271113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,64,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,64,balanced,0.031152000029881794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,64,balanced,0.02991466720898946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,64,balanced,0.031119999786218006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,64,balanced,0.03148266673088074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,64,balanced,0.03124266614516576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,64,balanced,0.03159466634194056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,64,balanced,0.031397332747777305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,64,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,64,balanced,0.03327466547489166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,64,balanced,0.033471999069054924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,64,balanced,0.031386665999889374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,64,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,64,balanced,0.03596800069014231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,64,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,64,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,64,balanced,0.037317333122094475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,64,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,64,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,64,balanced,0.0413973331451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,64,balanced,0.04572799801826477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,64,balanced,0.04555733501911163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,64,balanced,0.05561600128809611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,64,balanced,0.06012799839178721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,64,balanced,0.07240533332029979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,64,balanced,0.08236266672611237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,256,power_law_1.01,0.05869439840316772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,256,power_law_1.01,0.058297598361968996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,256,power_law_1.01,0.05882880091667175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,256,power_law_1.01,0.06110079884529114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,256,power_law_1.01,0.059961599111557004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,256,power_law_1.01,0.061260801553726194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,256,power_law_1.01,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,256,power_law_1.01,0.06266239881515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,256,power_law_1.01,0.06328319907188415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,256,power_law_1.01,0.063264000415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,256,power_law_1.01,0.06239359974861145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,16,power_law_1.01,0.073990398645401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,16,power_law_1.01,0.06989439725875854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,16,power_law_1.01,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,16,power_law_1.01,0.07059839963912964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,16,power_law_1.01,0.07181439995765686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,256,power_law_1.01,0.06419199705123901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,256,power_law_1.01,0.06458879709243774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,256,power_law_1.01,0.06414080262184144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,256,power_law_1.01,0.06778879761695862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,256,power_law_1.01,0.06789759993553161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,256,power_law_1.01,0.07022719979286193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,256,power_law_1.01,0.08533120155334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,256,power_law_1.01,0.08608639836311341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,256,power_law_1.01,0.11151360273361206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,256,power_law_1.01,0.13823360204696655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,256,power_law_1.01,0.183571195602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,256,power_law_1.01,0.22128000259399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,256,power_law_1.01,0.3126271963119507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,16,power_law_1.01,0.06895359754562377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,16,power_law_1.01,0.07360000014305115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,16,power_law_1.01,0.07215999960899352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,16,power_law_1.01,0.07303680181503296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,16,power_law_1.01,0.07227519750595093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,16,power_law_1.01,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,16,power_law_1.01,0.07509120106697083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,16,power_law_1.01,0.0745855987071991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,16,power_law_1.01,0.08085759878158569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,16,power_law_1.01,0.08716800212860107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,16,power_law_1.01,0.0884223997592926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,256,power_law_1.01,0.3937664031982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,16,power_law_1.01,0.09353600144386291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,16,power_law_1.01,0.10988160371780395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,256,power_law_1.01,0.5574079990386963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,16,power_law_1.01,0.11177599430084229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,256,power_law_1.01,0.7368127822875976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,16,power_law_1.01,0.12792320251464845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,16,power_law_1.01,0.14925440549850463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,16,power_law_1.01,0.18072320222854615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,16,power_law_1.01,0.22200961112976075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,16,power_law_1.01,0.2966655969619751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,16,power_law_1.01,0.3880000114440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,16,power_law_1.01,0.5181568145751954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,16,power_law_1.01,0.6115200042724609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,64,power_law_1.2,0.16702719926834106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,64,power_law_1.2,0.1643839955329895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,64,power_law_1.2,0.1608448028564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,64,power_law_1.2,0.1653887987136841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,64,power_law_1.2,0.15509120225906373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,64,power_law_1.2,0.14869760274887084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,64,power_law_1.2,0.13971199989318847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,64,power_law_1.2,0.1677440047264099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,2,power_law_1.01,0.07046399712562561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,64,power_law_1.2,0.15927679538726808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,2,power_law_1.01,0.08295040130615235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,2,power_law_1.01,0.08702719807624817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,64,power_law_1.2,0.15965440273284912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,2,power_law_1.01,0.12717440128326415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,64,power_law_1.2,0.14640640020370482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,2,power_law_1.01,0.1173248052597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,2,power_law_1.01,0.12570879459381104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,64,power_law_1.2,0.15911680459976196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,2,power_law_1.01,0.1314687967300415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,2,power_law_1.01,0.13328640460968016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,2,power_law_1.01,0.13586560487747193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,2,power_law_1.01,0.13441920280456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,2,power_law_1.01,0.141593599319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,2,power_law_1.01,0.14422399997711183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,2,power_law_1.01,0.15763839483261108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,2,power_law_1.01,0.1569408059120178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,2,power_law_1.01,0.16718080043792724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,64,power_law_1.2,0.17635200023651124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,2,power_law_1.01,0.17790080308914186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,64,power_law_1.2,0.16421760320663453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,64,power_law_1.2,0.17503360509872437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,2,power_law_1.01,0.19421440362930298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,64,power_law_1.2,0.17426559925079346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,64,power_law_1.2,0.17250560522079467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,2,power_law_1.01,0.2350464105606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,64,power_law_1.2,0.18625279664993286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,64,power_law_1.2,0.20833280086517333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,64,power_law_1.2,0.23431680202484131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,2,power_law_1.01,0.24264960289001464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,64,power_law_1.2,0.26598401069641114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,64,power_law_1.2,0.3478912115097046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,64,power_law_1.2,0.4310463905334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,64,power_law_1.2,0.5219776153564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,2,power_law_1.01,0.3315200090408325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,64,power_law_1.2,0.8010944366455078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,64,power_law_1.2,0.9788288116455078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,64,power_law_1.2,1.3081215858459472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,2,power_law_1.01,0.41137919425964353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,2,power_law_1.01,0.49171838760375974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,2,power_law_1.01,0.659276819229126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,2,power_law_1.01,0.954867172241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,2,power_law_1.01,1.2557439804077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,8,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,8,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,2,power_law_1.01,1.8442047119140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,8,balanced,0.047653332352638245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,8,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,8,balanced,0.05236800014972687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,8,balanced,0.0676800012588501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,8,balanced,0.08463999629020691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,8,balanced,0.08470400174458821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,8,balanced,0.08542399605115254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,8,balanced,0.08703466256459554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,8,balanced,0.0881866713364919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,8,balanced,0.0909493366877238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,8,balanced,0.09442666172981262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,8,balanced,0.09464533130327861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,8,balanced,0.10354133447011311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,8,balanced,0.10529067118962605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,8,balanced,0.11541333794593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,8,balanced,0.12974400321642557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,8,balanced,0.13894400000572205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,8,balanced,0.1660480002562205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,8,balanced,0.18979199727376303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,8,balanced,0.23796266317367554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,8,balanced,0.3104640046755473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,8,balanced,0.4114240010579427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,8,balanced,0.5329493284225464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,8,balanced,0.7600106398264567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,8,balanced,0.9647520383199056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,2,power_law_1.01,2.379814338684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,4,balanced,0.04427200059096018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,4,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,4,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,4,balanced,0.05789866546789805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,4,balanced,0.08010133107503255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,4,balanced,0.11449066797892253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,4,balanced,0.11563199758529663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,4,balanced,0.11559466520945232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,4,balanced,0.1163200040658315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,4,balanced,0.1188266674677531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,4,balanced,0.12038933237393697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,4,balanced,0.12379733721415202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,4,balanced,0.12837866942087808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,4,balanced,0.12814933061599731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,4,balanced,0.13873599966367087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,4,balanced,0.14402133226394653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,4,balanced,0.15094932913780212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,4,balanced,0.17047999302546182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,4,balanced,0.18196799357732138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,4,balanced,0.21442667643229166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,4,balanced,0.24175999561945596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,4,balanced,0.3244746724764506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,4,balanced,0.38702932993570965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,4,balanced,0.5394453207651774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,4,balanced,0.667365312576294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,4,balanced,0.9532480239868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,4,balanced,1.2287413279215496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,64,power_law_1.2,0.03221119940280914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,64,power_law_1.2,0.030963200330734252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,64,power_law_1.2,0.025843200087547303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,64,power_law_1.2,0.027871999144554137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,64,power_law_1.2,0.028313601016998292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,64,power_law_1.2,0.029337599873542786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,64,power_law_1.2,0.02953599989414215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,64,power_law_1.2,0.029580798745155335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,64,power_law_1.2,0.030067199468612672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,64,power_law_1.2,0.02971520125865936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,64,power_law_1.2,0.031219199299812317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,64,power_law_1.2,0.03111039996147156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,64,power_law_1.2,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,64,power_law_1.2,0.03315199911594391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,64,power_law_1.2,0.037011200189590455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,64,power_law_1.2,0.03744640052318573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,64,power_law_1.2,0.03893119990825653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,64,power_law_1.2,0.041177600622177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,64,power_law_1.2,0.0438975989818573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,64,power_law_1.2,0.052902400493621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,64,power_law_1.2,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,64,power_law_1.2,0.06684799790382386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,64,power_law_1.2,0.07968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,64,power_law_1.2,0.1051584005355835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,64,power_law_1.2,0.1320639967918396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,64,power_law_1.2,0.1779711961746216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,64,power_law_1.2,0.11340800523757935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,64,power_law_1.2,0.08339200019836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,64,power_law_1.2,0.08501759767532349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,64,power_law_1.2,0.08560640215873719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,64,power_law_1.2,0.07850880026817322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,64,power_law_1.2,0.08263040184974671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,64,power_law_1.2,0.07989119887351989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,64,power_law_1.2,0.08309760093688964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,64,power_law_1.2,0.08195199966430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,64,power_law_1.2,0.22016639709472657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,64,power_law_1.2,0.08067839741706848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,64,power_law_1.2,0.08515840172767639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,64,power_law_1.2,0.08904320001602173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,64,power_law_1.2,0.08897280097007751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,64,power_law_1.2,0.0893119990825653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,64,power_law_1.2,0.09653120040893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,64,power_law_1.2,0.09472000002861022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,64,power_law_1.2,0.10707199573516846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,64,power_law_1.2,0.11768319606781005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,64,power_law_1.2,0.1279744029045105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,64,power_law_1.2,0.15856640338897704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,64,power_law_1.2,0.1742591977119446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,64,power_law_1.2,0.2394047975540161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,64,power_law_1.2,0.25601279735565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,64,power_law_1.2,0.3573184013366699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,64,power_law_1.2,0.4175104141235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,64,power_law_1.2,0.5952896118164063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,64,power_law_1.2,0.8287039756774902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,64,balanced,0.05222400029500326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,64,balanced,0.04353066782156626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,64,balanced,0.043568000197410583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,64,balanced,0.04621333380540212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,64,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,64,balanced,0.05389333268006643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,64,balanced,0.053786665201187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,64,balanced,0.053472002347310386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,64,balanced,0.05453333258628845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,64,balanced,0.05445333321889242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,64,balanced,0.054560000697771706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,64,balanced,0.05593066910902659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,64,balanced,0.058864002426465355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,64,balanced,0.06166933476924896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,64,balanced,0.06875733534495036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,64,balanced,0.07220800220966339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,64,balanced,0.07878399888674419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,64,balanced,0.0897173285484314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,64,balanced,0.09836266438166301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,64,balanced,0.11379200220108032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,64,balanced,0.12915733456611633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,64,balanced,0.1625759998957316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,64,balanced,0.1954453388849894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,64,balanced,0.26027733087539673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,64,balanced,0.32365334033966064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,64,balanced,0.4541440010070801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,64,balanced,0.5758933226267496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,1,power_law_1.2,0.07058560252189636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,1,power_law_1.2,0.08209279775619507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,1,power_law_1.2,0.11334400177001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,1,power_law_1.2,0.1673087954521179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,1,power_law_1.2,0.21751039028167723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,1,power_law_1.2,0.2872704029083252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,1,power_law_1.2,0.4011648178100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,1,power_law_1.2,0.42545280456542967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,1,power_law_1.2,0.4422463893890381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,1,power_law_1.2,0.4550144195556641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,1,power_law_1.2,0.4731328010559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,1,power_law_1.2,0.48456320762634275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,1,power_law_1.2,0.49954562187194823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,1,power_law_1.2,0.5177728176116944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,1,power_law_1.2,0.5297920227050781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,1,power_law_1.2,0.550054407119751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,1,power_law_1.2,0.5732096195220947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,1,power_law_1.2,0.6439231872558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,1,power_law_1.2,0.658457612991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,1,power_law_1.2,0.7351935863494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,1,power_law_1.2,0.7639999866485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,1,power_law_1.2,0.8924991607666015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,1,power_law_1.2,1.0316224098205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,1,power_law_1.2,1.2903679847717284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,1,power_law_1.2,1.5449983596801757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,1,power_law_1.2,2.0561407089233397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,1,power_law_1.2,2.5934207916259764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,16,power_law_1.01,0.052339202165603636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,16,power_law_1.01,0.04835200011730194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,128,power_law_1.01,0.057651197910308837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,16,power_law_1.01,0.04565759897232056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,16,power_law_1.01,0.047577598690986635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,16,power_law_1.01,0.04904960095882416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,16,power_law_1.01,0.047200000286102294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,16,power_law_1.01,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,16,power_law_1.01,0.050316798686981204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,16,power_law_1.01,0.04962559938430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,16,power_law_1.01,0.050547200441360476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,128,power_law_1.01,0.06149119734764099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,16,power_law_1.01,0.05242879986763001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,16,power_law_1.01,0.0532800018787384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,16,power_law_1.01,0.05440639853477478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,16,power_law_1.01,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,16,power_law_1.01,0.0627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,16,power_law_1.01,0.06298239827156067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,16,power_law_1.01,0.06785920262336731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,16,power_law_1.01,0.07738239765167236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,128,power_law_1.01,0.049465599656105044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,128,power_law_1.01,0.0525056004524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,128,power_law_1.01,0.053862398862838744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,128,power_law_1.01,0.05589119791984558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,8,power_law_1.01,0.03635199964046478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,8,power_law_1.01,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,128,power_law_1.01,0.05839999914169312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,16,power_law_1.01,0.08249599933624267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,8,power_law_1.01,0.03293440043926239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,128,power_law_1.01,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,16,power_law_1.01,0.10274560451507568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,128,power_law_1.01,0.05548160076141358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,16,power_law_1.01,0.11382399797439575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,128,power_law_1.01,0.05625600218772888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,16,power_law_1.01,0.15386879444122314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,128,power_law_1.01,0.05737599730491638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,16,power_law_1.01,0.19089920520782472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,128,power_law_1.01,0.06478719711303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,16,power_law_1.01,0.25265278816223147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,128,power_law_1.01,0.06806399822235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,16,power_law_1.01,0.3101632118225098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,8,power_law_1.01,0.03294079899787903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,8,power_law_1.01,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,8,power_law_1.01,0.035071998834609985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,8,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,8,power_law_1.01,0.035334399342536925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,8,power_law_1.01,0.03549439907073974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,8,power_law_1.01,0.03580799996852875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,8,power_law_1.01,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,8,power_law_1.01,0.03671039938926697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,128,power_law_1.01,0.07160959839820862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,8,power_law_1.01,0.03904640078544617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,128,power_law_1.01,0.07910400032997131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,16,power_law_1.01,0.4495744228363037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,128,power_law_1.01,0.08109440207481385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,128,power_law_1.01,0.0850048005580902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,16,power_law_1.01,0.5673151969909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,128,power_law_1.01,0.10291839838027954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,128,power_law_1.01,0.10705920457839965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,128,power_law_1.01,0.13729920387268066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,128,power_law_1.01,0.15413119792938232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,8,power_law_1.01,0.039647999405860904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,8,power_law_1.01,0.043961599469184875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,8,power_law_1.01,0.04577920138835907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,8,power_law_1.01,0.04759039878845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,8,power_law_1.01,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,8,power_law_1.01,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,8,power_law_1.01,0.07244799733161926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,8,power_law_1.01,0.0889792025089264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,8,power_law_1.01,0.11469440460205078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,8,power_law_1.01,0.12981120347976685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,128,power_law_1.01,0.19748480319976808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,8,power_law_1.01,0.19576319456100463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,128,power_law_1.01,0.20460159778594972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,8,power_law_1.01,0.2265791893005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,8,power_law_1.01,0.34389119148254393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,8,power_law_1.01,0.4691328048706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,128,power_law_1.01,0.2569983959197998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,128,power_law_1.01,0.2954943895339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,128,power_law_1.01,0.4263807773590088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,128,power_law_1.01,0.48463997840881345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,32,power_law_1.01,0.04488320052623749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,32,power_law_1.01,0.04366720020771027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,32,power_law_1.01,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,32,power_law_1.01,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,32,power_law_1.01,0.047513601183891294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,32,power_law_1.01,0.04684160053730011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,32,power_law_1.01,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,32,power_law_1.01,0.04992640018463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,32,power_law_1.01,0.050329601764678954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,32,power_law_1.01,0.05007359981536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,32,power_law_1.01,0.05225600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,32,power_law_1.01,0.05305600166320801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,32,power_law_1.01,0.054771202802658084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,32,power_law_1.01,0.057011198997497556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,32,power_law_1.01,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,32,power_law_1.01,0.06412799954414368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,32,power_law_1.01,0.06965759992599488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,32,power_law_1.01,0.07841920256614685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,32,power_law_1.01,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,32,power_law_1.01,0.1155776023864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,32,power_law_1.01,0.1335360050201416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,32,power_law_1.01,0.17486079931259155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,32,power_law_1.01,0.22018558979034425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,32,power_law_1.01,0.2907968044281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,32,power_law_1.01,0.3706943988800049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,32,power_law_1.01,0.5274367809295655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,32,power_law_1.01,0.7166528224945068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,32,8,power_law_1.01,0.020211200416088104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,32,8,power_law_1.01,0.021555200219154358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,32,8,power_law_1.01,0.02260479927062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,32,8,power_law_1.01,0.023104000091552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,32,8,power_law_1.01,0.02456959933042526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,32,8,power_law_1.01,0.025971201062202454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,32,8,power_law_1.01,0.0281792014837265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,32,8,power_law_1.01,0.02842240035533905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,32,8,power_law_1.01,0.029382398724555968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,32,8,power_law_1.01,0.030348798632621764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,32,8,power_law_1.01,0.030988800525665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,32,8,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,32,8,power_law_1.01,0.03344640135765076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,32,8,power_law_1.01,0.03528960049152374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,32,8,power_law_1.01,0.03828479945659637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,32,8,power_law_1.01,0.04101119935512543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,32,8,power_law_1.01,0.0438400000333786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,32,8,power_law_1.01,0.051475197076797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,32,8,power_law_1.01,0.07584000229835511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,32,8,power_law_1.01,0.09105920195579528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,32,8,power_law_1.01,0.08826239705085755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,32,8,power_law_1.01,0.10866559743881225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,32,8,power_law_1.01,0.12376960515975952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,32,8,power_law_1.01,0.17717119455337524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,32,8,power_law_1.01,0.20663039684295653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,32,8,power_law_1.01,0.295142388343811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,32,8,power_law_1.01,0.3689280033111572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,1,power_law_1.2,0.04081920087337494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,1,power_law_1.2,0.04320639967918396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,1,power_law_1.2,0.04782719910144806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,1,power_law_1.2,0.052908802032470705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,1,power_law_1.2,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,1,power_law_1.2,0.05539839863777161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,1,power_law_1.2,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,1,power_law_1.2,0.055379199981689456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,1,power_law_1.2,0.054713600873947145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,1,power_law_1.2,0.0553600013256073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,1,power_law_1.2,0.05583999752998352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,1,power_law_1.2,0.0585536003112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,1,power_law_1.2,0.0597055971622467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,1,power_law_1.2,0.060185599327087405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,1,power_law_1.2,0.06940159797668458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,1,power_law_1.2,0.07084800004959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,1,power_law_1.2,0.07768959999084472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,1,power_law_1.2,0.09051520228385926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.2,0.10032639503479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,16,power_law_1.2,0.08298239707946778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,16,power_law_1.2,0.07011200189590454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,16,power_law_1.2,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,16,power_law_1.2,0.07066879868507385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,16,power_law_1.2,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,16,power_law_1.2,0.06934400200843811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,16,power_law_1.2,0.07008000016212464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,16,power_law_1.2,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,16,power_law_1.2,0.0738752007484436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.2,0.12385280132293701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,16,power_law_1.2,0.07479680180549622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,16,power_law_1.2,0.07294080257415772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,16,power_law_1.2,0.07548800110816956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,16,power_law_1.2,0.07671679854393006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,16,power_law_1.2,0.08255360126495362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,16,power_law_1.2,0.09132800102233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,16,power_law_1.2,0.09306880235671997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,16,power_law_1.2,0.09516800045967103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,16,power_law_1.2,0.11518720388412476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,16,power_law_1.2,0.11508480310440064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,16,power_law_1.2,0.13649920225143433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,16,power_law_1.2,0.1526528000831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,16,power_law_1.2,0.21431679725646974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,16,power_law_1.2,0.24410240650177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,16,power_law_1.2,0.328601598739624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,16,power_law_1.2,0.4247424125671387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.2,0.14442880153656007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,16,power_law_1.2,0.6324672222137451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,16,power_law_1.2,0.7631743907928467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.2,0.19349119663238526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.2,0.2430527925491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.2,0.33327360153198243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.2,0.4256768226623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,16,power_law_1.01,0.07098879814147949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,16,power_law_1.01,0.07093120217323304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,16,power_law_1.01,0.06846079826354981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,16,power_law_1.01,0.07377920150756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,16,power_law_1.01,0.0710591971874237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,16,power_law_1.01,0.07478399872779846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,16,power_law_1.01,0.07633919715881347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,16,power_law_1.01,0.07670400142669678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,16,power_law_1.01,0.07790719866752624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,16,power_law_1.01,0.07841920256614685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,16,power_law_1.01,0.07962239980697632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,16,power_law_1.01,0.08061439990997314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,16,power_law_1.01,0.08836479783058167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,16,power_law_1.01,0.0902079999446869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,16,power_law_1.01,0.09708799719810486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,4,power_law_1.2,0.05178239941596985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,4,power_law_1.2,0.05816320180892944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,4,power_law_1.2,0.07037439942359924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,4,power_law_1.2,0.0784063994884491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,4,power_law_1.2,0.09258880019187928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,4,power_law_1.2,0.10241919755935669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,4,power_law_1.2,0.11478400230407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,4,power_law_1.2,0.12042880058288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,4,power_law_1.2,0.12912000417709352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,16,power_law_1.01,0.09512959718704224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,4,power_law_1.2,0.1244223952293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,16,power_law_1.01,0.10317440032958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,4,power_law_1.2,0.12830079793930055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,16,power_law_1.01,0.10972800254821777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,4,power_law_1.2,0.14216959476470947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,16,power_law_1.01,0.1229375958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,4,power_law_1.2,0.12830719947814942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,16,power_law_1.01,0.1415552020072937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,4,power_law_1.2,0.14840320348739625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,16,power_law_1.01,0.16928000450134278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,16,power_law_1.01,0.22704639434814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,16,power_law_1.01,0.3055743932723999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,16,power_law_1.01,0.40220160484313966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,4,power_law_1.2,0.15475840568542482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,4,power_law_1.2,0.15947519540786742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,4,power_law_1.2,0.1705407977104187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,4,power_law_1.2,0.1883072018623352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,4,power_law_1.2,0.2160128116607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,4,power_law_1.2,0.26442880630493165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,16,power_law_1.01,0.5256383895874024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,4,power_law_1.2,0.2956415891647339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,16,power_law_1.01,0.7942848205566406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,16,power_law_1.01,1.0029696464538573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,4,power_law_1.2,0.3591104030609131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,4,power_law_1.2,0.4565824031829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,4,power_law_1.2,0.6248960018157959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,4,power_law_1.2,0.7853312015533447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,4,power_law_1.2,1.0606143951416016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,4,power_law_1.2,1.3627455711364747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.2,0.6082880020141601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.2,0.8010239601135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,4,balanced,0.045978665351867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,4,balanced,0.04790399968624115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,4,balanced,0.050474668542544045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,4,balanced,0.0609493354956309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,4,balanced,0.08035733302434285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,4,balanced,0.1128480037053426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,4,balanced,0.14828266700108847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,4,balanced,0.14595199624697366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,4,balanced,0.14494933684666952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,4,balanced,0.14839999874432883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,4,balanced,0.1467573344707489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,4,balanced,0.14962666233380637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,4,balanced,0.15009066462516785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,4,balanced,0.15593600273132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,4,balanced,0.1611146628856659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,4,balanced,0.16471999883651733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,4,balanced,0.17190933227539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,4,balanced,0.1878239909807841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,4,balanced,0.2056586742401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,4,balanced,0.23301867643992105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,4,balanced,0.26077866554260254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,4,balanced,0.3129013379414876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,4,balanced,0.3901866674423218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,4,balanced,0.49534932772318524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,4,balanced,0.6386239926020304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,4,balanced,0.8591039975484213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,4,balanced,1.1231199900309246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,2,power_law_1.01,0.09031680226325989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,2,power_law_1.01,0.1078976035118103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,2,power_law_1.01,0.11948800086975098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,2,power_law_1.01,0.14024959802627562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,2,power_law_1.01,0.16185599565505981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,2,power_law_1.01,0.1825152039527893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,2,power_law_1.01,0.20702080726623534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,2,power_law_1.01,0.21313920021057128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,8,power_law_1.01,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,2,power_law_1.01,0.21573119163513182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,2,power_law_1.01,0.22147839069366454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,2,power_law_1.01,0.22953600883483888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,2,power_law_1.01,0.2323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,2,power_law_1.01,0.2403264045715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,2,power_law_1.01,0.24671359062194825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,8,power_law_1.01,0.03552640080451965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,8,power_law_1.01,0.0364544004201889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,8,power_law_1.01,0.03847039937973022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,8,power_law_1.01,0.03980799913406372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,8,power_law_1.01,0.04268159866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,8,power_law_1.01,0.04336639940738678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,8,power_law_1.01,0.04464640021324158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,8,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,8,power_law_1.01,0.04769279956817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,8,power_law_1.01,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,8,balanced,0.06080000102519989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,8,balanced,0.06247999767462412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,8,balanced,0.066170667608579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,8,balanced,0.0804906686147054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,8,balanced,0.09570667147636414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,8,balanced,0.1302079955736796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,8,balanced,0.12961600224177042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,8,balanced,0.13088533282279968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,8,balanced,0.13321600357691446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,8,balanced,0.1309279998143514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,8,balanced,0.13296000162760416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,8,balanced,0.13678933183352152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,8,balanced,0.13537066181500754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,8,balanced,0.1386560002962748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,8,balanced,0.1472640037536621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,8,balanced,0.14831999937693277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,8,balanced,0.15267200271288553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,8,power_law_1.01,0.062080001831054686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,8,balanced,0.17220799128214517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,8,balanced,0.1783199906349182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,8,balanced,0.19647467136383057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,8,balanced,0.22178133328755698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,8,balanced,0.26469866434733075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,8,balanced,0.3097280065218608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,8,balanced,0.43146665891011554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,8,balanced,0.5247093439102173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,8,power_law_1.01,0.062015998363494876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,8,balanced,0.7484373251597086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,8,balanced,0.9676799774169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,8,power_law_1.01,0.06883839964866638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,8,power_law_1.01,0.08819199800491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,8,power_law_1.01,0.07820159792900086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,8,power_law_1.01,0.09259520173072815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,8,power_law_1.01,0.13176319599151612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,8,power_law_1.01,0.1403712034225464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,8,power_law_1.01,0.19631359577178956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,8,power_law_1.01,0.2611135959625244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,8,power_law_1.01,0.34533119201660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,8,power_law_1.01,0.4700607776641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,8,power_law_1.01,0.6576320171356201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,8,power_law_1.01,1.0049856185913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,4,power_law_1.01,0.05987840294837952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,4,power_law_1.01,0.06677759885787964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,4,power_law_1.01,0.07544959783554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,4,power_law_1.01,0.0843392014503479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,4,power_law_1.01,0.09281280040740966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,8,power_law_1.01,1.493286418914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,4,power_law_1.01,0.101254403591156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,4,power_law_1.01,0.10735360383987427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,4,power_law_1.01,0.10869120359420777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,4,power_law_1.01,0.1088703989982605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,4,power_law_1.01,0.10751999616622925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,4,power_law_1.01,0.11311359405517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,4,power_law_1.01,0.11150720119476318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,4,power_law_1.01,0.1123136043548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,4,power_law_1.01,0.11819519996643066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,4,power_law_1.01,0.12052479982376099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,4,power_law_1.01,0.12177280187606812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,4,power_law_1.01,0.12627840042114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,4,power_law_1.01,0.13609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,4,power_law_1.01,0.14839040040969848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,4,power_law_1.01,0.18095359802246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,4,power_law_1.01,0.20762240886688232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,4,power_law_1.01,0.24871039390563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,4,power_law_1.01,0.31646080017089845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,4,power_law_1.01,0.43213438987731934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,4,power_law_1.01,0.5783743858337402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,4,power_law_1.01,0.7960576057434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,4,power_law_1.01,1.0677632331848144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,8,power_law_1.01,1.8495744705200194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,2,power_law_1.2,0.07505919933319091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,2,power_law_1.2,0.09882879853248597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,2,power_law_1.2,0.12494080066680908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,2,power_law_1.2,0.16092159748077392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,2,power_law_1.2,0.2028287887573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,2,power_law_1.2,0.2759743928909302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,2,power_law_1.2,0.346617603302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,2,power_law_1.2,0.34707839488983155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,2,power_law_1.2,0.3742271900177002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,2,power_law_1.2,0.35523200035095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,2,power_law_1.2,0.39504640102386473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,2,power_law_1.2,0.39957120418548586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,2,power_law_1.2,0.3999232053756714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,2,power_law_1.2,0.4170112133026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,2,power_law_1.2,0.44124159812927244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,2,power_law_1.2,0.44754562377929685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,2,power_law_1.2,0.4727168083190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,2,power_law_1.2,0.5309247970581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,2,power_law_1.2,0.5551167964935303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,2,power_law_1.2,0.6481664180755615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,2,power_law_1.2,0.6818111896514892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,2,power_law_1.2,0.8073920249938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,2,power_law_1.2,0.9470463752746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,2,power_law_1.2,1.1784064292907714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,2,power_law_1.2,1.4367615699768066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,2,power_law_1.2,1.9280832290649415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,2,power_law_1.2,2.553817558288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,4,power_law_1.01,0.03457919955253601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,4,power_law_1.01,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,4,power_law_1.01,0.04023680090904236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,4,power_law_1.01,0.045535999536514285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,4,power_law_1.01,0.04830079972743988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,4,power_law_1.01,0.0578495979309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,4,power_law_1.01,0.059494400024414064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,4,power_law_1.01,0.060601598024368285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,4,power_law_1.01,0.06048640012741089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,4,power_law_1.01,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,4,power_law_1.01,0.06469759941101075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,4,power_law_1.01,0.06750079989433289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,4,power_law_1.01,0.06912000179290771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,4,power_law_1.01,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,4,power_law_1.01,0.07950720191001892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,4,power_law_1.01,0.081523197889328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,4,power_law_1.01,0.08833919763565064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,4,power_law_1.01,0.10120320320129395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,4,power_law_1.01,0.11542400121688842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,4,power_law_1.01,0.13688960075378417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,4,power_law_1.01,0.1581760048866272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,4,power_law_1.01,0.21107199192047119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,4,power_law_1.01,0.280947208404541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,4,power_law_1.01,0.35459198951721194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,4,power_law_1.01,0.4391488075256348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,4,power_law_1.01,0.6775360107421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,4,power_law_1.01,0.8643072128295899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,16,power_law_1.2,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,16,power_law_1.2,0.0699455976486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,16,power_law_1.2,0.06903039813041686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,16,power_law_1.2,0.07690240144729614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,16,power_law_1.2,0.0763647973537445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,16,power_law_1.2,0.06454399824142457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,16,power_law_1.2,0.07008000016212464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,16,power_law_1.2,0.07029119729995728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,16,power_law_1.2,0.07121919989585876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,16,power_law_1.2,0.07339519858360291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,16,power_law_1.2,0.07612159848213196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,16,power_law_1.2,0.08142719864845276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,16,power_law_1.2,0.08684160113334656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,16,power_law_1.2,0.08356480002403259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,16,power_law_1.2,0.09047039747238159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,16,power_law_1.2,0.08883200287818908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,16,power_law_1.2,0.09463679790496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,16,power_law_1.2,0.10591360330581664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,16,power_law_1.2,0.10700160264968872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,16,power_law_1.2,0.13150719404220582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,16,power_law_1.2,0.14321279525756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,16,power_law_1.2,0.174508798122406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,16,power_law_1.2,0.21237759590148925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,16,power_law_1.2,0.26382720470428467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,16,power_law_1.2,0.3672319889068604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,16,power_law_1.2,0.46286721229553224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,16,power_law_1.2,0.5537600040435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,8,balanced,0.03182400017976761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,8,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,8,balanced,0.031514666974544525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,8,balanced,0.05250666538874308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,8,balanced,0.06817600131034851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,8,balanced,0.08153066535790761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,8,balanced,0.0804319977760315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,8,balanced,0.07687999804814656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,8,balanced,0.0784746656815211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,8,balanced,0.07842666904131572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,8,balanced,0.07859200239181519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,8,balanced,0.07896533111731212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,8,balanced,0.07891733447710673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,8,balanced,0.0804319977760315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,8,balanced,0.0844533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,8,balanced,0.08506666620572408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,8,balanced,0.08483200271924336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,8,balanced,0.10404266913731892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,8,balanced,0.09277333815892537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,8,balanced,0.12612266341845194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,8,balanced,0.10288533568382263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,8,balanced,0.1337279975414276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,8,balanced,0.12689066926638284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,8,balanced,0.17463467518488565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,8,balanced,0.1923733353614807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,8,balanced,0.2741386691729228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,8,balanced,0.27224000295003253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,16,power_law_1.2,0.02620159983634949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,16,power_law_1.2,0.022572800517082214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,16,power_law_1.2,0.020883199572563172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,16,power_law_1.2,0.021408000588417055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,16,power_law_1.2,0.02186879962682724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,16,power_law_1.2,0.02378239929676056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,16,power_law_1.2,0.023814399540424348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,16,power_law_1.2,0.025459200143814087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,16,power_law_1.2,0.024716800451278685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,16,power_law_1.2,0.02588160037994385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,16,power_law_1.2,0.03076480031013489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,16,power_law_1.2,0.02932479977607727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,16,power_law_1.2,0.029875200986862183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,16,power_law_1.2,0.054201602935791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,16,power_law_1.2,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,16,power_law_1.2,0.05727999806404114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,16,power_law_1.2,0.05082240104675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,16,power_law_1.2,0.05204480290412903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,1,power_law_1.2,0.03237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,1,power_law_1.2,0.036582401394844054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,1,power_law_1.2,0.046137601137161255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,1,power_law_1.2,0.06228479743003845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,1,power_law_1.2,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,1,power_law_1.2,0.11104639768600463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,16,power_law_1.2,0.05622400045394897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,1,power_law_1.2,0.11204479932785034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,16,power_law_1.2,0.07294080257415772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,1,power_law_1.2,0.11994240283966065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,16,power_law_1.2,0.08915839791297912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,1,power_law_1.2,0.12434560060501099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,16,power_law_1.2,0.10806399583816528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,1,power_law_1.2,0.12595839500427247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,16,power_law_1.2,0.13401600122451782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,1,power_law_1.2,0.12817920446395875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,16,power_law_1.2,0.17002240419387818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,1,power_law_1.2,0.1303104043006897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,1,power_law_1.2,0.13073279857635497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,1,power_law_1.2,0.13557759523391724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,1,power_law_1.2,0.14343039989471434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,1,power_law_1.2,0.14826879501342774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,1,power_law_1.2,0.15569280385971068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,1,power_law_1.2,0.17004799842834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,1,power_law_1.2,0.18687360286712645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,1,power_law_1.2,0.22423040866851807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,1,power_law_1.2,0.2507839918136597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,1,power_law_1.2,0.3117824077606201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,1,power_law_1.2,0.3641599893569946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,1,power_law_1.2,0.47925119400024413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,16,power_law_1.2,0.22791678905487062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,16,power_law_1.2,0.29745280742645264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,1,power_law_1.2,0.5910463809967041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,1,power_law_1.2,0.822713565826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,16,power_law_1.2,0.4104576110839844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,1,power_law_1.2,1.0608768463134766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,8,power_law_1.01,0.07500159740447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,8,power_law_1.01,0.09472000002861022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,8,power_law_1.01,0.09657599925994872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,8,power_law_1.01,0.11477760076522828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,8,power_law_1.01,0.12295039892196655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,8,power_law_1.01,0.11509120464324951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,8,power_law_1.01,0.11980160474777221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,8,power_law_1.01,0.11610879898071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,8,power_law_1.01,0.12421120405197143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,8,power_law_1.01,0.11964800357818603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,8,power_law_1.01,0.11822079420089722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,8,power_law_1.01,0.11932159662246704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,4,power_law_1.2,0.09789440035820007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,8,power_law_1.01,0.12865279912948607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,4,power_law_1.2,0.12340480089187622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,8,power_law_1.01,0.12674560546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,4,power_law_1.2,0.11028480529785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,8,power_law_1.01,0.13564159870147705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,8,power_law_1.01,0.1308351993560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,8,power_law_1.01,0.13657599687576294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,8,power_law_1.01,0.14775680303573607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,8,power_law_1.01,0.16342400312423705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,8,power_law_1.01,0.18852479457855226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,8,power_law_1.01,0.2056191921234131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,4,power_law_1.2,0.13839999437332154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,8,power_law_1.01,0.27497599124908445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,4,power_law_1.2,0.139027202129364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,4,power_law_1.2,0.14172159433364867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,8,power_law_1.01,0.31724159717559813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,4,power_law_1.2,0.1816256046295166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,4,power_law_1.2,0.17227519750595094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,8,power_law_1.01,0.43837437629699705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,4,power_law_1.2,0.20428800582885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,8,power_law_1.01,0.549126386642456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,4,power_law_1.2,0.21269121170043945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,4,power_law_1.2,0.1593664050102234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,8,power_law_1.01,0.7704192161560058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,4,power_law_1.2,0.18027520179748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,8,power_law_1.01,0.9435327529907227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,4,power_law_1.2,0.1850175976753235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,4,power_law_1.2,0.19795199632644653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,4,power_law_1.2,0.21665918827056885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,4,power_law_1.2,0.22035839557647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,4,power_law_1.2,0.25294721126556396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,4,power_law_1.2,0.325708794593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,4,power_law_1.2,0.36281599998474123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,4,power_law_1.2,0.5561471939086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,4,power_law_1.2,0.6583360195159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,4,power_law_1.2,0.7494272232055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,4,power_law_1.2,1.05032320022583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,4,power_law_1.2,1.4164223670959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,1,balanced,0.0582826683918635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,1,balanced,0.07126399874687195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,1,balanced,0.09922132889429729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,1,balanced,0.15874133507410684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,1,balanced,0.25732266902923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,128,power_law_1.2,0.05240319967269898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,1,balanced,0.4615519841512044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,1,balanced,0.4595573345820109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,128,power_law_1.2,0.06346880197525025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,1,balanced,0.46375465393066406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,1,balanced,0.47419198354085285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,1,balanced,0.4743306636810303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,1,balanced,0.4814773400624593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,1,balanced,0.4880533218383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,1,balanced,0.49274134635925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,1,balanced,0.49910934766133624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,1,balanced,0.509114662806193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,1,balanced,0.5190293391545614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,1,balanced,0.5347573359807333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,1,balanced,0.5742559830347697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,1,balanced,0.6118293205897013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,1,balanced,0.6856586933135986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,1,balanced,0.7513759930928549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,1,balanced,0.9664426644643148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,128,power_law_1.2,0.04381439983844757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,1,balanced,1.074954668680827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,128,power_law_1.2,0.045772799849510194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,1,balanced,1.4855200449625652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,128,power_law_1.2,0.04673919975757599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,1,balanced,1.8063146273295085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,128,power_law_1.2,0.04733439981937408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,128,power_law_1.2,0.04749439954757691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,128,power_law_1.2,0.04802559912204742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,128,power_law_1.2,0.04860160052776337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,128,power_law_1.2,0.04986880123615265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,128,power_law_1.2,0.055283200740814206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,128,power_law_1.2,0.05583360195159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,128,power_law_1.2,0.05730559825897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,128,power_law_1.2,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,128,power_law_1.2,0.06843519806861878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,1,balanced,2.5386346181233725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,128,power_law_1.2,0.07000960111618042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,128,power_law_1.2,0.06895999908447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,128,power_law_1.2,0.07551360130310059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,1,balanced,3.1869227091471353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,128,power_law_1.2,0.08455680012702942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,128,power_law_1.2,0.10497280359268188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,4,power_law_1.2,1.8930368423461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,128,power_law_1.2,0.1088703989982605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,128,power_law_1.2,0.13427200317382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,128,power_law_1.2,0.18315520286560058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,128,power_law_1.2,0.23296639919281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,128,power_law_1.2,0.27538559436798093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,128,power_law_1.2,0.40302720069885256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,128,power_law_1.2,0.5386240005493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,4,power_law_1.2,2.713337516784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,4,power_law_1.2,3.5114559173583983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,1,power_law_1.2,0.07772799730300903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,2,power_law_1.01,0.10368000268936158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,1,power_law_1.2,0.09503359794616699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,1,power_law_1.2,0.1291327953338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,1,power_law_1.2,0.16942720413208007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,1,power_law_1.2,0.20989439487457276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,1,power_law_1.2,0.33025920391082764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,2,power_law_1.01,0.14559359550476075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,1,power_law_1.2,0.35447680950164795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,2,power_law_1.01,0.1840831995010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,1,power_law_1.2,0.36801919937133787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,2,power_law_1.01,0.24677119255065919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,1,power_law_1.2,0.3746432065963745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,2,power_law_1.01,0.31023359298706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,1,power_law_1.2,0.38071680068969727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,2,power_law_1.01,0.4715968132019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,1,power_law_1.2,0.4064640045166016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,2,power_law_1.01,0.49796481132507325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,1,power_law_1.2,0.4086463928222656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,2,power_law_1.01,0.50032639503479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,1,power_law_1.2,0.41843199729919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,2,power_law_1.01,0.521446418762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,1,power_law_1.2,0.43160319328308105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,2,power_law_1.01,0.5375296115875244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,1,power_law_1.2,0.4476287841796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,2,power_law_1.01,0.5128320217132568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,1,power_law_1.2,0.4653439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,1,power_law_1.2,0.49387521743774415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,1,power_law_1.2,0.5616384029388428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.2,0.6309567928314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.2,0.7763775825500489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,2,power_law_1.01,0.5389311790466309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,2,power_law_1.01,0.5683904170989991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.2,0.8242176055908204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,2,power_law_1.01,0.5836480140686036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.2,1.0511039733886718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,2,power_law_1.01,0.5957056045532226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,2,power_law_1.01,0.6016384124755859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.2,1.311398410797119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,2,power_law_1.01,0.6184256076812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,2,power_law_1.01,0.6737664222717286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,2,power_law_1.01,0.7279871940612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,2,power_law_1.01,0.8758079528808593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,2,power_law_1.01,0.996070384979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,2,power_law_1.01,1.2254464149475097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,2,power_law_1.01,1.1934207916259765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,2,power_law_1.01,1.4633600234985351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.2,1.7104320526123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,2,power_law_1.01,2.0608320236206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,2,power_law_1.01,2.623891258239746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,2,power_law_1.01,3.0224639892578127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.2,2.129055976867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,2,balanced,0.08763733506202698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,2,balanced,0.12678933143615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,2,balanced,0.19605332612991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,2,balanced,0.33902935187021893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,2,balanced,0.5962400039037069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,2,balanced,0.8504213492075602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,2,balanced,0.8514346281687418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.2,2.9545663833618163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,2,balanced,0.8513333002726237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,2,balanced,0.8558453718821207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,2,balanced,0.8600266774495443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,2,balanced,0.8661973476409912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,2,balanced,0.8730613390604655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,2,balanced,0.8773492972056071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,2,balanced,0.8894293308258057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,2,balanced,0.9007413387298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,2,balanced,0.909109354019165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,2,balanced,0.9296053250630697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,2,balanced,1.067834695180257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,2,balanced,1.0038452943166096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,2,balanced,1.2998932997385662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,2,balanced,1.1607360045115154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.2,3.7896575927734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,2,balanced,1.8562827110290527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,2,balanced,1.4658133188883464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,2,balanced,2.224773406982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,2,balanced,2.282778739929199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,2,balanced,3.6877759297688804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,2,balanced,3.7662668228149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,4,power_law_1.01,0.07537279725074768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,1,balanced,0.08083199958006541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,4,power_law_1.01,0.10901119709014892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,1,balanced,0.0831573357184728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,1,balanced,0.0872373382250468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,1,balanced,0.10379200180371602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,1,balanced,0.13429333766301474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,1,balanced,0.13470932841300964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,1,balanced,0.13736533125241598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,1,balanced,0.1381280024846395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,1,balanced,0.14005866646766663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,1,balanced,0.14199999968210855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,1,balanced,0.14410133163134256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,1,balanced,0.14865600069363913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,1,balanced,0.15040533741315207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,1,balanced,0.15432533621788025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,1,balanced,0.1507200002670288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,4,power_law_1.01,0.12784639596939087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,4,power_law_1.01,0.15960320234298705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,4,power_law_1.01,0.19146239757537842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,4,power_law_1.01,0.22028799057006837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,4,power_law_1.01,0.27992959022521974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,4,power_law_1.01,0.2645567893981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,4,power_law_1.01,0.26397440433502195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,4,power_law_1.01,0.24603519439697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,4,power_law_1.01,0.2705919981002808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,4,power_law_1.01,0.26398720741271975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,4,power_law_1.01,0.27178239822387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,1,balanced,0.15411733587582907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,1,balanced,0.1601759990056356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,4,power_law_1.01,0.2797631978988647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,1,balanced,0.17328532536824545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,4,power_law_1.01,0.2828671932220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,1,balanced,0.18713066975275675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,4,power_law_1.01,0.3020607948303223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,1,balanced,0.2190826733907064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,4,power_law_1.01,0.3120255947113037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,1,balanced,0.24870934089024863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,4,power_law_1.01,0.33879039287567136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,1,balanced,0.3564853270848592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,4,power_law_1.01,0.33919999599456785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,1,balanced,0.4254026810328166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,4,power_law_1.01,0.3906368017196655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,1,balanced,0.6004586617151896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,4,power_law_1.01,0.41269760131835936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,1,balanced,0.7768959999084473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,4,power_law_1.01,0.4946176052093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,1,balanced,1.124837319056193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,4,power_law_1.01,0.5185599803924561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,1,balanced,1.4808160463968914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,4,power_law_1.01,0.6671103954315185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,4,power_law_1.01,0.822163200378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,4,power_law_1.01,1.0504768371582032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,4,power_law_1.01,1.2729280471801758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,1,balanced,0.04821866750717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,1,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,1,balanced,0.05412266651789347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,1,balanced,0.07448533177375793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,1,balanced,0.10115733742713928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,1,balanced,0.14306132992108664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,1,balanced,0.14523733655611673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,1,balanced,0.1471946636835734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,1,balanced,0.14782933394114176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,1,balanced,0.14962666233380637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,1,balanced,0.15245333313941956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,1,balanced,0.1588213344415029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,1,balanced,0.16263467073440552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,1,balanced,0.164383997519811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,1,balanced,0.1759786605834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,1,balanced,0.181658665339152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,1,balanced,0.1927573283513387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,1,balanced,0.223962664604187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,1,balanced,0.24699733654658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,1,balanced,0.305242657661438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,1,balanced,0.34918399651845294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,1,balanced,0.4922933181126912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,1,balanced,0.5718453327814738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,1,balanced,0.8403253555297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,1,balanced,1.0524160067240398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,1,balanced,1.5274346669514973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,1,balanced,1.9764213562011719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,4,balanced,0.0685280015071233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,4,balanced,0.07844266792138417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,4,balanced,0.09885866443316142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,4,balanced,0.14232533176740012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,4,balanced,0.1998400092124939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,4,balanced,0.33896533648173016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,4,balanced,0.33074132601420086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,4,balanced,0.30102399984995526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,4,balanced,0.3258133331934611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,4,balanced,0.31191466252009076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,4,balanced,0.3104853431383769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,4,balanced,0.29950400193532306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,4,balanced,0.3139306704203288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,4,balanced,0.29658667246500653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,4,balanced,0.3107626636823018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,4,balanced,0.31836267312367755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,4,balanced,0.29951467116673786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,4,balanced,0.32523200909296673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,4,balanced,0.3335466782251994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,4,balanced,0.3482186794281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,4,balanced,0.37415464719136554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,4,balanced,0.4144586722056071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,4,balanced,0.47419198354085285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,4,balanced,0.7424586613972982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,4,balanced,0.8087200323740641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,4,balanced,1.1326506932576497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,4,balanced,1.479039986928304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,8,power_law_1.2,0.07216640114784241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,8,power_law_1.2,0.09975680112838745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,8,power_law_1.2,0.0995199978351593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,8,power_law_1.2,0.11727360486984253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,8,power_law_1.2,0.13479679822921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,8,power_law_1.2,0.13390079736709595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,8,power_law_1.2,0.14190080165863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,8,power_law_1.2,0.1284160017967224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,8,power_law_1.2,0.13749760389328003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,8,power_law_1.2,0.1375872015953064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,8,power_law_1.2,0.14228479862213134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,8,power_law_1.2,0.14496639966964722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,8,power_law_1.2,0.15048960447311402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,8,power_law_1.2,0.15247360467910767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,8,power_law_1.2,0.16357120275497436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,8,power_law_1.2,0.16853760480880736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,8,power_law_1.2,0.1796288013458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,8,power_law_1.2,0.20931200981140136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,8,power_law_1.2,0.22094080448150635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,8,power_law_1.2,0.2798527956008911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,8,power_law_1.2,0.3012671947479248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,8,power_law_1.2,0.37703680992126465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,8,power_law_1.2,0.4345600128173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,8,power_law_1.2,0.6108352184295655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,8,power_law_1.2,0.7595136165618896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,8,power_law_1.2,1.1329343795776368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,8,power_law_1.2,1.466521644592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,8,power_law_1.01,0.07847679853439331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,8,power_law_1.01,0.09722239971160888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,8,power_law_1.01,0.11445120573043824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,8,power_law_1.01,0.1122431993484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,8,power_law_1.01,0.1181056022644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,8,power_law_1.01,0.12545280456542968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,8,power_law_1.01,0.12688640356063843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,8,power_law_1.01,0.12760319709777831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,8,power_law_1.01,0.12993279695510865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,8,power_law_1.01,0.12492159605026246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,8,power_law_1.01,0.1287551999092102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,8,power_law_1.01,0.12981760501861572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,8,power_law_1.01,0.13397120237350463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,8,power_law_1.01,0.1352512001991272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,8,power_law_1.01,0.14396799802780152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,8,power_law_1.01,0.1490623950958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,8,power_law_1.01,0.15692800283432007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,8,power_law_1.01,0.1636288046836853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,8,power_law_1.01,0.18076800107955932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,8,power_law_1.01,0.21076478958129882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,8,power_law_1.01,0.25182719230651857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,8,power_law_1.01,0.30976641178131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,8,power_law_1.01,0.3871040105819702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,8,power_law_1.01,0.53853440284729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,8,power_law_1.01,0.6667456150054931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,8,power_law_1.01,0.9785599708557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,8,power_law_1.01,1.3404928207397462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,16,power_law_1.01,0.02051839977502823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,16,power_law_1.01,0.02112639993429184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,16,power_law_1.01,0.02147199958562851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,16,power_law_1.01,0.022009600698947907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,16,power_law_1.01,0.023027199506759643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,16,power_law_1.01,0.023123200237751006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,16,power_law_1.01,0.023820799589157105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,16,power_law_1.01,0.0239424005150795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,16,power_law_1.01,0.02913280129432678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,16,power_law_1.01,0.029183998703956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,16,power_law_1.01,0.029945600032806396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,16,power_law_1.01,0.052025598287582395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,16,power_law_1.01,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,16,power_law_1.01,0.05237759947776795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,16,power_law_1.01,0.04775039851665497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,16,power_law_1.01,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,16,power_law_1.01,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,16,power_law_1.01,0.052160000801086424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,16,power_law_1.01,0.05624319911003113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,16,power_law_1.01,0.06715520024299622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,16,power_law_1.01,0.07589120268821717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,16,power_law_1.01,0.09489920139312744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,16,power_law_1.01,0.1142016053199768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,16,power_law_1.01,0.1530943989753723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,16,power_law_1.01,0.18778879642486573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,16,power_law_1.01,0.261900806427002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,16,power_law_1.01,0.33609600067138673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,1,power_law_1.01,0.10317440032958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,1,power_law_1.01,0.11498880386352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,1,power_law_1.01,0.13319040536880494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,1,power_law_1.01,0.18409600257873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,1,power_law_1.01,0.22881920337677003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,1,power_law_1.01,0.2712127923965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,1,power_law_1.01,0.33210880756378175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,1,power_law_1.01,0.3538815975189209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,1,power_law_1.01,0.35204479694366453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,1,power_law_1.01,0.37325439453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,1,power_law_1.01,0.38549759387969973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,1,power_law_1.01,0.3827008008956909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,1,power_law_1.01,0.3981503963470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,1,power_law_1.01,0.4017024040222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,1,power_law_1.01,0.42244482040405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,1,power_law_1.01,0.42641282081604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,1,power_law_1.01,0.44689278602600097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,1,power_law_1.01,0.4974656105041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.01,0.5419072151184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.01,0.648134422302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.01,0.7376063823699951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.01,0.9435775756835938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.01,1.1380736351013183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.01,1.54268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.01,1.950387191772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.01,2.7472896575927734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.01,3.5341758728027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,16,power_law_1.01,0.050406402349472045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,16,power_law_1.01,0.06452479958534241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,16,power_law_1.01,0.0548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,16,power_law_1.01,0.05102720260620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,16,power_law_1.01,0.05136640071868896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,16,power_law_1.01,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,16,power_law_1.01,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,16,power_law_1.01,0.05485439896583557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,16,power_law_1.01,0.05477759838104248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,16,power_law_1.01,0.05618559718132019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,16,power_law_1.01,0.05708799958229065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,16,power_law_1.01,0.05994880199432373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,16,power_law_1.01,0.06083199977874756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,16,power_law_1.01,0.0643455982208252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,16,power_law_1.01,0.07022079825401306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,16,power_law_1.01,0.07160320281982421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,16,power_law_1.01,0.07635200023651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,16,power_law_1.01,0.09095039963722229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,16,power_law_1.01,0.10279680490493774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,16,power_law_1.01,0.13078399896621704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,16,power_law_1.01,0.15368959903717042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,16,power_law_1.01,0.20865280628204347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,16,power_law_1.01,0.25077118873596194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,16,power_law_1.01,0.35041921138763427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,16,power_law_1.01,0.4397247791290283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,16,power_law_1.01,0.666431999206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,16,power_law_1.01,0.8504768371582031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,4,balanced,0.019226666539907455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,4,balanced,0.019066666563351948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,4,balanced,0.02006400004029274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,4,balanced,0.019002666076024372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,4,balanced,0.020362666497627895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,4,balanced,0.020421333611011505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,4,balanced,0.023370665808518726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,4,balanced,0.023061332603295643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,4,balanced,0.04165333261092504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,4,balanced,0.041146665811538696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,4,balanced,0.04081599911053976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,4,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,4,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,4,balanced,0.03341866781314214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,4,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,4,balanced,0.034117333590984344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,4,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,4,balanced,0.047600001096725464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,4,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,4,balanced,0.06406400104363759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,4,balanced,0.07743466893831889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,4,balanced,0.0995146632194519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,4,balanced,0.12307733297348022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,4,balanced,0.1666826605796814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,4,balanced,0.20902933677037558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,4,balanced,0.2967680096626282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,4,balanced,0.38444801171620685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,1,power_law_1.2,0.07948799729347229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,1,power_law_1.2,0.08297600150108338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,1,power_law_1.2,0.08941439986228943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,1,power_law_1.2,0.10814720392227173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,1,power_law_1.2,0.1286080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,1,power_law_1.2,0.14617600440979003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,1,power_law_1.2,0.1804800033569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,1,power_law_1.2,0.1884160041809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,1,power_law_1.2,0.19189120531082154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,1,power_law_1.2,0.1965376019477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,1,power_law_1.2,0.20183680057525635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,1,power_law_1.2,0.21068160533905028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,1,power_law_1.2,0.21426560878753662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,1,power_law_1.2,0.22291839122772217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,1,power_law_1.2,0.23653120994567872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,1,power_law_1.2,0.24102399349212647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,1,power_law_1.2,0.2650752067565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,1,power_law_1.2,0.30161280632019044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,1,power_law_1.2,0.33794560432434084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,1,power_law_1.2,0.4158207893371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,1,power_law_1.2,0.482528018951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,1,power_law_1.2,0.6268352031707763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,1,power_law_1.2,0.7774720191955566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,1,power_law_1.2,1.0754624366760255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,1,power_law_1.2,1.363263988494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,1,power_law_1.2,1.954982376098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,1,power_law_1.2,2.538585662841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,16,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,16,balanced,0.04614933331807455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,16,balanced,0.04909333089987437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,4,power_law_1.01,0.09976320266723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,16,balanced,0.057999998331069946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,16,balanced,0.0652213344971339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,4,power_law_1.01,0.13519999980926514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,16,balanced,0.08081066608428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,16,balanced,0.07696000238259633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,16,balanced,0.07530133426189423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,16,balanced,0.07627200086911519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,16,balanced,0.07666133344173431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,16,balanced,0.07460266848405202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,16,balanced,0.07433066765467326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,16,balanced,0.07656000057856242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,16,balanced,0.0762613316377004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,16,balanced,0.08094933132330577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,16,balanced,0.08152000109354655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,16,balanced,0.086517333984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,4,power_law_1.01,0.137772798538208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,4,power_law_1.01,0.17527040243148803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,4,power_law_1.01,0.18312959671020507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,4,power_law_1.01,0.21294078826904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,16,balanced,0.09150933225949605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,16,balanced,0.09922132889429729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,16,balanced,0.10745599865913391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,16,balanced,0.11920000116030376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,16,balanced,0.14231999715169272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,16,balanced,0.16251200437545776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,16,balanced,0.22418133417765299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,16,balanced,0.2691626747449239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,16,balanced,0.372538685798645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,16,balanced,0.47430400053660077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,4,power_law_1.01,0.2514944076538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,4,power_law_1.01,0.2450495958328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,4,power_law_1.01,0.24988160133361817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,4,power_law_1.01,0.24380159378051758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,4,power_law_1.01,0.28346240520477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,4,power_law_1.01,0.26136319637298583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,4,power_law_1.01,0.2740288019180298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,4,power_law_1.01,0.27626240253448486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,4,power_law_1.01,0.2779200077056885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,4,power_law_1.01,0.28555519580841066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,4,power_law_1.01,0.302508807182312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,4,power_law_1.01,0.31741440296173096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,4,power_law_1.01,0.3470655918121338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,4,power_law_1.01,0.37998080253601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,4,power_law_1.01,0.43534078598022463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,4,power_law_1.01,0.5104447841644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,4,power_law_1.01,0.6199999809265136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,4,power_law_1.01,0.8708031654357911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,4,power_law_1.01,1.0619263648986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,8,power_law_1.01,0.032518398761749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,8,power_law_1.01,0.03177599906921387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,8,power_law_1.01,0.031699201464653014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,8,power_law_1.01,0.03416320085525513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,8,power_law_1.01,0.03484799861907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,8,power_law_1.01,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,8,power_law_1.01,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,8,power_law_1.01,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,8,power_law_1.01,0.03849599957466125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,8,power_law_1.01,0.038540801405906676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,8,power_law_1.01,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,8,power_law_1.01,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,8,power_law_1.01,0.04490880072116852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,8,power_law_1.01,0.050342398881912234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,8,power_law_1.01,0.055155199766159055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,8,power_law_1.01,0.06110079884529114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,8,power_law_1.01,0.06687999963760376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,8,power_law_1.01,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,8,power_law_1.01,0.11633919477462769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,8,power_law_1.01,0.15945600271224974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,8,power_law_1.01,0.19521280527114868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,8,power_law_1.01,0.2509567975997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,16,balanced,0.06238399942715963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,16,balanced,0.062496001521746315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,16,balanced,0.07404266794522603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,16,balanced,0.08885332942008972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,16,balanced,0.12566399574279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,16,balanced,0.13646933436393738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,16,balanced,0.1372106671333313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,16,balanced,0.13638400038083395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,16,balanced,0.1388053297996521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,16,balanced,0.13769599795341492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,16,balanced,0.14012266198794046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,16,balanced,0.14073066910107931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,16,balanced,0.14110933740933737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,16,balanced,0.14217066764831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,16,balanced,0.15036267042160034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,16,balanced,0.15341867009798685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,16,balanced,0.15544533729553223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,16,balanced,0.16485866904258728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,16,balanced,0.1696000099182129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,16,balanced,0.18000533183415732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,16,balanced,0.1954666574796041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,16,balanced,0.2622613310813904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,16,balanced,0.2878613273302714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,16,balanced,0.3731199900309245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,16,balanced,0.4816960096359253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,16,balanced,0.6330240170160929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,8,power_law_1.01,0.3821568012237549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,16,balanced,0.8309706846872965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,8,power_law_1.01,0.5494719982147217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,8,power_law_1.01,0.7220416069030762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,8,power_law_1.01,1.054694366455078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,8,power_law_1.01,1.370464038848877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,4,power_law_1.2,0.0339711993932724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,4,power_law_1.01,0.025856000185012818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,4,power_law_1.2,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,4,power_law_1.2,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,4,power_law_1.2,0.042803201079368594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,4,power_law_1.2,0.04454399943351746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,4,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,4,power_law_1.2,0.057920002937316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,4,power_law_1.2,0.059436798095703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,4,power_law_1.2,0.05990399718284607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,4,power_law_1.2,0.060262399911880496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,4,power_law_1.01,0.04487679898738861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,4,power_law_1.2,0.06130560040473938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,4,power_law_1.01,0.055692797899246214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,4,power_law_1.01,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,4,power_law_1.01,0.0838271975517273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,4,power_law_1.01,0.09212160110473633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,4,power_law_1.01,0.10249600410461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,4,power_law_1.01,0.10212479829788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,4,power_law_1.2,0.06335359811782837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,4,power_law_1.2,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,4,power_law_1.2,0.06934400200843811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,4,power_law_1.2,0.0752128005027771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,4,power_law_1.2,0.08169599771499633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,4,power_law_1.2,0.08235520124435425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,4,power_law_1.2,0.09070079922676086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,4,power_law_1.2,0.10209920406341552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,4,power_law_1.2,0.12937599420547485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,4,power_law_1.2,0.13731839656829833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,4,power_law_1.2,0.18151040077209474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,4,power_law_1.2,0.2352128028869629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,4,power_law_1.01,0.10832639932632446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,4,power_law_1.01,0.11050879955291748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,4,power_law_1.01,0.11634559631347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,4,power_law_1.01,0.11402239799499511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,4,power_law_1.01,0.11893119812011718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,4,power_law_1.01,0.12461440563201905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,4,power_law_1.01,0.13984639644622804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,4,power_law_1.2,0.32417280673980714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,4,power_law_1.01,0.13153280019760133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,4,power_law_1.01,0.1414720058441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,4,power_law_1.01,0.2552639961242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,4,power_law_1.01,0.2873087882995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,4,power_law_1.01,0.23084158897399903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,4,power_law_1.01,0.2386240005493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,4,power_law_1.01,0.2785664081573486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,4,power_law_1.01,0.3153343915939331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,4,power_law_1.01,0.39456639289855955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,4,power_law_1.01,0.46737918853759763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,4,power_law_1.01,0.6171775817871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,4,power_law_1.01,0.8369088172912598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,4,power_law_1.2,0.40563201904296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,2,power_law_1.01,0.05231999754905701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,2,power_law_1.01,0.06631039977073669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,2,power_law_1.01,0.07647359967231751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,4,power_law_1.2,0.5453440189361572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,2,power_law_1.01,0.09594240188598632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,2,power_law_1.01,0.10862079858779908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,2,power_law_1.01,0.12911360263824462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,2,power_law_1.01,0.15409280061721803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,2,power_law_1.01,0.16104960441589355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,2,power_law_1.01,0.15763200521469117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,2,power_law_1.01,0.16218240261077882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,2,power_law_1.01,0.1716928005218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,2,power_law_1.01,0.1787775993347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,2,power_law_1.01,0.17626880407333373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,2,power_law_1.01,0.1863808035850525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,2,power_law_1.01,0.1932736039161682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,2,power_law_1.01,0.19738240242004396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,2,power_law_1.01,0.21663360595703124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,2,power_law_1.01,0.24129281044006348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,2,power_law_1.01,0.2498944044113159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,2,power_law_1.01,0.2990976095199585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,2,power_law_1.01,0.33508479595184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,4,power_law_1.2,0.7050623893737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,2,power_law_1.01,0.4272448062896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,2,power_law_1.01,0.5057343959808349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,2,power_law_1.01,0.6716288089752197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,2,power_law_1.01,0.8574336051940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,2,power_law_1.01,1.1765312194824218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,2,power_law_1.01,1.5438528060913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,32,1,power_law_1.01,0.01788160055875778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,32,1,power_law_1.01,0.01679999977350235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,32,1,power_law_1.01,0.01690240055322647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,32,1,power_law_1.01,0.017798399925231932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,32,1,power_law_1.01,0.01825920045375824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,32,1,power_law_1.01,0.018361599743366243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,32,1,power_law_1.01,0.01875839978456497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,32,1,power_law_1.01,0.019385600090026857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,32,1,power_law_1.01,0.019840000569820403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,32,1,power_law_1.01,0.019980800151824952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,32,1,power_law_1.01,0.020230400562286376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,32,1,power_law_1.01,0.022873599827289582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,32,1,power_law_1.01,0.023244799673557283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,32,1,power_law_1.01,0.023904000222682954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,32,1,power_law_1.01,0.04142079949378967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,32,1,power_law_1.01,0.041971200704574586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,32,1,power_law_1.01,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,32,1,power_law_1.01,0.03785600066184998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.01,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.01,0.04864639937877655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.01,0.050291198492050174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.01,0.06935039758682252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.01,0.07972480058670044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.01,0.10720640420913696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.01,0.1340224027633667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.01,0.18863999843597412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.01,0.23549439907073974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,8,power_law_1.2,0.09503359794616699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,8,power_law_1.2,0.13994879722595216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,8,power_law_1.2,0.14403200149536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,8,power_law_1.2,0.17575680017471312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,8,power_law_1.2,0.20444800853729247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,8,power_law_1.2,0.20867199897766114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,8,power_law_1.2,0.22471039295196532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,8,power_law_1.2,0.21536641120910643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,8,power_law_1.2,0.2144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,8,power_law_1.2,0.2324608087539673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,8,power_law_1.2,0.23120639324188233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,8,power_law_1.2,0.22150399684906005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,8,power_law_1.2,0.24140799045562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,8,power_law_1.2,0.24080638885498046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,8,power_law_1.2,0.262937593460083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,8,power_law_1.2,0.2662911891937256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,8,power_law_1.2,0.28769280910491946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,8,power_law_1.2,0.3268480062484741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,8,power_law_1.2,0.344320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,1,power_law_1.2,0.06053760051727295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,8,power_law_1.2,0.4448512077331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,1,power_law_1.2,0.08209919929504395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,1,power_law_1.2,0.10410879850387574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,8,power_law_1.2,0.4238016128540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,1,power_law_1.2,0.15122560262680054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,8,power_law_1.2,0.5250368118286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,1,power_law_1.2,0.19393279552459716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,8,power_law_1.2,0.6104191780090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,1,power_law_1.2,0.2495807886123657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,1,power_law_1.2,0.34944639205932615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,1,power_law_1.2,0.3572160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,1,power_law_1.2,0.38978559970855714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,1,power_law_1.2,0.4023871898651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,1,power_law_1.2,0.41347198486328124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,8,power_law_1.2,0.8164416313171386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,1,power_law_1.2,0.4167168140411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,1,power_law_1.2,0.43710718154907224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,8,power_law_1.2,0.9959615707397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,1,power_law_1.2,0.45921921730041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,1,power_law_1.2,0.47869439125061036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,8,power_law_1.2,1.3447936058044434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,1,power_law_1.2,0.4971456050872803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,1,power_law_1.2,0.5382783889770508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,8,power_law_1.2,1.8459840774536134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,1,power_law_1.2,0.5984064102172851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,1,power_law_1.2,0.6475647926330567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,1,power_law_1.2,0.768832015991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,1,power_law_1.2,0.8591936111450196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,1,power_law_1.2,1.0740863800048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,1,power_law_1.2,1.2803711891174316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,1,power_law_1.2,1.6677888870239257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,1,power_law_1.2,2.0512063980102537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,1,power_law_1.2,2.837958335876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,1,power_law_1.2,3.6424320220947264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,8,power_law_1.01,0.08773120045661927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,8,power_law_1.01,0.12338559627532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,8,power_law_1.01,0.10752639770507813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,8,power_law_1.01,0.12654720544815062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,8,power_law_1.01,0.1411967992782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,8,power_law_1.01,0.14305280447006224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,8,power_law_1.01,0.13757439851760864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,8,power_law_1.01,0.15383679866790773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,8,power_law_1.01,0.14871679544448851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,8,power_law_1.01,0.14642560482025146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,8,power_law_1.01,0.15401599407196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,8,power_law_1.01,0.15448960065841674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,8,power_law_1.01,0.1553536057472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,8,power_law_1.01,0.15818239450454713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,8,power_law_1.01,0.1675968050956726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,8,power_law_1.01,0.17568639516830445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,8,power_law_1.01,0.18558080196380616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,8,power_law_1.01,0.21431679725646974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,8,power_law_1.01,0.23537919521331788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,8,power_law_1.01,0.28640639781951904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,8,power_law_1.01,0.3089344024658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,8,power_law_1.01,0.38680319786071776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,8,power_law_1.01,0.4267583847045898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,8,power_law_1.01,0.5708032131195069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,8,power_law_1.01,0.712063980102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,8,power_law_1.01,0.8873663902282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,8,power_law_1.01,1.157817554473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,8,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,8,balanced,0.04456533491611481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,8,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,8,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,8,balanced,0.051781331499417625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,8,balanced,0.06196799874305725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,8,balanced,0.06459199885527293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,8,balanced,0.0639573335647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,8,balanced,0.06629333396752675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,8,balanced,0.06773333251476288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,8,balanced,0.06635199983914693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,8,balanced,0.06809066732724507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,8,balanced,0.06976533432801564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,8,balanced,0.07180800040562947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,8,balanced,0.07947733501593272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,8,balanced,0.08250133196512859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,8,balanced,0.0853653351465861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,8,balanced,0.09574400385220845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,8,balanced,0.10226666927337646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,8,balanced,0.11938132842381795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,8,balanced,0.13210133711496988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,8,balanced,0.17360534270604452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,8,balanced,0.20710933208465576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,8,balanced,0.27856000264485675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,8,balanced,0.3426133394241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,8,balanced,0.4835466543833415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,8,balanced,0.6110293467839559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,16,balanced,0.03405333310365677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,16,balanced,0.031717332700888314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,16,balanced,0.031680000325044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,16,balanced,0.02975466599067052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,16,balanced,0.03301866600910822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,16,balanced,0.03586666782697042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,16,balanced,0.03749333322048187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,16,balanced,0.03457066665093104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,16,balanced,0.033999999364217125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,16,balanced,0.033759998778502144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,16,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,16,balanced,0.035786665976047516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,16,balanced,0.03480000048875809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,16,balanced,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,16,balanced,0.04002666721741358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,16,balanced,0.04013866682847341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,16,balanced,0.040021332601706185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,16,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,16,balanced,0.04204800228277842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,16,balanced,0.04809066653251648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,16,balanced,0.04782933493455251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,16,balanced,0.05666666726271311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,16,balanced,0.05801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,16,balanced,0.07808533310890198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,16,balanced,0.08457600076993306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,16,balanced,0.1070186694463094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,16,balanced,0.12801067034403482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,2,power_law_1.01,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,4,power_law_1.01,0.06231039762496948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,2,power_law_1.01,0.04009599983692169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,2,power_law_1.01,0.05812479853630066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,4,power_law_1.01,0.07291520237922669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,2,power_law_1.01,0.06714879870414733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,2,power_law_1.01,0.0797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,4,power_law_1.01,0.09786239862442017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,2,power_law_1.01,0.10593279600143432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,2,power_law_1.01,0.10929280519485474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,4,power_law_1.01,0.10430079698562622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,2,power_law_1.01,0.1155135989189148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,2,power_law_1.01,0.11756160259246826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,4,power_law_1.01,0.12136319875717164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,4,power_law_1.01,0.13987200260162352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,4,power_law_1.01,0.14458240270614625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,4,power_law_1.01,0.15690879821777343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,4,power_law_1.01,0.16981760263442994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,2,power_law_1.01,0.11701120138168335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,2,power_law_1.01,0.1195904016494751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,2,power_law_1.01,0.12247040271759033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,2,power_law_1.01,0.12394880056381226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,2,power_law_1.01,0.12904959917068481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,2,power_law_1.01,0.13169920444488525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,2,power_law_1.01,0.137772798538208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,2,power_law_1.01,0.14016000032424927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,2,power_law_1.01,0.15365760326385497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,4,power_law_1.01,0.16959359645843505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,2,power_law_1.01,0.16953599452972412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,4,power_law_1.01,0.1674880027770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,2,power_law_1.01,0.19070719480514525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,2,power_law_1.01,0.21146240234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,2,power_law_1.01,0.2583807945251465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,2,power_law_1.01,0.31804800033569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,2,power_law_1.01,0.4066944122314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,2,power_law_1.01,0.5011263847351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,4,power_law_1.01,0.16641279458999633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,2,power_law_1.01,0.7052544116973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,4,power_law_1.01,0.17748479843139647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,2,power_law_1.01,0.8704704284667969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,4,power_law_1.01,0.17493120431900025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,4,power_law_1.01,0.19129600524902343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,4,power_law_1.01,0.18818559646606445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,4,power_law_1.01,0.20677120685577394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,4,power_law_1.01,0.2222208023071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,4,power_law_1.01,0.23854079246520996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,4,power_law_1.01,0.2869247913360596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,4,power_law_1.01,0.3158143997192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,4,power_law_1.01,0.40604801177978517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,4,power_law_1.01,0.4829887866973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,4,power_law_1.01,0.6520832061767579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,4,power_law_1.01,0.8183487892150879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,4,power_law_1.01,1.1545215606689454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,4,power_law_1.01,1.5254655838012696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,1,balanced,0.06061866879463196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,1,balanced,0.0641599992911021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,1,balanced,0.07542400062084198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,1,balanced,0.09698133667310078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,1,balanced,0.14460800091425577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,1,balanced,0.2195626695950826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,1,balanced,0.2975306709607442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,1,balanced,0.2972533305486043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,1,balanced,0.30087467034657794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,1,balanced,0.3022720019022624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,1,balanced,0.30638400713602704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,1,balanced,0.31173332532246906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,1,balanced,0.31667200724283856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,1,balanced,0.3205653429031372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,1,balanced,0.33007999261220294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,1,balanced,0.3378239870071411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,32,balanced,0.03992533435424169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,1,balanced,0.3533546527226766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,1,balanced,0.3974826733271281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,1,balanced,0.42984533309936523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,1,balanced,0.5046186844507853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,1,balanced,0.5579306681950887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,256,power_law_1.01,0.051692801713943484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,1,balanced,0.702303965886434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,256,power_law_1.01,0.05139840245246887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,1,balanced,0.9085386594136556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,256,power_law_1.01,0.05204480290412903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,1,balanced,1.2063946723937988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,256,power_law_1.01,0.057011198997497556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,1,balanced,1.6051680246988933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,32,balanced,0.04163199911514918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,1,balanced,2.2751466433207193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,32,balanced,0.039818666875362396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,32,balanced,0.04775466521581014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,32,balanced,0.05808533231417338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,32,balanced,0.05509866774082184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,32,balanced,0.05420266588528951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,32,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,1,balanced,3.023834546407064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,32,balanced,0.05593066910902659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,32,balanced,0.056159997979799904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,32,balanced,0.05622933308283488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,32,balanced,0.057589332262674965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,32,balanced,0.05451733370621999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,32,balanced,0.05568000177542368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,32,balanced,0.05726400017738342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,32,balanced,0.060309335589408875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,32,balanced,0.05834133426348368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,32,balanced,0.060362666845321655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,32,balanced,0.06436266501744588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,32,balanced,0.06851733227570851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,32,balanced,0.07230400045712788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,32,balanced,0.09640533725420634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,32,balanced,0.10434133807818095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,32,balanced,0.13185600439707437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,32,balanced,0.1588640014330546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,32,balanced,0.2126026749610901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,32,balanced,0.26710400978724164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,256,power_law_1.01,0.06255360245704651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,256,power_law_1.01,0.07271680235862732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,256,power_law_1.01,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,256,power_law_1.01,0.07308160066604615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,256,power_law_1.01,0.07189760208129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,256,power_law_1.01,0.07418879866600037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,256,power_law_1.01,0.07502719759941101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,256,power_law_1.01,0.07882879972457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,256,power_law_1.01,0.07918720245361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,256,power_law_1.01,0.08071039915084839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,256,power_law_1.01,0.09011840224266052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,256,power_law_1.01,0.09193599820137024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,256,power_law_1.01,0.09882879853248597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,256,power_law_1.01,0.11784960031509399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,256,power_law_1.01,0.13308800458908082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,256,power_law_1.01,0.17493120431900025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,256,power_law_1.01,0.21789441108703614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,256,power_law_1.01,0.2838399887084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,256,power_law_1.01,0.3304383993148804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,16,balanced,0.05036800106366476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,16,balanced,0.054192001620928444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,256,power_law_1.01,0.47599358558654786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,16,balanced,0.0746666689713796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,16,balanced,0.10805867115656535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,16,balanced,0.14216533303260803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,16,balanced,0.15929067134857178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,256,power_law_1.01,0.6250944137573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,16,balanced,0.1602666676044464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,16,balanced,0.15991999705632529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,16,balanced,0.1609599987665812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,16,balanced,0.16246400276819864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,256,power_law_1.01,0.9056896209716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,16,balanced,0.16425599654515585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,16,balanced,0.16582399606704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,16,balanced,0.1667893330256144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,16,balanced,0.16856533288955688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,16,balanced,0.17534399032592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,256,power_law_1.01,1.1292160034179688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,16,balanced,0.17574934164683023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,16,balanced,0.18372799952824911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,16,balanced,0.2071733276049296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,16,balanced,0.2064853310585022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,16,balanced,0.26320000489552814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,16,balanced,0.2588160037994385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,16,balanced,0.3627893527348836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,16,balanced,0.3319946726163228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,16,balanced,0.4781866470972697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,16,balanced,0.5002133448918661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,16,balanced,0.726693312327067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,16,balanced,0.8199413617451986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,64,power_law_1.2,0.028729599714279175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,64,power_law_1.2,0.02858879864215851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,64,power_law_1.2,0.028364801406860353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,64,power_law_1.2,0.030483201146125793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,64,power_law_1.2,0.03126400113105774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,64,power_law_1.2,0.032179200649261476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,64,power_law_1.2,0.0321727991104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,64,power_law_1.2,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,64,power_law_1.2,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,64,power_law_1.2,0.032876798510551454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,64,power_law_1.2,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,64,power_law_1.2,0.035795199871063235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,64,power_law_1.2,0.03723520040512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,64,power_law_1.2,0.03929600119590759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,64,power_law_1.2,0.04631040096282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,64,power_law_1.2,0.044352000951766966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,64,power_law_1.2,0.049472001194953916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,64,power_law_1.2,0.05788159966468811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,64,power_law_1.2,0.06575359702110291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,64,power_law_1.01,0.01730560064315796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,64,power_law_1.01,0.016531200706958772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,64,power_law_1.01,0.01608320027589798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,64,power_law_1.01,0.017183999717235564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,64,power_law_1.01,0.017452800273895265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,64,power_law_1.01,0.0197952002286911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,64,power_law_1.01,0.020563200116157532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,64,power_law_1.2,0.09453439712524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,64,power_law_1.01,0.027647998929023743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,64,power_law_1.2,0.10703999996185302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,64,power_law_1.2,0.15048320293426515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,64,power_law_1.2,0.20628480911254882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,64,power_law_1.01,0.028147199749946596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,64,power_law_1.01,0.028070399165153505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,64,power_law_1.01,0.02800639867782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,64,power_law_1.01,0.028467199206352232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,64,power_law_1.01,0.02800639867782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,64,power_law_1.01,0.02776319980621338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,64,power_law_1.01,0.028883200883865357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,64,power_law_1.01,0.029235199093818665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,64,power_law_1.01,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,64,power_law_1.01,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,64,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,64,power_law_1.2,0.29935359954833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,64,power_law_1.01,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,64,power_law_1.01,0.04438399970531463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,64,power_law_1.01,0.05137919783592224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,64,power_law_1.01,0.0636672019958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,64,power_law_1.01,0.08375679850578308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,64,power_law_1.01,0.10216959714889526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,64,power_law_1.01,0.13983360528945923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,64,power_law_1.01,0.19158400297164918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,64,power_law_1.2,0.3631232023239136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,64,power_law_1.2,0.5968832015991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,64,power_law_1.2,0.890777587890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,4,power_law_1.2,0.0785152018070221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,4,power_law_1.2,0.08415359854698182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,4,power_law_1.2,0.09825279712677001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,4,power_law_1.2,0.12439039945602418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,4,power_law_1.2,0.12188800573348998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,4,power_law_1.2,0.16526080369949342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,4,power_law_1.2,0.16817920207977294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,4,power_law_1.2,0.16212480068206786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,4,power_law_1.2,0.16904959678649903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,4,power_law_1.2,0.1694975972175598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,4,power_law_1.2,0.1718016028404236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,4,power_law_1.2,0.17948800325393677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,4,power_law_1.2,0.18423680067062378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,4,power_law_1.2,0.19153280258178712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,4,power_law_1.2,0.20183680057525635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,4,power_law_1.2,0.21159040927886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,4,power_law_1.2,0.22076799869537353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,4,power_law_1.2,0.255513596534729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,4,power_law_1.2,0.2533951997756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,4,power_law_1.2,0.3106816053390503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,4,power_law_1.2,0.34816000461578367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,4,power_law_1.2,0.390720009803772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,4,power_law_1.2,0.4971519947052002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,4,power_law_1.2,0.6568128108978272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,4,power_law_1.2,0.8064512252807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,32,power_law_1.2,0.01793919950723648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,64,power_law_1.2,0.052934402227401735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,32,power_law_1.2,0.017420800030231477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,64,power_law_1.2,0.04389120042324066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,32,power_law_1.2,0.017759999632835387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,64,power_law_1.2,0.04230400025844574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,64,power_law_1.2,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,64,power_law_1.2,0.04657280147075653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,64,power_law_1.2,0.04467839896678925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,4,power_law_1.2,1.1465984344482423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,64,power_law_1.2,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,64,power_law_1.2,0.047513601183891294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,32,power_law_1.2,0.017401599884033205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,32,power_law_1.2,0.017990399897098542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,32,power_law_1.2,0.018291200697422027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,32,power_law_1.2,0.02305919975042343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,32,power_law_1.2,0.02322559952735901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,32,power_law_1.2,0.03661440014839172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,64,power_law_1.2,0.04776319861412048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,64,power_law_1.2,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,64,power_law_1.2,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,64,power_law_1.2,0.05317760109901428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,64,power_law_1.2,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,4,power_law_1.2,1.5272192001342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,64,power_law_1.2,0.05686399936676025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,64,power_law_1.2,0.06225280165672302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,64,power_law_1.2,0.06559360027313232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,32,power_law_1.2,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,64,power_law_1.2,0.07049599885940552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,64,power_law_1.2,0.08172799944877625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,64,power_law_1.2,0.08821759819984436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,64,power_law_1.2,0.11125119924545288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,64,power_law_1.2,0.12004480361938477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,64,power_law_1.2,0.1565824031829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,64,power_law_1.2,0.19791359901428224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,64,power_law_1.2,0.25692799091339114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,64,power_law_1.2,0.3446592092514038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,64,power_law_1.2,0.45495038032531737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,64,power_law_1.2,0.6175551891326905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,32,power_law_1.2,0.03692159950733185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,32,power_law_1.2,0.03338240087032318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,32,power_law_1.2,0.03361279964447021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,32,power_law_1.2,0.03362559974193573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,32,power_law_1.2,0.03475840091705322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,32,power_law_1.2,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,32,power_law_1.2,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,32,power_law_1.2,0.03646079897880554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,32,power_law_1.2,0.039852800965309146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,32,power_law_1.2,0.04366720020771027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,32,power_law_1.2,0.04806399941444397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,32,power_law_1.2,0.06035199761390686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,32,power_law_1.2,0.07359359860420227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,32,power_law_1.2,0.09676799774169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,32,power_law_1.2,0.1153216004371643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,32,power_law_1.2,0.1600640058517456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,32,power_law_1.2,0.20995841026306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,8,power_law_1.2,0.032876798510551454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,8,power_law_1.2,0.0328000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,8,power_law_1.2,0.033452799916267394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,8,power_law_1.2,0.035872000455856326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,8,power_law_1.2,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,8,power_law_1.2,0.03874559998512268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,8,power_law_1.2,0.03953920006752014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,8,power_law_1.2,0.04016000032424927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,8,power_law_1.2,0.03898879885673523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,8,power_law_1.2,0.03937920033931732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,8,power_law_1.2,0.039743998646736146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,8,power_law_1.2,0.04051840007305145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,8,power_law_1.2,0.04268159866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,8,power_law_1.2,0.04306559860706329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,8,power_law_1.2,0.0511680006980896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,8,power_law_1.2,0.051648002862930295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,8,power_law_1.2,0.05562880039215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,8,power_law_1.2,0.062431997060775755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,8,power_law_1.2,0.07009279727935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,8,power_law_1.2,0.08885120153427124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,8,power_law_1.2,0.10837759971618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,8,power_law_1.2,0.1384063959121704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,8,power_law_1.2,0.166867196559906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,16,power_law_1.2,0.11765120029449463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,16,power_law_1.2,0.13730560541152953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,8,power_law_1.2,0.2803584098815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,16,power_law_1.2,0.14353920221328736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,16,power_law_1.2,0.15590399503707886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,16,power_law_1.2,0.15082240104675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,16,power_law_1.2,0.1397503972053528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,16,power_law_1.2,0.14174720048904418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,16,power_law_1.2,0.14369280338287355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,16,power_law_1.2,0.14785280227661132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,16,power_law_1.2,0.14365439414978026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,16,power_law_1.2,0.15343999862670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,16,power_law_1.2,0.15356160402297975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,16,power_law_1.2,0.1593664050102234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,16,power_law_1.2,0.1733504056930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,8,power_law_1.2,0.3877824068069458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,16,power_law_1.2,0.173580801486969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,16,power_law_1.2,0.1804479956626892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,16,power_law_1.2,0.19589760303497314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,16,power_law_1.2,0.2275968074798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,16,power_law_1.2,0.24065918922424318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,16,power_law_1.2,0.2979072093963623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,16,power_law_1.2,0.3225663900375366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,16,power_law_1.2,0.41079039573669435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,16,power_law_1.2,0.46317439079284667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,16,power_law_1.2,0.5570303916931152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,16,power_law_1.2,0.7316480159759522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,8,power_law_1.2,0.501043176651001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,16,power_law_1.2,1.0897791862487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,16,power_law_1.2,1.2783231735229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,16,balanced,0.058490668733914696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,16,balanced,0.062352001667022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,16,balanced,0.07428800066312154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,16,balanced,0.0946560005346934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,16,balanced,0.12038399775822957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,16,balanced,0.12956800063451132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,16,balanced,0.13011733690897623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,16,balanced,0.13057600458463034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,16,balanced,0.13328533371289572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,16,balanced,0.13238933682441711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,16,balanced,0.13211199641227722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,16,balanced,0.13806399703025818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,16,balanced,0.13726400335629782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,16,balanced,0.1402186652024587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,16,balanced,0.14668266971906027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,16,balanced,0.14672533671061197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,16,balanced,0.15406933426856995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,16,balanced,0.16589867075284323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,16,balanced,0.17786665757497153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,16,balanced,0.22260799010594687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,16,balanced,0.2328746716181437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,16,balanced,0.3261706630388896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,16,balanced,0.3232693274815877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,16,balanced,0.45045868555704754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,16,balanced,0.5002079804738363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,16,balanced,0.7056000232696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,16,balanced,0.8542506694793701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,8,power_law_1.2,0.7609344005584717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,1,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,1,balanced,0.04744533201058706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,1,balanced,0.06234666705131531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,1,balanced,0.0660693347454071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,1,balanced,0.06821333368619283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,1,balanced,0.06811200082302094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,1,balanced,0.06844266752401988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,1,balanced,0.07010133564472198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,1,balanced,0.06916800141334534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,1,balanced,0.07014933228492737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,1,balanced,0.07025599976380666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,1,balanced,0.07444266478220622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,1,balanced,0.0747519979874293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,1,balanced,0.07718400160471599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,1,balanced,0.0842080016930898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,1,balanced,0.08813333511352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,1,balanced,0.08704533179601033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,1,balanced,0.11554666360219319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,1,balanced,0.12541332840919495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,1,balanced,0.15969066818555197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,1,balanced,0.19339199860890707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,1,balanced,0.290010670820872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,1,balanced,0.37015998363494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,1,balanced,0.5268160104751587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,1,balanced,0.6757919788360596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,1,balanced,0.9744853178660074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,1,balanced,1.281450668970744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,2,power_law_1.01,0.018246400356292724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,2,power_law_1.01,0.01875839978456497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,2,power_law_1.01,0.020883199572563172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,2,power_law_1.01,0.02499839961528778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,2,power_law_1.01,0.026624000072479247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,2,power_law_1.01,0.030611199140548707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,2,power_law_1.01,0.03150080144405365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,2,power_law_1.01,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,2,power_law_1.01,0.037401598691940305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,2,power_law_1.01,0.03805440068244934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,2,power_law_1.01,0.03924480080604553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,2,power_law_1.01,0.0392767995595932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,2,power_law_1.01,0.040012800693511964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,2,power_law_1.01,0.043270400166511534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,2,power_law_1.01,0.05132160186767578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,2,power_law_1.01,0.04992640018463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,2,power_law_1.01,0.05380480289459229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,2,power_law_1.01,0.09292160272598267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,2,power_law_1.01,0.10033279657363892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,2,power_law_1.01,0.09818239808082581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,2,power_law_1.01,0.1062656044960022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,2,power_law_1.01,0.1313088059425354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,2,power_law_1.01,0.14792319536209106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,256,balanced,0.06298666695753734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,2,power_law_1.01,0.18883199691772462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,2,power_law_1.01,0.23595519065856935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,2,power_law_1.01,0.32334721088409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,2,power_law_1.01,0.43102078437805175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,256,balanced,0.06225066880385081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,256,balanced,0.06413333117961884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,256,balanced,0.06620266536871593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,256,balanced,0.06462933123111725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,2,power_law_1.01,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,256,balanced,0.06824533144632976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,256,balanced,0.06825600067774455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,2,power_law_1.01,0.06270080208778381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,2,power_law_1.01,0.09354239702224731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,2,power_law_1.01,0.11276160478591919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,2,power_law_1.01,0.11909760236740112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,2,power_law_1.01,0.11939200162887573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,2,power_law_1.01,0.13397760391235353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,2,power_law_1.01,0.12785919904708862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,2,power_law_1.01,0.13006720542907715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,2,power_law_1.01,0.12954239845275878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,256,balanced,0.06814399858315785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,256,balanced,0.07145600020885468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,256,balanced,0.07029866675535838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,256,balanced,0.07022933165232341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,256,balanced,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,256,balanced,0.07291199763615926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,256,balanced,0.07215466598669688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,256,balanced,0.07867200175921123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,256,balanced,0.07892266909281413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,256,balanced,0.08046933511892955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,256,balanced,0.08850133419036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,256,balanced,0.09470400214195251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,256,balanced,0.1165173351764679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,256,balanced,0.13831999897956848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,256,balanced,0.1776640017827352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,256,balanced,0.2104640007019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,256,balanced,0.2799359957377116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,2,power_law_1.01,0.1279039978981018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,256,balanced,0.3498293161392212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,2,power_law_1.01,0.13449599742889404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,256,balanced,0.48712531725565594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,2,power_law_1.01,0.1368384003639221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,256,balanced,0.6223413149515787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,2,power_law_1.01,0.14766720533370972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,2,power_law_1.01,0.1575808048248291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,2,power_law_1.01,0.16248960494995118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,2,power_law_1.01,0.1833407998085022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,2,power_law_1.01,0.19576319456100463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,2,power_law_1.01,0.22444798946380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,2,power_law_1.01,0.3251456022262573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,2,power_law_1.01,0.349068808555603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,2,power_law_1.01,0.49956479072570803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,2,power_law_1.01,0.6024256229400635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,2,power_law_1.01,0.8045248031616211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,2,power_law_1.01,0.9949952125549316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,2,power_law_1.01,1.337228775024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,4,power_law_1.2,0.021721599996089934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,4,power_law_1.2,0.021484799683094025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,2,power_law_1.01,2.3590784072875977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,4,power_law_1.2,0.020953600108623505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,4,power_law_1.2,0.021619200706481934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,4,power_law_1.2,0.022342400252819063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,4,power_law_1.2,0.02308479994535446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,4,power_law_1.2,0.02959359884262085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,4,power_law_1.2,0.026335999369621277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,4,power_law_1.2,0.051481598615646364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,4,power_law_1.2,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,4,power_law_1.2,0.051846402883529666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,4,power_law_1.2,0.05140479803085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,4,power_law_1.2,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,4,power_law_1.2,0.04042240083217621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,4,power_law_1.2,0.06218240261077881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,4,power_law_1.2,0.0527616024017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,4,power_law_1.2,0.05926399827003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,4,power_law_1.2,0.07342720031738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,2,power_law_1.2,0.018361599743366243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,2,power_law_1.2,0.01727360039949417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,2,power_law_1.2,0.017791999876499175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,2,power_law_1.2,0.01912959963083267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,2,power_law_1.2,0.021299199759960176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.2,0.09101439714431762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,2,power_law_1.2,0.021401600539684297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,2,power_law_1.2,0.02197760045528412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,2,power_law_1.2,0.02250239998102188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,2,power_law_1.2,0.023846399784088135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,2,power_law_1.2,0.02327679991722107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,2,power_law_1.2,0.02369280010461807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,2,power_law_1.2,0.04176000058650971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,2,power_law_1.2,0.0417279988527298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,2,power_law_1.2,0.04229120016098022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,2,power_law_1.2,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.2,0.12462079524993896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,2,power_law_1.2,0.03370240032672882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,2,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,2,power_law_1.2,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.2,0.047916799783706665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.2,0.15101439952850343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.2,0.06317440271377564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.2,0.21196160316467286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,16,power_law_1.01,0.01642879992723465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,16,power_law_1.01,0.016012799739837647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.2,0.07672320008277893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,16,power_law_1.01,0.016076800227165223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,16,power_law_1.01,0.016441600024700166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,16,power_law_1.01,0.017715199291706084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,16,power_law_1.01,0.01791359931230545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,16,power_law_1.01,0.018931199610233308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.2,0.26720640659332273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,16,power_law_1.01,0.019968000054359437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,16,power_law_1.01,0.020204800367355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,16,power_law_1.01,0.020287999510765077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,16,power_law_1.01,0.0222463995218277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,16,power_law_1.01,0.02268799990415573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,16,power_law_1.01,0.02336000055074692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,16,power_law_1.01,0.031379199028015135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,16,power_law_1.01,0.031999999284744264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,16,power_law_1.01,0.032076799869537355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,16,power_law_1.01,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,16,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,16,power_law_1.01,0.037254399061203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,16,power_law_1.01,0.04168959856033325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,16,power_law_1.01,0.04523519873619079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,16,power_law_1.01,0.05571200251579285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,16,power_law_1.01,0.06332799792289734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,16,power_law_1.01,0.08455039858818054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,16,power_law_1.01,0.10857599973678589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,16,power_law_1.01,0.14227839708328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,16,power_law_1.01,0.1835904002189636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.2,0.09639040231704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,64,balanced,0.025583999852339428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,64,balanced,0.023210667073726654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,64,balanced,0.02312533309062322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,64,balanced,0.021365332106749218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,64,balanced,0.023061332603295643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,64,balanced,0.023050665855407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,64,balanced,0.02553066611289978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,64,balanced,0.027050666511058807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,64,balanced,0.0331839993596077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,64,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,64,balanced,0.04374399781227112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,64,balanced,0.044256001710891724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,64,balanced,0.04260799785455068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,64,balanced,0.043562665581703186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,64,balanced,0.05726400017738342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,64,balanced,0.0583840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,64,balanced,0.05169600248336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,64,balanced,0.05580266813437144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,64,balanced,0.06562666594982147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,64,balanced,0.09237333138783772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,64,balanced,0.10326932867368062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,64,balanced,0.14261866609255472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,64,balanced,0.18126932779947916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,64,balanced,0.25828800598780316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,64,balanced,0.3253386616706848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,64,balanced,0.478277325630188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,64,balanced,0.6405813296635946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.2,0.12152320146560669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.2,0.383078408241272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.2,0.1651584029197693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.2,0.4980031967163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.2,0.20768001079559326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.2,0.7297791957855224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.2,0.29506559371948243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,1,balanced,0.021829334398110706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,1,balanced,0.023141334454218548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,1,balanced,0.025413334369659424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,1,balanced,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,1,balanced,0.059936001896858215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,1,balanced,0.09684266646703084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,1,balanced,0.10096533099810283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,1,balanced,0.10345066587130229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,1,balanced,0.107232004404068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,1,balanced,0.10904000202814738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,1,balanced,0.11103467146555583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,1,balanced,0.11538666486740112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,1,balanced,0.1181706686814626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,1,balanced,0.1202186644077301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,1,balanced,0.12402666608492534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,1,balanced,0.12703999876976013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,1,balanced,0.14007467031478882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,1,balanced,0.16142933567365012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,1,balanced,0.17243733008702597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,1,balanced,0.21574934323628744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,1,balanced,0.2476053237915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,1,balanced,0.3307360013326009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,1,balanced,0.335477352142334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,1,balanced,0.4676320155461629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,1,balanced,0.5966240167617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,1,balanced,0.8835466702779134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,1,balanced,1.116085370381673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,128,balanced,0.03046933313210805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,128,balanced,0.030213333666324615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,128,balanced,0.029546665648619335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,128,balanced,0.03163733333349228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,128,balanced,0.03327466547489166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.2,0.38190720081329343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,128,balanced,0.03369066615899404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,128,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.2,0.9609343528747558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,128,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,128,balanced,0.0355679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,128,balanced,0.03572800010442734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,128,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,128,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,128,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,128,balanced,0.041663999358812966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,128,balanced,0.046623999873797096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,128,balanced,0.04589866598447164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,128,balanced,0.04821333289146423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,128,balanced,0.052000001072883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,128,balanced,0.05425600210825602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,128,balanced,0.060933331648508705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,128,balanced,0.06412800153096516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,128,balanced,0.08217599987983704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,128,balanced,0.09505599737167358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,128,balanced,0.12153599659601848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,128,balanced,0.14591466387112936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,128,balanced,0.19207467635472616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,128,balanced,0.2398080031077067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,8,power_law_1.01,0.05135359764099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,8,power_law_1.01,0.05149440169334411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,8,power_law_1.01,0.061587202548980716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,8,power_law_1.01,0.06727679967880248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,8,power_law_1.01,0.07214720249176025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,8,power_law_1.01,0.07167999744415283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,8,power_law_1.01,0.07451519966125489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,8,power_law_1.01,0.07004799842834472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,8,power_law_1.01,0.07539839744567871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,8,power_law_1.01,0.07536640167236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,8,power_law_1.01,0.07681919932365418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,8,power_law_1.01,0.08242560029029847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,8,power_law_1.01,0.085343998670578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,8,power_law_1.01,0.09112319946289063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,8,power_law_1.01,0.09964159727096558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,8,power_law_1.01,0.1017408013343811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,8,power_law_1.01,0.1082111954689026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,8,power_law_1.01,0.12194559574127198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,8,power_law_1.01,0.1346176028251648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,8,power_law_1.01,0.1600640058517456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,8,power_law_1.01,0.18371200561523438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,8,power_law_1.01,0.24853758811950682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,8,power_law_1.01,0.3070784091949463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,8,power_law_1.01,0.40470399856567385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,8,power_law_1.01,0.5255104064941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,8,power_law_1.01,0.7636223793029785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,8,power_law_1.01,0.9797887802124023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,128,power_law_1.2,0.01772159934043884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,128,power_law_1.2,0.017817600071430205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,128,power_law_1.2,0.01855359971523285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,128,power_law_1.2,0.020576000213623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,128,power_law_1.2,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,128,power_law_1.2,0.029177600145339967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,128,power_law_1.2,0.02951039969921112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,128,power_law_1.2,0.029651200771331786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,128,power_law_1.2,0.030073601007461547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,128,power_law_1.2,0.029900801181793214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,128,power_law_1.2,0.030406400561332703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,128,power_law_1.2,0.030720001459121703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,128,power_law_1.2,0.03079040050506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,128,power_law_1.2,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,128,power_law_1.2,0.03372800052165985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,128,power_law_1.2,0.03468160033226013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,128,power_law_1.2,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,128,power_law_1.2,0.03726080060005188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.2,0.040582400560379026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.2,0.04672000110149384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.2,0.04967679977416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,1,balanced,0.15461333592732748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.2,0.060627198219299315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,1,balanced,0.2598080039024353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,1,balanced,0.47046399116516113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,1,balanced,0.892143964767456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.2,0.07945600152015686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,1,balanced,1.7233866055806477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,1,balanced,2.1420106887817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,1,balanced,2.153733412424723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,1,balanced,2.1665333112080893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,1,balanced,2.182037353515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,1,balanced,2.181311925252279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,1,balanced,2.1943626403808594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,1,balanced,2.194986661275228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,1,balanced,2.2036479314168296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,1,balanced,2.2186293601989746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,1,balanced,2.2443092664082847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,1,balanced,2.2577706972757974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,1,balanced,2.2888693809509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,1,balanced,2.3407039642333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,1,balanced,2.393237272898356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,1,balanced,3.038090705871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,1,balanced,2.5962613423665366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,1,balanced,3.5335305531819663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,1,balanced,3.0238240559895835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.2,0.10297600030899048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,1,balanced,5.068304061889648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,1,balanced,4.855546633402507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,1,balanced,8.153008143107096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,1,balanced,9.11738141377767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.2,0.1289791941642761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.2,0.1784000039100647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,8,power_law_1.2,0.050995200872421265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,8,power_law_1.2,0.056403201818466184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,8,power_law_1.2,0.0645632028579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,8,power_law_1.2,0.08049280047416688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,8,power_law_1.2,0.07006719708442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,8,power_law_1.2,0.08097919821739197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,8,power_law_1.2,0.08235520124435425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,8,power_law_1.2,0.082150399684906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,8,power_law_1.2,0.08575999736785889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,8,power_law_1.2,0.08697599768638611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,8,power_law_1.2,0.08851839900016785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,8,power_law_1.2,0.09297919869422913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,8,power_law_1.2,0.09367679953575134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,8,power_law_1.2,0.10000640153884888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,8,power_law_1.2,0.10652159452438355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,8,power_law_1.2,0.10951679944992065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,8,power_law_1.2,0.11273599863052368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,8,power_law_1.2,0.12954239845275878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,8,power_law_1.2,0.15277440547943116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,8,power_law_1.2,0.186572802066803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,8,power_law_1.2,0.2248447895050049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,8,power_law_1.2,0.28718719482421873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,8,power_law_1.2,0.36462080478668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,8,power_law_1.2,0.4739840030670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,8,power_law_1.2,0.6908480167388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,1,power_law_1.01,0.0511680006980896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,1,power_law_1.01,0.08090879917144775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,1,power_law_1.01,0.1297600030899048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,8,power_law_1.2,0.7988096237182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,1,power_law_1.01,0.19898879528045654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,1,power_law_1.01,0.2744256019592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,1,power_law_1.01,0.42949762344360354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,1,power_law_1.01,0.47756161689758303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,1,power_law_1.01,0.4562943935394287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,1,power_law_1.01,0.504307222366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,1,power_law_1.01,0.5289152145385743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,1,power_law_1.01,0.5770624160766602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,1,power_law_1.01,0.5651584148406983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,1,power_law_1.01,0.540940809249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,1,power_law_1.01,0.5749504089355468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,1,power_law_1.01,0.6191679954528808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,1,power_law_1.01,0.5729343891143799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,1,power_law_1.01,0.6128767967224121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,16,power_law_1.01,0.04035199880599975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,1,power_law_1.01,0.6377408027648925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,16,power_law_1.01,0.038124799728393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.01,0.6797696113586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,16,power_law_1.01,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,16,power_law_1.01,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,16,power_law_1.01,0.03431679904460907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.01,1.6297855377197266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,16,power_law_1.01,0.036320000886917114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,16,power_law_1.01,0.036524799466133115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,16,power_law_1.01,0.036575999855995175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,16,power_law_1.01,0.037011200189590455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.01,1.878598403930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,16,power_law_1.01,0.03688960075378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,16,power_law_1.01,0.03829120099544525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,16,power_law_1.01,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,16,power_law_1.01,0.03888640105724335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,16,power_law_1.01,0.04201599955558777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,16,power_law_1.01,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,16,power_law_1.01,0.04875519871711731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,16,power_law_1.01,0.04928640127182007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,16,power_law_1.01,0.05334399938583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,16,power_law_1.01,0.05743359923362732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,16,power_law_1.01,0.06165120005607605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,16,power_law_1.01,0.07071359753608704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,16,power_law_1.01,0.08848000168800355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,16,power_law_1.01,0.1030079960823059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,16,power_law_1.01,0.15179519653320311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,16,power_law_1.01,0.18354560136795045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,16,power_law_1.01,0.27513599395751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.01,1.1321344375610352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,8,power_law_1.2,1.1183168411254882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,16,power_law_1.01,0.3176000118255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.01,1.2015744209289552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.01,1.4630720138549804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.01,1.7890047073364257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.01,2.365951919555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.01,2.958950424194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,128,balanced,0.03344533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,128,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,128,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,128,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,128,balanced,0.035599999129772186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,128,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,128,balanced,0.037231999138991036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,128,balanced,0.03841600070397059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,128,balanced,0.039605334401130676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,128,balanced,0.03756266583998998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,128,balanced,0.039077334105968475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,128,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,128,balanced,0.03972266614437103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,128,balanced,0.040063999593257904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,128,balanced,0.04262933135032654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,128,balanced,0.043920000394185386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,128,balanced,0.04413333535194397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,128,balanced,0.04599999884764353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,128,balanced,0.04693866769472758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,128,balanced,0.05203199883302053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,128,balanced,0.05410666763782501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,128,balanced,0.06017066538333893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,128,balanced,0.06592533489068349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,128,balanced,0.09286399682362874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,128,balanced,0.11866133411725362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,128,balanced,0.15467733144760132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,128,balanced,0.19131199518839517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.2,0.23116159439086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,8,power_law_1.01,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,8,power_law_1.01,0.05973119735717773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,8,power_law_1.01,0.06280959844589233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,8,power_law_1.01,0.06690559983253479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,8,power_law_1.01,0.07812479734420777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,8,power_law_1.01,0.07623680233955384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,8,power_law_1.01,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,8,power_law_1.01,0.08086400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,8,power_law_1.01,0.0791487991809845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,8,power_law_1.01,0.07942399978637696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,8,power_law_1.01,0.08245760202407837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,8,power_law_1.01,0.08432000279426574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,8,power_law_1.01,0.08541439771652222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,8,power_law_1.01,0.08977280259132385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,8,power_law_1.01,0.09375360012054443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,8,power_law_1.01,0.10127359628677368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,8,power_law_1.01,0.1059008002281189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,8,power_law_1.01,0.12536959648132323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,8,power_law_1.01,0.12710399627685548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,8,power_law_1.01,0.15846400260925292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,8,power_law_1.01,0.17386239767074585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,8,power_law_1.01,0.21789441108703614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,8,power_law_1.01,0.27628800868988035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,8,power_law_1.01,0.3739840030670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,8,power_law_1.01,0.43946242332458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,8,power_law_1.01,0.6109248161315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,8,power_law_1.01,0.8708095550537109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,2,power_law_1.2,0.04357120096683502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,2,power_law_1.2,0.04134399890899658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,2,power_law_1.2,0.04240640103816986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,2,power_law_1.2,0.04769279956817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,2,power_law_1.2,0.05097600221633911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,2,power_law_1.2,0.05898879766464234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,2,power_law_1.2,0.06119679808616638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,2,power_law_1.2,0.061401599645614625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,2,power_law_1.2,0.06327040195465088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,2,power_law_1.2,0.06392319798469544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,2,power_law_1.2,0.06540799736976624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,2,power_law_1.2,0.06922879815101624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,2,power_law_1.2,0.07066879868507385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,2,power_law_1.2,0.07221760153770447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,2,power_law_1.2,0.07672960162162781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,2,power_law_1.2,0.08062719702720642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,2,power_law_1.2,0.08907520174980163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,2,power_law_1.2,0.1034432053565979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,2,power_law_1.2,0.1189568042755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,2,power_law_1.2,0.14665600061416625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,2,power_law_1.2,0.172326397895813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,2,power_law_1.2,0.23496320247650146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,2,power_law_1.2,0.2794624090194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,2,power_law_1.01,0.17036160230636596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,2,power_law_1.01,0.19664000272750853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,2,power_law_1.01,0.2392575979232788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,2,power_law_1.01,0.28669440746307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,2,power_law_1.01,0.29577600955963135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,2,power_law_1.01,0.2734783887863159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,2,power_law_1.01,0.29261438846588134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,2,power_law_1.01,0.29109759330749513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,2,power_law_1.01,0.2869312047958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,2,power_law_1.01,0.26331520080566406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,2,power_law_1.01,0.2709824085235596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,2,power_law_1.01,0.3090240001678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,2,power_law_1.01,0.28512001037597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,2,power_law_1.01,0.2896640062332153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,2,power_law_1.01,0.3353791952133179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,2,power_law_1.01,0.3548991918563843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,2,power_law_1.01,0.4213888168334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,2,power_law_1.01,0.4569087982177734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.01,0.5393919944763184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,1,power_law_1.01,0.06546559929847717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,1,power_law_1.01,0.07864320278167725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,1,power_law_1.01,0.10858240127563476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,1,power_law_1.01,0.17165440320968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,1,power_law_1.01,0.24462080001831055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,1,power_law_1.01,0.3113215923309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,1,power_law_1.01,0.4171455860137939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.01,0.7708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,1,power_law_1.01,0.4341695785522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,1,power_law_1.01,0.4493760108947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,4,power_law_1.01,0.03115519881248474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,4,power_law_1.01,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,1,power_law_1.01,0.46048641204833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,4,power_law_1.01,0.028838399052619933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,4,power_law_1.01,0.03244799971580505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,1,power_law_1.01,0.4792191982269287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,2,power_law_1.2,0.4242368221282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,1,power_law_1.01,0.49286398887634275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,1,power_law_1.01,0.5016704082489014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,1,power_law_1.01,0.5135168075561524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,4,power_law_1.01,0.03351680040359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.01,0.9196160316467286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,1,power_law_1.01,0.5253376007080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,4,power_law_1.01,0.038975998759269714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,4,power_law_1.01,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,1,power_law_1.01,0.5401535987854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,4,power_law_1.01,0.04490239918231964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,4,power_law_1.01,0.06239359974861145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,1,power_law_1.01,0.5522624015808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,4,power_law_1.01,0.05057920217514038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,4,power_law_1.01,0.05585920214653015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,4,power_law_1.01,0.06967679858207702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.01,1.1774399757385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,4,power_law_1.01,0.07468799948692321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,4,power_law_1.01,0.07308800220489502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,4,power_law_1.01,0.1054527997970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,4,power_law_1.01,0.08371840119361877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,1,power_law_1.01,0.6152575969696045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,1,power_law_1.01,0.6250048160552979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,4,power_law_1.01,0.09807999730110169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,4,power_law_1.01,0.13819520473480223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,1,power_law_1.01,0.7056640148162842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,1,power_law_1.01,0.7346816062927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.01,0.16287360191345215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,1,power_law_1.01,0.8499584197998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,1,power_law_1.01,0.9938240051269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.01,1.4575872421264648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,1,power_law_1.01,1.2477439880371093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,1,power_law_1.01,1.4983039855957032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,1,power_law_1.01,1.9951040267944335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,2,power_law_1.2,0.5301439762115479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,1,power_law_1.01,2.5900991439819334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.01,0.22979199886322021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.01,0.29198079109191893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.01,0.41811838150024416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.01,2.0855871200561524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,2,power_law_1.2,0.7765247821807861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.01,0.5303999900817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.01,0.7820479869842529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.01,3.0961280822753907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.01,1.0182784080505372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.01,4.172441482543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.01,1.5136128425598145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,8,power_law_1.2,0.05806080102920532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,8,power_law_1.2,0.052153599262237546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,8,power_law_1.2,0.05255680084228516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,8,power_law_1.2,0.060166400671005246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,8,power_law_1.2,0.06351360082626342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,8,power_law_1.2,0.062489598989486694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,8,power_law_1.2,0.06437119841575623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,8,power_law_1.2,0.06261759996414185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,8,power_law_1.2,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,8,power_law_1.2,0.06614400148391723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,8,power_law_1.2,0.06875519752502442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,8,power_law_1.2,0.06894720196723939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,8,power_law_1.2,0.0735871970653534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,8,power_law_1.2,0.07617920041084289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,8,power_law_1.2,0.08279680013656616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,8,power_law_1.2,0.09343360066413879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,8,power_law_1.2,0.08840320110321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,8,power_law_1.2,0.10142719745635986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,8,power_law_1.2,0.1122431993484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,8,power_law_1.2,0.14558080434799195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,8,power_law_1.2,0.16083840131759644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,8,power_law_1.2,0.21297919750213623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,8,power_law_1.2,0.2795583963394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,8,power_law_1.2,0.37842559814453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,8,power_law_1.2,0.46028800010681153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,8,power_law_1.2,0.6345983982086182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,8,power_law_1.2,0.8788224220275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,2,power_law_1.2,1.0721728324890136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.01,1.9992063522338868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.01,5.713107299804688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,4,balanced,0.03603733330965042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,4,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,4,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,4,balanced,0.03979733337958654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,4,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,4,balanced,0.03868266691764196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,4,balanced,0.03765333443880081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,4,balanced,0.03861333429813385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,4,balanced,0.03993066648642222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,4,balanced,0.039701332648595176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,4,balanced,0.0415040006240209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,4,balanced,0.04249600072701772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,4,balanced,0.042730664213498436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,4,balanced,0.049642667174339294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,4,balanced,0.05604266623655955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,4,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,4,balanced,0.06471999982992808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,4,balanced,0.0740586668252945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,4,balanced,0.07800533374150594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,4,balanced,0.097680002450943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,4,balanced,0.10341866811116536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,4,balanced,0.13191999991734824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,4,balanced,0.1604373355706533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,4,balanced,0.2155253291130066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,4,balanced,0.27083200216293335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,4,balanced,0.3725493351618449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,4,balanced,0.4714826742808024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,2,power_law_1.2,0.050271999835968015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,2,power_law_1.2,0.06958079934120179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,2,power_law_1.2,0.07864320278167725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,2,power_law_1.2,0.115065598487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,2,power_law_1.2,0.13511040210723876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,2,power_law_1.2,0.1631168007850647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,2,power_law_1.2,0.20743041038513182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,2,power_law_1.2,0.21671679019927978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,2,power_law_1.2,0.2082240104675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,2,power_law_1.2,0.229369592666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,2,power_law_1.2,0.2367232084274292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,2,power_law_1.2,0.23891839981079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,2,power_law_1.2,0.24644479751586915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,2,power_law_1.2,0.2571007966995239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,2,power_law_1.2,0.27352960109710694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,2,power_law_1.2,0.27530879974365235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,2,power_law_1.2,0.3035520076751709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,2,power_law_1.2,0.33841919898986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,2,power_law_1.2,0.36214399337768555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,2,power_law_1.2,0.4207808017730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,2,power_law_1.2,0.4764927864074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,2,power_law_1.2,0.6009088039398194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,2,power_law_1.2,0.7187583923339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,2,power_law_1.2,0.9153792381286621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,2,power_law_1.2,1.1838975906372071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,2,power_law_1.2,1.6030527114868165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,2,power_law_1.2,1.9608192443847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,32,power_law_1.2,0.0964735984802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,32,power_law_1.2,0.09765120148658753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,32,power_law_1.2,0.09812480211257935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,32,power_law_1.2,0.10288000106811523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,32,power_law_1.2,0.0925055980682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,32,power_law_1.2,0.0944703996181488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,32,power_law_1.2,0.09660800099372864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,32,power_law_1.2,0.10145280361175538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,32,power_law_1.2,0.09961599707603455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,32,power_law_1.2,0.09896320104598999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,32,power_law_1.2,0.10396159887313842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,32,power_law_1.2,0.10884480476379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,32,power_law_1.2,0.10894080400466918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,32,power_law_1.2,0.11249920129776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,32,power_law_1.2,0.12104959487915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,32,power_law_1.2,0.12246400117874146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,32,power_law_1.2,0.1338688015937805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,32,power_law_1.2,0.16582399606704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,32,power_law_1.2,0.17459839582443237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,32,power_law_1.2,0.2266239881515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,32,power_law_1.2,0.2388159990310669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,32,power_law_1.2,0.30139520168304446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,32,power_law_1.2,0.2992703914642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,32,power_law_1.2,0.42410879135131835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,32,power_law_1.2,0.505344009399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,32,power_law_1.2,0.7675776004791259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,2,power_law_1.2,0.09035519957542419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,2,power_law_1.2,0.10520960092544555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,2,balanced,0.10427199800809224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,2,balanced,0.1415893336137136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,2,power_law_1.2,0.11520639657974244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,2,balanced,0.21395732959111533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,2,balanced,0.3351093530654907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,2,power_law_1.2,0.13256319761276245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,2,balanced,0.5889386733373007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,2,power_law_1.2,0.1537727952003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,2,power_law_1.2,0.1757248044013977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,2,power_law_1.2,0.21157760620117189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,32,power_law_1.2,0.839027214050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,2,power_law_1.2,0.2082240104675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,2,balanced,0.9863999684651693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,2,balanced,0.9726453622182211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,2,balanced,0.9466719627380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,2,balanced,0.9361386299133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,2,balanced,0.9220746358235677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,2,balanced,0.9106346766153971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,2,balanced,0.8997066815694174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,2,balanced,0.9379680156707764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,2,balanced,0.9262399673461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,2,balanced,0.9409759839375814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,2,balanced,0.9904692967732748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,2,balanced,0.9849706490834554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,2,balanced,1.0370720227559407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,2,power_law_1.2,0.21925759315490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,2,power_law_1.2,0.21903998851776124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,2,power_law_1.2,0.23233280181884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,2,power_law_1.2,0.23839359283447265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,2,power_law_1.2,0.2320512056350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,2,balanced,1.0122933387756348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,2,power_law_1.2,0.24179840087890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,2,balanced,1.001578648885091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,2,balanced,1.0179200172424316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,2,balanced,1.0715733369191487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,2,balanced,1.1636799971262615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,2,balanced,1.9327306747436523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,2,balanced,1.9939039548238118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,2,balanced,2.741994539896647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,2,balanced,3.5563252766927085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,16,power_law_1.01,0.04897280037403107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,16,power_law_1.01,0.06170240044593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,16,power_law_1.01,0.049344000220298764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,16,power_law_1.01,0.049267199635505673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,16,power_law_1.01,0.049235200881958006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,16,power_law_1.01,0.05063040256500244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,16,power_law_1.01,0.051020801067352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,16,power_law_1.01,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,16,power_law_1.01,0.05343359708786011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,16,power_law_1.01,0.05371519923210144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,16,power_law_1.01,0.0546239972114563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,16,power_law_1.01,0.056550401449203494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,16,power_law_1.01,0.05854719877243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,16,power_law_1.01,0.06033279895782471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,16,power_law_1.01,0.06770560145378113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,16,power_law_1.01,0.07125120162963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,16,power_law_1.01,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,16,power_law_1.01,0.08142079710960388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,16,power_law_1.01,0.09575039744377137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,16,power_law_1.01,0.12551679611206054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,16,power_law_1.01,0.1491328001022339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,16,power_law_1.01,0.1910591959953308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,16,power_law_1.01,0.23221120834350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,16,power_law_1.01,0.32568960189819335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,16,power_law_1.01,0.3999871969223022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,16,power_law_1.01,0.5821504116058349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,16,power_law_1.01,0.785529613494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,4,power_law_1.01,0.04842239916324616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,4,power_law_1.01,0.051283198595047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,4,power_law_1.01,0.05486080050468445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,4,power_law_1.01,0.06396160125732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,4,power_law_1.01,0.07052159905433655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,4,power_law_1.01,0.08332800269126892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,4,power_law_1.01,0.08710399866104127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,4,power_law_1.01,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,4,power_law_1.01,0.08573439717292786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,4,power_law_1.01,0.08798080086708068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,4,power_law_1.01,0.08837760090827942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,4,power_law_1.01,0.09288319945335388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,4,power_law_1.01,0.09011200070381165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,4,power_law_1.01,0.09912959933280945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,4,power_law_1.01,0.11015679836273193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,4,power_law_1.01,0.11457279920578003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,4,power_law_1.01,0.1206976056098938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,4,power_law_1.01,0.13566720485687256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,4,power_law_1.01,0.15687040090560914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,4,power_law_1.01,0.18480000495910645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,4,power_law_1.01,0.22096641063690187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,4,power_law_1.01,0.27292160987854003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,4,power_law_1.01,0.3491904020309448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,4,power_law_1.01,0.4734015941619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,4,power_law_1.01,0.5836991786956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,4,power_law_1.01,0.756063985824585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,4,power_law_1.01,1.0835200309753419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,8,power_law_1.2,0.032902398705482484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,8,power_law_1.2,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,8,power_law_1.2,0.032678401470184325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,8,power_law_1.2,0.03500800132751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,8,power_law_1.2,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,8,power_law_1.2,0.036108800768852235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,8,power_law_1.2,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,8,power_law_1.2,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,8,power_law_1.2,0.03665919899940491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,8,power_law_1.2,0.03784320056438446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,8,power_law_1.2,0.03827199935913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,4,balanced,0.037818667789300285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,4,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,4,balanced,0.04695466657479604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,4,balanced,0.0682666649421056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,4,balanced,0.0995360016822815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,4,balanced,0.09963200489679973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,4,balanced,0.09749866525332133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,4,balanced,0.10028800368309021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,4,balanced,0.1016319990158081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,4,balanced,0.10088533163070679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,4,balanced,0.10231467088063557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,4,balanced,0.10364799698193868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,4,balanced,0.10724266370137532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,4,balanced,0.1077280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,4,balanced,0.11344533165295918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,4,balanced,0.11550933122634888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,4,balanced,0.12108266353607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,4,balanced,0.130213330189387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,4,balanced,0.141866664091746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,4,balanced,0.1616320013999939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,4,balanced,0.17891200383504233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,4,balanced,0.23401067654291788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,4,balanced,0.2664053241411845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,4,balanced,0.36185598373413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,4,balanced,0.451909343401591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,4,balanced,0.6389866669972738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,4,balanced,0.820192019144694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,8,power_law_1.2,0.04174720048904419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,8,power_law_1.2,0.04264959990978241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,8,power_law_1.01,0.06899840235710145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,8,power_law_1.01,0.07891200184822082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,8,power_law_1.01,0.09736319780349731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,8,power_law_1.01,0.09918720126152039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,8,power_law_1.01,0.10063999891281128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,8,power_law_1.01,0.10969599485397338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,8,power_law_1.01,0.10636160373687745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,8,power_law_1.01,0.11230080127716065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,8,power_law_1.01,0.11006079912185669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,8,power_law_1.01,0.1081279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,8,power_law_1.01,0.1124735951423645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,8,power_law_1.01,0.11758719682693482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,8,power_law_1.01,0.11572480201721191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,8,power_law_1.01,0.11887999773025512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,8,power_law_1.01,0.1294592022895813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,8,power_law_1.01,0.12988799810409546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,8,power_law_1.01,0.13602559566497802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,8,power_law_1.01,0.14709759950637818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,8,power_law_1.01,0.16174720525741576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,8,power_law_1.01,0.1893056035041809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,8,power_law_1.01,0.22283520698547363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,8,power_law_1.01,0.29808640480041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,8,power_law_1.01,0.3677567958831787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,8,power_law_1.01,0.5222208023071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,8,power_law_1.01,0.644761610031128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,8,power_law_1.01,0.9387328147888183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,8,power_law_1.01,1.183859157562256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,8,power_law_1.2,0.04657920002937317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,8,power_law_1.01,0.059033602476119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,8,power_law_1.01,0.0580735981464386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,8,power_law_1.01,0.05877760052680969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,8,power_law_1.01,0.0625216007232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,8,power_law_1.01,0.06283519864082336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,8,power_law_1.01,0.06429439783096313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,8,power_law_1.01,0.06380159854888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,8,power_law_1.01,0.06524800062179566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,8,power_law_1.01,0.07067520022392274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,8,power_law_1.01,0.07195519804954528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,8,power_law_1.01,0.07148159742355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,8,power_law_1.01,0.08263040184974671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,8,power_law_1.01,0.09107840061187744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,8,power_law_1.01,0.09039360284805298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,8,power_law_1.01,0.11444480419158935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,8,power_law_1.01,0.10533759593963624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,8,power_law_1.01,0.12498559951782226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,8,power_law_1.01,0.17095040082931517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,8,power_law_1.01,0.20726399421691893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,8,power_law_1.01,0.2924992084503174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,8,power_law_1.01,0.3789695978164673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,8,power_law_1.01,0.5476480007171631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,8,power_law_1.01,0.7100224018096923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,8,power_law_1.2,0.053990399837493895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,8,power_law_1.01,1.0339520454406739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,8,power_law_1.2,0.05490559935569763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,8,power_law_1.2,0.061875200271606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,8,power_law_1.01,1.203731155395508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,8,power_law_1.01,1.918739128112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,2,power_law_1.01,0.02030719965696335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,2,power_law_1.01,0.02223999947309494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,2,power_law_1.01,0.026150399446487428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,2,power_law_1.01,0.03973119854927063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,2,power_law_1.01,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,2,power_law_1.01,0.05604479908943176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,2,power_law_1.01,0.06659200191497802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,2,power_law_1.01,0.06830719709396363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,2,power_law_1.01,0.07157760262489318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,2,power_law_1.01,0.07230079770088196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,2,power_law_1.01,0.07630720138549804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,2,power_law_1.01,0.07910400032997131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,2,power_law_1.01,0.08119680285453797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,2,power_law_1.01,0.08065279722213745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,2,power_law_1.01,0.08405759930610657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,2,power_law_1.01,0.08869119882583618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,2,power_law_1.01,0.09374079704284669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,2,power_law_1.01,0.11011840105056762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,2,power_law_1.01,0.12458239793777466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,2,power_law_1.01,0.15565439462661743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,2,power_law_1.01,0.16593919992446898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,2,power_law_1.01,0.21266560554504393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,2,power_law_1.01,0.24492158889770507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,2,power_law_1.01,0.3091327905654907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,2,power_law_1.01,0.37618560791015626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,2,power_law_1.01,0.500710391998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,2,power_law_1.01,0.6420735836029052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,8,power_law_1.2,0.07348480224609374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,8,power_law_1.01,2.721887969970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,8,power_law_1.2,0.09137920141220093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,32,power_law_1.2,0.018080000579357148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,32,power_law_1.2,0.018195199966430663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,32,power_law_1.2,0.018028800189495087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,32,power_law_1.2,0.017791999876499175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,32,power_law_1.2,0.018969599902629853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,32,power_law_1.2,0.02195200026035309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,32,power_law_1.2,0.038624000549316403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,32,power_law_1.2,0.039059200882911684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,32,power_law_1.2,0.03173120021820068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,32,power_law_1.2,0.031891199946403506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,32,power_law_1.2,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,32,power_law_1.2,0.032543998956680295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,32,power_law_1.2,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,32,power_law_1.2,0.03226880133152008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,32,power_law_1.2,0.03349120020866394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,32,power_law_1.2,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,32,power_law_1.2,0.03691520094871521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,32,power_law_1.2,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,32,power_law_1.2,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,32,power_law_1.2,0.048230400681495665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,32,power_law_1.2,0.05550720095634461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,32,power_law_1.2,0.06814720034599304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,32,power_law_1.2,0.08556159734725952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,32,power_law_1.2,0.11749119758605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,32,power_law_1.2,0.14235520362854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,32,power_law_1.2,0.20113921165466309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,8,power_law_1.2,0.12352639436721802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,8,power_law_1.2,0.14513280391693115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,8,power_law_1.2,0.2098623991012573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,8,power_law_1.2,0.26326398849487304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,32,power_law_1.2,0.2594304084777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,8,power_law_1.2,0.4009664058685303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,8,power_law_1.2,0.5202047824859619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,8,power_law_1.2,0.8220800399780274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,8,power_law_1.01,0.06344320178031922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,8,power_law_1.01,0.06254720091819763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,8,power_law_1.01,0.0615231990814209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,8,power_law_1.01,0.06641280055046081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,8,power_law_1.01,0.06720640063285828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,8,power_law_1.01,0.06792960166931153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,8,power_law_1.01,0.06822400093078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,8,power_law_1.01,0.06901760101318359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,8,power_law_1.01,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,8,power_law_1.01,0.07451519966125489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,8,power_law_1.01,0.07555840015411378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,8,power_law_1.01,0.08722559809684753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,8,power_law_1.01,0.08771839737892151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,8,power_law_1.01,0.09874560236930847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,2,balanced,0.07445333401362102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,2,balanced,0.0849226713180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,2,balanced,0.11291199922561646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,2,balanced,0.16242133577664694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,2,balanced,0.24054400126139322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,2,balanced,0.3706666628519694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,2,balanced,0.360698660214742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,2,balanced,0.35929067929585773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,2,balanced,0.37730133533477783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,2,balanced,0.3536906639734904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,2,balanced,0.34940266609191895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,2,balanced,0.3455839951833089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,2,balanced,0.3572053511937459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,2,balanced,0.35842665036519367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,2,balanced,0.3652533292770386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,2,balanced,0.37081066767374676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,2,balanced,0.37379733721415204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,2,balanced,0.39708268642425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,2,balanced,0.4158773422241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,2,balanced,0.4445013205210368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,2,balanced,0.48073065280914307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,2,balanced,0.5604906479517618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,8,power_law_1.01,0.11823999881744385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,2,balanced,0.644922653834025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,2,balanced,0.9868693351745605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,8,power_law_1.01,0.1345919966697693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,2,balanced,1.1441386540730794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,8,power_law_1.01,0.1379520058631897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,2,balanced,1.6290399233500164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,2,balanced,2.129690647125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,8,power_law_1.01,0.1942720055580139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,8,power_law_1.01,0.23884799480438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,8,power_law_1.01,0.3371648073196411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,8,power_law_1.01,0.3900928020477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,8,power_law_1.01,0.5458943843841553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,8,power_law_1.2,1.120486354827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,8,power_law_1.01,0.7632639884948731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,8,power_law_1.01,1.0952704429626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,8,power_law_1.01,1.5463744163513184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,8,power_law_1.01,1.9834175109863281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,1,power_law_1.01,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,1,power_law_1.01,0.05102720260620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,1,power_law_1.01,0.07720959782600403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,1,power_law_1.01,0.09806079864501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,1,power_law_1.01,0.11982719898223877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,1,power_law_1.01,0.17749119997024537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,1,power_law_1.01,0.18187520503997803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,1,power_law_1.01,0.19156479835510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,1,power_law_1.01,0.19349119663238526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,1,power_law_1.01,0.1959231972694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,1,power_law_1.01,0.20042879581451417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,1,power_law_1.01,0.204038405418396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,1,power_law_1.01,0.20709118843078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,1,power_law_1.01,0.21303679943084716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,1,power_law_1.01,0.22659199237823485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,1,power_law_1.01,0.23274240493774415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,1,power_law_1.01,0.2345792055130005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,1,power_law_1.01,0.2600640058517456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,1,power_law_1.01,0.2805056095123291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,1,power_law_1.01,0.32860798835754396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,1,power_law_1.01,0.35570559501647947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,1,power_law_1.01,0.4435391902923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,1,power_law_1.01,0.5278783798217773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,1,power_law_1.01,0.7017792224884033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,1,power_law_1.01,0.8760512351989747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,1,power_law_1.01,1.2042367935180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,1,power_law_1.01,1.5272704124450684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,128,power_law_1.2,0.05745919942855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,128,power_law_1.2,0.05626879930496216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,128,power_law_1.2,0.05648639798164368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,128,power_law_1.2,0.05731840133666992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,128,power_law_1.2,0.05764480233192444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,128,power_law_1.2,0.058355200290679934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,128,power_law_1.2,0.058790397644042966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,128,power_law_1.2,0.06059520244598389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,128,power_law_1.2,0.06085119843482971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,128,power_law_1.2,0.06102399826049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,128,power_law_1.2,0.06202239990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,128,power_law_1.2,0.06123520135879516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,128,power_law_1.2,0.06284800171852112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,128,power_law_1.2,0.06401280164718628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,128,power_law_1.2,0.06886399984359741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,128,power_law_1.2,0.06926720142364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,128,power_law_1.2,0.07371519804000855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,128,power_law_1.2,0.08588160276412964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,128,power_law_1.2,0.09486719965934753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,128,power_law_1.2,0.12129919528961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,128,power_law_1.2,0.14031360149383545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,128,power_law_1.2,0.19292160272598266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,128,power_law_1.2,0.23169920444488526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,128,power_law_1.2,0.3527168035507202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,128,power_law_1.2,0.43601279258728026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,128,power_law_1.2,0.6483136177062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,128,power_law_1.2,0.8623104095458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,8,power_law_1.01,3.039891242980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,4,balanced,0.037647999823093414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,4,balanced,0.03881600002447764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,4,balanced,0.03979733337958654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,4,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,4,balanced,0.056287998954455055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,4,balanced,0.057904000083605446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,4,balanced,0.0561706672112147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,4,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,4,balanced,0.05825600028038025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,4,balanced,0.05788266658782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,4,balanced,0.05904000004132589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,4,balanced,0.06010666489601135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,4,balanced,0.062368000547091164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,4,balanced,0.06401599943637848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,4,balanced,0.06739733119805653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,4,balanced,0.07044266661008199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,4,balanced,0.07425066828727722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,4,balanced,0.08065066734949748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,4,balanced,0.08714666962623596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,4,balanced,0.10519466797510783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,4,balanced,0.1216373344262441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,4,balanced,0.1625599960486094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,4,balanced,0.1902880072593689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,4,balanced,0.2598399917284648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,4,balanced,0.3350133498509725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,4,balanced,0.4840213457743327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,4,balanced,0.6348693370819092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,1,power_law_1.01,0.05534719824790955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,1,power_law_1.01,0.07222399711608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,1,power_law_1.01,0.0791167974472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,1,power_law_1.01,0.11560319662094116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,1,power_law_1.01,0.1173248052597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,1,power_law_1.01,0.12239359617233277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,1,power_law_1.01,0.12509440183639525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,1,power_law_1.01,0.12283519506454468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,1,power_law_1.01,0.12717440128326415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,1,power_law_1.01,0.13230079412460327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,1,power_law_1.01,0.13164160251617432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,1,power_law_1.01,0.13899519443511962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,1,power_law_1.01,0.14232319593429565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,1,power_law_1.01,0.14613759517669678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,1,power_law_1.01,0.16061439514160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,1,power_law_1.01,0.16853760480880736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,1,power_law_1.01,0.17592320442199708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,1,power_law_1.01,0.20731520652770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,1,power_law_1.01,0.2234816074371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,1,power_law_1.01,0.29168000221252444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,1,power_law_1.01,0.33917438983917236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,32,4,balanced,0.017269333203633625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,32,4,balanced,0.017658667018016178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,32,4,balanced,0.01716800034046173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,32,4,balanced,0.01823466643691063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,32,4,balanced,0.018901333212852478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,1,power_law_1.01,0.4581120014190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,32,4,balanced,0.019253333409627277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,32,4,balanced,0.021013334393501282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,32,4,balanced,0.021242665747801464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,32,4,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,32,4,balanced,0.03925866633653641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,32,4,balanced,0.03937066594759623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,32,4,balanced,0.031093334158261616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,32,4,balanced,0.031397332747777305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,32,4,balanced,0.03170666595300039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,32,4,balanced,0.03204799940188726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,32,4,balanced,0.033029332756996155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,32,4,balanced,0.03398400048414866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,32,4,balanced,0.03977599988381068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,32,4,balanced,0.04228266576925913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,32,4,balanced,0.04948799808820089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,32,4,balanced,0.05221333106358846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,32,4,balanced,0.07163199782371521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,32,4,balanced,0.08172266681989034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,32,4,balanced,0.10989333192507426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,32,4,balanced,0.13634666800498962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,32,4,balanced,0.19150932629903158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,32,4,balanced,0.23854933182398477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,2,balanced,0.01937599976857503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,2,balanced,0.021018666525681812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,2,balanced,0.021231998999913532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,2,balanced,0.025445332129796345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,2,balanced,0.031632001201311745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,2,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,2,balanced,0.03391999999682108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,2,balanced,0.03531199942032496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,2,balanced,0.0422986646493276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,2,balanced,0.03955200066169103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,2,balanced,0.04230933388074239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,2,balanced,0.05009066561857859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,2,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,2,balanced,0.04969066878159841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,2,balanced,0.06709866722424825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,2,balanced,0.06795200208822887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,2,balanced,0.06923200190067291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,2,balanced,0.09553600351015727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,2,balanced,0.10282666484514873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,2,balanced,0.12616533041000366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,2,balanced,0.13763200243314108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,2,balanced,0.14013333121935526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,2,balanced,0.15074666341145834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,2,balanced,0.24219733476638794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,2,balanced,0.26155734062194824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,2,balanced,0.37747732798258465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,2,balanced,0.49021867911020917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,1,power_law_1.01,0.5577856063842773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,1,power_law_1.01,0.7849664211273193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,1,power_law_1.01,0.9923839569091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,1,power_law_1.01,1.4095104217529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,1,power_law_1.01,1.82739200592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,64,power_law_1.01,0.07778559923171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,64,power_law_1.01,0.07670400142669678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,64,power_law_1.01,0.07767680287361145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,64,power_law_1.01,0.07827839851379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,64,power_law_1.01,0.07844480276107788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,64,power_law_1.01,0.08033279776573181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,64,power_law_1.01,0.0800000011920929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,64,power_law_1.01,0.08044800162315369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,64,power_law_1.01,0.08147839903831482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,64,power_law_1.01,0.08030719757080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,64,power_law_1.01,0.08153600096702576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,64,power_law_1.01,0.08602240085601806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,64,power_law_1.01,0.08914560079574585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,64,power_law_1.01,0.08857600092887878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,64,power_law_1.01,0.0995519995689392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,64,power_law_1.01,0.10204800367355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,64,power_law_1.01,0.10456960201263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,64,power_law_1.01,0.12211840152740479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,64,power_law_1.01,0.1335935950279236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,64,power_law_1.01,0.164409601688385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,64,power_law_1.01,0.187827205657959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,64,power_law_1.01,0.24265599250793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,64,power_law_1.01,0.25571839809417723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,64,power_law_1.01,0.32163200378417967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,64,power_law_1.01,0.42042880058288573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,64,power_law_1.01,0.5462719917297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,64,power_law_1.01,0.6867072105407714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,64,power_law_1.2,0.08424959778785705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,64,power_law_1.2,0.05822719931602478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,64,power_law_1.2,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,64,power_law_1.2,0.05749760270118713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,64,power_law_1.2,0.05518720149993896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,64,power_law_1.2,0.05916159749031067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,64,power_law_1.2,0.06039040088653565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,64,power_law_1.2,0.06050559878349304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,64,power_law_1.2,0.06268159747123718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,64,power_law_1.2,0.06428800225257873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,64,power_law_1.2,0.06466559767723083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,64,power_law_1.2,0.06350719928741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,64,power_law_1.2,0.06781439781188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,64,power_law_1.2,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,64,power_law_1.2,0.08035200238227844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,64,power_law_1.2,0.08421120047569275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,64,power_law_1.2,0.08840960264205933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,64,power_law_1.2,0.10379519462585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.2,0.1156607985496521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.2,0.1349503993988037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.2,0.15474560260772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.2,0.19900799989700318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.2,0.2475264072418213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.2,0.33676159381866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.2,0.4713600158691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.2,0.6809279918670654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.2,0.7446015834808349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,8,power_law_1.01,0.030803200602531434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,8,power_law_1.01,0.030022400617599487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,8,power_law_1.01,0.029471999406814574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,8,power_law_1.01,0.029228800535202028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,8,power_law_1.01,0.02999039888381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,8,power_law_1.01,0.030380800366401672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,8,power_law_1.01,0.030771198868751525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,8,power_law_1.01,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,8,power_law_1.01,0.032767999172210696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,8,power_law_1.01,0.0348800003528595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,8,power_law_1.01,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,8,power_law_1.01,0.03446399867534637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,8,power_law_1.01,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,8,power_law_1.01,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,8,power_law_1.01,0.04318079948425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,8,power_law_1.01,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,8,power_law_1.01,0.049670401215553286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,8,balanced,0.051034669081370033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,8,balanced,0.07049599786599477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,8,balanced,0.09122133255004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,8,power_law_1.01,0.05443199872970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,8,balanced,0.14065066973368326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,8,balanced,0.20624534289042154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,8,balanced,0.25724265972773236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,8,balanced,0.2558453281720479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,8,balanced,0.2595840096473694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,8,balanced,0.26264532407124835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,8,balanced,0.2619679967562358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,8,balanced,0.2615999976793925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,8,power_law_1.01,0.060115200281143186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,8,balanced,0.2667413353919983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,8,balanced,0.2640106678009033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,8,balanced,0.2683680057525635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,8,balanced,0.27662932872772217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,8,balanced,0.2779039939244588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,8,balanced,0.2842506567637126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,8,balanced,0.32441065708796185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,8,balanced,0.31335999568303424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,8,balanced,0.4020906686782837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,8,balanced,0.37755731741587323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,8,balanced,0.5768320163091024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,8,balanced,0.48261332511901855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,8,balanced,0.71615997950236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,8,power_law_1.01,0.07494400143623352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,8,balanced,0.7142186959584554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,8,balanced,1.034490664800008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,8,balanced,1.1355199813842773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,8,power_law_1.01,0.08813440203666686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,8,power_law_1.01,0.11278079748153687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,8,power_law_1.01,0.14933120012283324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,8,power_law_1.01,0.21262719631195068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,2,power_law_1.2,0.076665598154068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,2,power_law_1.2,0.08945279717445373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,2,power_law_1.2,0.11675519943237304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,2,power_law_1.2,0.15072640180587768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,8,power_law_1.01,0.26005120277404786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,2,power_law_1.2,0.17939200401306152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,2,power_law_1.2,0.26000640392303465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,2,power_law_1.2,0.2667072057723999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,2,power_law_1.2,0.27324159145355226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,2,power_law_1.2,0.29958400726318357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,2,power_law_1.2,0.28936960697174074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,2,power_law_1.2,0.31102719306945803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,2,power_law_1.2,0.3119935989379883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,2,power_law_1.2,0.3299839973449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,2,power_law_1.2,0.33559679985046387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,2,power_law_1.2,0.3508863925933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,2,power_law_1.2,0.3606528043746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,2,power_law_1.2,0.3858432054519653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,2,power_law_1.2,0.4522751808166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,2,power_law_1.2,0.45972480773925783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,2,power_law_1.2,0.5631167888641357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,2,power_law_1.2,0.5842944145202636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,2,power_law_1.2,0.7013504028320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,2,power_law_1.2,0.8959487915039063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,2,power_law_1.2,1.1540672302246093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,2,power_law_1.2,1.5736255645751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,8,power_law_1.01,0.35713920593261717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,2,power_law_1.2,2.041267204284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,2,power_law_1.2,2.776492881774902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,128,balanced,0.04041066765785217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,128,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,128,balanced,0.04164266586303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,128,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,128,balanced,0.04354133208592733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,128,balanced,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,128,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,128,balanced,0.0476800004641215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,128,balanced,0.05175999800364176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,128,balanced,0.049728001157442726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,128,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,128,balanced,0.0561653325955073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,128,balanced,0.051728000243504844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,128,balanced,0.05630933245023092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,128,balanced,0.06196266909440359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,128,balanced,0.060271998246510826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,128,balanced,0.06412800153096516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,128,balanced,0.07227733234564464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,128,balanced,0.0742986649274826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,128,balanced,0.08488532900810242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,128,balanced,0.09690666198730469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,128,balanced,0.11859200398127238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,128,balanced,0.14012266198794046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,128,balanced,0.18450133005777994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,128,balanced,0.22588266928990683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,128,balanced,0.3166240056355794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,128,balanced,0.3977546691894531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,8,power_law_1.01,0.4926976203918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,32,balanced,0.023152001202106476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,32,balanced,0.02309333284695943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,32,balanced,0.023077333966890972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,32,balanced,0.023024000227451324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,32,balanced,0.02496533344189326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,32,balanced,0.029461334149042766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,32,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,32,balanced,0.0517546683549881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,32,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,32,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,32,balanced,0.04905599852403005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,32,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,32,balanced,0.04555733501911163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,32,balanced,0.04399466514587402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,32,balanced,0.053455998500188194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,32,balanced,0.05277333160241445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,32,balanced,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,32,balanced,0.05533333122730255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,32,balanced,0.07411733269691467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,32,balanced,0.08684800068537395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,32,balanced,0.10931199789047241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,32,balanced,0.14358933766682944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,32,balanced,0.17762132485707602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,32,balanced,0.24698134263356528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,32,balanced,0.32412266731262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,32,balanced,0.4621173143386841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,32,balanced,0.6134399970372518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,2,power_law_1.2,0.15553280115127563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,2,power_law_1.2,0.24776959419250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,2,power_law_1.2,0.3745023965835571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,2,power_law_1.2,0.5963263988494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,2,power_law_1.2,0.8072575569152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,2,power_law_1.2,1.1722304344177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,2,power_law_1.2,1.2685888290405274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,2,power_law_1.2,1.6872703552246093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,2,power_law_1.2,1.8016447067260741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,2,power_law_1.2,1.8053056716918945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,2,power_law_1.2,1.7495744705200196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,2,power_law_1.2,1.9316543579101562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,2,power_law_1.2,1.944607925415039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,2,power_law_1.2,1.9870719909667969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,2,power_law_1.2,2.053759956359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,2,power_law_1.2,2.0627199172973634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,2,power_law_1.2,2.1454015731811524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,2,power_law_1.2,2.372831916809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,2,power_law_1.2,2.493280029296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,2,power_law_1.2,2.8603967666625976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,2,power_law_1.2,2.8831615447998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,2,power_law_1.2,3.4634689331054687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,2,power_law_1.2,3.293369674682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,2,power_law_1.2,3.8656192779541017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,2,power_law_1.2,4.4819904327392575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,2,power_law_1.2,5.6844032287597654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,2,power_law_1.2,7.545702362060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,8,power_law_1.01,0.047065600752830505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,8,power_law_1.01,0.05267840027809143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,8,power_law_1.01,0.07446399927139283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,8,power_law_1.01,0.07560319900512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,8,power_law_1.01,0.08858879804611205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,8,power_law_1.01,0.09475839734077454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,8,power_law_1.01,0.09856640100479126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,8,power_law_1.01,0.09431679844856262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,8,power_law_1.01,0.10011520385742187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,8,power_law_1.01,0.09036160111427308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,8,power_law_1.01,0.09626880288124084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,8,power_law_1.01,0.09984639883041382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,8,power_law_1.01,0.10713599920272827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,8,power_law_1.01,0.10540159940719604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,8,power_law_1.01,0.11682560443878173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,8,power_law_1.01,0.11890560388565063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,8,power_law_1.01,0.12961920499801635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,8,power_law_1.01,0.15267839431762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,8,power_law_1.01,0.15998079776763915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,8,power_law_1.01,0.19806720018386842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,8,power_law_1.01,0.22106239795684815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,8,power_law_1.01,0.29713919162750246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,8,power_law_1.01,0.37726080417633057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,8,power_law_1.01,0.5076608180999755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,1,balanced,0.03577066709597906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,1,balanced,0.04863466819127401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,1,balanced,0.0684746652841568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,1,balanced,0.1046506663163503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,1,balanced,0.16910932461420694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,1,balanced,0.2153706749280294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,1,balanced,0.21805866559346518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,1,balanced,0.21915199359258017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,1,balanced,0.21947733561197916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,1,balanced,0.21946666638056436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,1,balanced,0.223088006178538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,1,balanced,0.22419200340906778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,1,balanced,0.2273226579030355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,1,balanced,0.22978132963180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,8,power_law_1.01,0.6478975772857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,8,power_law_1.01,0.9160767555236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,1,power_law_1.01,0.053830397129058835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,8,power_law_1.01,1.1671808242797852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,1,power_law_1.01,0.05660799741744995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,1,balanced,0.23400533199310303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,1,balanced,0.23637866973876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,1,balanced,0.2429813345273336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,1,balanced,0.28355733553568524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,1,balanced,0.27057067553202313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,1,balanced,0.3367786804835002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,1,balanced,0.31144533554712933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,1,balanced,0.4384266535441081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,1,balanced,0.4212426741917928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,1,balanced,0.6107626756032308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,1,balanced,0.669648011525472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,1,balanced,1.054533322652181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,1,balanced,1.1333279609680176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,1,power_law_1.01,0.06981760263442993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,1,power_law_1.01,0.08810240030288696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,1,power_law_1.01,0.1018496036529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,1,power_law_1.01,0.13834880590438842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,1,power_law_1.01,0.14112639427185059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,1,power_law_1.01,0.14646400213241578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,1,power_law_1.01,0.14999040365219116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,1,power_law_1.01,0.1538879990577698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,1,power_law_1.01,0.16149120330810546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,1,power_law_1.01,0.1659775972366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,1,power_law_1.01,0.16974719762802123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,1,power_law_1.01,0.1761791944503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,1,power_law_1.01,0.1882688045501709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,1,power_law_1.01,0.19790719747543334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,1,power_law_1.01,0.21255040168762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,1,power_law_1.01,0.2618943929672241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,1,power_law_1.01,0.2983167886734009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,1,power_law_1.01,0.3851583957672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,1,power_law_1.01,0.45941758155822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,1,power_law_1.01,0.6133952140808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,1,power_law_1.01,0.7624320030212403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,1,power_law_1.01,1.021286392211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,1,power_law_1.01,1.3680447578430175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,1,power_law_1.01,1.8855167388916017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,1,power_law_1.01,2.3519744873046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,2,power_law_1.2,0.04514560103416443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,2,power_law_1.2,0.06807039976119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,2,power_law_1.2,0.09645439982414246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,2,power_law_1.2,0.1372480034828186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,2,power_law_1.2,0.19512959718704223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,2,power_law_1.2,0.2229696035385132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,2,power_law_1.2,0.30053119659423827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,2,power_law_1.2,0.3162496089935303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,2,power_law_1.2,0.32546560764312743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,2,power_law_1.2,0.347705602645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,2,power_law_1.2,0.36234240531921386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,2,power_law_1.2,0.3502847909927368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,2,power_law_1.2,0.3438656091690063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,2,power_law_1.2,0.39535999298095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,2,power_law_1.2,0.39310081005096437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,2,power_law_1.2,0.3924607992172241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,2,power_law_1.2,0.4180607795715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,2,power_law_1.2,0.45229439735412597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,2,power_law_1.2,0.5037631988525391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,2,power_law_1.2,1.1032832145690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,2,power_law_1.2,1.2472831726074218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,2,power_law_1.2,0.8054656028747559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,2,power_law_1.2,0.8833024024963378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,2,power_law_1.2,1.0241472244262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,2,power_law_1.2,1.2323776245117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,2,power_law_1.2,1.5738816261291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,2,power_law_1.2,1.8954496383666992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,1,balanced,0.09706133604049683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,1,balanced,0.107232004404068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,1,balanced,0.14034666617711386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,1,balanced,0.20371200640996298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,1,balanced,0.29788267612457275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,1,balanced,0.4957493146260579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,1,balanced,0.48704532782236737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,1,balanced,0.48030932744344074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,1,balanced,0.48283199469248456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,1,balanced,0.47489599386850995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,1,balanced,0.4824533462524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,1,balanced,0.483354647954305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,1,balanced,0.4964853525161743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,1,balanced,0.5036480029424032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,1,balanced,0.506437341372172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,1,balanced,0.5185653368631998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,1,balanced,0.5295360088348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,1,balanced,0.5670133431752523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,1,balanced,0.5966720183690389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,1,balanced,0.661296010017395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,1,balanced,0.7205653190612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,1,balanced,0.8668213685353597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,1,balanced,1.0099573135375977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,1,balanced,1.5420427322387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,1,balanced,1.835861365000407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,1,balanced,2.6543307304382324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,1,balanced,3.453136126200358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,4,balanced,0.05055999755859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,4,balanced,0.05018133421738943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,4,balanced,0.05248000224431356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,4,balanced,0.06180266539255778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,4,balanced,0.0788266658782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,4,balanced,0.10949333508809407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,4,balanced,0.10963199536005656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,4,balanced,0.1132426659266154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,4,balanced,0.11136533816655476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,4,balanced,0.11168000102043152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,4,balanced,0.1118986705938975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,4,balanced,0.11399466792742412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,4,balanced,0.1184213360150655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,4,balanced,0.11854400237401326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,4,balanced,0.1258026659488678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,4,balanced,0.12814933061599731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,4,balanced,0.13194666306177774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,4,balanced,0.1442293326059977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,32,power_law_1.2,0.052908802032470705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,4,balanced,0.15214932958285013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,4,balanced,0.17362666130065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,32,power_law_1.2,0.06438400149345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,4,balanced,0.1905440092086792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,4,balanced,0.23469332853953043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,4,balanced,0.27269866069157916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,4,balanced,0.3768639961878459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,4,balanced,0.4625813166300456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,4,balanced,0.6554346481959025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,4,balanced,0.8542133172353109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,32,power_law_1.2,0.06442239880561829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,32,power_law_1.2,0.06892799735069274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,32,power_law_1.2,0.06643840074539184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,32,power_law_1.2,0.06722559928894042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,32,power_law_1.2,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,32,power_law_1.2,0.06720640063285828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,32,power_law_1.2,0.06650239825248719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,32,power_law_1.2,0.06318719983100891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,32,power_law_1.2,0.06814720034599304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,32,power_law_1.2,0.07083520293235779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,32,power_law_1.2,0.07520639896392822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,32,power_law_1.2,0.07878400087356567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,32,power_law_1.2,0.08683519959449768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,32,power_law_1.2,0.09144319891929627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,32,power_law_1.2,0.0979200005531311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,32,power_law_1.2,0.11616640090942383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,32,power_law_1.2,0.12005120515823364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,32,power_law_1.2,0.15299199819564818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,32,power_law_1.2,0.18760960102081298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,32,power_law_1.2,0.2402559995651245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,32,power_law_1.2,0.28231039047241213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,32,power_law_1.2,0.395795202255249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,32,power_law_1.2,0.5116799831390381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,32,power_law_1.2,0.8267264366149902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,32,power_law_1.2,0.9706303596496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,16,power_law_1.01,0.02409600019454956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,16,power_law_1.01,0.02661119997501373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,16,power_law_1.01,0.025119999051094057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,16,power_law_1.01,0.02456959933042526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,16,power_law_1.01,0.028070399165153505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,16,power_law_1.01,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,16,power_law_1.01,0.04490880072116852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,16,power_law_1.01,0.04418559968471527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,16,power_law_1.01,0.04247680008411407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,2,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,2,balanced,0.04599999884764353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,2,balanced,0.059952000776926674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,2,balanced,0.08990933497746785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,2,balanced,0.14566933115323386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,2,balanced,0.1743519902229309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,2,balanced,0.17804799477259317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,2,balanced,0.1930826703707377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,2,balanced,0.2007946570714315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,2,balanced,0.20902933677037558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,2,balanced,0.1991999944051107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,2,balanced,0.2055786649386088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,2,balanced,0.21520533164342245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,2,balanced,0.2155359983444214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,2,balanced,0.22633065780003866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,2,balanced,0.2403200070063273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,2,balanced,0.29204267263412476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,2,balanced,0.2749333381652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,2,balanced,0.2722826600074768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,2,balanced,0.4647200107574463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,2,balanced,0.45734934012095135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,2,balanced,0.4861493508021037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,2,balanced,0.48846932252248126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,2,balanced,0.8218026955922445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,2,balanced,0.8822240034739176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,2,balanced,1.3029706478118896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,2,balanced,1.6853973070780437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,2,power_law_1.01,0.10211199522018433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,2,power_law_1.01,0.1371008038520813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,2,power_law_1.01,0.16292480230331421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,2,power_law_1.01,0.20633599758148194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,2,power_law_1.01,0.25816960334777833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,2,power_law_1.01,0.3393791913986206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,2,power_law_1.01,0.35054080486297606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,2,power_law_1.01,0.355347204208374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,2,power_law_1.01,0.35744640827178953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,2,power_law_1.01,0.3694911956787109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,2,power_law_1.01,0.3745471954345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,2,power_law_1.01,0.38226559162139895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,2,power_law_1.01,0.3696127891540527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,2,power_law_1.01,0.39130239486694335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,2,power_law_1.01,0.40117759704589845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,2,power_law_1.01,0.40340480804443357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,2,power_law_1.01,0.4092095851898193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,2,power_law_1.01,0.4473599910736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.01,0.49068799018859866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.01,0.5607039928436279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.01,0.6569791793823242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.01,0.8271231651306152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.01,0.9756671905517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.01,1.2920576095581056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.01,1.553593635559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.01,2.134457588195801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.01,2.8076416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,16,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,16,balanced,0.05182399849096934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,16,balanced,0.05279466509819031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,16,balanced,0.06326933205127716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,16,balanced,0.0803306649128596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,16,balanced,0.08490133285522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,16,balanced,0.0865226686000824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,16,balanced,0.0863146682580312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,16,balanced,0.08526399731636047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,16,balanced,0.08752533793449402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,16,balanced,0.08665600419044495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,16,balanced,0.09099200367927551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,16,balanced,0.08964266379674275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,16,balanced,0.0918880005677541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,16,balanced,0.09759466846783955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,16,balanced,0.09954133629798889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,16,balanced,0.1036959985891978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,16,balanced,0.12139733632405598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,16,balanced,0.12350933750470479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,16,balanced,0.15642133355140686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,16,balanced,0.1767680048942566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,16,balanced,0.23747734228769937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,16,balanced,0.21919999519983926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,16,balanced,0.2975200017293294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,16,balanced,0.3460853497187297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,16,balanced,0.4779520034790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,16,balanced,0.5748960177103678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,1,power_law_1.2,0.01935359984636307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,1,power_law_1.2,0.01895039975643158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,1,power_law_1.2,0.01940480023622513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,1,power_law_1.2,0.019513599574565887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,1,power_law_1.2,0.019417600333690645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,1,power_law_1.2,0.020710399746894835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,1,power_law_1.2,0.020582400262355804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,2,balanced,0.04388799766699473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,2,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,1,power_law_1.2,0.021452799439430237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,2,balanced,0.0487413356701533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,2,balanced,0.05137600004673004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,1,power_law_1.2,0.02162559926509857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,1,power_law_1.2,0.02194560021162033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,1,power_law_1.2,0.023129600286483764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,2,balanced,0.06022400160630544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,1,power_law_1.2,0.02476159930229187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,2,balanced,0.08340799808502197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,2,balanced,0.08451199531555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,2,balanced,0.08474133412043254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,2,balanced,0.08653333783149719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,2,balanced,0.08681600292523702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,1,power_law_1.2,0.025241601467132568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,1,power_law_1.2,0.027539199590682982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,1,power_law_1.2,0.050496000051498416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,2,balanced,0.086517333984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,1,power_law_1.2,0.05190399885177612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,2,balanced,0.08970133463541667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,2,balanced,0.09227200349171956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,2,balanced,0.09299199779828389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,2,balanced,0.09691199660301208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,2,balanced,0.10114133358001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,2,balanced,0.1039626697699229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,1,power_law_1.2,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,2,balanced,0.11728533109029134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,2,balanced,0.13182933131853738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,2,balanced,0.1519306699434916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,2,balanced,0.17457065979639688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,2,balanced,0.2355253299077352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,2,balanced,0.28144532442092896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,2,balanced,0.40937066078186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,2,balanced,0.49023465315500897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,2,balanced,0.707535982131958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,2,balanced,0.928762674331665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,1,power_law_1.2,0.04424319863319397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.2,0.046009600162506104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.2,0.05386880040168762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,1,balanced,0.09717333316802979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,1,balanced,0.12142399946848552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,1,balanced,0.16685332854588827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,1,balanced,0.2510133385658264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,2,balanced,0.05186666548252106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,1,balanced,0.3940906524658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,2,balanced,0.07216000060240428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,1,balanced,0.46730132897694904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,2,balanced,0.10002666711807251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,1,balanced,0.44276801745096844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,2,balanced,0.15876799821853638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,1,balanced,0.44145600001017254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,2,balanced,0.26097599665323895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,1,balanced,0.4490933418273926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,2,balanced,0.3171466588973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,1,balanced,0.44544001420338947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,2,balanced,0.31964266300201416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,1,balanced,0.43818668524424237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,2,balanced,0.3216480016708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,1,balanced,0.4455360174179077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,1,balanced,0.44757334391276044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,1,balanced,0.47554131348927814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.2,0.062009602785110474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,1,balanced,0.46326398849487305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,1,balanced,0.4687946637471517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,1,balanced,0.498741348584493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,1,balanced,0.5219839811325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,1,balanced,0.5485813220342001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,1,balanced,0.6081279913584391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,2,balanced,0.3229066729545593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,1,balanced,0.6761546929677328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,2,balanced,0.3264426589012146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,2,balanced,0.3279786705970764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,1,balanced,1.0965813000996907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,2,balanced,0.33188267548878986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,2,balanced,0.3370933135350545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,1,balanced,1.1990453402201335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,2,balanced,0.34188799063364667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,2,balanced,0.3518720070521037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,2,balanced,0.35624531904856366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,1,balanced,1.6936532656351726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,2,balanced,0.3648800055185954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,2,balanced,0.4172319968541463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,2,balanced,0.4082026481628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,1,balanced,2.2462026278177896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,2,balanced,0.5063466628392538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,2,balanced,0.512389341990153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,2,balanced,0.6588853200276693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,1,balanced,2.9488159815470376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,2,balanced,0.7218720118204752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,2,balanced,1.055461327234904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,2,balanced,1.2223412990570068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,1,balanced,4.046970685323079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,2,balanced,1.9740053812662761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,2,balanced,2.287973403930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.2,0.08643839955329895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.2,0.09969919919967651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,64,power_law_1.01,0.058534401655197146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.2,0.13544319868087767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,64,power_law_1.01,0.05866879820823669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,64,power_law_1.01,0.05825279951095581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,64,power_law_1.01,0.059487998485565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,64,power_law_1.01,0.058841598033905027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,64,power_law_1.01,0.05960959792137146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,64,power_law_1.01,0.060249602794647215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,64,power_law_1.01,0.06146559715270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,64,power_law_1.01,0.061440002918243405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,64,power_law_1.01,0.06197119951248169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,64,power_law_1.01,0.06213759779930115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,64,power_law_1.01,0.06530560255050659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,64,power_law_1.01,0.06602879762649536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,64,power_law_1.01,0.06714879870414733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,64,power_law_1.01,0.07315199971199035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,64,power_law_1.01,0.07698559761047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,64,power_law_1.01,0.08335999846458435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,64,power_law_1.01,0.09905920028686524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,64,power_law_1.01,0.1109760046005249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,64,power_law_1.01,0.1402176022529602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,64,power_law_1.01,0.16755839586257934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,64,power_law_1.01,0.21310720443725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,64,power_law_1.01,0.22867839336395263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,64,power_law_1.01,0.2908927917480469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,64,power_law_1.01,0.342739200592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,64,power_law_1.01,0.46943359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,64,power_law_1.01,0.6016511917114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.2,0.1715648055076599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.2,0.24506878852844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,2,power_law_1.01,0.08796799778938294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,2,power_law_1.01,0.12538880109786987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,2,power_law_1.01,0.14481279850006104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,2,power_law_1.01,0.20147840976715087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,2,power_law_1.01,0.25553920269012453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,2,power_law_1.01,0.370035195350647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,2,power_law_1.01,0.4021440029144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,2,power_law_1.01,0.40316162109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,2,power_law_1.01,0.4001920223236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,2,power_law_1.01,0.41672320365905763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,2,power_law_1.01,0.41775999069213865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,2,power_law_1.01,0.4280064105987549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,2,power_law_1.01,0.44260478019714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,2,power_law_1.01,0.4526656150817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,2,power_law_1.01,0.47619199752807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,2,power_law_1.01,0.49772157669067385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,2,power_law_1.01,0.5055744171142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,2,power_law_1.01,0.5609024047851563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,2,power_law_1.01,0.6135424137115478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,2,power_law_1.01,0.7479743957519531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,2,power_law_1.01,0.7699520111083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,2,power_law_1.01,0.9471808433532715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,2,power_law_1.01,1.126694393157959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,2,power_law_1.01,1.4432767868041991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,2,power_law_1.01,1.9192256927490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,2,power_law_1.01,2.5841344833374023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,2,power_law_1.01,3.201939010620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.2,0.3053056001663208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,32,balanced,0.042725334564844765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,32,balanced,0.04191466669241587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,32,balanced,0.04234133164087931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,32,balanced,0.046282668908437095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,32,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,32,balanced,0.054101333022117615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,32,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,32,balanced,0.05449600021044413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,32,balanced,0.05593599875768026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,32,balanced,0.056874667604764305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,32,balanced,0.056176001826922096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,32,balanced,0.05840533475081126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,32,balanced,0.06158933540185293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,32,balanced,0.062224000692367554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,32,balanced,0.0705386648575465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,32,balanced,0.07357333103815715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,32,balanced,0.08042133351167043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,32,balanced,0.09152533610661824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,32,balanced,0.09890133142471313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,32,balanced,0.11903466780980428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,32,balanced,0.13566399614016214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,32,balanced,0.17006933689117432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,32,balanced,0.20553600788116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,32,balanced,0.2749759952227275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,32,balanced,0.3435200055440267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,32,balanced,0.48210132122039795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,32,balanced,0.6189226706822714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,4,balanced,0.04804799954096476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,4,balanced,0.04785066843032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,4,balanced,0.051872000098228455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,4,balanced,0.07443200051784515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,4,balanced,0.10310399532318115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,4,balanced,0.15523733695348105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,4,balanced,0.15596266587575278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,4,balanced,0.1588106652100881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,4,balanced,0.15917332967122397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,4,balanced,0.1593546668688456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,4,balanced,0.16239999731381735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,4,balanced,0.1630880037943522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,4,balanced,0.1644213298956553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,4,balanced,0.16875199476877847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,4,balanced,0.17695466677347818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,4,balanced,0.1800266702969869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,4,balanced,0.18336000045140585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,4,balanced,0.20026665925979614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,4,balanced,0.20975999037424722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,4,balanced,0.23706134160359701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,4,balanced,0.2640693386395772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,4,balanced,0.32409600416819256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,4,balanced,0.3723893165588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,4,balanced,0.5043359994888306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,4,balanced,0.5862026611963908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,4,balanced,0.8280266920725504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,2,balanced,0.04892266790072123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,2,balanced,0.04821866750717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,2,balanced,0.05197866757710775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,2,balanced,0.06241600215435028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,2,balanced,0.0890773336092631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,2,balanced,0.12428266803423564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,2,balanced,0.12520533800125122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,4,balanced,1.0500586827596028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,2,balanced,0.12574399511019388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,2,balanced,0.12754666805267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,2,balanced,0.12949867049853006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,2,balanced,0.1299626628557841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,2,balanced,0.13429866234461466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,2,balanced,0.1379200021425883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,2,balanced,0.14140266180038452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,2,balanced,0.15069866180419922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,2,balanced,0.15481066703796387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,2,balanced,0.16306133071581522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,2,balanced,0.18780267238616943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,2,balanced,0.2055893341700236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,2,balanced,0.2473120093345642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,2,balanced,0.2841973304748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,2,balanced,0.40678401788075763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,2,balanced,0.4715253512064616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,2,balanced,0.6874667008717855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,2,balanced,0.852842648824056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,2,balanced,1.2353333632151287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,2,balanced,1.6133972803751628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,4,power_law_1.01,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,4,power_law_1.01,0.033318400382995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,4,power_law_1.01,0.0336896002292633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,4,power_law_1.01,0.03640959858894348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,4,power_law_1.01,0.038848000764846805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,4,power_law_1.01,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,4,power_law_1.01,0.04364160001277924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,4,power_law_1.01,0.044761601090431216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,4,power_law_1.01,0.04467200040817261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,4,power_law_1.01,0.04528000056743622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,4,power_law_1.01,0.04583680033683777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,4,power_law_1.01,0.04720639884471893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,4,power_law_1.01,0.050918400287628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,4,power_law_1.01,0.055193597078323366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,4,power_law_1.01,0.06323840022087097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,4,power_law_1.01,0.06750720143318176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,4,power_law_1.01,0.06651520133018493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,4,power_law_1.01,0.07859200239181519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,4,power_law_1.01,0.09007359743118286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,4,power_law_1.01,0.10926719903945922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,4,power_law_1.01,0.1314560055732727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,4,power_law_1.01,0.174835205078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,4,power_law_1.01,0.21735680103302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,4,power_law_1.01,0.2922879934310913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,4,power_law_1.01,0.3478143930435181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,4,power_law_1.01,0.5007167816162109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,4,power_law_1.01,0.629088020324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,32,power_law_1.01,0.021823999285697938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,32,power_law_1.01,0.020787200331687926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,32,power_law_1.01,0.02114560008049011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,32,power_law_1.01,0.022009600698947907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,32,power_law_1.01,0.021107199788093566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,32,power_law_1.01,0.02072319984436035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,32,power_law_1.01,0.022047999501228332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,32,power_law_1.01,0.02353920042514801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,32,power_law_1.01,0.024396799504756927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,32,power_law_1.01,0.024672000110149382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,32,power_law_1.01,0.02929919958114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,32,power_law_1.01,0.02945919930934906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,32,power_law_1.01,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,32,power_law_1.01,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,32,power_law_1.01,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,32,power_law_1.01,0.04359680116176605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,32,power_law_1.01,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,32,power_law_1.01,0.051744002103805545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,32,power_law_1.01,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,32,power_law_1.01,0.07564160227775574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,32,power_law_1.01,0.0578495979309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,32,power_law_1.01,0.0718720018863678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,32,power_law_1.01,0.0859008014202118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,32,power_law_1.01,0.11356159448623657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,32,power_law_1.01,0.14204800128936768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,32,power_law_1.01,0.19621119499206544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,32,power_law_1.01,0.25722880363464357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,1,balanced,0.09248532851537068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,1,balanced,0.1362879971663157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,1,balanced,0.22266666094462076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,1,balanced,0.3926080067952474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,1,balanced,0.7333599726359049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,1,balanced,1.4070666631062825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,1,balanced,1.411354700724284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,1,balanced,1.4164427121480305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,1,balanced,1.426037311553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,1,balanced,1.4277440706888835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,1,balanced,1.428976058959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,1,balanced,1.439626693725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,1,balanced,1.4488372802734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,1,balanced,1.4577172597249348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,1,balanced,1.468997319539388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,1,balanced,1.4813920656840007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,1,balanced,1.5024159749348958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,1,balanced,1.5517226854960124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,1,balanced,1.6123305956522624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,1,balanced,1.7054133415222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,1,balanced,1.7950773239135742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,1,balanced,1.9836427370707195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,1,balanced,2.104128042856852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,1,balanced,2.740351994832357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,1,balanced,2.842320124308268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,1,balanced,4.502079963684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,1,balanced,4.808239936828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,8,power_law_1.2,0.018662400543689728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,8,power_law_1.2,0.0197952002286911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,8,power_law_1.2,0.020640000700950623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,8,power_law_1.2,0.022303999960422517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,8,power_law_1.2,0.02306559979915619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,8,power_law_1.2,0.025164800882339477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,8,power_law_1.2,0.02730880081653595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,8,power_law_1.2,0.026745599508285523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,8,power_law_1.2,0.030054399371147157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,8,power_law_1.2,0.030451199412345885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,8,power_law_1.2,0.03052160143852234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,8,power_law_1.2,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,8,power_law_1.2,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,8,power_law_1.2,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,8,power_law_1.2,0.057580798864364624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,8,power_law_1.2,0.05801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,8,power_law_1.2,0.05883520245552063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,8,power_law_1.2,0.0635968029499054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,8,power_law_1.2,0.061926400661468504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,8,power_law_1.2,0.06934400200843811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,8,power_law_1.2,0.07392640113830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,8,power_law_1.2,0.08588799834251404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,8,power_law_1.2,0.09717119932174682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,8,power_law_1.2,0.12749439477920532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,8,power_law_1.2,0.15433599948883056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,8,power_law_1.2,0.20806400775909423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,8,power_law_1.2,0.2711551904678345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,32,balanced,0.0317493329445521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,32,balanced,0.02917333443959554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,32,balanced,0.02934933453798294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,32,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,32,balanced,0.0315733328461647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,32,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,32,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,32,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,32,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,32,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,32,balanced,0.03373866776625315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,32,balanced,0.03551999976237615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,32,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,32,balanced,0.03550933301448822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,32,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,32,balanced,0.0378506655494372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,32,balanced,0.039488000174363456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,32,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,32,balanced,0.04158399999141693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,32,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,32,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,32,balanced,0.052149335543314614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,32,balanced,0.058042665322621666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,32,balanced,0.07256000240643819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,32,balanced,0.08644266923268636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,32,balanced,0.10905599594116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,32,balanced,0.13095466295878092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,32,power_law_1.01,0.07346559762954712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,32,power_law_1.01,0.08373119831085205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,32,power_law_1.01,0.0722495973110199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,32,power_law_1.01,0.07735679745674133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,32,power_law_1.01,0.07464320063591004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,32,power_law_1.01,0.06888960003852844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,32,power_law_1.01,0.0717631995677948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,32,power_law_1.01,0.07495679855346679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,32,power_law_1.01,0.07575039863586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,32,power_law_1.01,0.07659519910812378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,32,power_law_1.01,0.07774080038070678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,32,power_law_1.01,0.08293759822845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,32,power_law_1.01,0.08117120265960694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,32,power_law_1.01,0.08760319948196411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,32,power_law_1.01,0.09081599712371827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,32,power_law_1.01,0.09169279932975768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,32,power_law_1.01,0.09178879857063293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,32,power_law_1.01,0.10573439598083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,32,power_law_1.01,0.10687999725341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,32,power_law_1.01,0.12285439968109131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,32,power_law_1.01,0.1411072015762329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,32,power_law_1.01,0.17272959947586058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,32,power_law_1.01,0.18202240467071534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,32,power_law_1.01,0.23843200206756593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,32,power_law_1.01,0.3086335897445679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,32,power_law_1.01,0.4374591827392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,32,power_law_1.01,0.522163200378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,1,power_law_1.01,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,1,power_law_1.01,0.0824832022190094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,1,power_law_1.01,0.10982400178909302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,1,power_law_1.01,0.17820160388946532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,1,power_law_1.01,0.24451839923858643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,1,power_law_1.01,0.3084800004959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,1,power_law_1.01,0.41695361137390136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,1,power_law_1.01,0.4540224075317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,1,power_law_1.01,0.44796161651611327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,1,power_law_1.01,0.475270414352417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,1,power_law_1.01,0.4928256034851074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,1,power_law_1.01,0.49823999404907227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,1,power_law_1.01,0.5147136211395263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,1,power_law_1.01,0.5216703891754151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,1,power_law_1.01,0.5328256130218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,1,power_law_1.01,0.5471295833587646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,1,power_law_1.01,0.5844031810760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,1,power_law_1.01,0.6557375907897949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,1,power_law_1.01,0.6687871932983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,1,power_law_1.01,0.7832511901855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,1,power_law_1.01,0.8071807861328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,1,power_law_1.01,0.96495361328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,1,power_law_1.01,1.1147968292236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,1,power_law_1.01,1.4553919792175294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,1,power_law_1.01,1.7890752792358398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,1,power_law_1.01,2.4422143936157226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,1,power_law_1.01,3.1214015960693358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,8,power_law_1.2,0.03665919899940491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,8,power_law_1.2,0.03612799942493439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,8,power_law_1.2,0.03455359935760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,8,power_law_1.2,0.03637759983539581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,8,power_law_1.2,0.03555839955806732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,8,power_law_1.2,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,8,power_law_1.2,0.03976320028305054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,8,power_law_1.2,0.0385343998670578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,8,power_law_1.2,0.04007039964199066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,8,power_law_1.2,0.041068801283836366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,8,power_law_1.2,0.04207360148429871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,8,power_law_1.2,0.04312959909439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,8,power_law_1.2,0.04619520008563995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,8,power_law_1.2,0.05114240050315857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,8,power_law_1.2,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,8,power_law_1.2,0.0627839982509613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,8,power_law_1.2,0.0585536003112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,8,power_law_1.2,0.0691968023777008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,8,power_law_1.2,0.08348159790039063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,8,power_law_1.2,0.10271999835968018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,8,power_law_1.2,0.11156480312347412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,8,power_law_1.2,0.14583679437637329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,8,power_law_1.2,0.1892416000366211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,8,power_law_1.2,0.20253438949584962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,8,power_law_1.2,0.2803967952728271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,8,power_law_1.2,0.4156799793243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,8,power_law_1.2,0.5893887996673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,2,balanced,0.044218664367993675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,2,balanced,0.045824001232783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,2,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,2,balanced,0.06029866635799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,2,balanced,0.08316266536712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,2,balanced,0.1260640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,2,balanced,0.12851199507713318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,2,balanced,0.12589333454767862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,2,balanced,0.12634666760762533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,2,balanced,0.12753599882125854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,2,balanced,0.12812800208727518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,2,balanced,0.132314662138621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,2,balanced,0.1323306659857432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,2,balanced,0.13436800241470337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,2,balanced,0.1413226624329885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,2,balanced,0.14430399735768637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,2,balanced,0.15243732929229736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,2,balanced,0.16772266228993735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,2,balanced,0.17641599973042807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,1,power_law_1.2,0.10716160535812377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,2,balanced,0.2039253314336141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,2,balanced,0.22755199670791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,1,power_law_1.2,0.13883520364761354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,2,balanced,0.29475732644399005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,1,power_law_1.2,0.18911999464035034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,1,power_law_1.2,0.2583807945251465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,1,power_law_1.2,0.315993595123291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,2,balanced,0.32445865869522095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,1,power_law_1.2,0.38840959072113035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,2,balanced,0.46059199174245197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,2,balanced,0.5650933186213175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,2,balanced,0.7984639803568522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,2,balanced,1.0179200172424316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,1,power_law_1.2,0.5230656147003174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,1,power_law_1.2,0.5432511806488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,1,power_law_1.2,0.5618432044982911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,1,power_law_1.2,0.5720191955566406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,1,power_law_1.2,0.5885695934295654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,1,power_law_1.2,0.6082496166229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,1,power_law_1.2,0.6187456130981446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,1,power_law_1.2,0.6331007957458497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,1,power_law_1.2,0.6499328136444091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,1,power_law_1.2,0.6533120155334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,1,power_law_1.2,0.6919104099273682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,1,power_law_1.2,0.7548672199249268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,1,power_law_1.2,0.7977727890014649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,1,power_law_1.2,0.9163135528564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,1,power_law_1.2,1.0300415992736816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,1,power_law_1.2,1.2338687896728515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,1,power_law_1.2,1.441823959350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,1,power_law_1.2,1.8395456314086913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,1,power_law_1.2,2.248543930053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,1,power_law_1.2,3.071283149719238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,1,power_law_1.2,3.889324951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,4,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,4,balanced,0.05225066840648651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,4,balanced,0.07497066756089528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,4,balanced,0.12170666456222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,4,balanced,0.2050666610399882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,4,balanced,0.20678933461507162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,4,balanced,0.20547733704249063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,4,balanced,0.20588799317677817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,4,balanced,0.2081013321876526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,4,balanced,0.20909333229064941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,4,balanced,0.2125920057296753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,4,balanced,0.21332265933354697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,4,balanced,0.21434666713078818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,4,balanced,0.21587733427683511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,4,balanced,0.22172266244888306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,4,balanced,0.22188800573349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,4,balanced,0.22589333852132162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,4,balanced,0.23270932833353677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,4,balanced,0.25172799825668335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,4,balanced,0.26661866903305054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,4,balanced,0.2789386709531148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,4,balanced,0.34020264943440753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,4,balanced,0.355840007464091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,4,balanced,0.4748426675796509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,4,balanced,0.53056534131368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,4,balanced,0.7188373406728109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,4,balanced,0.9095679918924967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,1,power_law_1.01,0.0523967981338501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,1,power_law_1.01,0.05665919780731201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,1,power_law_1.01,0.06480640172958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,1,power_law_1.01,0.08977919816970825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,1,power_law_1.01,0.11100159883499146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,1,power_law_1.01,0.13022079467773437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,1,power_law_1.01,0.1657088041305542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,1,power_law_1.01,0.16899839639663697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,1,power_law_1.01,0.17569279670715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,1,power_law_1.01,0.17904000282287597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,1,power_law_1.01,0.18372479677200318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,1,power_law_1.01,0.1941696047782898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,1,power_law_1.01,0.1973952054977417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,1,power_law_1.01,0.20829439163208008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,1,power_law_1.01,0.21795198917388917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,1,power_law_1.01,0.2312000036239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,1,power_law_1.01,0.2427903890609741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,1,power_law_1.01,0.2814527988433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,1,power_law_1.01,0.31059839725494387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,1,power_law_1.01,0.3791232109069824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,1,power_law_1.01,0.4316351890563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,1,power_law_1.01,0.5624767780303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,1,power_law_1.01,0.6726912021636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,1,power_law_1.01,0.9218815803527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,1,power_law_1.01,1.1637248039245605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,1,power_law_1.01,1.6415552139282226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,1,power_law_1.01,2.107833671569824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,4,power_law_1.2,0.04904960095882416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,4,power_law_1.2,0.07520639896392822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,4,power_law_1.2,0.09344000220298768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,4,power_law_1.2,0.12104959487915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,4,power_law_1.2,0.14695680141448975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,4,power_law_1.2,0.1876863956451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,4,power_law_1.2,0.21262719631195068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,4,power_law_1.2,0.21232640743255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,4,power_law_1.2,0.20727040767669677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,4,power_law_1.2,0.2284480094909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,4,power_law_1.2,0.22249600887298585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,4,power_law_1.2,0.22328319549560546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,4,power_law_1.2,0.23975040912628173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,4,power_law_1.2,0.2469696044921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,4,power_law_1.2,0.2651776075363159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,4,power_law_1.2,0.2699712038040161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,4,power_law_1.2,0.2883968114852905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,4,power_law_1.2,0.3096447944641113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.2,0.3305984020233154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.2,0.39738879203796384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.2,0.4714752197265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.2,0.5571584224700927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.2,0.6193408012390137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.2,0.805561637878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,16,balanced,0.04397333165009817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.2,0.9442624092102051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,16,balanced,0.04587199787298838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,16,balanced,0.04390933116277059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,16,balanced,0.050293331344922386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,16,balanced,0.06834133466084798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.2,1.2411775588989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,16,balanced,0.09513066212336223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,16,balanced,0.09573333462079366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,16,balanced,0.09605333209037781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,16,balanced,0.09752532839775085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,16,balanced,0.0981333355108897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.2,1.565875244140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,16,balanced,0.09831466277440389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,16,balanced,0.10014933347702026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,16,balanced,0.10285866260528564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,16,balanced,0.10319999853769939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,16,balanced,0.11040000120798747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,16,balanced,0.11239999532699585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,16,balanced,0.11461333433787028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,16,balanced,0.12531200051307678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,16,balanced,0.13184533516565958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,16,balanced,0.15057599544525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,16,balanced,0.16936532656351724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,16,balanced,0.2103253404299418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,16,balanced,0.24311999479929605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,16,balanced,0.3265119989713033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,16,balanced,0.40212265650431317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,16,balanced,0.5537919998168945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,16,balanced,0.6998666922251383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,8,power_law_1.01,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,8,power_law_1.01,0.048256000876426695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,8,power_law_1.01,0.050310397148132326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,8,power_law_1.01,0.05201280117034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,8,power_law_1.01,0.05013120174407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,8,power_law_1.01,0.053651201725006106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,8,power_law_1.01,0.053011202812194826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,8,power_law_1.01,0.05434880256652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,8,power_law_1.01,0.052716797590255736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,8,power_law_1.01,0.05304319858551025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,8,power_law_1.01,0.05470079779624939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,8,power_law_1.01,0.05578879714012146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,8,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,8,power_law_1.01,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,8,power_law_1.01,0.060755199193954466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,8,power_law_1.01,0.06199679970741272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,8,power_law_1.01,0.06607999801635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,8,power_law_1.01,0.07272319793701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,8,power_law_1.01,0.07643520236015319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,8,power_law_1.01,0.09565439820289612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,8,power_law_1.01,0.1082111954689026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,8,power_law_1.01,0.12469760179519654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,8,power_law_1.01,0.15697280168533326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,8,power_law_1.01,0.2062079906463623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,8,power_law_1.01,0.24723200798034667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,8,power_law_1.01,0.36154239177703856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,8,power_law_1.01,0.4472383975982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,8,power_law_1.01,0.051635199785232545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,8,power_law_1.01,0.06792960166931153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,8,power_law_1.01,0.06111360192298889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,8,power_law_1.01,0.07336320281028748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,8,power_law_1.01,0.0810368001461029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,8,power_law_1.01,0.0812287986278534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,8,power_law_1.01,0.08402559757232667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,8,power_law_1.01,0.08304640054702758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,8,power_law_1.01,0.0846783995628357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,8,power_law_1.01,0.08737919926643371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,8,power_law_1.01,0.08955519795417785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,8,power_law_1.01,0.09175040125846863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,8,power_law_1.01,0.09459199905395507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,8,power_law_1.01,0.09796479940414429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,8,power_law_1.01,0.10531200170516967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,8,power_law_1.01,0.11000319719314575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,8,power_law_1.01,0.11284480094909669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,8,power_law_1.01,0.12718080282211303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,8,power_law_1.01,0.14403200149536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,8,power_law_1.01,0.17582720518112183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,8,power_law_1.01,0.20408320426940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,8,power_law_1.01,0.2692863941192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,8,power_law_1.01,0.3434751987457275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,64,power_law_1.01,0.07982079982757569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,8,power_law_1.01,0.4592127799987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,64,power_law_1.01,0.07797759771347046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,8,power_law_1.01,0.6176576137542724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,64,power_law_1.01,0.07654399871826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,8,power_law_1.01,0.7672768115997315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,64,power_law_1.01,0.08151040077209473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,64,power_law_1.01,0.0803712010383606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,64,power_law_1.01,0.08012160062789916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,64,power_law_1.01,0.0846015989780426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,64,power_law_1.01,0.08558080196380616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,64,power_law_1.01,0.08561919927597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,8,power_law_1.01,0.9859071731567383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,64,power_law_1.01,0.0846015989780426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,64,power_law_1.01,0.08810240030288696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,64,power_law_1.01,0.09388800263404846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,64,power_law_1.01,0.09344000220298768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,64,power_law_1.01,0.09363840222358703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,64,power_law_1.01,0.10519039630889893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,64,power_law_1.01,0.10636160373687745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,64,power_law_1.01,0.10894720554351807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,64,power_law_1.01,0.1220736026763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,64,power_law_1.01,0.13073279857635497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,64,power_law_1.01,0.1651584029197693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,64,power_law_1.01,0.1863808035850525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,2,power_law_1.2,0.04899199903011322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,2,power_law_1.2,0.05447040200233459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,64,power_law_1.01,0.23209600448608397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,2,power_law_1.2,0.06101760268211365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,2,power_law_1.2,0.08368639945983887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,64,power_law_1.01,0.287987208366394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,2,power_law_1.2,0.09412479996681214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,64,power_law_1.01,0.39159040451049804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,2,power_law_1.2,0.11173759698867798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,2,power_law_1.2,0.13601919412612914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,64,power_law_1.01,0.5008063793182373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,2,power_law_1.2,0.13946239948272704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,2,power_law_1.2,0.1447872042655945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,64,power_law_1.01,0.733459186553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,2,power_law_1.2,0.14049279689788818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,2,power_law_1.2,0.15406080484390258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,64,power_law_1.01,1.0145407676696778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,2,power_law_1.2,0.1580415964126587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,2,power_law_1.2,0.16926720142364501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,2,power_law_1.2,0.1694975972175598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,2,power_law_1.2,0.18283519744873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,2,power_law_1.2,0.18903679847717286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,2,power_law_1.2,0.19511040449142455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,2,power_law_1.2,0.22257280349731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,2,power_law_1.2,0.2514879941940308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,2,power_law_1.2,0.2993727922439575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,2,power_law_1.2,0.3263040065765381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,2,power_law_1.2,0.4297215938568115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,2,power_law_1.2,0.5292736053466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,2,power_law_1.2,0.7065023899078369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,2,power_law_1.2,0.8840576171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,2,power_law_1.2,1.2613120079040527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,2,power_law_1.2,1.5738431930541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,32,power_law_1.01,0.0889855980873108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,32,power_law_1.01,0.08952959775924682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,32,power_law_1.01,0.08465920090675354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,32,power_law_1.01,0.08823040127754211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,32,power_law_1.01,0.0858560025691986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,32,power_law_1.01,0.08597760200500489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,32,power_law_1.01,0.08257920145988465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,32,power_law_1.01,0.08863999843597412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,32,power_law_1.01,0.08829439878463745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,32,power_law_1.01,0.09200639724731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,32,power_law_1.01,0.09132800102233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,32,power_law_1.01,0.0906112015247345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,32,power_law_1.01,0.09104639887809754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,32,power_law_1.01,0.09841920137405395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,32,power_law_1.01,0.10086400508880615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,32,power_law_1.01,0.10317440032958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,32,power_law_1.01,0.10915199518203736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,32,power_law_1.01,0.12093440294265748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,32,power_law_1.01,0.12248320579528808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,32,power_law_1.01,0.1444159984588623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,32,power_law_1.01,0.1579327940940857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,32,power_law_1.01,0.19676159620285033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,32,power_law_1.01,0.22493441104888917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,32,power_law_1.01,0.3061631917953491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,32,power_law_1.01,0.37499520778656004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,32,power_law_1.01,0.4925695896148682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,32,power_law_1.01,0.614463996887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,8,power_law_1.01,0.03270399868488312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,8,power_law_1.01,0.03320319950580597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,8,power_law_1.01,0.03294720053672791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,8,power_law_1.01,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,8,power_law_1.01,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,8,power_law_1.01,0.0355648010969162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,8,power_law_1.01,0.03588480055332184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,8,power_law_1.01,0.0361407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,8,power_law_1.01,0.03612799942493439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,8,power_law_1.01,0.037049600481987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,8,power_law_1.01,0.03666560053825378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,8,power_law_1.01,0.03758080005645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,8,power_law_1.01,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,8,power_law_1.01,0.038975998759269714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,8,power_law_1.01,0.042656001448631284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,8,power_law_1.01,0.04356479942798615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,8,power_law_1.01,0.045388799905776975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,8,power_law_1.01,0.051596802473068235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,8,power_law_1.01,0.05508480072021484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,8,power_law_1.01,0.06559360027313232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,8,power_law_1.01,0.07406719923019409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,8,power_law_1.01,0.09425280094146729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,8,power_law_1.01,0.11029119491577148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,8,power_law_1.01,0.14616960287094116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,8,power_law_1.01,0.18262399435043336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,1,power_law_1.2,0.04991360008716583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,1,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,1,power_law_1.2,0.12323199510574341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,1,power_law_1.2,0.20364160537719728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,1,power_law_1.2,0.2958143949508667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,1,power_law_1.2,0.41591677665710447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,1,power_law_1.2,0.5787583827972412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,1,power_law_1.2,0.6067135810852051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,1,power_law_1.2,0.6328703880310058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,1,power_law_1.2,0.649567985534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,1,power_law_1.2,0.7036159992218017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,1,power_law_1.2,0.7124671936035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,1,power_law_1.2,0.740883207321167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,1,power_law_1.2,0.7599167823791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,1,power_law_1.2,0.7773824214935303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,1,power_law_1.2,0.7827648162841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,1,power_law_1.2,0.8652992248535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,1,power_law_1.2,0.9474240303039551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.2,0.9592703819274903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.2,1.155846405029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.2,1.3162303924560548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.2,1.6766912460327148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.2,1.5574336051940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.2,1.9569664001464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,64,power_law_1.01,0.016415999829769136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,64,power_law_1.01,0.01581439971923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.2,2.449990463256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,64,power_law_1.01,0.01669120043516159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,64,power_law_1.01,0.017107200622558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,64,power_law_1.01,0.019468800723552705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,64,power_law_1.01,0.03134720027446747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,64,power_law_1.01,0.02632960081100464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.2,3.2687614440917967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,64,power_law_1.01,0.026067200303077697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,64,power_law_1.01,0.026726400852203368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,64,power_law_1.01,0.026899200677871705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,64,power_law_1.01,0.026739200949668883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,64,power_law_1.01,0.02720000147819519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,64,power_law_1.01,0.02771199941635132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.2,4.160851287841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,64,power_law_1.01,0.027551999688148497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,64,power_law_1.01,0.028352001309394838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,64,power_law_1.01,0.028896000981330872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,64,power_law_1.01,0.030272001028060914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,64,power_law_1.01,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,64,power_law_1.01,0.03363839983940124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,64,power_law_1.01,0.038124799728393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,64,power_law_1.01,0.042028799653053284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,64,power_law_1.01,0.04835200011730194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,64,power_law_1.01,0.05631999969482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,64,power_law_1.01,0.07927039861679078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,64,power_law_1.01,0.09571200013160705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,64,power_law_1.01,0.12342400550842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,8,power_law_1.01,0.257804799079895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,64,power_law_1.01,0.15751680135726928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,8,power_law_1.01,0.3321280002593994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,4,balanced,0.08049599826335907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,4,balanced,0.08286400139331818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,4,balanced,0.0960693359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,4,balanced,0.1144480009873708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,4,balanced,0.1495626668135325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,4,balanced,0.2066239913304647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,4,balanced,0.2063573400179545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,4,balanced,0.20570667584737143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,4,balanced,0.20843199888865152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,4,balanced,0.20728532473246256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,4,balanced,0.20913066466649374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,4,balanced,0.2155359983444214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,4,balanced,0.2177493373552958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,4,balanced,0.22177066405614218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,4,balanced,0.22868265708287558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,4,balanced,0.22933334112167358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,4,balanced,0.23597866296768188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,4,balanced,0.25311466058095294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,4,balanced,0.2642293373743693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,4,balanced,0.3064853350321452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,4,balanced,0.33986131350199383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,64,balanced,0.03399466723203659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,4,balanced,0.4095466534296672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,64,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,64,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,64,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,64,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,64,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,64,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,64,balanced,0.039066667358080544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,64,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,64,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,64,balanced,0.03944533318281174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,64,balanced,0.03932266682386398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,64,balanced,0.0414986660083135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,64,balanced,0.040864000717798867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,64,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,64,balanced,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,64,balanced,0.044735997915267944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,64,balanced,0.04570133487383524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,64,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,64,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,64,balanced,0.050442665815353394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,4,balanced,0.4804533322652181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,4,balanced,0.6741226514180502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,4,balanced,0.8140213489532471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,4,balanced,1.159989356994629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,4,balanced,1.517397403717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,64,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,64,balanced,0.06005333364009857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,64,balanced,0.07669866581757863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,64,balanced,0.09145599603652954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,64,balanced,0.11489599943161011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,64,balanced,0.1388746698697408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,8,power_law_1.2,0.08305280208587647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,8,power_law_1.2,0.08899199962615967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,8,power_law_1.2,0.08727040290832519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,8,power_law_1.2,0.08492799997329711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,8,power_law_1.2,0.08591359853744507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,8,power_law_1.2,0.09144319891929627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,8,power_law_1.2,0.09595519900321961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,8,power_law_1.2,0.1245311975479126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,8,power_law_1.2,0.13371520042419432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,8,power_law_1.2,0.16822400093078613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,8,power_law_1.2,0.14453760385513306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,8,power_law_1.2,0.1832576036453247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,8,power_law_1.2,0.19263360500335694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,8,power_law_1.2,0.12840319871902467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,8,power_law_1.2,0.15218559503555298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,8,power_law_1.2,0.15363839864730836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,8,power_law_1.2,0.1958400011062622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,8,power_law_1.2,0.2500288009643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,8,power_law_1.2,0.30699520111083983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,8,power_law_1.2,0.4449471950531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,8,power_law_1.2,0.4964416027069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,8,power_law_1.2,0.7077631950378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,8,power_law_1.2,0.8520064353942871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,8,power_law_1.2,1.3460991859436036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,8,power_law_1.2,1.4442943572998046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,8,power_law_1.2,2.3221952438354494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,4,power_law_1.2,0.06257920265197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,4,power_law_1.2,0.08101119995117187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,4,power_law_1.2,0.08113920092582702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,4,power_law_1.2,0.10730880498886108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,4,power_law_1.2,0.12117760181427002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,4,power_law_1.2,0.1280832052230835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,4,power_law_1.2,0.13939199447631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,4,power_law_1.2,0.15301120281219482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,4,power_law_1.2,0.14746240377426148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,4,power_law_1.2,0.15864959955215455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,4,power_law_1.2,0.15907200574874877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,4,power_law_1.2,0.1692479968070984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,4,power_law_1.2,0.16935039758682252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,4,power_law_1.2,0.173363196849823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,4,power_law_1.2,0.18407679796218873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,16,balanced,0.052517334620157875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,4,power_law_1.2,0.19029760360717773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,16,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,16,balanced,0.057477335135142006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,4,power_law_1.2,0.1975167989730835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,4,power_law_1.2,0.22210559844970704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,4,power_law_1.2,0.25661439895629884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,4,power_law_1.2,0.3005824089050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,4,power_law_1.2,0.33708159923553466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,4,power_law_1.2,0.40631680488586425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,16,balanced,0.07212799787521362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,16,balanced,0.09196266531944275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,4,power_law_1.2,0.4952256202697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,4,power_law_1.2,0.6624063968658447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,4,power_law_1.2,0.8203904151916503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,4,power_law_1.2,1.1789055824279786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,16,balanced,0.09799999992052714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,16,balanced,0.09902933239936829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,16,balanced,0.09878399968147278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,4,power_law_1.2,1.5880831718444823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,16,balanced,0.0995253324508667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,16,balanced,0.09989866614341736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,16,balanced,0.10316266616185506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,16,balanced,0.10620799660682678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,16,balanced,0.10706667105356853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,16,balanced,0.10730666915575664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,16,balanced,0.11385066310564677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,16,balanced,0.11687466502189636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,16,balanced,0.12198932965596516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,16,balanced,0.13706666231155396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,16,balanced,0.14863999684651694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,16,balanced,0.199072003364563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,16,balanced,0.20159467061360678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,16,balanced,0.27430933713912964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,16,balanced,0.2800053358078003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,16,balanced,0.3936320145924886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,16,balanced,0.4356853167215983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,16,balanced,0.6130346854527792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,16,balanced,0.7584799925486246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,8,power_law_1.01,0.040012800693511964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,8,power_law_1.01,0.04246399998664856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,8,power_law_1.01,0.04380159974098206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,8,power_law_1.01,0.04647679924964905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,8,power_law_1.01,0.04844799935817719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,8,power_law_1.01,0.04947839975357056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,8,power_law_1.01,0.05055999755859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,8,power_law_1.01,0.051507198810577394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,8,power_law_1.01,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,8,power_law_1.01,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,8,power_law_1.01,0.0536575973033905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,8,power_law_1.01,0.05430399775505066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,8,power_law_1.01,0.056569600105285646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,8,power_law_1.01,0.05985919833183288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,8,power_law_1.01,0.0657151997089386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,8,power_law_1.01,0.06731520295143127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,8,power_law_1.01,0.07085440158843995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,8,power_law_1.01,0.08174719810485839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,8,power_law_1.2,3.3147518157958986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,8,power_law_1.01,0.08430079817771911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,8,power_law_1.01,0.10591360330581664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,8,power_law_1.01,0.12452479600906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,8,power_law_1.01,0.16528639793395997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,8,power_law_1.01,0.2120896100997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,8,power_law_1.01,0.2844480037689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,8,power_law_1.01,0.37322239875793456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,8,power_law_1.01,0.5107647895812988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,8,power_law_1.01,0.6586495876312256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,8,balanced,0.04728533327579498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,8,balanced,0.04557866851488749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,8,balanced,0.04640000065167745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,8,balanced,0.048997332652409874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,8,balanced,0.048384000857671104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,8,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,8,balanced,0.047637333472569786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,8,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,8,balanced,0.049839998284975685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,8,balanced,0.0487360010544459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,8,balanced,0.047925333182017006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,8,balanced,0.0458186666170756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,8,balanced,0.047983999053637184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,2,balanced,0.021541332205136616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,2,balanced,0.02311466634273529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,2,balanced,0.028618666032950085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,2,balanced,0.03997333347797394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,2,balanced,0.062021334966023765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,2,balanced,0.09574400385220845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,2,balanced,0.0990666647752126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,2,balanced,0.10203199585278828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,2,balanced,0.10469333330790202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,2,balanced,0.10570666193962097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,2,balanced,0.11327466368675232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,2,balanced,0.10734933614730835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,2,balanced,0.11102933684984843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,2,balanced,0.11563199758529663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,8,balanced,0.047930667797724404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,2,balanced,0.13332800070444742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,8,balanced,0.048698668678601585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,2,balanced,0.13196266690889993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,8,balanced,0.050714666644732155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,2,balanced,0.1321386694908142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,2,balanced,0.18580265839894614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,2,balanced,0.17940799395243326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,2,balanced,0.26993600527445477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,2,balanced,0.2984586755434672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,2,balanced,0.31568000713984173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,2,balanced,0.32889066139856976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,2,balanced,0.3187733292579651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,2,balanced,0.3306720058123271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,2,balanced,0.5957173506418864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,8,balanced,0.05037866532802582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,8,balanced,0.05994666616121928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,2,balanced,0.5977280139923096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,8,balanced,0.05792533357938131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,8,balanced,0.07018133501211803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,8,balanced,0.07800533374150594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,8,balanced,0.10044800241788228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,8,balanced,0.11750933527946472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,8,balanced,0.152319997549057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,8,balanced,0.18625599145889282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,8,balanced,0.25622934103012085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,8,balanced,0.32496533791224164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,64,power_law_1.2,0.08677759766578674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,64,power_law_1.2,0.08265600204467774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,64,power_law_1.2,0.08241919875144958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,64,power_law_1.2,0.08444799780845642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,64,power_law_1.2,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,64,power_law_1.2,0.08061439990997314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,64,power_law_1.2,0.08115839958190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,64,power_law_1.2,0.08254079818725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,64,power_law_1.2,0.08468480110168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,64,power_law_1.2,0.08638079762458802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,64,power_law_1.2,0.08685439825057983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,64,power_law_1.2,0.08826239705085755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,64,power_law_1.2,0.0884223997592926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,64,power_law_1.2,0.09168000221252441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,64,power_law_1.2,0.09371520280838012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,64,power_law_1.2,0.09679999947547913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,64,power_law_1.2,0.10041600465774536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,64,power_law_1.2,0.11071360111236572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,64,power_law_1.2,0.11804800033569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,64,power_law_1.2,0.13948800563812255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,64,power_law_1.2,0.15932159423828124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,64,power_law_1.2,0.1927616000175476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,64,power_law_1.2,0.24551041126251222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,64,power_law_1.2,0.31962881088256834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,64,power_law_1.2,0.40517759323120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,64,power_law_1.2,0.5892864227294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,64,power_law_1.2,0.7460288047790528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,32,balanced,0.05397333204746246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,32,balanced,0.043578664461771645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,32,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,32,balanced,0.04774933556715647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,32,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,32,balanced,0.0528053343296051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,32,balanced,0.054586668809254967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,32,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,32,balanced,0.05388799806435903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,32,balanced,0.05593599875768026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,2,power_law_1.2,0.0915008008480072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,2,power_law_1.2,0.1114240050315857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,2,power_law_1.2,0.13790080547332764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,2,power_law_1.2,0.1916416049003601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,32,balanced,0.05598400036493937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,32,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,32,balanced,0.05852800110975901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,2,power_law_1.2,0.24408960342407227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,32,balanced,0.061343997716903687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,32,balanced,0.06614933411280315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,32,balanced,0.06653866668542226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,32,balanced,0.07091733316580455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,2,power_law_1.2,0.37606399059295653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,2,power_law_1.2,0.3898751974105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,2,power_law_1.2,0.40935678482055665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,2,power_law_1.2,0.4127808094024658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,2,power_law_1.2,0.41585278511047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,32,balanced,0.07867200175921123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,32,balanced,0.08264000217119853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,32,balanced,0.09481066465377808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,32,balanced,0.1072746713956197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,32,balanced,0.12717866897583008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,32,balanced,0.14697600404421488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,32,balanced,0.1856693426767985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,32,balanced,0.22483734289805093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,32,balanced,0.3155253330866496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,32,balanced,0.3824586470921834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,2,power_law_1.2,0.42371201515197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,2,power_law_1.2,0.4420159816741943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,2,power_law_1.2,0.45503997802734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,2,power_law_1.2,0.46355838775634767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,2,power_law_1.2,0.47392640113830564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,2,power_law_1.2,0.5027071952819824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,2,power_law_1.2,0.5417727947235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,2,power_law_1.2,0.6100671768188477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,2,power_law_1.2,0.630182409286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,2,power_law_1.2,0.7474688053131103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,2,power_law_1.2,0.7874688148498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,2,power_law_1.2,0.9842816352844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,2,power_law_1.2,1.1616127967834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,2,power_law_1.2,1.474623966217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,2,power_law_1.2,1.8899263381958007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,2,power_law_1.2,2.6757503509521485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,2,power_law_1.2,3.205120086669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,32,power_law_1.2,0.056959998607635495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,32,power_law_1.2,0.059308797121047974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,32,power_law_1.2,0.056383997201919556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,32,power_law_1.2,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,32,power_law_1.2,0.06264320015907288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,32,power_law_1.2,0.06405760049819946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,32,power_law_1.2,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,32,power_law_1.2,0.06345599889755249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,32,power_law_1.2,0.06501759886741638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,32,power_law_1.2,0.06584320068359376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,32,power_law_1.2,0.06584320068359376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,32,power_law_1.2,0.06835200190544129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,32,power_law_1.2,0.0701312005519867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,32,power_law_1.2,0.07180160284042358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,32,power_law_1.2,0.0787775993347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,32,power_law_1.2,0.08007680177688599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,32,power_law_1.2,0.0918079972267151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,32,power_law_1.2,0.11020159721374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,32,power_law_1.2,0.11398400068283081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,32,power_law_1.2,0.1407487988471985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,32,power_law_1.2,0.17274240255355836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,32,power_law_1.2,0.2189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,32,power_law_1.2,0.23569281101226808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,32,power_law_1.2,0.3251327991485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,32,power_law_1.2,0.37973120212554934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,32,balanced,0.04435733457406362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,32,balanced,0.043978666265805565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,32,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,32,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,32,balanced,0.0476693312327067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,32,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,32,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,32,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,32,balanced,0.05203199883302053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,32,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,32,balanced,0.05203199883302053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,32,power_law_1.2,0.5231808185577392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,32,balanced,0.05416533350944519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,32,balanced,0.05421333511670431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,32,balanced,0.05610666672388712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,32,balanced,0.0602453351020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,32,power_law_1.2,0.6648960113525391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,32,balanced,0.05995733539263407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,32,balanced,0.06429333488146464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,32,balanced,0.07040533423423767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,32,balanced,0.0749120016892751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,32,balanced,0.08566400408744812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,32,balanced,0.09725866715113322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,32,balanced,0.12145066261291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,32,balanced,0.14340800046920776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,32,balanced,0.19130132595698038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,32,balanced,0.2367039918899536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,32,balanced,0.3299520015716553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,32,balanced,0.4157973527908325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,8,power_law_1.01,0.06709120273590088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,8,power_law_1.01,0.055615997314453124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,8,power_law_1.01,0.05425919890403748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,8,power_law_1.01,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,8,power_law_1.01,0.05486720204353333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,8,power_law_1.01,0.055193597078323366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,8,power_law_1.01,0.05860480070114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,8,power_law_1.01,0.05983359813690185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,8,power_law_1.01,0.0633791983127594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,8,power_law_1.01,0.06475520133972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,8,power_law_1.01,0.06392319798469544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,8,power_law_1.01,0.07277439832687378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,8,power_law_1.01,0.07045120000839233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,8,power_law_1.01,0.0748416006565094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,8,power_law_1.01,0.08409600257873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,8,power_law_1.01,0.08844159841537476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,8,power_law_1.01,0.0995136022567749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,8,power_law_1.01,0.12113920450210572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,8,power_law_1.01,0.14161280393600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,8,power_law_1.01,0.19856640100479125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,8,power_law_1.01,0.23077120780944824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,8,power_law_1.01,0.28176639080047605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,8,power_law_1.01,0.39530880451202394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,8,balanced,0.029546665648619335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,8,balanced,0.02943466603755951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,8,balanced,0.029029332101345062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,8,balanced,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,8,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,8,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,8,balanced,0.03346133232116699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,8,balanced,0.03549333413441976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,8,balanced,0.03348266581694285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,8,balanced,0.03514133393764496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,8,balanced,0.033674667278925575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,8,balanced,0.03576533248027166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,8,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,8,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,128,balanced,0.028389332195123036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,128,balanced,0.026133333643277485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,128,balanced,0.026533332963784535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,128,balanced,0.02900800108909607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,128,balanced,0.029440000653266907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,8,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,8,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,8,balanced,0.044031997521718345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,8,balanced,0.05106666684150696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,8,balanced,0.05783466498057047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,8,balanced,0.06193066636721293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,128,balanced,0.029493334392706554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,8,balanced,0.07022933165232341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,8,balanced,0.08866666754086812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,128,balanced,0.029616000751654308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,8,balanced,0.1029866635799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,8,balanced,0.1439413328965505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,128,balanced,0.03161066770553589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,8,balanced,0.17493865887324014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,8,balanced,0.24431467056274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,8,balanced,0.3136319915453593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,8,power_law_1.01,0.5971968173980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,128,balanced,0.02940800040960312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,128,balanced,0.030533333619435627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,128,balanced,0.031514666974544525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,128,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,128,balanced,0.031471999982992806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,8,power_law_1.01,0.726304006576538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,128,balanced,0.03201599915822347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,128,balanced,0.03600533306598663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,128,balanced,0.03939733405907949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,128,balanced,0.04009599983692169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,128,balanced,0.040976000328858696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,128,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,128,balanced,0.05002133548259735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,128,balanced,0.052042668064435325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,128,balanced,0.06425066788991292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,8,power_law_1.01,1.1194368362426759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,128,balanced,0.07260266443093617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,128,balanced,0.09122133255004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,128,balanced,0.10818666219711304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,128,balanced,0.1464906632900238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,128,balanced,0.18441067139307657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,32,1,power_law_1.2,0.018751999735832213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,32,1,power_law_1.2,0.01754239946603775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,32,1,power_law_1.2,0.01785600036382675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,32,1,power_law_1.2,0.01733759939670563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,32,1,power_law_1.2,0.0176704004406929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,32,1,power_law_1.2,0.018540799617767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,32,1,power_law_1.2,0.018764799833297728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,32,1,power_law_1.2,0.01937279999256134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,32,1,power_law_1.2,0.019865599274635316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,32,1,power_law_1.2,0.019776000082492827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,32,1,power_law_1.2,0.02101760059595108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,32,1,power_law_1.2,0.022438399493694305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,32,1,power_law_1.2,0.02327679991722107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,32,1,power_law_1.2,0.02372480034828186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,32,1,power_law_1.2,0.041440001130104064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,32,1,power_law_1.2,0.04261119961738587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,32,1,power_law_1.2,0.04353919923305512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,4,balanced,0.03803733239571253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,4,balanced,0.03777066618204117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,4,balanced,0.04391466577847799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,4,balanced,0.05009066561857859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,4,balanced,0.06613333523273468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,4,balanced,0.06313066681226094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,4,balanced,0.06005866825580597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,4,balanced,0.06438399851322174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,4,balanced,0.06066666543483734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,4,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,4,balanced,0.062314664324124656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,4,balanced,0.06072533130645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,4,balanced,0.06619200110435486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,4,balanced,0.06613866488138835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,4,balanced,0.0682773341735204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,4,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,4,balanced,0.07052266597747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,4,balanced,0.045935998360315956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,4,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,4,balanced,0.05624000231424967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,4,balanced,0.06503466765085857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,4,balanced,0.09678933024406433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,4,balanced,0.11961600184440613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,4,balanced,0.11336533228556316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,4,balanced,0.1151093343893687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,4,balanced,0.11618666847546895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,4,balanced,0.1169653336207072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,4,balanced,0.1216213305791219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,4,balanced,0.12573333581288657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,4,balanced,0.1260426640510559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,4,balanced,0.1356160044670105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,4,balanced,0.07259733478228252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,4,balanced,0.07812266548474629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,4,balanced,0.08121599753697713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,4,balanced,0.09667199850082397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,4,balanced,0.10797866185506184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,4,balanced,0.14751467108726501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,4,balanced,0.1706506609916687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,4,balanced,0.23413866758346558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,4,balanced,0.29833600918451947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,4,balanced,0.4241493145624797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,4,balanced,0.1383840044339498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,4,balanced,0.548245350519816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,4,balanced,0.1499626636505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,32,1,power_law_1.2,0.037484800815582274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,4,balanced,0.16870933771133423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,4,balanced,0.18453866243362427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,4,balanced,0.2181653380393982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,4,balanced,0.24965333938598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,4,balanced,0.31965865691502887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,4,balanced,0.4011146624883016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,4,balanced,0.5395946502685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,4,balanced,0.7179520130157471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,4,balanced,1.0122506618499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.2,0.03930239975452423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,4,balanced,1.3209866682688396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.2,0.04679679870605469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.2,0.050400000810623166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.2,0.07021440267562866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,8,power_law_1.01,1.4980416297912598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.2,0.07926399707794189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.2,0.10725760459899902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,4,balanced,0.05825066566467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,4,balanced,0.06062933305899302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,4,balanced,0.06353066861629486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,4,balanced,0.07281066477298737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,4,balanced,0.0825386643409729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,4,balanced,0.10744532942771912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,4,balanced,0.12877866625785828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,4,balanced,0.13062933087348938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,4,balanced,0.1297706663608551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,4,balanced,0.13182933131853738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,4,balanced,0.12548800309499106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,4,balanced,0.13053866227467856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,4,balanced,0.1339413324991862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,4,balanced,0.13635200262069702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,4,balanced,0.14527466893196106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,4,balanced,0.14691733320554098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,4,balanced,0.15609600146611533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,4,balanced,0.1725333333015442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,4,balanced,0.1855306625366211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,4,balanced,0.21288534005482992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,4,balanced,0.24003199736277261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,4,balanced,0.3035999933878581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,4,balanced,0.36531198024749756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,4,balanced,0.49399467309316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,4,balanced,0.6652906735738119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,4,balanced,0.9276320139567057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,4,balanced,1.2254613240559895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,1,balanced,0.054042667150497437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,1,balanced,0.057909334699312844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,1,balanced,0.06673066814740498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,1,balanced,0.09474133451779683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,1,balanced,0.14147200187047324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,1,balanced,0.22005333503087363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,1,balanced,0.22124799092610678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,1,balanced,0.2250666618347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,1,balanced,0.22784000635147095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,1,balanced,0.22988800207773843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,1,balanced,0.23174933592478433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,1,balanced,0.23880000909169516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,1,balanced,0.24016533295313516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,1,balanced,0.2435893416404724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,1,balanced,0.24728532632191977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,1,balanced,0.2556053400039673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,1,balanced,0.27366934219996136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,1,balanced,0.31547733147939044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,1,balanced,0.33799465497334796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,1,balanced,0.4122186501820882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,1,balanced,0.4829973379770915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,1,balanced,0.6758080323537191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,1,balanced,0.7801173528035482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,1,balanced,1.1595306396484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.2,0.13358080387115479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,1,balanced,1.4468426704406738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,1,balanced,2.0884159406026206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,1,balanced,2.698213259379069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.2,0.1884608030319214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.2,0.23547520637512206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,32,2,power_law_1.01,0.017113600671291352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,32,2,power_law_1.01,0.016864000260829924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,32,2,power_law_1.01,0.017081600427627564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,32,2,power_law_1.01,0.0173567995429039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,32,2,power_law_1.01,0.017420800030231477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,32,2,power_law_1.01,0.018848000466823576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,32,2,power_law_1.01,0.01852799952030182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,32,2,power_law_1.01,0.01958400011062622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,32,2,power_law_1.01,0.021491199731826782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,32,2,power_law_1.01,0.021376000344753267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,32,2,power_law_1.01,0.021324799954891206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,32,2,power_law_1.01,0.03904640078544617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,32,2,power_law_1.01,0.03875199854373932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,32,2,power_law_1.01,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,32,2,power_law_1.01,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,32,2,power_law_1.01,0.03167999982833862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,32,2,power_law_1.01,0.03255040049552917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,32,2,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,32,2,power_law_1.01,0.040300801396369934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,32,2,power_law_1.01,0.04659200012683869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,32,2,power_law_1.01,0.05023360252380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,32,2,power_law_1.01,0.06952319741249084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,32,2,power_law_1.01,0.07975680232048035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,32,2,power_law_1.01,0.10739840269088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,32,2,power_law_1.01,0.13393919467926024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,2,power_law_1.2,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,2,power_law_1.2,0.03828479945659637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,2,power_law_1.2,0.04986239969730377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,2,power_law_1.2,0.14431359767913818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,2,power_law_1.2,0.06331520080566407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,2,power_law_1.2,0.07374719977378845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,2,power_law_1.2,0.10355839729309083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,2,power_law_1.2,0.10591360330581664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,2,power_law_1.2,0.11356799602508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,2,power_law_1.2,0.1126911997795105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,2,power_law_1.2,0.1170240044593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,2,power_law_1.2,0.11800320148468017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,2,power_law_1.2,0.12494720220565796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,2,power_law_1.2,0.12448639869689941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,2,power_law_1.2,0.2421247959136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,2,power_law_1.2,0.1268928050994873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,2,power_law_1.2,0.13437440395355224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,2,power_law_1.2,0.14295680522918702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,2,power_law_1.2,0.14981759786605836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,2,power_law_1.2,0.1691648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,2,power_law_1.2,0.16596479415893556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,2,power_law_1.2,0.196288001537323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,2,power_law_1.2,0.22696959972381592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,2,power_law_1.2,0.266592001914978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,2,power_law_1.2,0.3362368106842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,2,power_law_1.2,0.28134400844573976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,2,power_law_1.2,0.4992703914642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,2,power_law_1.2,0.3514111995697021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,2,power_law_1.2,0.4343679904937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,2,power_law_1.2,0.5952576160430908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,2,power_law_1.2,0.6721920013427735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,2,power_law_1.2,0.8070976257324218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,2,power_law_1.2,1.211903953552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,2,power_law_1.2,1.1853695869445802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,2,power_law_1.2,1.2313535690307618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,16,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,16,balanced,0.035786665976047516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,16,balanced,0.03549333413441976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,16,balanced,0.03939733405907949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,2,power_law_1.2,1.2656191825866698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,2,power_law_1.2,1.3603391647338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,2,power_law_1.2,1.396224021911621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,16,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,16,balanced,0.046298667788505554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,16,balanced,0.04600533346335093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,16,balanced,0.04757866760094961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,16,balanced,0.04776533444722494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,16,balanced,0.04791999856630961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,16,balanced,0.049685334165891014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,16,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,16,balanced,0.05006400247414907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,16,balanced,0.05187733471393585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,16,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,16,balanced,0.0562720000743866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,16,balanced,0.05952533086140951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,16,balanced,0.06609066824118297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,2,power_law_1.2,1.455519962310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,16,balanced,0.07826666533946991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,16,balanced,0.08684266606966655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,16,balanced,0.08685866991678874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,16,balanced,0.1033066709836324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,2,power_law_1.2,1.4836031913757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,2,power_law_1.2,1.5314559936523438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,2,power_law_1.2,1.5571136474609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,2,power_law_1.2,0.7581183910369873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,16,balanced,0.11355732878049214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,2,power_law_1.2,1.6852607727050781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,16,balanced,0.1460693379243215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,16,balanced,0.1809813380241394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,2,power_law_1.2,1.8722047805786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,16,balanced,0.24125866095225015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,16,balanced,0.292410671710968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,32,2,power_law_1.01,0.18878079652786256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,2,power_law_1.2,1.892019271850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,2,power_law_1.2,2.1694463729858398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,2,power_law_1.2,1.9948863983154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,2,power_law_1.2,2.306380844116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,2,power_law_1.2,2.7117120742797853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,2,power_law_1.2,3.1560575485229494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,2,power_law_1.2,3.794540786743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,2,power_law_1.2,5.136761474609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,2,power_law_1.2,5.898495864868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,1,balanced,0.07337066531181335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,1,balanced,0.09012266993522644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,1,balanced,0.1295840044816335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,1,balanced,0.18826133012771606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,1,balanced,0.3189280033111572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,1,balanced,0.288975993792216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,1,balanced,0.27460267146428424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,1,balanced,0.2824373245239258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,1,balanced,0.283135990301768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,1,balanced,0.28356266021728516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,1,balanced,0.28733332951863605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,1,balanced,0.2780746618906657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,1,balanced,0.29075199365615845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,1,balanced,0.28464533885320026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,1,balanced,0.296725332736969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,1,balanced,0.3088480035463969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,1,balanced,0.3124106725056966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,1,balanced,0.33268266916275024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,1,balanced,0.3549813429514567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,1,balanced,0.40831466515858966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,1,balanced,0.45716265837351483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,1,balanced,0.7239306767781576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,1,balanced,0.8254079818725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,1,balanced,1.1663946310679119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,1,balanced,1.4869227409362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,1,balanced,2.160831928253174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,1,balanced,2.789557456970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,16,power_law_1.01,0.024076800048351287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,16,power_law_1.01,0.029631999135017396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,16,power_law_1.01,0.024460799992084503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,16,power_law_1.01,0.02314240038394928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,16,power_law_1.01,0.023980799317359924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,16,power_law_1.01,0.025760000944137572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,16,power_law_1.01,0.02659200131893158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,16,power_law_1.01,0.028384000062942505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,16,power_law_1.01,0.028703999519348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,16,power_law_1.01,0.030214399099349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,16,power_law_1.01,0.04014720022678375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,16,power_law_1.01,0.040031999349594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,16,power_law_1.01,0.041171199083328246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,16,power_law_1.01,0.05792639851570129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,16,power_law_1.01,0.05978879928588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,16,power_law_1.01,0.05982720255851746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,16,power_law_1.01,0.08344320058822632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,16,power_law_1.01,0.08932480216026306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.01,0.06990079879760742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.01,0.08358399868011475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.01,0.09057919979095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.01,0.11450879573822022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.01,0.13841279745101928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.01,0.18263039588928223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.01,0.23445119857788085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.01,0.3272576093673706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.01,0.4331520080566406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,32,2,power_law_1.01,0.23621759414672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,16,8,power_law_1.01,0.06054400205612183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,16,8,power_law_1.01,0.06570879817008972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,16,8,power_law_1.01,0.06969599723815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,16,8,power_law_1.01,0.0762943983078003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,16,8,power_law_1.01,0.08010879755020142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,16,8,power_law_1.01,0.07749119997024537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,16,8,power_law_1.01,0.0811456024646759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,16,8,power_law_1.01,0.08164479732513427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,16,8,power_law_1.01,0.08402559757232667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,16,8,power_law_1.01,0.08328319787979126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,16,8,power_law_1.01,0.08599039912223816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,16,8,power_law_1.01,0.08707839846611024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,16,8,power_law_1.01,0.08936960101127625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,16,8,power_law_1.01,0.09258239865303039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,16,8,power_law_1.01,0.09874560236930847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,16,8,power_law_1.01,0.09918720126152039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,16,8,power_law_1.01,0.10389759540557861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,16,8,power_law_1.01,0.12302080392837525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,16,8,power_law_1.01,0.1332159996032715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,16,8,power_law_1.01,0.15363839864730836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,16,8,power_law_1.01,0.18499200344085692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,16,8,power_law_1.01,0.25128319263458254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,16,8,power_law_1.01,0.2986752033233643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,16,8,power_law_1.01,0.42332801818847654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,16,8,power_law_1.01,0.5889472007751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,16,8,power_law_1.01,0.7671743869781494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,16,8,power_law_1.01,1.1118399620056152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,2,balanced,0.06657599906126659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,2,balanced,0.06844266752401988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,2,balanced,0.07433600227038066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,2,balanced,0.08936533331871033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,2,balanced,0.11494933565457661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,2,balanced,0.16169599692026773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,2,balanced,0.16103466351826987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,2,balanced,0.16038399934768677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,2,balanced,0.16056000192960104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,2,balanced,0.1609119971593221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,2,balanced,0.16208533445994058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,2,balanced,0.1622880001862844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,2,balanced,0.16542933384577432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,2,balanced,0.1670773426691691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,2,balanced,0.17322667439778647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,2,balanced,0.17439999183019003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,2,balanced,0.18243199586868286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,2,balanced,0.20082666476567587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,2,balanced,0.21391467253367105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,2,balanced,0.2436479926109314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,2,balanced,0.27588266134262085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,2,balanced,0.34147198994954425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,2,balanced,0.40854398409525555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,2,balanced,0.6075573364893595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,2,balanced,0.7436213493347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,2,balanced,1.0749066670735676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,2,balanced,1.3972320556640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,64,balanced,0.05213333169619242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,64,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,64,balanced,0.03994133323431015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,64,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,64,balanced,0.04219200213750204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,64,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,64,balanced,0.049679999550183616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,64,balanced,0.04970133304595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,64,balanced,0.04974400003751119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,64,balanced,0.04868266483147939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,64,balanced,0.05095999936262766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,64,balanced,0.05136533578236898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,64,balanced,0.05067733426888784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,64,balanced,0.0525439977645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,64,balanced,0.057215998570124306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,64,balanced,0.05665066838264465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,64,balanced,0.05614933371543884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,64,balanced,0.061535999178886414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,64,balanced,0.0624533345301946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,64,balanced,0.07042133311430614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,64,balanced,0.07845333218574524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,64,balanced,0.09630399942398071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,64,balanced,0.11427733302116394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,64,balanced,0.14703466494878134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,64,balanced,0.18286399046579996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,64,balanced,0.23614400625228882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,64,balanced,0.2942506670951843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,2,power_law_1.01,0.021465599536895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,2,power_law_1.01,0.022489599883556366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,2,power_law_1.01,0.022086399793624877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,2,power_law_1.01,0.023039999604225158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,2,power_law_1.01,0.028940799832344054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,2,power_law_1.01,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,2,power_law_1.01,0.029996800422668456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,2,power_law_1.01,0.03173120021820068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,2,power_law_1.01,0.03384959995746613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,2,power_law_1.01,0.0318015992641449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,2,power_law_1.01,0.0356799989938736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,2,power_law_1.01,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,2,power_law_1.01,0.07534719705581665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,2,power_law_1.01,0.0794368028640747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,2,power_law_1.01,0.05923200249671936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,2,power_law_1.01,0.052671998739242554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,2,power_law_1.01,0.05928959846496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,2,power_law_1.01,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.01,0.09067519903182983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.01,0.1259775996208191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.01,0.1510208010673523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.01,0.211680006980896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.01,0.26721279621124266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.01,0.381932806968689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.01,0.49790081977844236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.01,0.7294655799865722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,1,power_law_1.01,0.02481919974088669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,1,power_law_1.01,0.03751679956912994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,1,power_law_1.01,0.055251199007034305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,1,power_law_1.01,0.08376320004463196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,1,power_law_1.01,0.11505919694900513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,1,power_law_1.01,0.1558527946472168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,1,power_law_1.01,0.20265600681304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,1,power_law_1.01,0.21401600837707518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,1,power_law_1.01,0.21738240718841553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,1,power_law_1.01,0.22867839336395263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,1,power_law_1.01,0.23742079734802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,1,power_law_1.01,0.246726393699646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,1,power_law_1.01,0.2561919927597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,64,power_law_1.2,0.03190400004386902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,64,power_law_1.2,0.027558401226997375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,64,power_law_1.2,0.027193599939346315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,64,power_law_1.2,0.029388800263404846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,64,power_law_1.2,0.030246400833129884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,64,power_law_1.2,0.030982398986816408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,64,power_law_1.2,0.030995199084281923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,64,power_law_1.2,0.03139199912548065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,64,power_law_1.2,0.03178240060806274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,64,power_law_1.2,0.031718400120735166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,64,power_law_1.2,0.03253119885921478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,64,power_law_1.2,0.03287039995193482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,64,power_law_1.2,0.033529600501060484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,64,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,64,power_law_1.2,0.03871999979019165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,64,power_law_1.2,0.039289599657058714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,1,power_law_1.01,0.25931520462036134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,64,power_law_1.2,0.03979519903659821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,1,power_law_1.01,0.26807680130004885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,64,power_law_1.2,0.0431551992893219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,64,power_law_1.2,0.04806399941444397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,64,power_law_1.2,0.05430399775505066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,64,power_law_1.2,0.06257280111312866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,64,power_law_1.2,0.06963199973106385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,1,power_law_1.01,0.28078720569610593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,1,power_law_1.01,0.29485440254211426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,1,power_law_1.01,0.33696000576019286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,1,power_law_1.01,0.34292480945587156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,1,power_law_1.01,0.428326416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,1,power_law_1.01,0.44069762229919435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,1,power_law_1.01,0.5681151866912841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,1,power_law_1.01,0.5410240173339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,1,power_law_1.01,0.6890175819396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,64,power_law_1.2,0.08780800104141236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,1,power_law_1.01,0.8580160140991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,1,power_law_1.01,1.192627239227295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,1,power_law_1.01,1.5448384284973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,64,power_law_1.2,0.12471040487289428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.01,0.9650495529174805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,64,power_law_1.2,0.16198400259017945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,64,power_law_1.2,0.22446079254150392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,64,power_law_1.2,0.2860480070114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,8,balanced,0.019424000134070713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,8,balanced,0.01921066641807556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,8,balanced,0.019071999937295914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,8,balanced,0.02093333254257838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,8,balanced,0.022133332987626392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,8,balanced,0.023007998863856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,8,balanced,0.02516266703605652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,8,balanced,0.025194667279720306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,8,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,8,balanced,0.03156266609827677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,8,balanced,0.03086400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,8,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,8,balanced,0.04586133360862732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,8,balanced,0.04747733473777771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,8,balanced,0.04955733319123586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,8,balanced,0.049839998284975685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,8,balanced,0.04870399832725525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,8,balanced,0.055231998364130654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,8,balanced,0.062224000692367554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,8,balanced,0.07135466734568278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,8,balanced,0.08224533498287201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,8,balanced,0.09802666306495667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,8,balanced,0.11981866757074992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,8,balanced,0.16635200381278992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,8,balanced,0.20119466384251913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,8,balanced,0.28097599744796753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,8,balanced,0.36180798212687176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,2,power_law_1.01,0.02101760059595108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,2,power_law_1.01,0.024780799448490144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,2,power_law_1.01,0.0395583987236023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,2,power_law_1.01,0.05467519760131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,2,power_law_1.01,0.07070720195770264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,2,power_law_1.01,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,2,power_law_1.01,0.10051200389862061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,2,power_law_1.01,0.10353280305862426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,2,power_law_1.01,0.10808320045471191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,2,power_law_1.01,0.11048959493637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,2,power_law_1.01,0.11698559522628785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,2,power_law_1.01,0.11476479768753052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,2,power_law_1.01,0.11632640361785888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,2,power_law_1.01,0.1268288016319275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,2,power_law_1.01,0.12581119537353516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,2,power_law_1.01,0.12874879837036132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,2,power_law_1.01,0.13624320030212403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,2,power_law_1.01,0.15004160404205322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,2,power_law_1.01,0.15954560041427612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,2,power_law_1.01,0.31683199405670165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,2,power_law_1.01,0.35282559394836427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,2,power_law_1.01,0.28758399486541747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,2,power_law_1.01,0.32766079902648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,2,power_law_1.01,0.39032959938049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,2,power_law_1.01,0.4720128059387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,2,power_law_1.01,0.6222784042358398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,2,power_law_1.01,0.7824192047119141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,32,power_law_1.2,0.026265600323677064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,32,power_law_1.2,0.030847999453544616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,32,power_law_1.2,0.023865599930286408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,32,power_law_1.2,0.0243136003613472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,32,power_law_1.2,0.025088000297546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,32,power_law_1.2,0.02717440128326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,32,power_law_1.2,0.030105599761009218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,32,power_law_1.2,0.03091199994087219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,32,power_law_1.2,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,32,power_law_1.2,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,32,power_law_1.2,0.039001598954200745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,32,power_law_1.2,0.043859198689460754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,32,power_law_1.2,0.045388799905776975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,32,power_law_1.2,0.06451200246810913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,32,power_law_1.2,0.07294080257415772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,32,power_law_1.2,0.07126399874687195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,32,power_law_1.2,0.08403840065002441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,32,power_law_1.2,0.09125120043754578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,32,power_law_1.2,0.11561599969863892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,32,power_law_1.2,0.14635519981384276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,32,power_law_1.2,0.09963520169258118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,32,power_law_1.2,0.12684160470962524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,32,power_law_1.2,0.16189440488815307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,32,power_law_1.2,0.2202752113342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,32,power_law_1.2,0.2660223960876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,32,power_law_1.2,0.3977279901504517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,32,power_law_1.2,0.5195712089538574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,4,power_law_1.01,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,4,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,4,power_law_1.01,0.03307519853115082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,4,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,4,power_law_1.01,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,4,power_law_1.01,0.0385343998670578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,4,power_law_1.01,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,4,power_law_1.01,0.038924801349639895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,4,power_law_1.01,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,4,power_law_1.01,0.04049279987812042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,4,power_law_1.01,0.04077439904212952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,4,power_law_1.01,0.04166400134563446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,4,power_law_1.01,0.043110400438308716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,4,power_law_1.01,0.04565759897232056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,4,power_law_1.01,0.05111680030822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,4,power_law_1.01,0.05508480072021484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,4,power_law_1.01,0.059436798095703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,4,power_law_1.01,0.06989439725875854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,4,power_law_1.01,0.07757440209388733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,128,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,4,power_law_1.01,0.09870079755783082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,4,power_law_1.01,0.10780800580978393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,4,power_law_1.01,0.14746880531311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,4,power_law_1.01,0.20561280250549316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,4,power_law_1.01,0.3007999897003174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,4,power_law_1.01,0.4051648139953613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,4,power_law_1.01,0.48311681747436525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,4,power_law_1.01,0.6260543823242187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,128,balanced,0.03158933420976003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,128,balanced,0.031583999594052635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,128,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,128,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,128,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,128,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,128,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,128,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,128,balanced,0.03649600098530451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,128,balanced,0.03734400123357773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,128,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,128,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,128,balanced,0.039434666434923805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,128,balanced,0.03982933362325033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,128,balanced,0.041893333196640015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,128,balanced,0.042863999803860985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,128,balanced,0.045408000548680626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,128,balanced,0.04394666850566864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,128,balanced,0.04866666595141093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,128,balanced,0.050101334849993386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,128,balanced,0.05608533322811127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,128,balanced,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,128,balanced,0.07675200204054515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,128,balanced,0.09845866759618123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,128,balanced,0.13555733362833658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,128,balanced,0.1668000022570292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,16,balanced,0.05622399846712748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,16,balanced,0.05621333420276642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,16,balanced,0.06201066573460897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,16,balanced,0.07503466804822286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,16,balanced,0.08648000160853068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,16,balanced,0.0962506632010142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,16,balanced,0.09687466422716777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,16,balanced,0.09610133369763692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,16,balanced,0.09918933113416036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,16,balanced,0.09724799791971843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,16,balanced,0.09879466891288757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,16,balanced,0.10113599896430969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,16,balanced,0.10130666693051656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,16,balanced,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,16,balanced,0.11106133460998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,32,power_law_1.01,0.16792960166931153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,32,power_law_1.01,0.16643840074539185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,32,power_law_1.01,0.1606528043746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,32,power_law_1.01,0.16562559604644775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,32,power_law_1.01,0.1709247946739197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,16,balanced,0.11101866761843364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,32,power_law_1.01,0.16755199432373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,16,balanced,0.11296533544858296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,16,balanced,0.12158399820327759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,16,balanced,0.12804800271987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,32,power_law_1.01,0.16024960279464723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,16,balanced,0.13783466815948486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,32,power_law_1.01,0.16603519916534423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,16,balanced,0.15251200397809347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,32,power_law_1.01,0.17004159688949586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,16,balanced,0.199946661790212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,32,power_law_1.01,0.16463359594345092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,16,balanced,0.22366400559743246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,32,power_law_1.01,0.17211519479751586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,32,power_law_1.01,0.17359999418258668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,16,balanced,0.30248000224431354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,32,power_law_1.01,0.17262719869613646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,16,balanced,0.3842666546503703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,32,power_law_1.01,0.17040640115737915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,16,balanced,0.5210346778233846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,16,balanced,0.6766453584035238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,32,power_law_1.01,0.16846079826354982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,32,power_law_1.01,0.17644799947738649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,32,power_law_1.01,0.18258559703826904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,32,power_law_1.01,0.19392000436782836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,32,power_law_1.01,0.22420480251312255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,32,power_law_1.01,0.22982399463653563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,32,power_law_1.01,0.25824000835418703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,32,power_law_1.01,0.3256704092025757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,32,power_law_1.01,0.38045439720153806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,32,power_law_1.01,0.5180352210998536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,32,power_law_1.01,0.655027198791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,32,power_law_1.01,0.8401984214782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,32,power_law_1.01,1.1497407913208009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,1,power_law_1.01,0.04128639996051788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,1,power_law_1.01,0.04394879937171936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,1,power_law_1.01,0.05113599896430969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,1,power_law_1.01,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,1,power_law_1.01,0.06470400094985962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,1,power_law_1.01,0.0658623993396759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,1,power_law_1.01,0.0658623993396759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,1,power_law_1.01,0.06702079772949218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,1,power_law_1.01,0.067084801197052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,1,power_law_1.01,0.06666880249977111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,1,power_law_1.01,0.06876159906387329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,1,power_law_1.01,0.07283200025558471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,1,power_law_1.01,0.07361279726028443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,1,power_law_1.01,0.07886080145835876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,1,power_law_1.01,0.09566720128059387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,1,power_law_1.01,0.09971200227737427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,1,power_law_1.01,0.1038591980934143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,1,power_law_1.01,0.1193727970123291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.01,0.14574719667434693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.01,0.1904960036277771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.01,0.22986879348754882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.01,0.3264703989028931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.01,0.4273344039916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.01,0.6055935859680176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.01,0.7750656127929687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.01,1.1110015869140626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.01,1.516486358642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,2,power_law_1.01,0.030880001187324525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,2,power_law_1.01,0.03036159873008728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,2,power_law_1.01,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,2,power_law_1.01,0.03372159898281098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,2,power_law_1.01,0.033580800890922545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,2,power_law_1.01,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,2,power_law_1.01,0.03722879886627197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,2,power_law_1.01,0.037222400307655334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,2,power_law_1.01,0.04067200124263763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,2,power_law_1.01,0.04255360066890716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,2,power_law_1.01,0.041126400232315063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,2,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,2,power_law_1.01,0.050380802154541014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,2,power_law_1.01,0.04897280037403107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,2,power_law_1.01,0.05911679863929749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,2,power_law_1.01,0.06284800171852112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,2,power_law_1.01,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,2,power_law_1.01,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,2,power_law_1.01,0.08091520071029663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,2,power_law_1.01,0.0936191976070404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,2,power_law_1.01,0.11399040222167969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,2,power_law_1.01,0.1529855966567993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,2,power_law_1.01,0.1725376009941101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,2,power_law_1.01,0.245849609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,2,power_law_1.01,0.2811840057373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,2,power_law_1.01,0.3959680080413818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,2,power_law_1.01,0.5121471881866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,8,balanced,0.04248000184694926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,8,balanced,0.043738668163617454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,8,balanced,0.048058668772379555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,8,balanced,0.06651199857393901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,8,balanced,0.09427199761072795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,8,balanced,0.09451199571291606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,8,balanced,0.09483733773231506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,8,balanced,0.10310399532318115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,8,balanced,0.09505066275596619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,8,balanced,0.09533333778381348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,8,balanced,0.09497066338857015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,8,balanced,0.09915199875831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,8,balanced,0.09841600060462952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,8,balanced,0.09898666540781657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,8,balanced,0.107013334830602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,8,balanced,0.10778133074442546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,8,balanced,0.109333336353302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,8,balanced,0.11794666449228923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,8,balanced,0.12191466490427653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,8,balanced,0.13321600357691446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,8,balanced,0.1442400018374125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,8,balanced,0.19218132893244425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,8,balanced,0.21135467290878296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,8,balanced,0.28143999973932904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,8,balanced,0.36666667461395264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,8,balanced,0.5175679922103882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,8,balanced,0.6746666431427002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,1,power_law_1.2,0.16575360298156738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,1,power_law_1.2,0.19937280416488648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,1,power_law_1.2,0.28153600692749026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,1,power_law_1.2,0.3885951995849609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,1,power_law_1.2,0.5150464057922364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,1,power_law_1.2,0.8826687812805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,1,power_law_1.2,0.9467455863952636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,1,power_law_1.2,0.9870528221130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,1,power_law_1.2,1.0269760131835937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,1,power_law_1.2,1.0499008178710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,1,power_law_1.2,1.067232036590576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,1,power_law_1.2,1.1033791542053222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,1,power_law_1.2,1.146022415161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,1,power_law_1.2,1.1877504348754884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,1,power_law_1.2,1.2374655723571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,1,power_law_1.2,1.2666048049926757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,1,power_law_1.2,1.327564811706543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,1,power_law_1.2,1.495910358428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.2,1.6204160690307616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.2,1.9629247665405274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.2,2.155558395385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.2,2.690118408203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.2,3.277088165283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.2,4.359148788452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.2,5.424563217163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,1,power_law_1.2,7.637510681152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,1,power_law_1.2,9.766284942626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,8,balanced,0.04171733558177948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,8,balanced,0.043552001317342125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,8,balanced,0.04432533184687296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,8,balanced,0.06074133515357971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,8,balanced,0.07557866473992665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,8,balanced,0.11702932914098103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,8,balanced,0.1174720029036204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,8,balanced,0.11529067158699036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,8,balanced,0.11321600278218587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,8,balanced,0.11513599753379822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,8,balanced,0.114847997824351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,8,balanced,0.11572800079981486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,8,balanced,0.11553600430488586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,8,balanced,0.11891200145085652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,8,balanced,0.12193600336710612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,8,balanced,0.1237386663754781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,8,balanced,0.12685333689053854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,8,balanced,0.13212266564369202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,8,balanced,0.1383519967397054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,8,balanced,0.15223999818166098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,8,balanced,0.16267733772595724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,8,balanced,0.19653334220250449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,8,balanced,0.21241599321365356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,8,balanced,0.27265600363413495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,8,balanced,0.32359999418258667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,8,balanced,0.440341313680013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,8,balanced,0.5407840013504028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,64,power_law_1.2,0.07973120212554932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,64,power_law_1.2,0.07701119780540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,64,power_law_1.2,0.07585920095443725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,64,power_law_1.2,0.07920640110969543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,64,power_law_1.2,0.07667199969291687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,64,power_law_1.2,0.0753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,64,power_law_1.2,0.07625600099563598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,64,power_law_1.2,0.07998080253601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,64,power_law_1.2,0.08049280047416688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,64,power_law_1.2,0.08110079765319825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,64,power_law_1.2,0.08277119994163513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,64,power_law_1.2,0.08435199856758117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,64,power_law_1.2,0.08595839738845826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,64,power_law_1.2,0.08664320111274719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,64,power_law_1.2,0.09507840275764465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,64,power_law_1.2,0.09717119932174682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,64,power_law_1.2,0.10433919429779052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,64,power_law_1.2,0.11708799600601197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,64,power_law_1.2,0.12982399463653566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,64,power_law_1.2,0.15854079723358155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,64,power_law_1.2,0.1851263999938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,64,power_law_1.2,0.24060161113739015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,64,power_law_1.2,0.3100928068161011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,64,power_law_1.2,0.40165119171142577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,64,power_law_1.2,0.5197504043579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,64,power_law_1.2,0.7863999843597412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,64,power_law_1.2,1.0616767883300782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,2,balanced,0.022533332308133442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,2,balanced,0.027210667729377747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,2,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,2,balanced,0.05506666501363119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,2,balanced,0.08245866497357686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,2,balanced,0.10256000359853108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,2,balanced,0.10296000043551128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,2,balanced,0.10528000195821126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,2,balanced,0.1070186694463094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2688,1856,6,128,1,2,balanced,0.10972266395886739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2688,1856,6,128,1,2,balanced,0.10669333736101787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2688,1856,6,128,1,2,balanced,0.1158026655515035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2688,1856,6,128,1,2,balanced,0.11532266934712727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2688,1856,6,128,1,2,balanced,0.12079999844233195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2688,1856,6,128,1,2,balanced,0.12962133685747781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2688,1856,6,128,1,2,balanced,0.16234133640925089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2688,1856,6,128,1,2,balanced,0.12972266475359598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,8,power_law_1.01,0.140774405002594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,8,power_law_1.01,0.22094719409942626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,8,power_law_1.01,0.26039040088653564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,8,power_law_1.01,0.27485439777374265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,8,power_law_1.01,0.34163200855255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,8,power_law_1.01,0.2950079917907715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,8,power_law_1.01,0.37173120975494384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,8,power_law_1.01,0.3130176067352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,8,power_law_1.01,0.33223040103912355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,8,power_law_1.01,0.346617603302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,8,power_law_1.01,0.3472896099090576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,8,power_law_1.01,0.34812800884246825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,8,power_law_1.01,0.34570879936218263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,8,power_law_1.01,0.3531455993652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,8,power_law_1.01,0.3912575960159302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,8,power_law_1.01,0.39095680713653563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,8,power_law_1.01,0.40643839836120604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,8,power_law_1.01,0.4340928077697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,8,power_law_1.01,0.443174409866333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,8,power_law_1.01,0.5192319869995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,8,power_law_1.01,0.5089983940124512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,8,power_law_1.01,0.6216127872467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,8,power_law_1.01,0.6934720039367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,8,power_law_1.01,0.8175168037414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,8,power_law_1.01,0.99551362991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,8,power_law_1.01,1.3275391578674316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,8,power_law_1.01,1.688435173034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,64,power_law_1.01,0.08110719919204712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,64,power_law_1.01,0.08042240142822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,64,power_law_1.01,0.07826560139656066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,64,power_law_1.01,0.08036479949951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,64,power_law_1.01,0.07681919932365418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,64,power_law_1.01,0.08082559704780579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,64,power_law_1.01,0.07866880297660828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,64,power_law_1.01,0.08006399869918823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,64,power_law_1.01,0.08123520016670227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,64,power_law_1.01,0.081523197889328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,64,power_law_1.01,0.08298879861831665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,64,power_law_1.01,0.0849407970905304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,64,power_law_1.01,0.08632320165634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,64,power_law_1.01,0.08610560297966004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,64,power_law_1.01,0.09165440201759338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,64,power_law_1.01,0.09523199796676636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,64,power_law_1.01,0.10314879417419434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,64,power_law_1.01,0.1183743953704834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,64,power_law_1.01,0.12658560276031494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,64,power_law_1.01,0.1529088020324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,64,power_law_1.01,0.16808320283889772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,64,power_law_1.01,0.22350718975067138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,64,power_law_1.01,0.22402560710906982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,64,power_law_1.01,0.2877568006515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,64,power_law_1.01,0.37166719436645507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,64,power_law_1.01,0.4825727939605713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,64,power_law_1.01,0.5574975967407226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,1,power_law_1.2,0.05397760272026062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,1,power_law_1.2,0.059059202671051025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,1,power_law_1.2,0.07612159848213196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,1,power_law_1.2,0.10478719472885131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,1,power_law_1.2,0.13880959749221802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,1,power_law_1.2,0.17329920530319215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,1,power_law_1.2,0.23322880268096924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,1,power_law_1.2,0.24454400539398194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,1,power_law_1.2,0.251366400718689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,1,power_law_1.2,0.26204159259796145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,1,power_law_1.2,0.2652672052383423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,1,power_law_1.2,0.2803136110305786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,1,power_law_1.2,0.28590080738067625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,1,power_law_1.2,0.29789440631866454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,1,power_law_1.2,0.3170367956161499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,1,power_law_1.2,0.3307071924209595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,1,power_law_1.2,0.34764800071716306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,1,power_law_1.2,0.4093696117401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,1,power_law_1.2,0.44352002143859864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,1,power_law_1.2,0.5365695953369141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,1,power_law_1.2,0.604966402053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,1,power_law_1.2,0.7720640182495118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,1,power_law_1.2,0.9660160064697265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,1,power_law_1.2,1.2904895782470702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,1,power_law_1.2,1.5990655899047852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,1,power_law_1.2,2.2429887771606447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,1,power_law_1.2,2.8775104522705077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,256,power_law_1.2,0.057740801572799684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,256,power_law_1.2,0.05726720094680786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,256,power_law_1.2,0.05772799849510193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,256,power_law_1.2,0.05969280004501343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,256,power_law_1.2,0.059680002927780154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,256,power_law_1.2,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,256,power_law_1.2,0.06290559768676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,256,power_law_1.2,0.06291199922561645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,256,power_law_1.2,0.06410880088806152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,256,power_law_1.2,0.06266239881515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,256,power_law_1.2,0.062054401636123656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,256,power_law_1.2,0.06339840292930603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,256,power_law_1.2,0.06279680132865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,256,power_law_1.2,0.06252800226211548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,256,power_law_1.2,0.06657919883728028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,256,power_law_1.2,0.06638720035552978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,256,power_law_1.2,0.07529600262641907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,256,power_law_1.2,0.07983360290527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,256,power_law_1.2,0.09845759868621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,256,power_law_1.2,0.12766079902648925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,256,power_law_1.2,0.1518847942352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,256,power_law_1.2,0.20706560611724853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,256,power_law_1.2,0.25622398853302003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,256,power_law_1.2,0.35987200736999514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,256,power_law_1.2,0.4663871765136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,256,power_law_1.2,0.6829887866973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,256,power_law_1.2,0.9850111961364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,16,power_law_1.01,0.056652802228927615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,16,power_law_1.01,0.06678400039672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,16,power_law_1.01,0.06356480121612548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,16,power_law_1.01,0.06778879761695862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,16,power_law_1.01,0.06769919991493226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,16,power_law_1.01,0.06652160286903382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,16,power_law_1.01,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,16,power_law_1.01,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,16,power_law_1.01,0.06677759885787964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,16,power_law_1.01,0.06734079718589783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,16,power_law_1.01,0.07253119945526124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,16,power_law_1.01,0.07148159742355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,16,power_law_1.01,0.07474560141563416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,16,power_law_1.01,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,16,power_law_1.01,0.08239359855651855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,16,power_law_1.01,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,16,power_law_1.01,0.09262080192565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,16,power_law_1.01,0.10540800094604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,16,power_law_1.01,0.10952320098876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,16,power_law_1.01,0.13313280344009398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,16,power_law_1.01,0.15217280387878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,16,power_law_1.01,0.18735359907150267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,16,power_law_1.01,0.24058239459991454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,16,power_law_1.01,0.3152575969696045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,16,power_law_1.01,0.38874239921569825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,16,power_law_1.01,0.5600063800811768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,16,power_law_1.01,0.7090047836303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,32,power_law_1.2,0.1128383994102478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,32,power_law_1.2,0.03208959996700287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,32,power_law_1.2,0.031251201033592226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,32,power_law_1.2,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,32,power_law_1.2,0.02988159954547882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,32,power_law_1.2,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,32,power_law_1.2,0.03207040131092072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,32,power_law_1.2,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,32,power_law_1.2,0.03331199884414673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,32,power_law_1.2,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,32,power_law_1.2,0.03375360071659088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,32,power_law_1.2,0.03489919900894165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,32,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,32,power_law_1.2,0.03639039993286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,32,power_law_1.2,0.040166398882865904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,32,power_law_1.2,0.041740798950195314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,32,power_law_1.2,0.043584001064300534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,32,power_law_1.2,0.04832639992237091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,32,power_law_1.2,0.052249598503112796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,32,power_law_1.2,0.05715199708938599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,32,power_law_1.2,0.06204800009727478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,32,power_law_1.2,0.0775488018989563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,32,power_law_1.2,0.09835519790649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,32,power_law_1.2,0.13708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,32,power_law_1.2,0.16551040410995482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,32,power_law_1.2,0.22111361026763915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,32,power_law_1.2,0.3019455909729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,1,power_law_1.01,0.030739200115203858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,1,power_law_1.01,0.030585598945617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,1,power_law_1.01,0.03335680067539215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,1,power_law_1.01,0.037574398517608645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,1,power_law_1.01,0.03722879886627197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,1,power_law_1.01,0.03918719887733459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,1,power_law_1.01,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,4,power_law_1.2,0.060166400671005246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,1,power_law_1.01,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,4,power_law_1.2,0.0688704013824463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,1,power_law_1.01,0.04477440118789673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,1,power_law_1.01,0.0466623991727829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,1,power_law_1.01,0.04616959989070892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,1,power_law_1.01,0.049004799127578734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,1,power_law_1.01,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,4,power_law_1.2,0.06845440268516541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,4,power_law_1.2,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,4,power_law_1.2,0.0778880000114441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,1,power_law_1.01,0.05360640287399292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,4,power_law_1.2,0.07800319790840149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,1,power_law_1.01,0.06499199867248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,4,power_law_1.2,0.0762880027294159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,4,power_law_1.2,0.07834879755973816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,4,power_law_1.2,0.07752959728240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,4,power_law_1.2,0.07555199861526489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,4,power_law_1.2,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,4,power_law_1.2,0.09080960154533387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,4,power_law_1.2,0.08734080195426941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,4,power_law_1.2,0.08800640106201171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,1,power_law_1.01,0.06786559820175171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,1,power_law_1.01,0.0744704008102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,4,power_law_1.2,0.09868159890174866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,1,power_law_1.01,0.09155840277671815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,4,power_law_1.2,0.1017408013343811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,1,power_law_1.01,0.09534080028533935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,1,power_law_1.01,0.12421120405197143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,1,power_law_1.01,0.1397696018218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,4,power_law_1.2,0.10947200059890747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,4,power_law_1.2,0.12317440509796143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.2,0.14762879610061647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.2,0.1782464027404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,1,power_law_1.01,0.18033280372619628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,1,power_law_1.01,0.2246527910232544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.2,0.24858241081237792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,1,power_law_1.01,0.31461760997772215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.2,0.31746559143066405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.2,0.3967423915863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,1,power_law_1.01,0.40790400505065916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.2,0.5551936149597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,1,power_law_1.01,0.5826560020446777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.2,0.7331136226654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,64,balanced,0.039434666434923805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,64,balanced,0.03841066608826319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,64,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,64,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,64,balanced,0.04148799926042557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,64,balanced,0.041663999358812966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,64,balanced,0.043807998299598694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,64,balanced,0.04317333300908407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,64,balanced,0.04374399781227112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,64,balanced,0.04358933369318644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,64,balanced,0.04532266656557719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,64,balanced,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,64,balanced,0.04553600152333578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,64,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,64,balanced,0.051957334081331887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,64,balanced,0.0518506666024526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,64,balanced,0.053904001911481224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,64,balanced,0.05986666679382324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,64,balanced,0.062047998110453285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,64,balanced,0.06833066542943318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,64,balanced,0.07457600037256877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,64,balanced,0.09270399808883667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,64,balanced,0.103301336367925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,64,balanced,0.1360640029112498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,64,balanced,0.1693920095761617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,64,balanced,0.22660799821217856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,64,balanced,0.2848213315010071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.2,0.9803263664245605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,1,power_law_1.01,0.7563519954681397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,8,power_law_1.01,0.09489279985427856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,8,power_law_1.01,0.12903039455413817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,8,power_law_1.01,0.14601600170135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,8,power_law_1.01,0.14742399454116822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,8,power_law_1.01,0.1642303943634033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,8,power_law_1.01,0.17964160442352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,8,power_law_1.01,0.1691648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,8,power_law_1.01,0.17519999742507936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,8,power_law_1.01,0.17517440319061278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,8,power_law_1.01,0.1755328059196472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,8,power_law_1.01,0.18061439990997313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,8,power_law_1.01,0.18330880403518676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,8,power_law_1.01,0.18279039859771729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,8,power_law_1.01,0.18735359907150267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,8,power_law_1.01,0.19583359956741334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,8,power_law_1.01,0.2012415885925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,8,power_law_1.01,0.20296320915222169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,8,power_law_1.01,0.23141119480133057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,8,power_law_1.01,0.23595519065856935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,8,power_law_1.01,0.269324803352356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,8,power_law_1.01,0.3101887941360474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,8,power_law_1.01,0.3937279939651489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,8,power_law_1.01,0.4832767963409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,8,power_law_1.01,0.658406400680542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,8,power_law_1.01,0.8506303787231445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,8,power_law_1.01,1.2596096038818358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,8,power_law_1.01,1.5308480262756348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.2,1.474086380004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,4,power_law_1.2,0.10028799772262573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,4,power_law_1.2,0.14654719829559326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,4,power_law_1.2,0.15247360467910767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,4,power_law_1.2,0.22145280838012696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,4,power_law_1.2,0.2641024112701416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,4,power_law_1.2,0.3050816059112549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,4,power_law_1.2,0.3978944063186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,4,power_law_1.2,0.3397183895111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,4,power_law_1.2,0.39475839138031005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,4,power_law_1.2,0.3780992031097412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,4,power_law_1.2,0.39483520984649656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,4,power_law_1.2,0.3829184055328369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,4,power_law_1.2,0.39317760467529295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,4,power_law_1.2,0.42463998794555663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,4,power_law_1.2,0.4246272087097168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,4,power_law_1.2,0.436191987991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,4,power_law_1.2,0.4710720062255859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,4,power_law_1.2,0.5209727764129639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,4,power_law_1.2,0.5325759887695313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,32,balanced,0.05236266553401947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,32,balanced,0.052527998884518944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,32,balanced,0.053930665055910744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,4,power_law_1.2,0.6436287879943847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,32,balanced,0.06323199967543285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,32,balanced,0.07851199805736542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,4,power_law_1.2,0.6146111965179444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,32,balanced,0.10754133264223735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,32,balanced,0.10917333761850993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,32,balanced,0.10674666364987691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,32,balanced,0.1114453375339508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,32,balanced,0.10871466994285583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,32,balanced,0.1102186640103658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,4,power_law_1.2,0.7007808208465576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,4,power_law_1.2,0.8460800170898437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,32,balanced,0.11492799719174702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,32,balanced,0.11494933565457661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,32,balanced,0.1169706682364146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,4,power_law_1.2,1.073414421081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,32,balanced,0.13532800475756326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,32,balanced,0.13531200091044107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,32,balanced,0.13652799526850382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,32,balanced,0.1490079959233602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,32,balanced,0.15455999970436096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,32,balanced,0.16850133736928305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,32,balanced,0.18288532892862955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,32,balanced,0.2125813364982605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,32,balanced,0.23808000485102335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,32,balanced,0.30853333075841266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,4,power_law_1.2,1.3264575958251954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,32,balanced,0.3588586648305257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,32,balanced,0.4901866515477498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,4,power_law_1.2,1.6986175537109376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,32,balanced,0.6238346497217814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,4,power_law_1.2,2.219852828979492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,8,power_law_1.2,0.06465280055999756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,8,power_law_1.2,0.055078399181365964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,8,power_law_1.2,0.051801598072052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,8,power_law_1.2,0.05488640069961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,8,power_law_1.2,0.054201602935791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,8,power_law_1.2,0.05488640069961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,8,power_law_1.2,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,8,power_law_1.2,0.05968639850616455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,8,power_law_1.2,0.06483200192451477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,32,balanced,0.06706133484840393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,8,power_law_1.2,0.06559360027313232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,8,power_law_1.2,0.06419839859008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,8,power_law_1.2,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,8,power_law_1.2,0.06869120001792908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,8,power_law_1.2,0.07455360293388366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,8,power_law_1.2,0.08893439769744874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,32,balanced,0.0800853321949641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,32,balanced,0.09013866384824117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,32,balanced,0.12593066692352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,8,power_law_1.2,0.08886399865150452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,32,balanced,0.15877866744995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,32,balanced,0.15928533673286438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,8,power_law_1.2,0.09648640155792236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,8,power_law_1.2,0.1249343991279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,32,balanced,0.1588213344415029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,32,balanced,0.16074666380882263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,32,balanced,0.15904532869656882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,32,balanced,0.16430399815241495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,32,balanced,0.16581333676973978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,32,balanced,0.16689600547154745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,32,balanced,0.17092265685399374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,32,balanced,0.17087999979654947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,32,balanced,0.1776426633199056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,32,balanced,0.1803200046221415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,32,balanced,0.18661866585413614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,32,balanced,0.20188266038894653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,32,balanced,0.2146186629931132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,32,balanced,0.24531733989715576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,32,balanced,0.2726080020268758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,32,balanced,0.3691360155741374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,32,balanced,0.3823893467585246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,8,power_law_1.2,0.14529919624328613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,32,balanced,0.5006453196207682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,32,balanced,0.5883253415425619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,32,balanced,0.8485439618428549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,32,balanced,1.0738666852315266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,8,power_law_1.2,0.19855999946594238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,8,power_law_1.2,0.22257919311523439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,8,power_law_1.2,0.3090496063232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,8,power_law_1.2,0.3896127939224243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,8,power_law_1.2,0.6162879943847657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,32,power_law_1.2,0.01969279944896698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,32,power_law_1.2,0.019148799777030944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,32,power_law_1.2,0.018534399569034576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,32,power_law_1.2,0.018092800676822663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,32,power_law_1.2,0.018137599527835845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,32,power_law_1.2,0.019206400215625762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,32,power_law_1.2,0.0190528005361557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,32,power_law_1.2,0.019840000569820403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,32,power_law_1.2,0.020729599893093108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,32,power_law_1.2,0.021049599349498748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,32,power_law_1.2,0.02391680032014847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,32,power_law_1.2,0.023865599930286408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,32,power_law_1.2,0.024268800020217897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,32,power_law_1.2,0.03560320138931274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,32,power_law_1.2,0.03675520122051239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,32,power_law_1.2,0.03739520013332367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,32,power_law_1.2,0.03741439878940582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,32,power_law_1.2,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,32,power_law_1.2,0.04217599928379059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,32,power_law_1.2,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,32,power_law_1.2,0.0496832013130188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,32,power_law_1.2,0.06663680076599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,32,power_law_1.2,0.07722240090370178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,32,power_law_1.2,0.10229760408401489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,32,power_law_1.2,0.13185919523239137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,32,power_law_1.2,0.17953920364379883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,32,power_law_1.2,0.22878720760345458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,8,power_law_1.2,0.8185088157653808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,8,power_law_1.2,1.2448639869689941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,1,balanced,0.08944533268610637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,1,balanced,0.13309866189956665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,1,balanced,0.2167253295580546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,1,balanced,0.3868639866511027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,1,balanced,0.7222506999969482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,1,balanced,1.3965973854064941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,1,balanced,1.4030613899230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,1,balanced,1.404842694600423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,1,balanced,1.40665070215861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,1,balanced,1.413365364074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,1,balanced,1.4194933573404949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,1,balanced,1.4270826975504558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,1,balanced,1.4368640581766765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,1,balanced,1.4434773127237956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,1,balanced,1.4533012708028157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,1,balanced,1.4665973981221516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,1,balanced,1.4871360460917156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,1,balanced,1.5389866828918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,1,balanced,1.577829360961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,1,balanced,1.6647413571675618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,1,balanced,1.7349440256754558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,1,balanced,1.8618292808532715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,1,balanced,2.0397866566975913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,1,balanced,2.400181293487549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,1,balanced,2.608282725016276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,1,balanced,4.111728032430013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,1,balanced,4.202080090840657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,8,power_law_1.2,1.5195648193359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,4,power_law_1.2,0.07121919989585876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,4,power_law_1.2,0.07973120212554932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,4,power_law_1.2,0.07564799785614014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,4,power_law_1.2,0.09724159836769104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,4,power_law_1.2,0.1093824028968811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,4,power_law_1.2,0.11395200490951538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,4,power_law_1.2,0.14006400108337402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,4,power_law_1.2,0.12635519504547119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,4,power_law_1.2,0.1356160044670105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,4,power_law_1.2,0.1364351987838745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,4,power_law_1.2,0.1373568058013916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,4,power_law_1.2,0.13453439474105836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,4,power_law_1.2,0.13810559511184692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,4,power_law_1.2,0.13768320083618163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,4,power_law_1.2,0.15102720260620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,4,power_law_1.2,0.14933120012283324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,4,power_law_1.2,0.15870720148086548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,4,power_law_1.2,0.1735551953315735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,4,power_law_1.2,0.19319039583206177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,4,power_law_1.2,0.22799360752105713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,4,power_law_1.2,0.2673919916152954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,4,power_law_1.2,0.34002559185028075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,4,power_law_1.2,0.40728321075439455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,4,power_law_1.2,0.5630144119262696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,4,power_law_1.2,0.6992512226104737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,4,power_law_1.2,0.9724864006042481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,4,power_law_1.2,1.325984001159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,32,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,32,balanced,0.029258665939172108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,32,balanced,0.029631999631722767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,32,balanced,0.031541332602500916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,32,balanced,0.031541332602500916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,32,balanced,0.03369600077470144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,32,balanced,0.03384533276160558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,32,balanced,0.03477866699298223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,32,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,32,balanced,0.033733333150545754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,32,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,32,balanced,0.03738666574160258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,32,balanced,0.03734400123357773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,32,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,32,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,32,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,32,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,32,balanced,0.043807998299598694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,32,balanced,0.04557333389918009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,32,balanced,0.05242133140563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,32,balanced,0.057962665955225624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,32,balanced,0.07738133271535237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,32,balanced,0.09891200065612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,32,balanced,0.12846400340398154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,32,balanced,0.16265599926312765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,32,balanced,0.21896000703175864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,32,balanced,0.27669866879781085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,2,power_law_1.2,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,2,power_law_1.2,0.039654400944709775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,2,power_law_1.2,0.056729602813720706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,2,power_law_1.2,0.06805760264396668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,2,power_law_1.2,0.07943040132522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,2,power_law_1.2,0.11344640254974366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,2,power_law_1.2,0.11622400283813476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,2,power_law_1.2,0.11981439590454102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,2,power_law_1.2,0.12064000368118286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,2,power_law_1.2,0.12357120513916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,2,power_law_1.2,0.1264256000518799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,2,power_law_1.2,0.13061120510101318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,2,power_law_1.2,0.13303680419921876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,2,power_law_1.2,0.1399616003036499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,2,power_law_1.2,0.14627200365066528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,2,power_law_1.2,0.15155199766159058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,2,power_law_1.2,0.15805439949035643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,2,power_law_1.2,0.17562880516052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,2,power_law_1.2,0.19423359632492065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,2,power_law_1.2,0.2375040054321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,2,power_law_1.2,0.2651583909988403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,2,power_law_1.2,0.3374592065811157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,2,power_law_1.2,0.43742079734802247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,2,power_law_1.2,0.5921984195709229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,32,power_law_1.2,0.032595199346542356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,32,power_law_1.2,0.031436800956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,32,power_law_1.2,0.030048000812530517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,32,power_law_1.2,0.030112001299858093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,32,power_law_1.2,0.03048959970474243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,32,power_law_1.2,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,32,power_law_1.2,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,32,power_law_1.2,0.03223679959774017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,32,power_law_1.2,0.03229439854621887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,32,power_law_1.2,0.03273600041866302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,32,power_law_1.2,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,32,power_law_1.2,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,32,power_law_1.2,0.03473280072212219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,32,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,32,power_law_1.2,0.04067200124263763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,32,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,32,power_law_1.2,0.04126079976558685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,32,power_law_1.2,0.04402559995651245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,32,power_law_1.2,0.048390400409698484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,32,power_law_1.2,0.05822719931602478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,32,power_law_1.2,0.06312320232391358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,2,power_law_1.2,0.7186880111694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,32,power_law_1.2,0.08085759878158569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,32,power_law_1.2,0.09881600141525268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,32,power_law_1.2,0.1496448040008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,32,power_law_1.2,0.16355839967727662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,32,power_law_1.2,0.25157120227813723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,2,power_law_1.2,0.9624704360961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,2,power_law_1.2,1.2377280235290526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,32,power_law_1.2,0.3377471923828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,64,power_law_1.2,0.060659199953079224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,64,power_law_1.2,0.05759360194206238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,64,power_law_1.2,0.05137919783592224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,64,power_law_1.2,0.05305600166320801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,64,power_law_1.2,0.0573311984539032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,64,power_law_1.2,0.06026880145072937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,64,power_law_1.2,0.05780479907989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,64,power_law_1.2,0.05863680243492127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,64,power_law_1.2,0.058822399377822875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,64,power_law_1.2,0.0582144021987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,64,power_law_1.2,0.0591808021068573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,64,power_law_1.2,0.06599680185317994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,64,power_law_1.2,0.06822400093078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,64,power_law_1.2,0.07017599940299987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,64,power_law_1.2,0.08208000063896179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,64,power_law_1.2,0.0815936028957367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,64,power_law_1.2,0.08520320057868958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,64,power_law_1.2,0.10480639934539795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,64,power_law_1.2,0.10794240236282349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,64,power_law_1.2,0.13934719562530518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,64,power_law_1.2,0.157478404045105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,64,power_law_1.2,0.2066431999206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,64,power_law_1.2,0.22598400115966796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,64,power_law_1.2,0.2827136039733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,64,power_law_1.2,0.36969599723815916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,64,power_law_1.2,0.49932160377502444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,64,power_law_1.2,0.6136640071868896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,16,power_law_1.01,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,16,power_law_1.01,0.04291839897632599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,16,power_law_1.01,0.04087679982185364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,16,power_law_1.01,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,16,power_law_1.01,0.04296320080757141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,16,power_law_1.01,0.04604800045490265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,16,power_law_1.01,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,16,power_law_1.01,0.04591360092163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,16,power_law_1.01,0.045900800824165346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,16,power_law_1.01,0.04824959933757782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,16,power_law_1.01,0.049446401000022885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,16,power_law_1.01,0.05061759948730469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,16,power_law_1.01,0.05135359764099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,16,power_law_1.01,0.05445759892463684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,16,power_law_1.01,0.060262399911880496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,16,power_law_1.01,0.06232960224151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,16,power_law_1.01,0.06867200136184692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,16,power_law_1.01,0.08121600151062011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.01,0.09445760250091553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.01,0.10350079536437988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.01,0.10913920402526855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.01,0.13452800512313842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.01,0.16614400148391723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.01,0.21571199893951415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.01,0.32381439208984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.01,0.4131455898284912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.01,0.47830400466918943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,32,power_law_1.2,0.020473599433898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,32,power_law_1.2,0.020230400562286376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,32,power_law_1.2,0.019859200716018675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,32,power_law_1.2,0.021062399446964263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,32,power_law_1.2,0.023974399268627166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,2,balanced,0.06253866851329803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,32,power_law_1.2,0.032051199674606325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,2,balanced,0.06776000062624614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,2,balanced,0.08774399757385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,2,balanced,0.12372266252835591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,2,balanced,0.17450666427612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,2,balanced,0.28407466411590576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,2,balanced,0.2852693398793538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,2,balanced,0.29361599683761597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,2,balanced,0.291701336701711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,2,balanced,0.2961440086364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,2,balanced,0.29844266176223755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,2,balanced,0.3031733234723409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,2,balanced,0.3062719901402791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,2,balanced,0.3100053270657857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,2,balanced,0.3194719950358073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,2,balanced,0.3264373342196147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,2,balanced,0.33382399876912433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,2,balanced,0.3604053258895874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,2,balanced,0.3847200075785319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,2,balanced,0.4301760196685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,2,balanced,0.46800533930460614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,2,balanced,0.5598133405049642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,2,balanced,0.6451626618703207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,2,balanced,0.9072693188985189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,2,balanced,1.0861866474151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,2,balanced,1.5495360692342122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,2,balanced,1.9660587310791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,4,power_law_1.01,0.04318079948425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,4,power_law_1.01,0.04664320051670075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,4,power_law_1.01,0.047788798809051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,4,power_law_1.01,0.04843519926071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,4,power_law_1.01,0.05304960012435913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,4,power_law_1.01,0.05373439788818359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,4,power_law_1.01,0.060102397203445436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,4,power_law_1.01,0.05982080101966858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,4,power_law_1.01,0.05996800065040588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,4,power_law_1.01,0.06135680079460144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,4,power_law_1.01,0.0633791983127594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,4,power_law_1.01,0.06600959897041321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,4,power_law_1.01,0.06762880086898804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,4,power_law_1.01,0.07267199754714966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,4,power_law_1.01,0.08245760202407837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,4,power_law_1.01,0.08422399759292602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,4,power_law_1.01,0.08615679740905761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,4,power_law_1.01,0.09920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,4,power_law_1.01,0.10984959602355956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,4,power_law_1.01,0.13616000413894652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,4,power_law_1.01,0.163481605052948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,4,power_law_1.01,0.20531840324401857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,4,power_law_1.01,0.2543936014175415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,4,power_law_1.01,0.34780158996582033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,4,power_law_1.01,0.46862077713012695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,4,power_law_1.01,0.6847743988037109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,4,power_law_1.01,0.8607808113098144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,16,power_law_1.01,0.05010560154914856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,16,power_law_1.01,0.04764159917831421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,16,power_law_1.01,0.045849600434303285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,16,power_law_1.01,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,16,power_law_1.01,0.04563199877738953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,16,power_law_1.01,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,16,power_law_1.01,0.05017600059509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,16,power_law_1.01,0.04849919974803925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,16,power_law_1.01,0.04945279955863953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,16,power_law_1.01,0.04925439953804016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,16,power_law_1.01,0.04966399967670441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,16,power_law_1.01,0.050483202934265135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,16,power_law_1.01,0.05013120174407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,16,power_law_1.01,0.051558399200439455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,16,power_law_1.01,0.055212801694869994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,16,power_law_1.01,0.054425597190856934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,16,power_law_1.01,0.057715201377868654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,16,power_law_1.01,0.06808320283889771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.01,0.0713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.01,0.0852735996246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.01,0.09991040229797363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.01,0.1233407974243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.01,0.14765440225601195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.01,0.1942080020904541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.01,0.2541183948516846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.01,0.3467200040817261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.01,0.4926976203918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,1,power_law_1.2,0.03544960021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,1,power_law_1.2,0.04015359878540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,1,power_law_1.2,0.04926080107688904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,1,power_law_1.2,0.06285439729690552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,1,power_law_1.2,0.07891839742660522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,1,power_law_1.2,0.1123136043548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,1,power_law_1.2,0.11887999773025512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,1,power_law_1.2,0.1227903962135315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,1,power_law_1.2,0.12391040325164795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,1,power_law_1.2,0.12643840312957763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,1,power_law_1.2,0.13354239463806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,1,power_law_1.2,0.13782399892807007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,1,power_law_1.2,0.14302719831466676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,1,power_law_1.2,0.14049279689788818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,1,power_law_1.2,0.1474176049232483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,1,power_law_1.2,0.15181440114974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,1,power_law_1.2,0.1581120014190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,4,power_law_1.01,0.04545280039310455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,1,power_law_1.2,0.17596800327301027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,4,power_law_1.01,0.0496832013130188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,1,power_law_1.2,0.19626879692077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,4,power_law_1.01,0.05367040038108826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,4,power_law_1.01,0.058387202024459836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,4,power_law_1.01,0.06280959844589233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,4,power_law_1.01,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,4,power_law_1.01,0.07645440101623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,4,power_law_1.01,0.07946239709854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,4,power_law_1.01,0.08244479894638061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,4,power_law_1.01,0.08031359910964966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,4,power_law_1.01,0.08556159734725952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,4,power_law_1.01,0.08862079977989197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,4,power_law_1.01,0.09420160055160523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,4,power_law_1.01,0.09936000108718872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,4,power_law_1.01,0.11233919858932495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,1,power_law_1.2,0.23618559837341307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,4,power_law_1.01,0.11787519454956055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,1,power_law_1.2,0.25827200412750245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,4,power_law_1.01,0.11992319822311401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,1,power_law_1.2,0.334553599357605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,4,power_law_1.01,0.14022400379180908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,4,power_law_1.01,0.15285760164260864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,4,power_law_1.01,0.1999295949935913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,4,power_law_1.01,0.23945600986480714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,4,power_law_1.01,0.29910399913787844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,1,power_law_1.2,0.4142399787902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,4,power_law_1.01,0.3935296058654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,4,power_law_1.01,0.54301438331604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,4,power_law_1.01,0.6716671943664551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,1,power_law_1.2,0.5513984203338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,4,power_law_1.01,1.0087167739868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,4,power_law_1.01,1.2972415924072265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,1,power_law_1.2,0.6927552223205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,1,power_law_1.2,0.9677951812744141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,4,power_law_1.2,0.019123199582099914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,4,power_law_1.2,0.0584384024143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,4,power_law_1.2,0.01863040030002594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,4,power_law_1.2,0.07336320281028748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,4,power_law_1.2,0.07534719705581665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,4,power_law_1.2,0.09488639831542969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,4,power_law_1.2,0.12215679883956909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,4,power_law_1.2,0.13105920553207398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,4,power_law_1.2,0.15448960065841674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,4,power_law_1.2,0.019014400243759156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,4,power_law_1.2,0.15485440492630004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,2,power_law_1.2,0.021401600539684297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,4,power_law_1.2,0.15044480562210083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,2,power_law_1.2,0.022681599855422972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,4,power_law_1.2,0.15283199548721313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,2,power_law_1.2,0.025273600220680238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,4,power_law_1.2,0.15119359493255616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,2,power_law_1.2,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,4,power_law_1.2,0.1670912027359009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,2,power_law_1.2,0.04211840033531189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,4,power_law_1.2,0.16005120277404786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,2,power_law_1.2,0.047328001260757445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,4,power_law_1.2,0.17199360132217406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,2,power_law_1.2,0.05861120223999024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,4,power_law_1.2,0.019014400243759156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,4,power_law_1.2,0.01912959963083267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,4,power_law_1.2,0.020172800123691558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,4,power_law_1.2,0.02298240065574646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,4,power_law_1.2,0.02306559979915619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,4,power_law_1.2,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,4,power_law_1.2,0.04713599979877472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,4,power_law_1.2,0.04721280038356781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,2,power_law_1.2,0.06049280166625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,4,power_law_1.2,0.03656319975852966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,2,power_law_1.2,0.06407039761543273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,4,power_law_1.2,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,2,power_law_1.2,0.0661184012889862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,2,power_law_1.2,0.07311999797821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,2,power_law_1.2,0.06737279891967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,2,power_law_1.2,0.07169280052185059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,2,power_law_1.2,0.07959039807319641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,4,power_law_1.2,0.17836159467697144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,2,power_law_1.2,0.07772160172462464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,4,power_law_1.2,0.18472959995269775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,2,power_law_1.2,0.08116480112075805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,4,power_law_1.2,0.1906559944152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,2,power_law_1.2,0.08855680227279664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,4,power_law_1.2,0.21342079639434813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,2,power_law_1.2,0.10703359842300415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,4,power_law_1.2,0.21821439266204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.2,0.11904640197753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,4,power_law_1.2,0.24861440658569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.2,0.1891263961791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,4,power_law_1.2,0.28713600635528563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.2,0.20008959770202636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,4,power_law_1.2,0.36348800659179686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.2,0.23102080821990967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,4,power_law_1.2,0.378438401222229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.2,0.2559551954269409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,4,power_law_1.2,0.505350399017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.2,0.32468481063842775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.2,0.38949759006500245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,4,power_law_1.2,0.6291520118713378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.2,0.5373760223388672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,4,power_law_1.2,0.8103360176086426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,4,power_law_1.2,0.035488000512123107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.2,0.6653567790985108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,4,power_law_1.2,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,4,power_law_1.2,1.0426624298095704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,4,power_law_1.2,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,4,power_law_1.2,0.04136320054531097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,1,power_law_1.2,1.2494208335876464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,4,power_law_1.2,0.043724799156188966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.2,0.04607360064983368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.2,0.0664896011352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.2,0.07528319954872131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.2,0.09941759705543518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.2,0.12451200485229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.2,0.17148799896240235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,8,power_law_1.2,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,8,power_law_1.2,0.06784639954566955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,8,power_law_1.2,0.06783360242843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,8,power_law_1.2,0.07938560247421264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,8,power_law_1.2,0.08586239814758301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,8,power_law_1.2,0.08399999737739564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,8,power_law_1.2,0.09425280094146729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,8,power_law_1.2,0.0932416021823883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,8,power_law_1.2,0.08977280259132385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,8,power_law_1.2,0.09252480268478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,8,power_law_1.2,0.0958079993724823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,8,power_law_1.2,0.09503999948501587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,8,power_law_1.2,0.10069119930267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,8,power_law_1.2,0.1027519941329956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,8,power_law_1.2,0.11395839452743531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,8,power_law_1.2,0.11542400121688842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,8,power_law_1.2,0.12560640573501586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,8,power_law_1.2,0.14856959581375123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.2,0.16129280328750611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.2,0.20446081161499025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.2,0.20184319019317626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.2,0.2667072057723999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.2,0.3025536060333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.2,0.42298240661621095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.2,0.5446720123291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.2,0.8412351608276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.2,1.1260479927062987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.2,0.21068799495697021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.2,0.3071104049682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,4,balanced,0.04811733464399973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,4,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,4,balanced,0.059989333152770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,4,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,4,balanced,0.06597333153088887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.2,0.3924736022949219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,4,balanced,0.06598400076230367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,4,balanced,0.0662720004717509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,4,balanced,0.06776000062624614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,4,balanced,0.06692266464233398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,4,balanced,0.06679466863473256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,4,balanced,0.06643199920654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,4,balanced,0.07071466743946075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,4,balanced,0.0703306645154953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,4,balanced,0.07111999889214833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,4,balanced,0.08486933509508769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,4,balanced,0.08681066830952962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,4,balanced,0.08063999811808269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,4,balanced,0.1027733286221822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,4,balanced,0.10179199775060017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,4,balanced,0.1302826702594757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,4,balanced,0.17129600048065186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,4,balanced,0.21620800097783408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,4,balanced,0.2653493285179138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,4,balanced,0.3686773379643758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,4,balanced,0.4656053384145101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,4,balanced,0.6865013440450033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,4,balanced,0.8943999608357748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,2,power_law_1.2,0.04955520033836365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,2,power_law_1.2,0.07564160227775574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,2,power_law_1.2,0.10969599485397338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,2,power_law_1.2,0.16195839643478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,2,power_law_1.2,0.2057663917541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,2,power_law_1.2,0.2488192081451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,2,power_law_1.2,0.3718656063079834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,2,power_law_1.2,0.35618560314178466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,2,power_law_1.2,0.39209599494934083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,2,power_law_1.2,0.4016255855560303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,2,power_law_1.2,0.4078400135040283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,2,power_law_1.2,0.4114816188812256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,2,power_law_1.2,0.44012160301208497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,2,power_law_1.2,0.4386176109313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,2,power_law_1.2,0.45061120986938474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,2,power_law_1.2,0.4561408042907715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,2,power_law_1.2,0.49278721809387205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,2,power_law_1.2,0.5615808010101319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.2,0.6092544078826905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.2,0.744704008102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.2,0.7090112209320069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.2,0.8863039970397949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.2,1.0528448104858399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.2,1.1901951789855958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.2,1.3816512107849122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,2,power_law_1.2,0.07362560033798218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.2,1.8808256149291993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,2,power_law_1.2,0.08392320275306701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,2,power_law_1.2,0.10086400508880615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,2,power_law_1.2,0.12568960189819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.2,2.268492889404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,2,power_law_1.2,0.15595519542694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,2,power_law_1.2,0.1757248044013977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,2,power_law_1.2,0.19485440254211425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,2,power_law_1.2,0.24944000244140624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,2,power_law_1.2,0.26311678886413575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,2,power_law_1.2,0.2754496097564697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,2,power_law_1.2,0.26499199867248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,2,power_law_1.2,0.28485119342803955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,2,power_law_1.2,0.28497920036315916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,2,power_law_1.2,0.2991552114486694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,2,power_law_1.2,0.3190592050552368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,2,power_law_1.2,0.3071104049682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,2,power_law_1.2,0.333407998085022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,2,power_law_1.2,0.358841609954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,2,power_law_1.2,0.4111616134643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,2,power_law_1.2,0.48081278800964355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,2,power_law_1.2,0.5478720188140869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,2,power_law_1.2,0.6754559993743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,2,power_law_1.2,0.7932096004486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,2,power_law_1.2,1.0289600372314454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,2,power_law_1.2,1.3294464111328126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,64,balanced,0.05820266902446747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,2,power_law_1.2,1.8414976119995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,64,balanced,0.06218666831652323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,64,balanced,0.06965866684913635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,64,balanced,0.09842666983604431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,64,balanced,0.1070240040620168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,64,balanced,0.10748799641927083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,2,power_law_1.2,2.234310340881348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,2,power_law_1.2,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,2,power_law_1.2,0.06144639849662781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,2,power_law_1.2,0.07133439779281617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,64,balanced,0.11523200074831645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,2,power_law_1.2,0.08929920196533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,64,balanced,0.1172320048014323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,64,balanced,0.11519466837247212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,2,power_law_1.2,0.11090559959411621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,2,power_law_1.2,0.17753599882125853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,2,power_law_1.2,0.1860479950904846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,2,power_law_1.2,0.18934400081634523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,2,power_law_1.2,0.19352960586547852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,2,power_law_1.2,0.19553279876708984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,2,power_law_1.2,0.20028159618377686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,64,balanced,0.11965866883595784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,64,balanced,0.12147733569145203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,64,balanced,0.12270933389663696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,64,balanced,0.12411733468373616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,64,balanced,0.12686933080355325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,64,balanced,0.13467199603716531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,64,balanced,0.13678399721781412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,64,balanced,0.1435413360595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,64,balanced,0.15662399927775064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,64,balanced,0.16725333531697592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,64,balanced,0.19508800903956094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,2,power_law_1.2,0.2068095922470093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,2,power_law_1.2,0.21336960792541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,2,power_law_1.2,0.21697919368743895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,2,power_law_1.2,0.23192319869995118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,2,power_law_1.2,0.23841919898986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,2,power_law_1.2,0.23633279800415039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,2,power_law_1.2,0.27171199321746825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,2,power_law_1.2,0.27282559871673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,64,balanced,0.2205173373222351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,64,balanced,0.29253333806991577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,64,balanced,0.2996480067571004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,64,balanced,0.39000535011291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,64,balanced,0.46251734097798664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,64,balanced,0.6532373428344727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,2,power_law_1.2,0.3241472005844116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,2,power_law_1.2,0.35562241077423096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,64,balanced,0.8197706540425619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,2,power_law_1.2,0.4336703777313232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,2,power_law_1.2,0.45168638229370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,2,power_law_1.2,0.5398528099060058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,2,power_law_1.2,0.6873536109924316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,2,power_law_1.2,0.9253312110900879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,2,power_law_1.2,1.101203155517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,4,balanced,0.1083679993947347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,4,balanced,0.09059733152389526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,4,balanced,0.14891200264294943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,4,balanced,0.14131733775138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,4,balanced,0.1376213332017263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,4,balanced,0.14030399918556213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,4,balanced,0.14056000113487244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,4,balanced,0.14167466759681702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,4,balanced,0.14797332882881165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,4,balanced,0.15262400110562643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,4,balanced,0.14843733112017313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,4,balanced,0.1439573367436727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,4,balanced,0.14542933305104574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,4,balanced,0.15546133120854697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,4,balanced,0.16862932840983072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,4,balanced,0.17105066776275635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,4,balanced,0.17746132612228394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,4,balanced,0.2015519936879476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,4,balanced,0.21853333711624146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,4,balanced,0.27956799666086835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,4,balanced,0.29729066292444867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,4,balanced,0.39771731694539386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,4,balanced,0.4777333339055379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,4,balanced,0.6747519969940186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,4,balanced,0.8768906593322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,4,balanced,1.277402639389038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,4,balanced,1.627423922220866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,2,power_law_1.01,0.07461119890213012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,2,power_law_1.01,0.08376960158348083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,2,power_law_1.01,0.09098880290985108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,2,power_law_1.01,0.1152959942817688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,2,power_law_1.01,0.12855039834976195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,2,power_law_1.01,0.13749760389328003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,2,power_law_1.01,0.1666432023048401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,2,power_law_1.01,0.16681599617004395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,2,power_law_1.01,0.16171519756317138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,2,power_law_1.01,0.16654720306396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,2,power_law_1.01,0.17592320442199708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,2,power_law_1.01,0.175545597076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,2,power_law_1.01,0.17668479681015015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,2,power_law_1.01,0.18362239599227906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,2,power_law_1.01,0.19031039476394654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,2,power_law_1.01,0.19100159406661987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,2,power_law_1.01,0.19640320539474487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,2,power_law_1.01,0.21719679832458497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,2,power_law_1.01,0.24584319591522216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,2,power_law_1.01,0.29231359958648684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,2,power_law_1.01,0.3260159969329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,2,power_law_1.01,0.4324031829833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,2,power_law_1.01,0.5300352096557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,2,power_law_1.01,0.7048192024230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,1,balanced,0.035946667194366455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,1,balanced,0.04078399886687597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,1,balanced,0.05758399764696757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,1,balanced,0.08786666393280029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,1,balanced,0.13492266337076822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,1,balanced,0.13981866836547852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,1,balanced,0.14014400045077005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,1,balanced,0.1437013347943624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,1,balanced,0.1441386640071869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,1,balanced,0.1523413360118866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,1,balanced,0.1516266663869222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,1,balanced,0.1571999986966451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,1,balanced,0.15980799992879233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,1,balanced,0.16645866632461548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,1,balanced,0.17458132902781168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,1,balanced,0.18107199668884277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,1,balanced,0.19409600893656412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,1,balanced,0.22164799769719443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,1,balanced,0.25062400102615356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,1,balanced,0.3261599938074748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,1,balanced,0.3598346710205078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,1,balanced,0.505946675936381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,1,balanced,0.6065920193990072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,1,balanced,0.8677120208740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,2,power_law_1.01,0.8988479614257813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,1,balanced,1.1051572958628337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,2,power_law_1.01,1.2355648040771485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,1,balanced,1.6088213920593262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,1,balanced,2.08406400680542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,4,balanced,0.03857066730658213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,4,balanced,0.03782933453718821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,4,balanced,0.05663466453552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,4,balanced,0.05580266813437144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,4,balanced,0.056101332108179726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,4,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,2,power_law_1.01,1.6220544815063476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,4,balanced,0.058245331048965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,4,balanced,0.058634668588638306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,4,balanced,0.05856533348560333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,4,balanced,0.060517330964406334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,4,balanced,0.062021334966023765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,4,balanced,0.062224000692367554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,4,balanced,0.06256533165772755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,4,balanced,0.06609066824118297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,4,balanced,0.07457600037256877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,4,balanced,0.07531733314196269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,4,balanced,0.07673066854476929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,4,balanced,0.09602133433024089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,4,balanced,0.09737599889437358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,4,balanced,0.1195199986298879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,4,balanced,0.13717333475748697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,4,balanced,0.17947733402252197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,4,balanced,0.21986132860183716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,4,balanced,0.2990933259328206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,4,balanced,0.37015998363494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,4,balanced,0.527071992556254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,4,balanced,0.6740960280100504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,16,balanced,0.02420266717672348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,16,balanced,0.023557332654794056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,16,balanced,0.023408000667889912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,16,balanced,0.02510400116443634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,16,balanced,0.03562133262554804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,16,balanced,0.052144000927607216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,16,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,16,balanced,0.05950400233268738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,16,balanced,0.06192533175150553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,16,balanced,0.06408533453941345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,16,balanced,0.05991999804973602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,16,balanced,0.06485866506894429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,16,balanced,0.06773866713047028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,16,balanced,0.07426133255163829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,16,balanced,0.07925333579381307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,16,balanced,0.08450133601824443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,16,balanced,0.1368053356806437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,16,balanced,0.15501866738001505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,16,balanced,0.1879840095837911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,16,balanced,0.25044800837834674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,16,balanced,0.1733013391494751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,16,balanced,0.21726399660110474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,16,balanced,0.2650986711184184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,16,balanced,0.34141866366068524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,16,balanced,0.42344534397125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,16,balanced,0.6391786734263102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,16,balanced,0.8148159980773926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,1,power_law_1.01,0.18655999898910522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,1,power_law_1.01,0.24789121150970458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,1,power_law_1.01,0.32766079902648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,1,power_law_1.01,0.5066175937652588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,1,power_law_1.01,0.5316287994384765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,1,power_law_1.01,0.5247807979583741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,1,power_law_1.01,0.5240896224975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,1,power_law_1.01,0.5224832057952881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,1,power_law_1.01,0.5110464096069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,1,power_law_1.01,0.5038271903991699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,1,power_law_1.01,0.5111487865447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,1,power_law_1.01,0.490169620513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,1,power_law_1.01,0.5234496116638183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,1,power_law_1.01,0.5501376152038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,1,power_law_1.01,0.5679359912872315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,1,power_law_1.01,0.5889535903930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,1,power_law_1.01,0.7250112056732178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,1,power_law_1.01,0.7425727844238281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.01,0.9201919555664062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,256,power_law_1.01,0.046911999583244324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,256,power_law_1.01,0.04720639884471893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,256,power_law_1.01,0.04734080135822296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.01,1.1188672065734864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,256,power_law_1.01,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,256,power_law_1.01,0.05135359764099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.01,1.3836480140686036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.01,1.902707290649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,256,power_law_1.01,0.054995197057724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,256,power_law_1.01,0.056953597068786624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.01,2.3230592727661135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,256,power_law_1.01,0.06034560203552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,256,power_law_1.01,0.06034560203552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,256,power_law_1.01,0.06412799954414368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,256,power_law_1.01,0.06207360029220581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.01,3.2172161102294923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,256,power_law_1.01,0.06362239718437195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,16,power_law_1.01,0.10012799501419067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,16,power_law_1.01,0.11117440462112427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,256,power_law_1.01,0.06710399985313416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,16,power_law_1.01,0.10176639556884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,256,power_law_1.01,0.06536319851875305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,16,power_law_1.01,0.11403520107269287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,256,power_law_1.01,0.07702400088310242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,16,power_law_1.01,0.10696959495544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,16,power_law_1.01,0.10810240507125854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,16,power_law_1.01,0.10485759973526002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,16,power_law_1.01,0.11153279542922974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,16,power_law_1.01,0.11592960357666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,16,power_law_1.01,0.11282559633255004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,256,power_law_1.01,0.07902079820632935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,16,power_law_1.01,0.11931519508361817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,256,power_law_1.01,0.08234239816665649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,256,power_law_1.01,0.09587839841842652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,256,power_law_1.01,0.1037824034690857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,256,power_law_1.01,0.12848000526428222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,256,power_law_1.01,0.15020159482955933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,16,power_law_1.01,0.12052479982376099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,16,power_law_1.01,0.11832959651947021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,256,power_law_1.01,0.1966655969619751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,16,power_law_1.01,0.12252800464630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,256,power_law_1.01,0.2501823902130127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,16,power_law_1.01,0.13523199558258056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,256,power_law_1.01,0.33441920280456544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,16,power_law_1.01,0.13752959966659545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,16,power_law_1.01,0.1447424054145813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,256,power_law_1.01,0.45833601951599123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,16,power_law_1.01,0.16282880306243896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,256,power_law_1.01,0.6430848121643067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,16,power_law_1.01,0.1985856056213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,16,power_law_1.01,0.2407167911529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,256,power_law_1.01,0.8663488388061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,16,power_law_1.01,0.2360447883605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,16,power_law_1.01,0.30821120738983154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,16,power_law_1.01,0.3738431930541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,16,power_law_1.01,0.4859903812408447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.01,4.207238388061524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,16,power_law_1.01,0.6226624011993408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,16,power_law_1.01,0.8995264053344727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,16,power_law_1.01,1.1315584182739258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.01,5.997875213623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.01,7.956793975830078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,8,balanced,0.04523199796676636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,8,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,8,balanced,0.05176533261934916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,8,balanced,0.0726506660381953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,8,balanced,0.09790933132171631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,8,balanced,0.151829332113266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,8,balanced,0.2019253373146057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,8,balanced,0.202074666817983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,8,balanced,0.2023093303044637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,8,balanced,0.2034613291422526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,8,balanced,0.2043786644935608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,8,balanced,0.20665067434310913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,8,balanced,0.2104640007019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,8,balanced,0.214026669661204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,8,balanced,0.2155946691830953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,8,balanced,0.21868266661961874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,8,balanced,0.22453866402308145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,8,balanced,0.23748799165089926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,8,balanced,0.24540799856185913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,8,balanced,0.26790400346120197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,8,balanced,0.28730666637420654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,8,balanced,0.3304426670074463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,8,balanced,0.37690667311350506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,8,balanced,0.456218679745992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,8,balanced,0.5454773505528768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,16,power_law_1.01,0.033817601203918454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,16,power_law_1.01,0.03281919956207276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,8,balanced,0.710640033086141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,16,power_law_1.01,0.029203200340270997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,8,balanced,0.9068106810251871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,16,power_law_1.01,0.029900801181793214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,16,power_law_1.01,0.03184640109539032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,16,power_law_1.01,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,16,power_law_1.01,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,16,power_law_1.01,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,16,power_law_1.01,0.03310079872608185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,16,power_law_1.01,0.033036801218986514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,16,power_law_1.01,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,16,power_law_1.01,0.03671039938926697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,16,power_law_1.01,0.036575999855995175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,16,power_law_1.01,0.03909760117530823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,16,power_law_1.01,0.04312320053577423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,16,power_law_1.01,0.043968001008033754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,16,power_law_1.01,0.04636160135269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,16,power_law_1.01,0.05130879878997803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,16,power_law_1.01,0.05791360139846802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,16,power_law_1.01,0.0691648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,16,power_law_1.01,0.07034879922866821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,16,power_law_1.01,0.08961920142173767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,16,power_law_1.01,0.12163840532302857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,16,power_law_1.01,0.1614464044570923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,16,power_law_1.01,0.2020479917526245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,16,power_law_1.01,0.26796159744262693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,16,power_law_1.01,0.33909759521484373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,8,power_law_1.2,0.04827519953250885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,8,power_law_1.2,0.0654591977596283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,8,power_law_1.2,0.06598399877548218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,8,power_law_1.2,0.07921280264854431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,8,power_law_1.2,0.08700159788131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,8,power_law_1.2,0.09201279878616334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,8,power_law_1.2,0.08947839736938476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,8,power_law_1.2,0.09472000002861022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,8,power_law_1.2,0.09345279932022095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,8,power_law_1.2,0.09493119716644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,8,power_law_1.2,0.09624959826469422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,8,power_law_1.2,0.09694079756736755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,8,power_law_1.2,0.10578559637069702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,8,power_law_1.2,0.10967680215835571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,8,power_law_1.2,0.11819519996643066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,8,power_law_1.2,0.1197119951248169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,8,power_law_1.2,0.1256767988204956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,8,power_law_1.2,0.1439296007156372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,8,power_law_1.2,0.15738879442214965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,8,power_law_1.2,0.18796160221099853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,8,power_law_1.2,0.21492478847503663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,8,power_law_1.2,0.2833856105804443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,8,power_law_1.2,0.38166399002075196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,8,power_law_1.2,0.4863232135772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,8,power_law_1.2,0.6100351810455322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,8,power_law_1.2,0.8347904205322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,8,power_law_1.2,1.1845696449279786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,4,power_law_1.01,0.1220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,4,power_law_1.01,0.18702720403671264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,4,power_law_1.01,0.18517119884490968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,16,power_law_1.01,0.027603200078010558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,16,power_law_1.01,0.049619200825691226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,16,power_law_1.01,0.04798080027103424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,16,power_law_1.01,0.04978559911251068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,16,power_law_1.01,0.049881601333618165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,16,power_law_1.01,0.04881280064582825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,4,power_law_1.01,0.253222393989563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,4,power_law_1.01,0.2998719930648804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,4,power_law_1.01,0.4001664161682129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,16,power_law_1.01,0.0554751992225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,16,power_law_1.01,0.054054397344589236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,4,power_law_1.01,0.4009407997131348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,16,power_law_1.01,0.06761599779129028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,4,power_law_1.01,0.39483520984649656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,16,power_law_1.01,0.06772480010986329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,16,power_law_1.01,0.0670527994632721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,16,power_law_1.01,0.0967423975467682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,16,power_law_1.01,0.09634559750556945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,16,power_law_1.01,0.09745919704437256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,16,power_law_1.01,0.10459519624710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,16,power_law_1.01,0.10974719524383544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,4,power_law_1.01,0.39672958850860596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,4,power_law_1.01,0.42879362106323243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,4,power_law_1.01,0.4208576202392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,4,power_law_1.01,0.41841278076171873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,4,power_law_1.01,0.4346816062927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,4,power_law_1.01,0.4427840232849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,4,power_law_1.01,0.46544637680053713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,4,power_law_1.01,0.48017277717590334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,4,power_law_1.01,0.5199359893798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,4,power_law_1.01,0.592307186126709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,16,power_law_1.01,0.1049280047416687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,4,power_law_1.01,0.5807424068450928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,16,power_law_1.01,0.10880639553070068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,4,power_law_1.01,0.6890880107879639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.01,0.10050560235977173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.01,0.11934720277786255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,4,power_law_1.01,0.7733759880065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.01,0.12995200157165526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.01,0.16028159856796265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,4,power_law_1.01,0.9502911567687988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.01,0.18775039911270142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,4,power_law_1.01,0.9394880294799804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.01,0.2680704116821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.01,0.31893119812011717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,4,power_law_1.01,1.1808575630187987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.01,0.4710336208343506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.01,0.5987135887145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,4,power_law_1.01,1.442835235595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,4,power_law_1.01,2.1078271865844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,4,power_law_1.01,2.3120000839233397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,8,power_law_1.2,0.020844799280166627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,8,power_law_1.2,0.019167999923229217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,8,power_law_1.2,0.019763199985027312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,8,power_law_1.2,0.020127999782562255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,8,power_law_1.2,0.02046079933643341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,8,power_law_1.2,0.022252799570560457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,8,power_law_1.2,0.025241601467132568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,8,power_law_1.2,0.02366719990968704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,8,power_law_1.2,0.027993598580360414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,8,power_law_1.2,0.02805120050907135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,8,power_law_1.2,0.02534399926662445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,8,power_law_1.2,0.027456000447273254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,8,power_law_1.2,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,8,power_law_1.2,0.030137598514556885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,8,power_law_1.2,0.03164800107479095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,8,power_law_1.2,0.0335999995470047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,8,power_law_1.2,0.05568000078201294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,8,power_law_1.2,0.06738560199737549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.2,0.07555840015411378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.2,0.08936319947242737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.2,0.07427840232849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.2,0.0941760003566742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.2,0.111353600025177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.2,0.1424191951751709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,1,power_law_1.2,0.07256320118904114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,1,power_law_1.2,0.08784000277519226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,1,power_law_1.2,0.11741440296173096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,1,power_law_1.2,0.1501312017440796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,1,power_law_1.2,0.16105600595474243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,1,power_law_1.2,0.16597119569778443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,1,power_law_1.2,0.16276479959487916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,1,power_law_1.2,0.16802560091018676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.2,0.18448640108108522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,1,power_law_1.2,0.16892160177230836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,1,power_law_1.2,0.17146240472793578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,1,power_law_1.2,0.1779904007911682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,1,power_law_1.2,0.18364160060882567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,1,power_law_1.2,0.18858879804611206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,1,power_law_1.2,0.19723520278930665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,1,power_law_1.2,0.21084799766540527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,1,power_law_1.2,0.22100479602813722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,1,power_law_1.2,0.23725440502166747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,1,power_law_1.2,0.2813119888305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.2,0.3244927883148193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.2,0.40416641235351564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.2,0.4979519844055176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.2,0.6595327854156494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.2,0.2533760070800781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,4,power_law_1.2,0.04490239918231964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,4,power_law_1.2,0.04576640129089356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,4,power_law_1.2,0.04769279956817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,4,power_law_1.2,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,4,power_law_1.2,0.05226879715919495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,4,power_law_1.2,0.05355520248413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,4,power_law_1.2,0.05303679704666138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,4,power_law_1.2,0.05554559826850891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.2,0.8384127616882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,4,power_law_1.2,0.05436800122261047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,4,power_law_1.2,0.05392000079154968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,4,power_law_1.2,0.041843199729919435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,4,power_law_1.2,0.05305600166320801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,4,power_law_1.2,0.04334079921245575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,4,power_law_1.2,0.06030719876289368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,4,power_law_1.2,0.047513601183891294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,4,power_law_1.2,0.05708799958229065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,4,power_law_1.2,0.05603839755058289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,4,power_law_1.2,0.06224640011787415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,4,power_law_1.2,0.0732479989528656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,4,power_law_1.2,0.0738431990146637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,4,power_law_1.2,0.08426880240440368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,4,power_law_1.2,0.060819202661514284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,4,power_law_1.2,0.07054719924926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,4,power_law_1.2,0.06964480280876159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,4,power_law_1.2,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,4,power_law_1.2,0.09467520117759705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,4,power_law_1.2,0.07147520184516906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,4,power_law_1.2,0.07008000016212464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,4,power_law_1.2,0.07371519804000855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,4,power_law_1.2,0.07496320009231568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.2,1.205894374847412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,4,power_law_1.2,0.07768319845199585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,4,power_law_1.2,0.07882239818572997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,4,power_law_1.2,0.11425280570983887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,4,power_law_1.2,0.08075519800186157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,4,power_law_1.2,0.08277119994163513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,4,power_law_1.2,0.09074559807777405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,4,power_law_1.2,0.10332159996032715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,4,power_law_1.2,0.11096320152282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,4,power_law_1.2,0.13375359773635864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,4,power_law_1.2,0.13490560054779052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,4,power_law_1.2,0.1704256057739258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,4,power_law_1.2,0.21802239418029784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,4,power_law_1.2,0.31541759967803956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,4,power_law_1.2,0.16151679754257203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,4,power_law_1.2,0.2307584047317505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,4,power_law_1.2,0.28279039859771726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,4,power_law_1.2,0.42610559463500974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.2,1.6021568298339843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,4,power_law_1.2,0.5402751922607422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,4,power_law_1.2,0.8020799636840821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.2,0.3370048046112061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,4,power_law_1.2,0.4082496166229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.2,2.376063919067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,4,power_law_1.2,1.010316753387451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,4,power_law_1.2,0.48241281509399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,1,power_law_1.2,0.021817600727081297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,1,power_law_1.2,0.025702399015426636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,1,power_law_1.2,0.04229120016098022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,1,power_law_1.2,0.06253439784049988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,1,power_law_1.2,0.08410879969596863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,1,power_law_1.2,0.10997120141983033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,1,power_law_1.2,0.15256320238113402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,1,power_law_1.2,0.15777920484542846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,1,power_law_1.2,0.15828479528427125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,1,power_law_1.2,0.16691839694976807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,8,power_law_1.01,0.06743680238723755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,1,power_law_1.2,0.1724544048309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,1,power_law_1.2,0.1874176025390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,1,power_law_1.2,0.19684480428695678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,1,power_law_1.2,0.21409280300140382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,1,power_law_1.2,0.2077120065689087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,1,power_law_1.2,0.21898880004882812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,1,power_law_1.2,0.23927040100097657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,1,power_law_1.2,0.23855359554290773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,8,power_law_1.01,0.08195840120315552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,1,power_law_1.2,0.26099200248718263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,8,power_law_1.01,0.07348480224609374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,1,power_law_1.2,0.2817408084869385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,8,power_law_1.01,0.08799999952316284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,8,power_law_1.01,0.08960639834403991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,8,power_law_1.01,0.09064319729804993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,8,power_law_1.01,0.09436799883842469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,8,power_law_1.01,0.09628159999847412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,8,power_law_1.01,0.09593600034713745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,8,power_law_1.01,0.09617279767990113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,8,power_law_1.01,0.09757440090179444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,8,power_law_1.01,0.09983360171318054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,1,power_law_1.2,0.3076544046401978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,8,power_law_1.01,0.10264960527420045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,8,power_law_1.01,0.10450559854507446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,8,power_law_1.01,0.11340160369873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,8,power_law_1.01,0.1162559986114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,8,power_law_1.01,0.12602880001068115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,8,power_law_1.01,0.14056320190429689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.01,0.153056001663208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.01,0.1953727960586548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,4,power_law_1.2,0.7614912033081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.01,0.20275840759277344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.01,0.264352011680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.01,0.3386559963226318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.01,0.4212800025939941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.01,0.58373122215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,1,power_law_1.2,0.6825407981872559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.01,0.7787712097167969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,1,power_law_1.2,0.7985343933105469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,1,power_law_1.2,0.6002111911773682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,1,power_law_1.2,0.6902592182159424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,1,power_law_1.2,0.9048447608947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,1,power_law_1.2,1.0935168266296387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.01,0.9384639739990235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.2,3.0406591415405275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,4,power_law_1.2,1.0050047874450683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,8,power_law_1.01,0.02928000092506409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,8,power_law_1.01,0.029606398940086365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,8,power_law_1.01,0.029049599170684816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,8,power_law_1.01,0.028947201371192933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,8,power_law_1.01,0.028998398780822755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,8,power_law_1.01,0.03028480112552643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,8,power_law_1.01,0.03108479976654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,8,power_law_1.01,0.03152639865875244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,8,power_law_1.01,0.03185920119285583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,8,power_law_1.01,0.03273600041866302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,8,power_law_1.01,0.033004799485206605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,8,power_law_1.01,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,8,power_law_1.01,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,8,power_law_1.01,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,8,power_law_1.01,0.04095999896526337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,8,power_law_1.01,0.04266240000724793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,8,power_law_1.01,0.043424001336097716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,8,power_law_1.01,0.049420800805091855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,8,power_law_1.01,0.05576320290565491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,8,power_law_1.01,0.06433280110359192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,8,power_law_1.01,0.07745919823646545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,8,power_law_1.01,0.10648959875106812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,8,power_law_1.01,0.14067200422286988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,2,balanced,0.09725866715113322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,2,balanced,0.14563733339309692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,2,balanced,0.24316267172495523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,2,balanced,0.4438186486562093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,2,balanced,0.8346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,2,balanced,1.6220213572184246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,2,balanced,2.408448060353597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,8,power_law_1.01,0.18330880403518676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,2,balanced,2.4068640073140464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,2,balanced,2.419706662495931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,2,balanced,2.4217920303344727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,2,balanced,2.4019254048665366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,2,balanced,2.413834730784098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,2,balanced,2.421328067779541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,2,balanced,2.456927935282389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,2,balanced,2.4720212618509927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,2,balanced,2.479584058125814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,2,balanced,2.4803199768066406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,2,balanced,2.5210347175598145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,2,balanced,2.5344746907552085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,2,balanced,2.601360003153483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,2,balanced,2.6955998738606772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,2,balanced,2.834255854288737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,8,power_law_1.01,0.2378688097000122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,2,balanced,2.9146080017089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,8,power_law_1.01,0.2972415924072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,2,balanced,3.1373440424601235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,2,balanced,3.435685475667318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,2,balanced,3.8069547017415366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,2,balanced,5.333840052286784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,8,power_law_1.01,0.4395264148712158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,32,power_law_1.2,0.10451840162277222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,32,power_law_1.2,0.10103679895401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,32,power_law_1.2,0.10030720233917237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,32,power_law_1.2,0.10462080240249634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,32,power_law_1.2,0.10317440032958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,32,power_law_1.2,0.09409279823303222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,32,power_law_1.2,0.10175360441207885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,32,power_law_1.2,0.10553599596023559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,32,power_law_1.2,0.10581760406494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,32,power_law_1.2,0.10531200170516967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,1,balanced,0.060922667384147644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,1,balanced,0.07493333518505096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,1,balanced,0.10460799932479858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,1,balanced,0.16135467092196146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,1,balanced,0.26268800099690753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,1,balanced,0.46776533126831055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,1,balanced,0.6734826564788818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,1,balanced,0.676736036936442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,32,power_law_1.2,0.10549119710922242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,1,balanced,0.6791733105977377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,1,balanced,0.6833226680755615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,1,balanced,0.6916373570760092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,1,balanced,0.702298641204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,1,balanced,0.7091306845347086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,1,balanced,0.7142453193664551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,1,balanced,0.7251946926116943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,1,balanced,0.7390080293019613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,1,balanced,0.7530986467997233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,1,balanced,0.8039466540018717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,1,balanced,0.8462773164113363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,1,balanced,0.9265173276265463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,1,balanced,1.0071039994557698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,32,power_law_1.2,0.11429120302200317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,32,power_law_1.2,0.11160320043563843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,1,balanced,1.1823413372039795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,32,power_law_1.2,0.11580159664154052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,1,balanced,1.3991039594014485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,1,balanced,1.7421013514200847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,1,balanced,2.2493759791056314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,1,balanced,2.9205652872721353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,32,power_law_1.2,0.13185919523239137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,32,power_law_1.2,0.12923519611358641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,32,power_law_1.2,0.13563519716262817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,1,balanced,4.027722676595052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,32,power_law_1.2,0.14559999704360962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,32,power_law_1.2,0.15820159912109374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,32,power_law_1.2,0.18963199853897095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,32,power_law_1.2,0.22287359237670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,32,power_law_1.2,0.2911168098449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,32,power_law_1.2,0.3858112096786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,32,power_law_1.2,0.5087744235992432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,32,power_law_1.2,0.706060791015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,64,balanced,0.05861333509286245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,32,power_law_1.2,1.051865577697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,64,balanced,0.05823466678460439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,64,balanced,0.059024001161257424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,64,balanced,0.06198933223883311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,64,balanced,0.07266666491826375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,32,power_law_1.2,1.3241600036621093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,64,balanced,0.09259733557701111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,64,balanced,0.10094400246938069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,64,balanced,0.10117333134015401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,64,balanced,0.1032319962978363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,64,balanced,0.10089600086212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,64,balanced,0.1060693363348643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,64,balanced,0.10936533411343892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,64,balanced,0.1092639962832133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,64,balanced,0.11117333173751831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,64,balanced,0.12194666266441345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,64,balanced,0.12158399820327759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,64,balanced,0.12774399916330972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,64,balanced,0.13668266932169595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,64,balanced,0.14381866653760275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,64,balanced,0.15992533167203268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,64,balanced,0.1769333283106486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,64,balanced,0.21625600258509317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,64,balanced,0.2500693400700887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,64,balanced,0.312885324160258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,64,balanced,0.39665599664052326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,64,balanced,0.5213706493377686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,64,balanced,0.6714666684468588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,64,balanced,0.058506667613983154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,64,balanced,0.0580266664425532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,64,balanced,0.05801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,64,balanced,0.06028266747792562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,64,balanced,0.06846933563550313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,64,balanced,0.08524800340334575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,64,balanced,0.08665066957473755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,64,balanced,0.08589333295822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,64,balanced,0.08905067046483357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,64,balanced,0.08658666412035625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,64,balanced,0.08872000376383464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,64,balanced,0.09577600161234538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,64,balanced,0.09512533744176228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,64,balanced,0.09311999877293904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,64,balanced,0.10538666447003682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,64,balanced,0.10667733351389568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,64,balanced,0.10909866293271382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,64,balanced,0.11768533786137898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,64,balanced,0.12272000312805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,64,balanced,0.13609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,64,balanced,0.14937067031860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,64,balanced,0.18320000171661377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,64,balanced,0.2091253399848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,64,balanced,0.28170667092005414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,64,balanced,0.32939199606577557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,64,balanced,0.4350186586380005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,64,balanced,0.5565386613210043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,128,power_law_1.2,0.030476799607276915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,128,power_law_1.2,0.02991360127925873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,128,power_law_1.2,0.03020159900188446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,128,power_law_1.2,0.031814399361610415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,128,power_law_1.2,0.03246079981327057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,128,power_law_1.2,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,128,power_law_1.2,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,128,power_law_1.2,0.033843201398849485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,128,power_law_1.2,0.03375360071659088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,128,power_law_1.2,0.03431679904460907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,128,power_law_1.2,0.0350271999835968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,128,power_law_1.2,0.035648000240325925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,128,power_law_1.2,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,128,power_law_1.2,0.03642880022525787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,128,power_law_1.2,0.038815999031066896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,128,power_law_1.2,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,128,power_law_1.2,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,128,power_law_1.2,0.04132480025291443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,128,power_law_1.2,0.04302720129489899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,128,power_law_1.2,0.04865919947624207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,128,power_law_1.2,0.0525056004524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,128,power_law_1.2,0.0592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,128,power_law_1.2,0.07523840069770812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,1,power_law_1.2,0.02041600048542023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,1,power_law_1.2,0.020313599705696107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,1,power_law_1.2,0.02099200040102005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,1,power_law_1.2,0.021414400637149812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,1,power_law_1.2,0.02399359941482544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,1,power_law_1.2,0.02871040105819702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,1,power_law_1.2,0.027929601073265076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,1,power_law_1.2,0.028703999519348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,1,power_law_1.2,0.02752000093460083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,1,power_law_1.2,0.028863999247550964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,1,power_law_1.2,0.031097599864006044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,1,power_law_1.2,0.03059200048446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,1,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,1,power_law_1.2,0.03470720052719116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,128,power_law_1.2,0.10250240564346313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,1,power_law_1.2,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,1,power_law_1.2,0.07566720247268677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,1,power_law_1.2,0.07888640165328979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,128,power_law_1.2,0.1211135983467102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,1,power_law_1.2,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.2,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.2,0.08168960213661194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.2,0.10128639936447144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.2,0.13657599687576294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,128,power_law_1.2,0.1656448006629944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.2,0.16542719602584838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,64,power_law_1.01,0.06497920155525208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,64,power_law_1.01,0.06211199760437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,64,power_law_1.01,0.060684800148010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,64,power_law_1.01,0.06450560092926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,64,power_law_1.01,0.06083199977874756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,64,power_law_1.01,0.06438400149345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,64,power_law_1.01,0.06339840292930603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,64,power_law_1.01,0.06504960060119629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,64,power_law_1.01,0.06506239771842956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,64,power_law_1.01,0.0657920002937317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,64,power_law_1.01,0.06575359702110291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,64,power_law_1.01,0.06735360026359558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,64,power_law_1.01,0.06751999855041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,64,power_law_1.01,0.06747519969940186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,64,power_law_1.01,0.07328640222549439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,64,power_law_1.01,0.07423359751701356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,64,power_law_1.01,0.07749119997024537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,1,power_law_1.2,0.102457594871521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,64,power_law_1.01,0.08698239922523499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,1,power_law_1.2,0.11431679725646973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.01,0.09423360228538513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,1,power_law_1.2,0.13639039993286134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.01,0.11125119924545288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,1,power_law_1.2,0.17723519802093507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.01,0.13037439584732055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,1,power_law_1.2,0.21393918991088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.01,0.16551040410995482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,1,power_law_1.2,0.2545023918151855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.01,0.20359039306640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,1,power_law_1.2,0.3315711975097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.01,0.2652992010116577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,1,power_law_1.2,0.09734399914741516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,1,power_law_1.2,0.3503040075302124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.2,0.23659520149230956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.01,0.33639678955078123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,1,power_law_1.2,0.3581183910369873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.01,0.4747776031494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.01,0.6221824169158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,1,power_law_1.2,0.10965759754180908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,1,power_law_1.2,0.37183361053466796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,1,power_law_1.2,0.1284160017967224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,1,power_law_1.2,0.37460479736328123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,1,power_law_1.2,0.17180800437927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,1,power_law_1.2,0.3845184087753296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,1,power_law_1.2,0.22762880325317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,1,power_law_1.2,0.3965888023376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,1,power_law_1.2,0.285862398147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,1,power_law_1.2,0.40734081268310546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,1,power_law_1.2,0.4284095764160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,1,power_law_1.2,0.4333631992340088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,1,power_law_1.2,0.44805121421813965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,1,power_law_1.2,0.5148416042327881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.2,0.5606719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,1,power_law_1.2,0.3265919923782349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.2,0.6688320159912109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,1,power_law_1.2,0.4147776126861572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,1,power_law_1.2,0.43089919090270995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,1,power_law_1.2,0.44654078483581544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,1,power_law_1.2,0.4570303916931152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.2,0.7486015796661377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,1,power_law_1.2,0.48798718452453616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,1,power_law_1.2,0.49276161193847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,1,power_law_1.2,0.5198783874511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,1,power_law_1.2,0.5335296154022217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,1,power_law_1.2,0.5501632213592529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,1,power_law_1.2,0.5824319839477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.2,0.9559935569763184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,1,power_law_1.2,0.6375040054321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.2,1.153388786315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.2,0.7050047874450683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.2,0.8292032241821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.2,0.9370495796203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.2,1.1777728080749512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.2,1.3333375930786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.2,1.5563712120056152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.2,1.7504640579223634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.2,1.9611007690429687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,1,power_law_1.2,2.3022079467773438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.2,2.7554176330566404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,1,power_law_1.2,3.115577507019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.2,3.5426239013671874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,1,power_law_1.2,4.1047615051269535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,128,power_law_1.2,0.2250432014465332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.2,0.3033024072647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,2,balanced,0.0706879993279775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,2,balanced,0.08389332890510559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,2,balanced,0.10946666200955708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,2,balanced,0.16062399744987488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,2,balanced,0.23602133989334106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,2,balanced,0.27027734120686847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,2,balanced,0.25988266865412396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,2,balanced,0.2660213311513265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,2,balanced,0.2585973342259725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,2,balanced,0.251306672890981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,2,balanced,0.25654399394989014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,2,balanced,0.2648266752560933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,2,balanced,0.2627626657485962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,2,balanced,0.2675679922103882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,2,balanced,0.26816533009211224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,2,balanced,0.2632960081100464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,2,balanced,0.2739466627438863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,2,balanced,0.29876800378163654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,2,balanced,0.30869332949320477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,2,balanced,0.34386134147644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,2,balanced,0.3778560161590576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,2,balanced,0.5981599887212118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,2,balanced,0.6534613370895386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,2,balanced,0.9204426606496176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,2,balanced,1.211301326751709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,2,balanced,1.5832427342732747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,2,balanced,2.1363253593444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.2,0.43752322196960447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,1,power_law_1.2,0.024831999838352204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,1,power_law_1.2,0.04408319890499115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,1,power_law_1.2,0.06695680022239685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,1,power_law_1.2,0.096697598695755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,1,power_law_1.2,0.1358847975730896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,1,power_law_1.2,0.19254399538040162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,1,power_law_1.2,0.2587712049484253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,1,power_law_1.2,0.2855168104171753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,1,power_law_1.2,0.29709439277648925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,1,power_law_1.2,0.2928960084915161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,1,power_law_1.2,0.3167167901992798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,1,power_law_1.2,0.3282367944717407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,1,power_law_1.2,0.33280000686645506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,1,power_law_1.2,0.3424896001815796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,1,power_law_1.2,0.34022400379180906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,1,power_law_1.2,0.3667072057723999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,1,power_law_1.2,0.35196800231933595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.2,0.5730879783630372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,1,power_law_1.2,0.42227840423583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,1,power_law_1.2,0.44021120071411135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,1,power_law_1.2,0.5204864025115967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,1,power_law_1.2,0.6010240077972412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,1,power_law_1.2,0.7801023960113526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,1,power_law_1.2,0.7162879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,1,power_law_1.2,0.9105664253234863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,1,power_law_1.2,1.140505599975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,1,power_law_1.2,1.5722623825073243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,1,power_law_1.2,1.9688640594482423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,2,power_law_1.2,0.07914239764213563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,2,power_law_1.2,0.093612802028656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,2,power_law_1.2,0.10089600086212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,2,power_law_1.2,0.12257920503616333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,2,power_law_1.2,0.1522112011909485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,2,power_law_1.2,0.17598719596862794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,2,power_law_1.2,0.20808959007263184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,2,power_law_1.2,0.21057920455932616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,2,power_law_1.2,0.21857280731201173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,2,power_law_1.2,0.21560320854187012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,2,power_law_1.2,0.22136321067810058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,2,power_law_1.2,0.2319808006286621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,2,power_law_1.2,0.23274240493774415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,2,power_law_1.2,0.23118720054626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,2,power_law_1.2,0.2511679887771606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,2,power_law_1.2,0.254035210609436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,2,power_law_1.2,0.2645247936248779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,2,power_law_1.2,0.2998719930648804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,2,power_law_1.2,0.32359039783477783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,2,power_law_1.2,0.3691776037216187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,2,power_law_1.2,0.4201344013214111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,2,power_law_1.2,0.5619455814361572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,2,power_law_1.2,0.6540351867675781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,2,balanced,0.04201066493988037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,2,balanced,0.042549331982930504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,2,power_law_1.2,0.8733951568603515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,2,balanced,0.04785599807898203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,2,balanced,0.06018666426340739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,2,balanced,0.09089066584904988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,2,balanced,0.09366933504740398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,2,balanced,0.09321066737174988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,2,balanced,0.09479999542236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,2,balanced,0.09501866499582927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,2,balanced,0.09763733545939128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,2,balanced,0.09730133414268494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,2,balanced,0.09942932923634847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,2,balanced,0.10246400038401286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,2,balanced,0.10419733325640361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,2,balanced,0.10845333337783813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,2,balanced,0.11275200049082439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,2,balanced,0.11829333504041036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,2,balanced,0.1343839963277181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,2,balanced,0.14858667055765787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,2,balanced,0.17698132991790771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,2,balanced,0.20336000124613443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,2,balanced,0.29224000374476117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,2,power_law_1.2,1.1187904357910157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,2,balanced,0.35235198338826496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,2,balanced,0.5058720111846924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,2,balanced,0.6507413387298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,2,balanced,0.9444320201873779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,2,balanced,1.2418560187021892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,2,power_law_1.2,1.467244815826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,2,power_law_1.2,2.0015743255615233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,64,balanced,0.056554665168126426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,64,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,64,balanced,0.044591998060544334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,64,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,64,balanced,0.050479998191197716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,64,balanced,0.07351999978224437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,64,balanced,0.07498666644096375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,64,balanced,0.07454399764537811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,64,balanced,0.07421333094437917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,64,balanced,0.07621333499749501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,64,balanced,0.075914666056633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,64,balanced,0.07666133344173431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,64,balanced,0.07806933422883351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,64,balanced,0.08077866832415263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,64,balanced,0.08343999584515889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,64,balanced,0.08454400300979614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,32,power_law_1.2,0.14419840574264525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,64,balanced,0.08682666222254436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,64,balanced,0.09200533231099446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,64,balanced,0.09508267045021057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,32,power_law_1.2,0.14341119527816773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,64,balanced,0.10499733686447144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,64,balanced,0.11496532956759135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,32,power_law_1.2,0.14024959802627562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,32,power_law_1.2,0.14497920274734497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,32,power_law_1.2,0.13656320571899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,32,power_law_1.2,0.12168960571289063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,32,power_law_1.2,0.13185280561447144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,32,power_law_1.2,0.14264960289001466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,64,balanced,0.1312373379866282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,32,power_law_1.2,0.14648319482803346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,32,power_law_1.2,0.14926079511642457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,32,power_law_1.2,0.13870079517364503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,32,power_law_1.2,0.1462656021118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,32,power_law_1.2,0.14054399728775024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,32,power_law_1.2,0.14751360416412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,64,balanced,0.15150933464368185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,64,balanced,0.18837867180506387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,64,balanced,0.2267626722653707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,64,balanced,0.3049973249435425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,64,balanced,0.3684693177541097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,32,power_law_1.2,0.15205760002136232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,32,power_law_1.2,0.16444799900054932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,32,power_law_1.2,0.1844607949256897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,32,power_law_1.2,0.21659519672393798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,32,power_law_1.2,0.2037247896194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,32,power_law_1.2,0.24374399185180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,32,power_law_1.2,0.26302080154418944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,32,power_law_1.2,0.3392319917678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,32,power_law_1.2,0.43160319328308105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,32,power_law_1.2,0.5666495800018311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,32,power_law_1.2,0.7173056125640869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,32,power_law_1.2,0.9700991630554199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,32,power_law_1.2,1.1852736473083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,4,power_law_1.01,0.10208640098571778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,4,power_law_1.01,0.11105279922485352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,4,power_law_1.01,0.12339199781417846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,4,power_law_1.01,0.12643840312957763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,4,power_law_1.01,0.12675199508666993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,4,power_law_1.01,0.13192319869995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,4,power_law_1.01,0.139628803730011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,4,power_law_1.01,0.13968640565872192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,4,power_law_1.01,0.15946240425109864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,4,power_law_1.01,0.150764799118042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,4,power_law_1.01,0.15362559556961058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,4,power_law_1.01,0.16722559928894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,4,power_law_1.01,0.17235200405120848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,4,power_law_1.01,0.17148159742355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,4,power_law_1.01,0.1978495955467224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,4,power_law_1.01,0.2118783950805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,4,power_law_1.01,0.23752319812774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,4,power_law_1.01,0.2747391939163208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,4,power_law_1.01,0.32817280292510986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,4,power_law_1.01,0.469158411026001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,4,power_law_1.01,0.5237631797790527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,4,power_law_1.01,0.8116928100585937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,4,power_law_1.01,0.981107234954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,4,power_law_1.01,1.3397631645202637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,4,power_law_1.01,1.9213760375976563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,128,balanced,0.04846400022506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,128,balanced,0.04852266609668732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,128,balanced,0.04781333108743032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,128,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,128,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,128,balanced,0.05829866727193197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,128,balanced,0.05811200042565664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,128,balanced,0.05994133154551188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,128,balanced,0.05806933343410492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,128,balanced,0.0601440022389094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,128,balanced,0.06050133208433787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,128,balanced,0.06426133215427399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,128,balanced,0.06425599753856659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,128,balanced,0.0660159985224406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,128,balanced,0.07285333176453908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,128,balanced,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,128,balanced,0.08210133512814839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,128,balanced,0.10188800096511841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,128,balanced,0.10547199845314026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,128,balanced,0.13190933068593344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,128,balanced,0.14678933223088583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,128,balanced,0.1926506757736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,128,balanced,0.19347200791041055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,128,balanced,0.2490346630414327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,128,balanced,0.28153600295384723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,128,balanced,0.37753065427144367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,128,balanced,0.46564801534016925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,4,power_law_1.01,2.9548479080200196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,4,power_law_1.01,3.7528575897216796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,1,power_law_1.2,0.01892479956150055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,1,power_law_1.2,0.019385600090026857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,1,power_law_1.2,0.01907840073108673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,1,power_law_1.2,0.019840000569820403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,1,power_law_1.2,0.02120320051908493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,1,power_law_1.2,0.02328319996595383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,1,power_law_1.2,0.023129600286483764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,1,power_law_1.2,0.023852799832820893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,1,power_law_1.2,0.023948800563812257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,1,power_law_1.2,0.02449280023574829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,1,power_law_1.2,0.02600319981575012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,1,power_law_1.2,0.026553601026535034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,1,power_law_1.2,0.02686080038547516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,1,power_law_1.2,0.028703999519348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,1,power_law_1.2,0.05192959904670715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,1,power_law_1.2,0.05356159806251526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,1,power_law_1.2,0.055257600545883176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,128,power_law_1.2,0.03953920006752014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,128,power_law_1.2,0.03882879912853241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,128,power_law_1.2,0.03920640051364899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,128,power_law_1.2,0.041075199842453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,128,power_law_1.2,0.04211840033531189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,1,power_law_1.2,0.0456959992647171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,128,power_law_1.2,0.043609601259231565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,128,power_law_1.2,0.04234879910945892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,128,power_law_1.2,0.04140160083770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,128,power_law_1.2,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,128,power_law_1.2,0.043033599853515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,128,power_law_1.2,0.04222080111503601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,128,power_law_1.2,0.04240640103816986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,128,power_law_1.2,0.04174720048904419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,128,power_law_1.2,0.04195840060710907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.2,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,128,power_law_1.2,0.0487744003534317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,128,power_law_1.2,0.04860160052776337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,128,power_law_1.2,0.049779200553894044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,128,power_law_1.2,0.055404800176620486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.2,0.06369280219078063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.2,0.07714560031890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.2,0.09160959720611572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.2,0.11703039407730102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.2,0.06780800223350525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.2,0.14750720262527467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.2,0.07643520236015319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.2,0.09948800206184387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,1,power_law_1.2,0.1831936001777649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,1,power_law_1.2,0.24816639423370362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,1,power_law_1.2,0.3280767917633057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,1,power_law_1.2,0.5059455871582031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,1,power_law_1.2,0.50381441116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,1,power_law_1.2,0.5113471984863281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,1,power_law_1.2,0.5109312057495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,1,power_law_1.2,0.5146431922912598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,1,power_law_1.2,0.4774208068847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,1,power_law_1.2,0.4890495777130127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,1,power_law_1.2,0.4938623905181885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,1,power_law_1.2,0.5195839881896973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.2,0.12414079904556274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,1,power_law_1.2,0.5117631912231445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,1,power_law_1.2,0.5493375778198242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,1,power_law_1.2,0.5968768119812011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,1,power_law_1.2,0.5766528129577637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,1,power_law_1.2,0.6288896083831788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.2,0.17209600210189818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,1,power_law_1.2,0.7306431770324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.2,0.21150081157684325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.2,0.8643391609191895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.2,1.099788761138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.2,0.26728320121765137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.2,1.3440704345703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.2,1.7869504928588866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.2,0.21031041145324708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.2,0.3707456111907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.2,2.268716812133789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.2,3.2574142456054687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.2,0.49989118576049807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.2,4.163187026977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.2,0.3076159954071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,2,power_law_1.2,0.042131200432777405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,2,power_law_1.2,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,2,power_law_1.2,0.04769920110702515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,2,power_law_1.2,0.05472639799118042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,2,power_law_1.2,0.05958399772644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,2,power_law_1.2,0.0754047989845276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,2,power_law_1.2,0.07847679853439331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,2,power_law_1.2,0.07898240089416504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,2,power_law_1.2,0.08118399977684021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,2,power_law_1.2,0.08268799781799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,2,power_law_1.2,0.08478080034255982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,2,power_law_1.2,0.0871936023235321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,2,power_law_1.2,0.08952959775924682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,2,power_law_1.2,0.09314560294151306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,2,power_law_1.2,0.10014079809188843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,2,power_law_1.2,0.10755200386047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,2,power_law_1.2,0.11544959545135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,2,power_law_1.2,0.13082239627838135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,2,power_law_1.2,0.1504320025444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,2,power_law_1.2,0.1857599973678589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,2,power_law_1.2,0.21996800899505614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,2,power_law_1.2,0.3317823886871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,2,power_law_1.2,0.37782399654388427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,2,power_law_1.2,0.5775360107421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.2,6.06959342956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,8,balanced,0.052341332038243614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,8,balanced,0.07137066622575124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,8,balanced,0.09062400460243225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,8,balanced,0.14192533493041992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,8,balanced,0.20844266812006632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,8,balanced,0.2582293351491292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,8,balanced,0.26146666208902997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,8,balanced,0.2627466718355815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,8,balanced,0.26266666253407794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,8,balanced,0.26451200246810913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,8,balanced,0.2630026737848918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,8,balanced,0.2664639949798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,8,balanced,0.2687573234240214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,8,balanced,0.27190933624903363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,8,balanced,0.2780906756718953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,8,balanced,0.28009599447250366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,8,balanced,0.28732800483703613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,8,balanced,0.3481280008951823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,8,balanced,0.31854933500289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,8,balanced,0.4021386702855428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,2,power_law_1.2,0.68853759765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,8,balanced,0.38490132490793866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,8,balanced,0.6157439947128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,8,balanced,0.4888853232065837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,8,balanced,0.7466773192087809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,8,balanced,0.7498613198598226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,8,balanced,1.1143306891123455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,8,balanced,1.2082293033599854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,4,power_law_1.2,0.06310399770736694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,4,power_law_1.2,0.07027199864387512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.2,0.3947200059890747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,4,power_law_1.2,0.07866880297660828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,4,power_law_1.2,0.08946560025215149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,4,power_law_1.2,0.1076159954071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,4,power_law_1.2,0.11610239744186401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,4,power_law_1.2,0.11324160099029541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,4,power_law_1.2,0.14762879610061647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,4,power_law_1.2,0.1463039994239807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,4,power_law_1.2,0.14650239944458007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,4,power_law_1.2,0.14724479913711547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,4,power_law_1.2,0.14450559616088868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,4,power_law_1.2,0.1443519949913025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,4,power_law_1.2,0.14801280498504638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,4,power_law_1.2,0.1640128016471863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,4,power_law_1.2,0.16090879440307618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,4,power_law_1.2,0.17469439506530762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,4,power_law_1.2,0.19340159893035888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,4,power_law_1.2,0.20940799713134767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,4,power_law_1.2,0.24916479587554932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,4,power_law_1.2,0.29148800373077394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,4,power_law_1.2,0.37344000339508054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,4,power_law_1.2,0.4462399959564209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,4,power_law_1.2,0.5960192203521728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,4,power_law_1.2,0.8146304130554199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,4,power_law_1.2,1.2065152168273925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,4,power_law_1.2,1.4001472473144532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,2,power_law_1.2,1.0217535972595215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.2,7.977049255371094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,2,power_law_1.2,1.2774080276489257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,8,balanced,0.05580799778302511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,8,balanced,0.05607999861240387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,8,balanced,0.055455997586250305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,8,balanced,0.05682666599750519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,8,balanced,0.05886933207511902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,8,balanced,0.05986666679382324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,8,balanced,0.05826666454474131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,8,balanced,0.060549333691596985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,8,balanced,0.06132799883683523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,8,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,8,balanced,0.06190933287143707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,8,balanced,0.062234664956728615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,8,balanced,0.06426133215427399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,8,balanced,0.06570133566856384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,8,balanced,0.07180800040562947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,8,balanced,0.07344000041484833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,8,balanced,0.07594666878382365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,8,balanced,0.08060800035794576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,8,balanced,0.09092266360918681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,8,balanced,0.10460799932479858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,8,balanced,0.1076693336168925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,8,balanced,0.13779200116793314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,8,balanced,0.1614720026652018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,8,balanced,0.21305600802103677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,8,balanced,0.2680799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,8,balanced,0.3643360137939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,8,balanced,0.47068798542022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,8,balanced,0.03316266586383184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,8,balanced,0.02924799919128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,8,balanced,0.03049066662788391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,8,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,8,balanced,0.04631466666857401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,8,balanced,0.0580320010582606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,8,balanced,0.05826666454474131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,8,balanced,0.056885331869125366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,8,balanced,0.057962665955225624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,8,balanced,0.058101331194241844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,8,balanced,0.05816533168156942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,8,balanced,0.05878399809201559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,8,balanced,0.05826666454474131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,8,balanced,0.059978668888409935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,8,balanced,0.06432533264160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,8,balanced,0.06338133414586385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,8,balanced,0.06396799782911937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,8,balanced,0.06673066814740498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,8,balanced,0.06851733227570851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,8,balanced,0.07816533247629802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,8,balanced,0.07840533554553986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,8,balanced,0.09273067116737366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,8,balanced,0.0972213347752889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,8,balanced,0.13565867145856222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,8,balanced,0.14666666587193808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,8,balanced,0.18710933128992716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,8,balanced,0.21355199813842773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,1,power_law_1.01,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,1,power_law_1.01,0.03873920142650604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,1,power_law_1.01,0.04628480076789856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,1,power_law_1.01,0.05653759837150574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,1,power_law_1.01,0.05697280168533325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,1,power_law_1.01,0.05824000239372253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,1,power_law_1.01,0.05761280059814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,1,power_law_1.01,0.05928320288658142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,1,power_law_1.01,0.05902079939842224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,1,power_law_1.01,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,1,power_law_1.01,0.06100479960441589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,1,power_law_1.01,0.06584960222244263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,1,power_law_1.01,0.06803839802742004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,1,power_law_1.01,0.07333760261535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,1,power_law_1.01,0.08236160278320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,1,power_law_1.01,0.08986240029335021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,1,power_law_1.01,0.08118399977684021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,1,power_law_1.01,0.09505280256271362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,1,power_law_1.01,0.1001471996307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,1,power_law_1.01,0.12394880056381226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,1,power_law_1.01,0.14840960502624512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,1,power_law_1.01,0.18862719535827638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,1,power_law_1.01,0.22879359722137452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,1,power_law_1.01,0.30616960525512693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,1,power_law_1.01,0.37636480331420896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,1,power_law_1.01,0.5308032035827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,1,power_law_1.2,0.040454399585723874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,1,power_law_1.2,0.04796159863471985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,1,power_law_1.2,0.05847039818763733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,1,power_law_1.2,0.06727679967880248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,1,power_law_1.2,0.06955519914627076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,1,power_law_1.2,0.0705024003982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,1,power_law_1.2,0.07128959894180298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,1,power_law_1.2,0.07130240201950074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,1,power_law_1.2,0.07182080149650574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,1,power_law_1.2,0.07260159850120544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,1,power_law_1.2,0.07402880191802978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,1,power_law_1.2,0.0785535991191864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,1,power_law_1.2,0.07945600152015686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,1,power_law_1.2,0.08412160277366638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,1,power_law_1.2,0.09255040287971497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,1,power_law_1.2,0.0955456018447876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,1,power_law_1.2,0.10243840217590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,1,power_law_1.2,0.11807359457015991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.2,0.13270399570465088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,1,power_law_1.01,0.682528018951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,4,power_law_1.2,0.01855359971523285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,4,power_law_1.2,0.019571200013160706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,4,power_law_1.2,0.021817600727081297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,4,power_law_1.2,0.022598400712013245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.2,0.1638335943222046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,4,power_law_1.2,0.024684800207614897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,4,power_law_1.2,0.026335999369621277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,4,power_law_1.2,0.027724799513816834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,4,power_law_1.2,0.02871040105819702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,4,power_law_1.2,0.03258880078792572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,4,power_law_1.2,0.03224959969520569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,4,power_law_1.2,0.03235200047492981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,4,power_law_1.2,0.04069119989871979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,4,power_law_1.2,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,8,balanced,0.041834667325019836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,8,balanced,0.04567466676235199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,4,power_law_1.2,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,8,balanced,0.04863466819127401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,8,balanced,0.05907199780146281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,4,power_law_1.2,0.058956801891326904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,4,power_law_1.2,0.06040319800376892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.2,0.20188159942626954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,4,power_law_1.2,0.06383360028266907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,4,power_law_1.2,0.07095680236816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,4,power_law_1.2,0.06429439783096313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,4,power_law_1.2,0.07246080040931702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,4,power_law_1.2,0.07866880297660828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,8,balanced,0.07964266836643219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,8,balanced,0.11315733194351196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.2,0.26300160884857177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,4,power_law_1.2,0.09393280148506164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,4,power_law_1.2,0.10706559419631959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,8,balanced,0.1150933305422465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,8,balanced,0.1163200040658315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,8,balanced,0.11538133025169373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,8,balanced,0.11629866560300191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,8,balanced,0.11871467034022014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,8,balanced,0.11853333314259847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,8,balanced,0.11952533324559529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,4,power_law_1.2,0.1494271993637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,8,balanced,0.12156800429026286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,8,balanced,0.12909332911173502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,8,balanced,0.13013866543769836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,8,balanced,0.13318399588267008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,8,balanced,0.14255467057228088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,8,balanced,0.14969066778818765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,8,balanced,0.16679465770721436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,8,balanced,0.18407466014226279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,8,balanced,0.22179200251897177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,8,balanced,0.2498133381207784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,8,balanced,0.3267199993133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,4,power_law_1.2,0.1825279951095581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.2,0.32442240715026854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,8,balanced,0.39971200625101727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,8,balanced,0.5592586596806844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,8,balanced,0.6915573279062907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.2,0.44625282287597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,4,power_law_1.2,0.24206080436706542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.2,0.5643712043762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,4,power_law_1.2,0.3246527910232544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,1,power_law_1.01,0.02025599926710129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,1,power_law_1.01,0.02375040054321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,1,power_law_1.01,0.03255679905414581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,1,power_law_1.01,0.04560000002384186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,1,power_law_1.01,0.05984640121459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,1,power_law_1.01,0.07882239818572997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,1,power_law_1.01,0.08646399974822998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,1,power_law_1.01,0.08880640268325805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,1,power_law_1.01,0.09477120041847228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,1,power_law_1.01,0.09761279821395874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,1,power_law_1.01,0.09869440197944641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,1,power_law_1.01,0.10558079481124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,1,power_law_1.01,0.10718079805374145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,1,power_law_1.01,0.10899840593338013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,1,power_law_1.01,0.11518720388412476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,1,power_law_1.01,0.11953920125961304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,1,power_law_1.01,0.12996480464935303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,1,power_law_1.01,0.1538815975189209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.01,0.18120319843292237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.01,0.2405247926712036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.01,0.22426879405975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.01,0.29605119228363036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.01,0.3478528022766113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.01,0.44200959205627444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.01,0.5369919776916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.01,0.7155839920043945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.01,0.9127424240112305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.2,0.8084287643432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,16,balanced,0.050554667909940086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,16,balanced,0.05100266635417938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,16,balanced,0.05161599814891815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,16,balanced,0.05690133571624756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,16,balanced,0.06488533318042755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,16,balanced,0.08959999680519104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,16,balanced,0.10594666997591655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,16,balanced,0.10742933551470439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,16,balanced,0.10782399773597717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,16,balanced,0.11061867078145345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,16,balanced,0.1092693308989207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,16,balanced,0.11212266484896342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,16,balanced,0.11316800117492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,16,balanced,0.11364799737930298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,16,balanced,0.12071466445922852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,16,balanced,0.12175466616948445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,16,balanced,0.12828266620635986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,16,balanced,0.13760000467300415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,16,balanced,0.14601600170135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,16,balanced,0.16476266582806906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,16,balanced,0.18263467152913412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,16,balanced,0.22236265738805136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,16,balanced,0.25571733713150024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,16,balanced,0.3264906605084737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,16,balanced,0.4220053354899089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,16,balanced,0.5534079869588217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,16,balanced,0.7176799774169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,32,power_law_1.2,0.025177600979804992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,32,power_law_1.2,0.024607999622821806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,32,power_law_1.2,0.02439039945602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,32,power_law_1.2,0.02393600046634674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,32,power_law_1.2,0.024320000410079957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,32,power_law_1.2,0.027852800488471986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,32,power_law_1.2,0.036313599348068236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,32,power_law_1.2,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,32,power_law_1.2,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,32,power_law_1.2,0.05245440006256104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,32,power_law_1.2,0.05226879715919495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,32,power_law_1.2,0.05631359815597534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,32,power_law_1.2,0.05370879769325256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,32,power_law_1.2,0.05591679811477661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,32,power_law_1.2,0.052902400493621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,32,power_law_1.2,0.05252479910850525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,32,power_law_1.2,0.05596799850463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,32,power_law_1.2,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,32,power_law_1.2,0.05432320237159729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,32,power_law_1.2,0.06538879871368408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,32,power_law_1.2,0.07844480276107788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,32,power_law_1.2,0.09656959772109985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,32,power_law_1.2,0.11372159719467163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,32,power_law_1.2,0.14926079511642457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,32,power_law_1.2,0.19090559482574462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,32,power_law_1.2,0.28296959400177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,32,power_law_1.2,0.3673151969909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.2,1.0828672409057618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,16,power_law_1.01,0.08620799779891967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,16,power_law_1.01,0.08943359851837158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,16,power_law_1.01,0.08772479891777038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,16,power_law_1.01,0.09057919979095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,16,power_law_1.01,0.08611840009689331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,16,power_law_1.01,0.09344639778137206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,16,power_law_1.01,0.09471359848976135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,16,power_law_1.01,0.09296000003814697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,16,power_law_1.01,0.09743360280990601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,16,power_law_1.01,0.09681919813156128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,16,power_law_1.01,0.09655680060386658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,16,power_law_1.01,0.09781759977340698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,16,power_law_1.01,0.09924479722976684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,16,power_law_1.01,0.10248960256576538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,16,power_law_1.01,0.10962560176849365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,16,power_law_1.01,0.11139199733734131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,16,power_law_1.01,0.12492799758911133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,16,power_law_1.01,0.14051200151443483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,16,power_law_1.01,0.15119999647140503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,16,power_law_1.01,0.1728384017944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,16,power_law_1.01,0.20364799499511718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,16,power_law_1.01,0.2585344076156616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,16,power_law_1.01,0.3385215997695923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,16,power_law_1.01,0.48618879318237307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,16,power_law_1.01,0.6042111873626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,16,power_law_1.01,0.8981056213378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,16,power_law_1.01,1.2116415977478028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,128,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,128,balanced,0.046666666865348816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,128,balanced,0.04772266745567322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,128,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,128,balanced,0.04985600213209788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,128,balanced,0.05595199763774872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,128,balanced,0.0621066689491272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,128,balanced,0.06223999957243601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,128,balanced,0.0666133314371109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,128,balanced,0.06681600213050842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,128,balanced,0.06478400031725566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,128,balanced,0.06826133529345195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,128,balanced,0.07276799778143565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,128,balanced,0.06824000179767609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,128,balanced,0.08398399750391643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,128,balanced,0.0844586690266927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,128,balanced,0.08752000331878662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,128,balanced,0.09558932979901631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,128,balanced,0.10299733281135559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,128,balanced,0.11585600177447002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,128,balanced,0.12822399536768594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,128,balanced,0.16109333435694376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,128,balanced,0.18532266219456991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,128,balanced,0.23663467168807983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,128,balanced,0.28563199440638226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,128,balanced,0.38787734508514404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,128,balanced,0.497658650080363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,16,power_law_1.01,0.041152000427246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,16,power_law_1.01,0.046758401393890384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,16,power_law_1.01,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,16,power_law_1.01,0.046060800552368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,16,power_law_1.01,0.04559360146522522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,16,power_law_1.01,0.04591360092163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,16,power_law_1.01,0.04645119905471802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,16,power_law_1.01,0.04732159972190857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,16,power_law_1.01,0.04785279929637909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,16,power_law_1.01,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,16,power_law_1.01,0.05008640289306641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,16,power_law_1.01,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,16,power_law_1.01,0.05633280277252197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,16,power_law_1.01,0.05711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,16,power_law_1.01,0.06398079991340637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,16,power_law_1.01,0.06197119951248169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,16,power_law_1.01,0.06551039814949036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,16,power_law_1.01,0.07714560031890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,16,power_law_1.01,0.07784960269927979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,16,power_law_1.01,0.09194239974021912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,16,power_law_1.01,0.10300159454345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,16,power_law_1.01,0.12792960405349732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,16,power_law_1.01,0.15371520519256593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,16,power_law_1.01,0.2138751983642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,16,power_law_1.01,0.2454591989517212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,16,power_law_1.01,0.34302079677581787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,16,power_law_1.01,0.45949440002441405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,8,power_law_1.01,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,8,power_law_1.01,0.03696640133857727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,8,power_law_1.01,0.03727999925613403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,8,power_law_1.01,0.03944959938526153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,8,power_law_1.01,0.04033919870853424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,8,power_law_1.01,0.04056319892406464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,8,power_law_1.01,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,8,power_law_1.01,0.0409280002117157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,8,power_law_1.01,0.041152000427246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,8,power_law_1.01,0.04118399918079376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,8,power_law_1.01,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,8,power_law_1.01,0.04311679899692535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,8,power_law_1.01,0.042777600884437564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,8,power_law_1.01,0.043731200695037845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,8,power_law_1.01,0.05015680193901062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,8,power_law_1.01,0.050457602739334105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,8,power_law_1.01,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,8,power_law_1.01,0.05841919779777527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,8,power_law_1.01,0.06592640280723572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,8,power_law_1.01,0.08635519742965699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,8,power_law_1.01,0.09917439818382263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,8,power_law_1.01,0.13653119802474975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,8,power_law_1.01,0.17496960163116454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,8,power_law_1.01,0.2473599910736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,8,power_law_1.01,0.318393611907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,8,balanced,0.039936001102129616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,8,balanced,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,8,balanced,0.04215466479460398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,8,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,8,balanced,0.059952000776926674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,8,balanced,0.08338666955629985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,8,balanced,0.08291199803352356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,8,balanced,0.08001066744327545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,8,balanced,0.0790880024433136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,8,balanced,0.08042133351167043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,8,balanced,0.07941333452860515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,8,balanced,0.08250133196512859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,8,balanced,0.08196266492207845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,8,balanced,0.08379200100898743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,8,balanced,0.08814932902654012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,8,balanced,0.09044800202051799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,8,balanced,0.09492266178131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,8,balanced,0.10290666421254475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,8,balanced,0.10589333375295003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,8,balanced,0.1204159955183665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,8,balanced,0.12984533111254373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,8,balanced,0.15808533628781637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,8,balanced,0.18383467197418213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,8,balanced,0.23391467332839966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,8,balanced,0.28748265902201336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,8,balanced,0.387717326482137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,8,balanced,0.48843733469645184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,8,power_law_1.01,0.45690240859985354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,8,power_law_1.01,0.6048575878143311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,4,balanced,0.09849599997202556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,4,balanced,0.07857066889603932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,4,balanced,0.1291093329588572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,4,balanced,0.1206826666990916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,4,balanced,0.11953600247701009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,4,balanced,0.12219732999801636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,4,balanced,0.12197333574295044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,4,balanced,0.12537599603335062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,4,balanced,0.12577066818873087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,4,balanced,0.12735467155774435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,4,balanced,0.12746133406956991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,4,balanced,0.13620266318321228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,4,balanced,0.13622933626174927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,4,balanced,0.13453867038091025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,4,balanced,0.15056000153223673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,4,balanced,0.1495306690533956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,4,balanced,0.14618133505185446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,4,balanced,0.1918720006942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,4,balanced,0.18001067638397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,4,balanced,0.24659200509389242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,4,balanced,0.26321067412694293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,4,balanced,0.348474661509196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,4,balanced,0.4077119827270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,4,balanced,0.57478400071462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,4,balanced,0.7582879861195883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,4,balanced,1.0768746534983318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,4,balanced,1.4391786257425945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,4,balanced,0.04821866750717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,4,balanced,0.053904001911481224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,4,balanced,0.0682239979505539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,4,balanced,0.10040000081062317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,4,balanced,0.15110400319099426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,4,balanced,0.24874132871627808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,4,balanced,0.3487360080083211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,4,balanced,0.34835731983184814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,4,balanced,0.34753600756327313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,4,balanced,0.35100265343983966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,4,balanced,0.3517119884490967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,4,balanced,0.35397334893544513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,4,balanced,0.35918935139973956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,4,balanced,0.3636159896850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,4,balanced,0.37138664722442627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,4,balanced,0.372538685798645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,4,balanced,0.3824640115102132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,4,balanced,0.397055983543396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,4,balanced,0.41120533148447674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,4,balanced,0.4431893428166707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,4,balanced,0.46672534942626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,4,balanced,0.5355786482493082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,4,balanced,0.5901439984639486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,4,balanced,0.7009867032368978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,4,balanced,0.8671466509501139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,4,balanced,1.1186827023824055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,4,balanced,1.438053290049235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,2,power_law_1.01,0.08554880023002624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,2,power_law_1.01,0.11260160207748413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,2,power_law_1.01,0.12222720384597778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,2,power_law_1.01,0.15657600164413452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,2,power_law_1.01,0.18434560298919678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,2,power_law_1.01,0.2422271966934204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,2,power_law_1.01,0.24437119960784912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,2,power_law_1.01,0.2519295930862427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,2,power_law_1.01,0.24374399185180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,2,power_law_1.01,0.24942080974578856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,2,power_law_1.01,0.254751992225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,2,power_law_1.01,0.24946560859680175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,2,power_law_1.01,0.2558079957962036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,2,power_law_1.01,0.25966720581054686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,2,power_law_1.01,0.2733247995376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,2,power_law_1.01,0.2745856046676636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,2,power_law_1.01,0.2884160041809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,2,power_law_1.01,0.3176192045211792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.01,0.34459519386291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.01,0.4104447841644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.01,0.474348783493042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.01,0.6229184150695801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.01,0.712281608581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.01,1.020684814453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.01,1.347231960296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.01,1.8586496353149413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.01,2.2761407852172852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,1,balanced,0.08518933256467183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,1,balanced,0.11999467015266418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,1,balanced,0.1890933314959208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,1,balanced,0.3351093530654907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,1,balanced,0.601744016011556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,1,balanced,0.86516801516215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,1,balanced,0.8678507010142008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,1,balanced,0.870848019917806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,1,balanced,0.8817439874013265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,1,balanced,0.8821173508961996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,1,balanced,0.8928213119506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,1,balanced,0.9043733278910319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,1,balanced,0.9108959833780924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,1,balanced,0.9229013125101725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,1,balanced,0.9378453095753988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,1,balanced,0.9518453280131022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,1,balanced,0.9808106422424316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,1,balanced,1.0460373560587566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,1,balanced,1.0991466840108235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,1,balanced,1.4360586802164714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,1,balanced,1.3445013364156086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,1,balanced,1.8105227152506511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,1,balanced,1.8815306027730305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,8,balanced,0.029909332593282063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,8,balanced,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,8,balanced,0.031557333966096245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,1,balanced,2.824415842692057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,1,balanced,3.10918394724528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,8,balanced,0.0356480007370313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,8,balanced,0.04012266546487808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,1,balanced,4.951338768005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,1,balanced,5.630181630452474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,8,balanced,0.03993066648642222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,8,balanced,0.04171733558177948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,8,balanced,0.04190933207670847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,8,balanced,0.042463997999827065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,8,balanced,0.0415040006240209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,8,balanced,0.04364266494909922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,8,balanced,0.04489600161711375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,8,balanced,0.04438933233420054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,8,balanced,0.046944002310434975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,8,balanced,0.052101333936055504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,8,balanced,0.054042667150497437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,8,balanced,0.05211733281612396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,8,balanced,0.05830933153629303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,8,balanced,0.07649066547552745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,8,balanced,0.09158933162689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,8,balanced,0.09890133142471313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,8,balanced,0.12680000066757202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,8,balanced,0.15878933668136597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,8,balanced,0.20454400777816772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,8,balanced,0.25650133689244586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,8,balanced,0.3599253495534261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,8,balanced,0.4596266746520996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,16,power_law_1.01,0.0809664011001587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,16,power_law_1.01,0.09773439764976502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,16,power_law_1.01,0.09749760031700135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,16,power_law_1.01,0.10377600193023681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,16,power_law_1.01,0.10737919807434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,16,power_law_1.01,0.09253119826316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,16,power_law_1.01,0.09760000109672547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,16,power_law_1.01,0.10146559476852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,16,power_law_1.01,0.09982720017433167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,16,power_law_1.01,0.10153599977493286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,16,power_law_1.01,0.10462080240249634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,16,power_law_1.01,0.10586880445480347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,16,power_law_1.01,0.11017600297927857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,16,power_law_1.01,0.1106495976448059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,16,power_law_1.01,0.1191167950630188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,16,power_law_1.01,0.12424960136413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,16,power_law_1.01,0.12471040487289428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,16,power_law_1.01,0.14485759735107423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,16,power_law_1.01,0.15661439895629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,16,power_law_1.01,0.2024384021759033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,16,power_law_1.01,0.22772479057312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,16,power_law_1.01,0.29745919704437257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,16,power_law_1.01,0.303276801109314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,16,power_law_1.01,0.4009535789489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,16,power_law_1.01,0.5221951961517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,16,power_law_1.01,0.6539904117584229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,16,power_law_1.01,0.8361215591430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,2,power_law_1.01,0.048825600743293764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,2,power_law_1.01,0.051635199785232545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,2,power_law_1.01,0.05856000185012818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,2,power_law_1.01,0.06528639793395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,2,power_law_1.01,0.06810879707336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,2,power_law_1.01,0.0835968017578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,2,power_law_1.01,0.08595839738845826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,2,power_law_1.01,0.08528000116348267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,2,power_law_1.01,0.08688639998435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,2,power_law_1.01,0.08858240246772767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,2,power_law_1.01,0.088128000497818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,2,power_law_1.01,0.0898751974105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,2,power_law_1.01,0.0921343982219696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,2,power_law_1.01,0.09473279714584351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,2,power_law_1.01,0.10008319616317748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,2,power_law_1.01,0.10325759649276733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,2,power_law_1.01,0.11117440462112427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,2,power_law_1.01,0.12479360103607177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,2,power_law_1.01,0.14345599412918092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,2,power_law_1.01,0.17293440103530883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,2,power_law_1.01,0.20729598999023438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,2,power_law_1.01,0.27066240310668943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,2,power_law_1.01,0.33802878856658936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,2,power_law_1.01,0.4735231876373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,2,power_law_1.01,0.6304128170013428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,2,power_law_1.01,0.8706111907958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,2,power_law_1.01,1.1240256309509278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,16,power_law_1.2,0.06785920262336731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,16,power_law_1.2,0.0802623987197876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,16,power_law_1.2,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,16,power_law_1.2,0.08427519798278808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,16,power_law_1.2,0.0836736023426056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,16,power_law_1.2,0.07954559922218322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,16,power_law_1.2,0.0818560004234314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,16,power_law_1.2,0.08379520177841186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,16,power_law_1.2,0.0857151985168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,16,power_law_1.2,0.08676480054855347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,16,power_law_1.2,0.08508800268173218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,16,power_law_1.2,0.08777599930763244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,16,power_law_1.2,0.08907520174980163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,16,power_law_1.2,0.09125760197639465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,16,power_law_1.2,0.09916800260543823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,16,power_law_1.2,0.0991424024105072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,16,power_law_1.2,0.10614399909973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,16,power_law_1.2,0.12161920070648194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,16,power_law_1.2,0.13274240493774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,16,power_law_1.2,0.16047999858856202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,16,power_law_1.2,0.1853376030921936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,16,power_law_1.2,0.24500479698181152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,16,power_law_1.2,0.3112191915512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,16,power_law_1.2,0.45688958168029786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,16,power_law_1.2,0.6283391952514649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,16,power_law_1.2,0.9280320167541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,16,power_law_1.2,1.198476791381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,128,power_law_1.2,0.021881599724292756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,128,power_law_1.2,0.022284799814224245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,128,power_law_1.2,0.023526400327682495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,128,power_law_1.2,0.02680320143699646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,128,power_law_1.2,0.03331199884414673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,128,power_law_1.2,0.05381119847297668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,128,power_law_1.2,0.05018879771232605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,128,power_law_1.2,0.0500544011592865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,128,power_law_1.2,0.05063679814338684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,128,power_law_1.2,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,128,power_law_1.2,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,128,power_law_1.2,0.05061119794845581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,128,power_law_1.2,0.049798399209976196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,128,power_law_1.2,0.04734080135822296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,128,power_law_1.2,0.045228800177574156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,128,power_law_1.2,0.04568319916725159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,128,power_law_1.2,0.0551360011100769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,128,power_law_1.2,0.05619840025901794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.2,0.07216640114784241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.2,0.08618879914283753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.2,0.11103999614715576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.2,0.1422271966934204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.2,0.17539199590682983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.2,0.259334397315979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.2,0.34628479480743407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.2,0.5269375801086426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.2,0.7230080127716064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,8,power_law_1.2,0.05135999917984009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,8,power_law_1.2,0.053504002094268796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,8,power_law_1.2,0.05325440168380737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,8,power_law_1.2,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,8,power_law_1.2,0.061459201574325564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,8,power_law_1.2,0.06268159747123718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,8,power_law_1.2,0.06391680240631104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,8,power_law_1.2,0.06496639847755432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,8,power_law_1.2,0.06622719764709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,8,power_law_1.2,0.06644480228424073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,8,power_law_1.2,0.06682239770889283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,8,power_law_1.2,0.06680319905281067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,8,power_law_1.2,0.07182719707489013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,8,power_law_1.2,0.07304959893226623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,8,power_law_1.2,0.07935360074043274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,8,power_law_1.2,0.08113920092582702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,8,power_law_1.2,0.09054080247879029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,8,power_law_1.2,0.10227839946746826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,8,power_law_1.2,0.11249279975891113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,8,power_law_1.2,0.13761279582977295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,8,power_law_1.2,0.16417920589447021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,8,power_law_1.2,0.20412158966064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,8,power_law_1.2,0.24176640510559083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,8,power_law_1.2,0.33930881023406984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,8,power_law_1.2,0.3993535995483398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,8,power_law_1.2,0.6015872001647949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,8,power_law_1.2,0.7208384037017822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,128,balanced,0.05356266597906748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,128,balanced,0.053930665055910744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,128,balanced,0.053802669048309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,128,balanced,0.055957332253456116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,128,balanced,0.054431999723116554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,128,balanced,0.06376533210277557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,128,balanced,0.07562666634718578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,128,balanced,0.08460799853006999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,128,balanced,0.08520533641179402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,128,balanced,0.08381866415341695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,128,balanced,0.08211733400821686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,128,balanced,0.09000533819198608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,128,balanced,0.08960533142089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,128,balanced,0.09107200304667155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,128,balanced,0.1032319962978363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,128,balanced,0.10125333070755005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,128,balanced,0.10735999544461568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,128,balanced,0.11646399895350139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,128,balanced,0.12366400162378947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,128,balanced,0.14046933253606161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,128,balanced,0.15648000439008078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,128,balanced,0.1941493352254232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,128,balanced,0.22856533527374268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,128,balanced,0.2919680078824361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,128,balanced,0.3598080078760783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,128,balanced,0.48454399903615314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,128,balanced,0.6255679925282797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,1,balanced,0.05738133192062378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,1,balanced,0.06328000128269196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,1,balanced,0.07833066582679749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,1,balanced,0.11496000488599141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,1,balanced,0.17759466171264648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,1,balanced,0.2971786657969157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,1,balanced,0.41467201709747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,1,balanced,0.41067731380462646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,1,balanced,0.412992000579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,1,balanced,0.41781866550445557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,1,balanced,0.4201013247172038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,1,balanced,0.4270666837692261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,1,balanced,0.4299466609954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,1,balanced,0.4413919846216838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,1,balanced,0.44758399327596027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,1,balanced,0.4586346546808879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,1,balanced,0.47069867451985675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,1,balanced,0.5171093146006266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,1,balanced,0.5545440117518107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,1,balanced,0.6372213363647461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,1,balanced,0.7022559642791748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,1,balanced,0.8574026425679525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,1,balanced,1.0530880292256672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,1,balanced,1.3607199986775715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,1,balanced,1.7902240753173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,1,balanced,2.4039360682169595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,1,balanced,3.244389216105143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,1,balanced,0.18241600195566812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,1,balanced,0.29097066322962445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,1,balanced,0.5087039868036906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,1,balanced,0.5028426647186279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,1,balanced,0.4861813386281331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,1,balanced,0.5196693340937296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,1,balanced,0.5115360021591187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,1,balanced,0.5164746840794882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,1,balanced,0.4940106471379598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,1,balanced,0.48496532440185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,1,balanced,0.490389347076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,1,balanced,0.46117866039276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,1,balanced,0.4610186815261841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,1,balanced,0.45026667912801105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,1,balanced,0.4082506497701009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,1,balanced,0.37645332018534344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,1,balanced,0.3537173271179199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,1,balanced,0.74563201268514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,1,balanced,0.6016159852345785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,1,balanced,0.8476373354593912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,1,balanced,1.0815520286560059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,1,balanced,1.558224042256673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,1,balanced,2.0358293851216636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,1,balanced,2.9694932301839194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,1,balanced,3.916960080464681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,1,balanced,5.79579226175944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,1,balanced,7.728634516398112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,16,balanced,0.04474133253097534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,16,balanced,0.04181866844495138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,16,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,16,balanced,0.04953599969546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,16,balanced,0.05993066728115082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,16,balanced,0.0631573349237442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,16,balanced,0.0701386680205663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,16,balanced,0.06834666430950165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,16,balanced,0.07664533456166585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,16,balanced,0.07778666913509369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,16,balanced,0.08265066643555959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,16,balanced,0.09571733077367146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,16,balanced,0.0958666702111562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,16,balanced,0.1467626690864563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,16,balanced,0.1732906699180603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,16,balanced,0.17762132485707602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,16,balanced,0.22702932357788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,8,power_law_1.2,0.026502400636672974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,16,balanced,0.291866660118103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,8,power_law_1.2,0.025753599405288697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,16,balanced,0.35210132598876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,8,power_law_1.2,0.027116799354553224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,16,balanced,0.49259201685587567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,16,balanced,0.26711465915044147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,8,power_law_1.2,0.027507200837135315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,16,balanced,0.3393973509470622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,8,power_law_1.2,0.030508801341056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,16,balanced,0.41367467244466144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,8,power_law_1.2,0.03242239952087402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,8,power_law_1.2,0.03216640055179596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,8,power_law_1.2,0.03152639865875244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,8,power_law_1.2,0.03237119913101196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,8,power_law_1.2,0.034227201342582704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,8,power_law_1.2,0.03858560025691986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,8,power_law_1.2,0.04211840033531189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,8,power_law_1.2,0.041254401206970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,8,power_law_1.2,0.04213759899139404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,8,power_law_1.2,0.04614399969577789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,16,balanced,0.6142186721165975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,8,power_law_1.2,0.05106559991836548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,16,balanced,0.7815519968668619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,16,balanced,1.1504426797231038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,16,balanced,1.5157334009806316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,8,power_law_1.2,0.05109120011329651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,8,power_law_1.2,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,8,power_law_1.2,0.0664255976676941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,8,power_law_1.2,0.07447680234909057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,8,power_law_1.2,0.08776959776878357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,8,power_law_1.2,0.10578559637069702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,8,power_law_1.2,0.12688640356063843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,8,power_law_1.2,0.17345279455184937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,8,power_law_1.2,0.2188352108001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,16,power_law_1.01,0.06788480281829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,16,power_law_1.01,0.09393280148506164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,16,power_law_1.01,0.09189119935035706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,16,power_law_1.01,0.09660800099372864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,16,power_law_1.01,0.0951744019985199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,16,power_law_1.01,0.09111040234565734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,16,power_law_1.01,0.09291520118713378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,16,power_law_1.01,0.09168000221252441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,16,power_law_1.01,0.0942463994026184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,16,power_law_1.01,0.09230719804763794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,16,power_law_1.01,0.09994239807128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,16,power_law_1.01,0.0963648021221161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,16,power_law_1.01,0.10162559747695923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,16,power_law_1.01,0.10528639554977418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,16,power_law_1.01,0.11072640419006348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,16,power_law_1.01,0.11441279649734497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,16,power_law_1.01,0.11768959760665894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,16,power_law_1.01,0.12982399463653566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,16,power_law_1.01,0.1388800024986267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,16,power_law_1.01,0.15904639959335326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,16,power_law_1.01,0.18195199966430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,16,power_law_1.01,0.22055680751800538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,16,power_law_1.01,0.2658623933792114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,16,power_law_1.01,0.32709760665893556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,16,power_law_1.01,0.41132798194885256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,16,power_law_1.01,0.5693696022033692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,16,power_law_1.01,0.7299327850341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,8,power_law_1.2,0.3185920000076294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,16,power_law_1.01,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,16,power_law_1.01,0.06838399767875672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,16,power_law_1.01,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,16,power_law_1.01,0.06899200081825256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,16,power_law_1.01,0.07018240094184876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,16,power_law_1.01,0.06741120219230652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,16,power_law_1.01,0.07046399712562561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,16,power_law_1.01,0.07030400037765502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,16,power_law_1.01,0.07160320281982421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,16,power_law_1.01,0.07274879813194275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,16,power_law_1.01,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,16,power_law_1.01,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,16,power_law_1.01,0.07726719975471497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,16,power_law_1.01,0.07783039808273315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,16,power_law_1.01,0.08291199803352356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,16,power_law_1.01,0.08466560244560242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,16,power_law_1.01,0.08698239922523499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,16,power_law_1.01,0.09730560183525086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,16,power_law_1.01,0.10491520166397095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,16,power_law_1.01,0.12138240337371826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,16,power_law_1.01,0.14419840574264525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,16,power_law_1.01,0.1776703953742981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,16,power_law_1.01,0.21283841133117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,16,power_law_1.01,0.28431999683380127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,16,power_law_1.01,0.40357117652893065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,16,power_law_1.01,0.5490943908691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,16,power_law_1.01,0.6909696102142334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,64,power_law_1.2,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,64,power_law_1.2,0.07141759991645813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,64,power_law_1.2,0.07065600156784058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,64,power_law_1.2,0.07596160173416137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,64,power_law_1.2,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,64,power_law_1.2,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,64,power_law_1.2,0.07886719703674316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,64,power_law_1.2,0.07800959944725036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,64,power_law_1.2,0.07881600260734559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,64,power_law_1.2,0.0809216022491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,64,power_law_1.2,0.08339840173721313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,64,power_law_1.2,0.08677759766578674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,8,power_law_1.2,0.38202879428863523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,64,power_law_1.2,0.0883520007133484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,64,power_law_1.2,0.09058560132980346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,64,power_law_1.2,0.09715200066566468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,64,power_law_1.2,0.1068992018699646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,64,power_law_1.2,0.11027840375900269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,64,power_law_1.2,0.1332927942276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,64,power_law_1.2,0.13939839601516724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,64,power_law_1.2,0.17852799892425536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,64,power_law_1.2,0.19622399806976318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,64,power_law_1.2,0.25905280113220214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,64,power_law_1.2,0.2768127918243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,64,power_law_1.2,0.3601279973983765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,64,power_law_1.2,0.5040448188781739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,64,power_law_1.2,0.7498688220977783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,64,power_law_1.2,0.7979199886322021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,2,balanced,0.029605334003766377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,2,balanced,0.030405332644780476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,2,balanced,0.03126399964094162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,2,balanced,0.03159466634194056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,2,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,2,balanced,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,2,balanced,0.03341866781314214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,2,balanced,0.0353973334034284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,2,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,2,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,2,balanced,0.03534399966398875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,2,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,2,balanced,0.03752533346414566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,2,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,2,balanced,0.04576533536116282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,2,balanced,0.04666133224964142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,2,balanced,0.046256000796953835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,2,balanced,0.05454933146635691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,2,balanced,0.056314667065938316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,2,balanced,0.07227199772993724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,2,balanced,0.08627733588218689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,2,balanced,0.1241333285967509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,2,balanced,0.14155200123786926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,2,balanced,0.19606399536132812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,2,balanced,0.24124799172083536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,2,balanced,0.3409279982248942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,2,balanced,0.45341332753499347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,32,power_law_1.2,0.09897599816322326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,32,power_law_1.2,0.07404159903526306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,32,power_law_1.2,0.07423999905586243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,32,power_law_1.2,0.07137920260429383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,32,power_law_1.2,0.07081599831581116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,32,power_law_1.2,0.07799680233001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,32,power_law_1.2,0.08037760257720947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,32,power_law_1.2,0.07920640110969543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,32,power_law_1.2,0.08135039806365967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,32,power_law_1.2,0.08027520179748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,32,power_law_1.2,0.08249599933624267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,32,power_law_1.2,0.08373119831085205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,32,power_law_1.2,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,32,power_law_1.2,0.09329919815063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,32,power_law_1.2,0.09664000272750854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,32,power_law_1.2,0.10239360332489014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,32,power_law_1.2,0.10536320209503174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,32,power_law_1.2,0.11932799816131592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,32,power_law_1.2,0.13393280506134034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,32,power_law_1.2,0.15905280113220216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,32,power_law_1.2,0.18197120428085328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,32,power_law_1.2,0.22542719841003417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,32,power_law_1.2,0.28418560028076173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,32,power_law_1.2,0.363321590423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,32,power_law_1.2,0.4455296039581299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,32,power_law_1.2,0.7442175865173339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,32,power_law_1.2,0.9542847633361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,2,balanced,0.07060799996058147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,2,balanced,0.07455466687679291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,2,balanced,0.08063999811808269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,2,balanced,0.09507200121879578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,2,balanced,0.12030933300654094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,2,balanced,0.18307199080785116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,2,balanced,0.2371413310368856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,2,balanced,0.2243786652882894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,2,balanced,0.22450133164723715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,2,balanced,0.227728009223938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,2,balanced,0.22874132792154947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,2,balanced,0.23266667127609253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,2,balanced,0.23595199982325235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,2,balanced,0.23943465948104858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,2,balanced,0.24835199117660522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,2,balanced,0.25094934304555255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,2,balanced,0.26235200961430866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,2,balanced,0.2892426649729411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,2,balanced,0.3110239903132121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,2,balanced,0.360154668490092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,2,balanced,0.39865068594614667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,2,balanced,0.4987306594848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,2,balanced,0.6019893487294515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,2,balanced,0.8120799859364828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,2,balanced,1.0998666286468506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,2,balanced,1.524629275004069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,2,balanced,2.017061392466227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,1,balanced,0.07844799757003784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,1,balanced,0.12056533495585124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,1,balanced,0.21596799294153848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,1,balanced,0.2097813288370768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,1,balanced,0.20917334159215292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,1,balanced,0.2069173256556193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,1,balanced,0.20797866582870483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,1,balanced,0.21043733755747476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,1,balanced,0.2127359906832377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,1,balanced,0.21051732699076334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,1,balanced,0.2182826598485311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,1,balanced,0.21650133530298868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,1,balanced,0.2181546688079834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,1,balanced,0.2325013279914856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,1,balanced,0.24726400772730509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,1,balanced,0.2503359913825989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,1,balanced,0.24995734294255575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,1,balanced,0.31108800570170086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,1,balanced,0.3181653420130412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,1,balanced,0.3962186574935913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,1,balanced,0.4400906562805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,1,balanced,0.6134719848632812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,1,balanced,0.7894773483276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,1,balanced,1.1202826499938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,1,balanced,1.4528266588846843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,1,balanced,2.125061353047689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,1,balanced,2.776026725769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,32,4,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,32,4,balanced,0.06025066475073496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,32,4,balanced,0.06551466882228851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,32,4,balanced,0.0764213353395462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,32,4,balanced,0.09162132938702901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,32,4,balanced,0.1313973367214203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,32,4,balanced,0.12999467055002847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,32,4,balanced,0.1338879962762197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,32,4,balanced,0.1320373316605886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,32,4,balanced,0.13322133819262186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,32,4,balanced,0.1346826652685801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,32,4,balanced,0.13733866810798645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,32,4,balanced,0.13741866747538248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,32,4,balanced,0.13489066561063132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,32,4,balanced,0.1420693298180898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,32,4,balanced,0.1415839989980062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,32,4,balanced,0.1476586659749349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,32,4,balanced,0.16205333669980368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,32,4,balanced,0.17076265811920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,32,4,balanced,0.1995946764945984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,32,4,balanced,0.21925334135691324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,32,4,balanced,0.2693866689999898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,32,4,balanced,0.31730133295059204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,32,4,balanced,0.4625120162963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,32,4,balanced,0.5633279879887899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,32,4,balanced,0.7931520144144694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,32,4,balanced,1.0372587045033772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,32,balanced,0.05113600194454193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,32,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,32,balanced,0.05175999800364176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,32,balanced,0.05436266462008158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,32,balanced,0.0555626650651296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,32,balanced,0.07030933101971944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,32,balanced,0.07160533467928569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,32,balanced,0.07062933345635732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,32,balanced,0.07316799958546956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,32,balanced,0.07186133166154225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,32,balanced,0.0745119998852412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,32,balanced,0.07515733440717061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,32,balanced,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,32,balanced,0.0766293356815974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,32,balanced,0.08130133152008057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,32,balanced,0.08073066671689351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,32,balanced,0.08307200173536937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,32,balanced,0.09296533465385437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,32,balanced,0.09474666913350423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,32,balanced,0.10594666997591655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,32,balanced,0.11748799681663513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,32,balanced,0.1413279970486959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,32,balanced,0.16355199615160623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,32,balanced,0.21814932425816855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,32,balanced,0.26149866978327435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,32,balanced,0.3635520140329997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,32,balanced,0.46056532859802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,2,power_law_1.01,0.02001280039548874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,2,power_law_1.01,0.024991999566555022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,2,power_law_1.01,0.032620799541473386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,2,power_law_1.01,0.04410240054130554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,2,power_law_1.01,0.05798400044441223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,2,power_law_1.01,0.06455039978027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,2,power_law_1.01,0.07913600206375122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,2,power_law_1.01,0.08110079765319825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,2,power_law_1.01,0.07902719974517822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,2,power_law_1.01,0.08261119723320007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,4,power_law_1.01,0.04312320053577423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,2,power_law_1.01,0.08785920143127442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,4,power_law_1.01,0.044588801264762876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,2,power_law_1.01,0.08950399756431579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,2,power_law_1.01,0.09029120206832886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,2,power_law_1.01,0.09607040286064147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,2,power_law_1.01,0.09772160053253173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,2,power_law_1.01,0.09342719912528992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,2,power_law_1.01,0.10616320371627808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,2,power_law_1.01,0.11409920454025269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,4,power_law_1.01,0.04479359984397888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,2,power_law_1.01,0.11530879735946656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,4,power_law_1.01,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,2,power_law_1.01,0.23451519012451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,4,power_law_1.01,0.05107839703559876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,2,power_law_1.01,0.25610239505767823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,4,power_law_1.01,0.05694720149040222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,2,power_law_1.01,0.1966912031173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,4,power_law_1.01,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,2,power_law_1.01,0.21930880546569825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,4,power_law_1.01,0.057843202352523805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,2,power_law_1.01,0.2629184007644653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,2,power_law_1.01,0.31191039085388184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,2,power_law_1.01,0.413267183303833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,2,power_law_1.01,0.5170688152313232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,4,power_law_1.01,0.056435197591781616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,4,power_law_1.01,0.05804160237312317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,4,power_law_1.01,0.060083198547363284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,4,power_law_1.01,0.060710400342941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,4,power_law_1.01,0.060070401430130003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,4,power_law_1.01,0.063372802734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,4,power_law_1.01,0.06602240204811097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,4,power_law_1.01,0.06999679803848266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,4,power_law_1.01,0.07070720195770264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,4,power_law_1.01,0.0790336012840271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,4,power_law_1.01,0.08547840118408204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,4,power_law_1.01,0.10703999996185302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,4,power_law_1.01,0.12467199563980103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,4,power_law_1.01,0.1556671977043152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,4,power_law_1.01,0.19114240407943725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,4,power_law_1.01,0.2657792091369629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,4,power_law_1.01,0.33256959915161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,4,power_law_1.01,0.44163198471069337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,4,power_law_1.01,0.6495744228363037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,8,power_law_1.2,0.07679359912872315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,8,power_law_1.2,0.09455360174179077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,8,power_law_1.2,0.10057599544525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,8,power_law_1.2,0.1059008002281189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,8,power_law_1.2,0.11975040435791015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,8,power_law_1.2,0.12092159986495972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,8,power_law_1.2,0.13313920497894288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,8,power_law_1.2,0.12569600343704224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,8,power_law_1.2,0.1301695942878723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,8,power_law_1.2,0.126528000831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,8,power_law_1.2,0.13281919956207275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,8,power_law_1.2,0.13219840526580812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,8,power_law_1.2,0.12961280345916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,8,power_law_1.2,0.13653119802474975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,8,power_law_1.2,0.14280320405960084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,8,power_law_1.2,0.15211520195007325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,8,power_law_1.2,0.15205760002136232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,8,power_law_1.2,0.1690816044807434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,8,power_law_1.2,0.1820479989051819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,8,power_law_1.2,0.23221120834350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,8,power_law_1.2,0.2526911973953247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,8,power_law_1.2,0.3322175979614258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,8,power_law_1.2,0.41028480529785155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,8,power_law_1.2,0.6075263977050781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,8,power_law_1.2,0.734278392791748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,8,power_law_1.2,1.1366399765014648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,8,power_law_1.2,1.6312000274658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,16,power_law_1.01,0.04876160025596619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,16,power_law_1.01,0.061510401964187625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,16,power_law_1.01,0.05861759781837463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,16,power_law_1.01,0.06268799901008607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,16,power_law_1.01,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,16,power_law_1.01,0.05251839756965637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,16,power_law_1.01,0.06108160018920898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,16,power_law_1.01,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,16,power_law_1.01,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,16,power_law_1.01,0.06135680079460144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,16,power_law_1.01,0.0648256003856659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,16,power_law_1.01,0.06551039814949036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,16,power_law_1.01,0.06810879707336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,16,power_law_1.01,0.06881920099258423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,16,power_law_1.01,0.07457280158996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,16,power_law_1.01,0.07853440046310425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,16,power_law_1.01,0.0836736023426056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,16,power_law_1.01,0.09457280039787293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,16,power_law_1.01,0.10015360116958619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,16,power_law_1.01,0.12058240175247192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,16,power_law_1.01,0.13006080389022828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,16,power_law_1.01,0.16728320121765136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,16,power_law_1.01,0.19922560453414917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,16,power_law_1.01,0.26431999206542967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,16,power_law_1.01,0.3283071994781494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,16,power_law_1.01,0.4809728145599365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,16,power_law_1.01,0.6222335815429687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,8,balanced,0.03035733352104823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,8,balanced,0.02962133288383484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,8,balanced,0.029322666426499683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,8,balanced,0.029946667452653248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,8,balanced,0.03192000091075897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,8,balanced,0.03146666785081228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,8,balanced,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,8,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,8,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,8,balanced,0.03162666658560435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,8,balanced,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,8,balanced,0.03331200033426285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,8,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,8,balanced,0.03527999917666117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,8,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,8,balanced,0.039477333426475525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,8,balanced,0.039818666875362396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,8,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,8,balanced,0.043791999419530235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,8,balanced,0.05383466680844625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,8,balanced,0.05602133274078369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,8,balanced,0.07640533149242401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,8,balanced,0.093231995900472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,8,balanced,0.1317866643269857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,8,balanced,0.16591466466585794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,8,balanced,0.23153066635131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,8,balanced,0.27532267570495605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,32,balanced,0.045168002446492515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,32,balanced,0.04534400006135305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,32,balanced,0.045994664231936135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,32,balanced,0.05212266743183136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,32,balanced,0.07229866584142049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,32,balanced,0.10233066479365031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,32,balanced,0.10253866513570149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,32,balanced,0.10328533252080281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,32,balanced,0.1030506690343221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,32,balanced,0.10458133618036906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,32,balanced,0.10392000277837117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,32,balanced,0.10662399729092915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,32,balanced,0.10928533474604289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,32,balanced,0.10976533095041911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,32,balanced,0.11690666278203328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,32,balanced,0.11760000387827556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,32,balanced,0.12101333340009053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,32,balanced,0.13176000118255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,32,balanced,0.14044800400733948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,32,balanced,0.1614400049050649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,32,balanced,0.176256000995636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,32,balanced,0.21565866470336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,32,balanced,0.24740799268086752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,32,balanced,0.3269173304239909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,32,balanced,0.39242132504781085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,32,balanced,0.5420480171839396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,32,balanced,0.6851733525594076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,8,power_law_1.2,0.04151679873466492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,8,power_law_1.2,0.0412416011095047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,8,power_law_1.2,0.04095360040664673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,8,power_law_1.2,0.043161600828170776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,8,power_law_1.2,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,8,power_law_1.2,0.0455808013677597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,8,power_law_1.2,0.04631040096282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,8,power_law_1.2,0.047679999470710756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,8,power_law_1.2,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,8,power_law_1.2,0.05615360140800476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,8,power_law_1.2,0.056543999910354616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,64,power_law_1.2,0.05943040251731872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,64,power_law_1.2,0.05782399773597717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,64,power_law_1.2,0.05748479962348938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,8,power_law_1.2,0.06394879817962647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,64,power_law_1.2,0.06004480123519897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,64,power_law_1.2,0.0574783980846405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,8,power_law_1.2,0.06487680077552796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,64,power_law_1.2,0.058246397972106935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,64,power_law_1.2,0.05980799794197082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,64,power_law_1.2,0.06151679754257202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,64,power_law_1.2,0.060550397634506224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,64,power_law_1.2,0.06164479851722717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,64,power_law_1.2,0.060736000537872314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,64,power_law_1.2,0.06357759833335877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,8,power_law_1.2,0.06731520295143127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,64,power_law_1.2,0.06474879980087281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,64,power_law_1.2,0.06477439999580384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,64,power_law_1.2,0.06967679858207702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,64,power_law_1.2,0.06977919936180114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,64,power_law_1.2,0.07538560032844543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,64,power_law_1.2,0.08446720242500305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,64,power_law_1.2,0.09148799777030944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,64,power_law_1.2,0.11077120304107665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,64,power_law_1.2,0.13044480085372925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,64,power_law_1.2,0.17292799949645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,64,power_law_1.2,0.19944319725036622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,64,power_law_1.2,0.27198081016540526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,64,power_law_1.2,0.33844480514526365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,8,power_law_1.2,0.08249599933624267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,64,power_law_1.2,0.491103982925415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,64,power_law_1.2,0.6478015899658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,8,power_law_1.2,0.09474560022354125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,8,power_law_1.2,0.09595519900321961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,8,power_law_1.2,0.1296704053878784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,8,power_law_1.2,0.1582208037376404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,8,power_law_1.2,0.21763200759887696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,8,power_law_1.2,0.26673920154571534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,8,power_law_1.2,0.39426560401916505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,8,power_law_1.2,0.5349760055541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,8,power_law_1.2,0.6909759998321533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,8,power_law_1.2,0.9808064460754394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,64,power_law_1.2,0.08401280045509338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,64,power_law_1.2,0.08428159952163697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,64,power_law_1.2,0.08124799728393554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,64,power_law_1.2,0.08138239979743958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,64,power_law_1.2,0.08080000281333924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,64,power_law_1.2,0.08423680067062378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,64,power_law_1.2,0.08449919819831848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,64,power_law_1.2,0.08522239923477173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,64,power_law_1.2,0.08551679849624634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,64,power_law_1.2,0.0864512026309967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,64,power_law_1.2,0.08365439772605895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,64,power_law_1.2,0.08378880023956299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,64,power_law_1.2,0.08910719752311706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,64,power_law_1.2,0.08929920196533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,64,power_law_1.2,0.09155840277671815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,64,power_law_1.2,0.09379839897155762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,64,power_law_1.2,0.0997376024723053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,32,4,power_law_1.01,0.01817599982023239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,64,power_law_1.2,0.11455999612808228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,32,4,power_law_1.01,0.017484800517559053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,32,4,power_law_1.01,0.018188799917697906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,32,4,power_law_1.01,0.01785600036382675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,32,4,power_law_1.01,0.018611200153827667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,32,4,power_law_1.01,0.018982400000095368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,32,4,power_law_1.01,0.022368000447750093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,32,4,power_law_1.01,0.021881599724292756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,32,4,power_law_1.01,0.03966079950332642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,32,4,power_law_1.01,0.03941119909286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,32,4,power_law_1.01,0.03953279852867127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,32,4,power_law_1.01,0.030169600248336793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,32,4,power_law_1.01,0.030131199955940248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,32,4,power_law_1.01,0.03089280128479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,32,4,power_law_1.01,0.03158400058746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,64,power_law_1.2,0.12375680208206177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,32,4,power_law_1.01,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,64,power_law_1.2,0.1522112011909485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,64,power_law_1.2,0.17089279890060424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,32,4,power_law_1.01,0.032595199346542356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,64,power_law_1.2,0.219596791267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,64,power_law_1.2,0.2802047967910767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,32,4,power_law_1.01,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,64,power_law_1.2,0.3961472034454346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,64,power_law_1.2,0.48200321197509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,64,power_law_1.2,0.6635327816009522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,64,power_law_1.2,0.9251199722290039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,32,4,power_law_1.01,0.03835520148277283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,32,4,power_law_1.01,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,8,power_law_1.2,1.4980352401733399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,1,power_law_1.01,0.10994559526443481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,1,power_law_1.01,0.16129280328750611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,1,power_law_1.01,0.25407359600067136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,1,power_law_1.01,0.4463168144226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,1,power_law_1.01,0.7412992000579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,1,power_law_1.01,1.0478400230407714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,32,4,power_law_1.01,0.04690560102462769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,1,power_law_1.01,1.2634431838989257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,32,4,power_law_1.01,0.05015680193901062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,32,4,power_law_1.01,0.05201280117034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,32,4,power_law_1.01,0.0687936007976532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,32,4,power_law_1.01,0.053420799970626834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,1,power_law_1.01,1.6241151809692382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,32,4,power_law_1.01,0.05546240210533142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,32,4,power_law_1.01,0.060115200281143186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,32,4,power_law_1.01,0.06758400201797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,1,power_law_1.01,1.6898176193237304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,32,4,power_law_1.01,0.07100800275802613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,32,4,power_law_1.01,0.07388799786567687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,1,power_law_1.01,1.758880043029785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,32,4,power_law_1.01,0.07507839798927307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,32,4,power_law_1.01,0.07616000175476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,32,4,power_law_1.01,0.07923200130462646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,1,power_law_1.01,1.8176448822021485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,32,4,power_law_1.01,0.08481280207633972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,32,4,power_law_1.01,0.08618239760398864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,32,4,power_law_1.01,0.07898880243301391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,32,4,power_law_1.01,0.09160959720611572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,1,power_law_1.01,1.8869504928588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,32,4,power_law_1.01,0.10180480480194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,32,4,power_law_1.01,0.10619519948959351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,32,4,power_law_1.01,0.10728960037231446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,1,power_law_1.01,1.8964927673339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,32,4,power_law_1.01,0.12874239683151245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,32,4,power_law_1.01,0.14458880424499512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,32,4,power_law_1.01,0.18095999956130981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,1,power_law_1.01,1.9941247940063476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,32,4,power_law_1.01,0.2236095905303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,32,4,power_law_1.01,0.2794624090194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,32,4,power_law_1.01,0.3590143918991089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,1,power_law_1.01,2.057708740234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,32,4,power_law_1.01,0.4960319995880127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,32,4,power_law_1.01,0.6265088081359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,1,power_law_1.01,2.0852287292480467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,32,4,power_law_1.01,0.8991168022155762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,1,power_law_1.01,2.138489532470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,1,power_law_1.01,2.255571174621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,1,power_law_1.01,2.434982490539551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,32,4,power_law_1.01,1.1974335670471192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,32,4,power_law_1.01,0.10690560340881347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,1,power_law_1.01,2.7340736389160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,1,power_law_1.01,2.613862419128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,1,power_law_1.01,2.876032066345215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,1,power_law_1.01,3.567705535888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,32,4,power_law_1.01,0.13409919738769532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,1,power_law_1.01,4.292121505737304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,1,power_law_1.01,4.392755126953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,8,power_law_1.2,1.7505279541015626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,1,power_law_1.01,5.544268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,1,power_law_1.01,7.556108856201172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,32,4,power_law_1.01,0.18957439661026002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,32,4,power_law_1.01,0.23554561138153077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,8,balanced,0.0521919975678126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,8,balanced,0.053802669048309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,8,balanced,0.06195733447869619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,8,balanced,0.08750933408737183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,8,balanced,0.11657599608103435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,8,balanced,0.17801066239674887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,8,balanced,0.23938133319218954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,8,balanced,0.23919999599456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,8,balanced,0.24009066820144653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,8,balanced,0.2426933248837789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,8,balanced,0.24235733350118002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,8,balanced,0.24553600947062174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,8,balanced,0.24860266844431558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,8,balanced,0.2502773404121399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,8,balanced,0.25763734181722003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,8,balanced,0.26023467381795246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,8,balanced,0.26605866352717084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,8,balanced,0.2789493401845296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,8,balanced,0.28993600606918335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,8,balanced,0.3120853304862976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,8,balanced,0.3341546853383382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,8,balanced,0.3915253480275472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,8,balanced,0.43562134106953937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,8,balanced,0.5346239805221558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,8,balanced,0.6769226392110189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,8,balanced,0.8559040228525797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,8,balanced,1.119599978129069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,1,balanced,0.08264000217119853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,1,balanced,0.13578133781750998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,1,balanced,0.23806933561960855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,1,balanced,0.2453546722730001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,1,balanced,0.23915199438730875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,1,balanced,0.24044267336527506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,1,balanced,0.24052266279856363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,1,balanced,0.24095465739568075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,1,balanced,0.24247999986012778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,1,balanced,0.24427199363708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,1,balanced,0.24876266717910767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,1,balanced,0.24758400519688925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,1,balanced,0.24807467063268027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,1,balanced,0.2702239950497945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,1,balanced,0.27821866671244305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,1,balanced,0.27898667256037396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,1,balanced,0.2892693281173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,1,balanced,0.35611732800801593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,1,balanced,0.3553599913914998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,1,balanced,0.4914986689885457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,1,balanced,0.5496746699015299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,1,balanced,0.7519466876983643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,1,balanced,0.9237653414408366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,1,balanced,1.3482294082641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,1,balanced,1.6793386141459148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,1,balanced,2.45686403910319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,1,balanced,3.2880373001098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,1,power_law_1.2,0.04869759976863861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,1,power_law_1.2,0.05255039930343628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,1,power_law_1.2,0.061977601051330565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,1,power_law_1.2,0.08302080035209655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,1,power_law_1.2,0.09585919976234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,1,power_law_1.2,0.13535360097885132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,1,power_law_1.2,0.14177919626235963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,1,power_law_1.2,0.14455039501190187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,1,power_law_1.2,0.14988160133361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,1,power_law_1.2,0.1537343978881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,1,power_law_1.2,0.1610368013381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,1,power_law_1.2,0.16795519590377808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,1,power_law_1.2,0.1711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,1,power_law_1.2,0.17851519584655762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,1,power_law_1.2,0.18922879695892333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,1,power_law_1.2,0.20074241161346434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,1,power_law_1.2,0.21890559196472167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,1,power_law_1.2,0.2624000072479248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,1,power_law_1.2,0.302291202545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,1,power_law_1.2,0.3923072099685669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,1,power_law_1.2,0.4701119899749756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,1,power_law_1.2,0.6196991920471191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,1,power_law_1.2,0.7686463832855225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,1,power_law_1.2,1.0278911590576172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,1,power_law_1.2,1.392249584197998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,8,power_law_1.01,0.08929920196533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,8,power_law_1.01,0.1295807957649231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,8,power_law_1.01,0.14191999435424804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,1,power_law_1.2,1.899737548828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,8,power_law_1.01,0.17058559656143188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,8,power_law_1.01,0.18116480112075806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,8,power_law_1.01,0.14631680250167847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,8,power_law_1.01,0.1942463994026184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,8,power_law_1.01,0.21008000373840333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,8,power_law_1.01,0.20010240077972413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,8,power_law_1.01,0.19018239974975587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,8,power_law_1.01,0.20315520763397216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,8,power_law_1.01,0.20207359790802001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,8,power_law_1.01,0.19850879907608032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,8,power_law_1.01,0.20393600463867187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,8,power_law_1.01,0.2173504114151001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,8,power_law_1.01,0.22797439098358155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,1,balanced,0.026506667335828144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,1,balanced,0.02942933390537898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,1,balanced,0.04375466704368591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,1,balanced,0.0618399977684021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,1,balanced,0.09742933511734009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,1,balanced,0.17223467429478964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,1,balanced,0.17583467562993368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,1,balanced,0.18103466431299844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,1,balanced,0.1857653260231018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,1,balanced,0.1913386583328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,1,balanced,0.19287467002868652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,1,balanced,0.20005333423614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,1,balanced,0.2072640061378479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,1,balanced,0.20763200521469116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,1,balanced,0.2174826661745707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,1,balanced,0.2273013393084208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,1,balanced,0.24453334013621011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,8,power_law_1.01,0.2299583911895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,8,power_law_1.01,0.26116480827331545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,8,power_law_1.01,0.2575808048248291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,8,power_law_1.01,0.3088128089904785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,8,power_law_1.01,0.31592960357666017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,8,power_law_1.01,0.3890367984771729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,8,power_law_1.2,0.034220799803733826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,8,power_law_1.2,0.034272000193595886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,8,power_law_1.2,0.03379839956760407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,8,power_law_1.2,0.03666560053825378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,8,power_law_1.2,0.03642880022525787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,8,power_law_1.2,0.03695360124111176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,1,balanced,0.2874720096588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,1,balanced,0.2834986646970113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,1,balanced,0.3819413185119629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,1,balanced,0.35684800148010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,1,balanced,0.4690186580022176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,1,balanced,0.43106667200724286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,8,power_law_1.2,0.03697920143604279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,1,balanced,0.586031993230184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,1,balanced,0.7459200223286947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,1,balanced,1.0838027000427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,1,balanced,1.4235626856486003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,8,power_law_1.01,0.43992958068847654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,8,power_law_1.2,0.036595198512077334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,8,power_law_1.2,0.03664000034332275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,8,power_law_1.2,0.03694719970226288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,8,power_law_1.2,0.03688960075378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,8,power_law_1.2,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,8,power_law_1.2,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,8,power_law_1.2,0.03950720131397247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,8,power_law_1.01,0.5394432067871093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,8,power_law_1.01,0.6500991821289063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,8,power_law_1.01,0.861580753326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,8,power_law_1.2,0.04457600116729736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,8,power_law_1.2,0.04485760033130646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,8,power_law_1.2,0.046675199270248414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,8,power_law_1.01,1.1486080169677735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,8,power_law_1.2,0.0565887987613678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.2,0.060249602794647215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.2,0.07264639735221863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.2,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,1,power_law_1.2,2.3514495849609376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.2,0.10932480096817017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.2,0.1322816014289856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,1,balanced,0.08878399928410848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,1,balanced,0.1051093339920044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,1,balanced,0.14287466804186502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,1,balanced,0.19898132483164468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,1,balanced,0.29822399218877155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,1,balanced,0.3496640125910441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,1,balanced,0.3306506673494975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,1,balanced,0.33444801966349286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,1,balanced,0.33083200454711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,1,balanced,0.33187200625737506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,1,balanced,0.32764265934626263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,1,balanced,0.33463998635609943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,1,balanced,0.34142935276031494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,1,balanced,0.3487519820531209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,1,balanced,0.3538186550140381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,1,balanced,0.3575413227081299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,1,balanced,0.3778453270594279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,1,balanced,0.41338666280110675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,1,balanced,0.44236799081166583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,1,balanced,0.5135200023651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,1,balanced,0.5781226555506388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,1,balanced,0.8975839614868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,1,balanced,1.0156213442484539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,1,balanced,1.4229547182718914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,1,balanced,1.863434632619222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,1,balanced,2.5340587298075357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,1,balanced,3.376218795776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,4,power_law_1.01,0.07609599828720093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,4,power_law_1.01,0.08894079923629761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,4,power_law_1.01,0.09737600088119507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,4,power_law_1.01,0.12033920288085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,4,power_law_1.01,0.12887040376663209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,4,power_law_1.01,0.14538880586624145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,4,power_law_1.01,0.16307200193405152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,4,power_law_1.01,0.15934720039367675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,4,power_law_1.01,0.15255039930343628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,4,power_law_1.01,0.1588544011116028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,4,power_law_1.01,0.15641599893569946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,4,power_law_1.01,0.15960960388183593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,4,power_law_1.01,0.16252800226211547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,4,power_law_1.01,0.16547199487686157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,4,power_law_1.01,0.17325439453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,4,power_law_1.01,0.174399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,4,power_law_1.01,0.18407679796218873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,4,power_law_1.01,0.20202879905700682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.01,0.21758720874786378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.01,0.24615681171417236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.01,0.29125759601593015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.01,0.36837120056152345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.01,0.4381375789642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.01,0.5848447799682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.2,0.1723199963569641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.01,0.7362880229949951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.01,1.030451202392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.01,1.3196096420288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.2,0.2189311981201172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.2,0.3256063938140869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,1,power_law_1.01,0.022720000147819518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,1,power_law_1.01,0.025939199328422546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,1,power_law_1.01,0.043647998571395875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,1,power_law_1.01,0.062015998363494876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,1,power_law_1.01,0.08543999791145325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,1,power_law_1.01,0.11727360486984253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,1,power_law_1.01,0.15399680137634278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,1,power_law_1.01,0.16016000509262085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,1,power_law_1.01,0.164409601688385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,1,power_law_1.01,0.1688639998435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,1,power_law_1.01,0.17950719594955444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,1,power_law_1.01,0.18973439931869507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,1,power_law_1.01,0.19622399806976318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,1,power_law_1.01,0.21132159233093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,1,power_law_1.01,0.2008960008621216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,1,power_law_1.01,0.21508479118347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,1,power_law_1.01,0.23432960510253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,1,power_law_1.01,0.22695040702819824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,1,power_law_1.01,0.2545599937438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,1,power_law_1.01,0.27035520076751707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,1,power_law_1.01,0.3023616075515747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,1,power_law_1.01,0.6635263919830322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,1,power_law_1.01,0.7765312194824219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,1,power_law_1.01,0.577126407623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,1,power_law_1.01,0.6567679882049561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,1,power_law_1.01,0.8798527717590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,1,power_law_1.01,1.095353603363037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.2,0.42074241638183596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,1,balanced,0.0414986660083135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,1,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,1,balanced,0.05795733133951823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,1,balanced,0.05681600173314413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,1,balanced,0.06001066664854685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,1,balanced,0.05994666616121928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,1,balanced,0.06188266475995382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,1,balanced,0.05580266813437144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,1,balanced,0.05749866863091787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,1,balanced,0.05607999861240387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,1,balanced,0.057861333092053734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,1,balanced,0.05649599929650625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,1,balanced,0.05835733314355215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,1,balanced,0.05840533475081126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,1,balanced,0.06247466802597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,1,balanced,0.06459733347098033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,1,balanced,0.06692266464233398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,1,balanced,0.08807466427485149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,1,balanced,0.09218133489290874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,1,balanced,0.11848533153533936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,1,balanced,0.13760000467300415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,1,balanced,0.18336532513300577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,1,balanced,0.22871466477711996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,1,balanced,0.3205920060475667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,1,balanced,0.41168534755706787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,1,balanced,0.5868853330612183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,1,balanced,0.7685120105743408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,8,balanced,0.021141332884629566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,8,balanced,0.023034666975339253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,8,balanced,0.023376000424226124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,8,balanced,0.02526933451493581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,8,balanced,0.025663999219735462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,8,balanced,0.03510933369398117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,8,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,8,balanced,0.037690666814645134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,8,balanced,0.03966933240493139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,8,balanced,0.04237333436806997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,8,balanced,0.042170668641726174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,8,balanced,0.045653333266576133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,8,balanced,0.04785599807898203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,8,balanced,0.056277334690093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,8,balanced,0.05852800110975901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,8,balanced,0.057818666100502014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,8,balanced,0.09093333284060161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,8,balanced,0.09642133116722107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,8,balanced,0.12436266740163167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,8,balanced,0.14170666535695395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,8,balanced,0.13160000244776407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,8,balanced,0.15639467040697733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,8,balanced,0.17971199750900269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,8,balanced,0.2662559946378072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,8,balanced,0.3224479953447978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,8,balanced,0.4633599917093913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,8,balanced,0.6005226771036783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,4,power_law_1.01,0.017824000120162962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,4,power_law_1.01,0.018918399512767792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,4,power_law_1.01,0.020044800639152528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,4,power_law_1.01,0.02234880030155182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,4,power_law_1.01,0.0236735999584198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,4,power_law_1.01,0.02570880055427551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,64,power_law_1.2,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,4,power_law_1.01,0.027187201380729675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,64,power_law_1.2,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,4,power_law_1.01,0.02746880054473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,64,power_law_1.2,0.030847999453544616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,4,power_law_1.01,0.02908799946308136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,64,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,4,power_law_1.01,0.03054080009460449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,64,power_law_1.2,0.03400320112705231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,64,power_law_1.2,0.035129600763320924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,4,power_law_1.01,0.03172479867935181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,4,power_law_1.01,0.032902398705482484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,4,power_law_1.01,0.03505280017852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,4,power_law_1.01,0.03526400029659271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,4,power_law_1.01,0.03656960129737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,4,power_law_1.01,0.03835520148277283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,4,power_law_1.01,0.04485119879245758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,4,power_law_1.01,0.04859519898891449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,4,power_law_1.01,0.052665597200393675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,64,power_law_1.2,0.03592959940433502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,4,power_law_1.01,0.06577919721603394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,64,power_law_1.2,0.036582401394844054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,4,power_law_1.01,0.08183040022850037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,4,power_law_1.01,0.09411200284957885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,4,power_law_1.01,0.10696959495544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,4,power_law_1.01,0.1332800030708313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,4,power_law_1.01,0.16391040086746217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,4,power_law_1.01,0.21447041034698486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,64,power_law_1.2,0.037376001477241516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,4,power_law_1.01,0.272710394859314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,64,power_law_1.2,0.037222400307655334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,64,power_law_1.2,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,64,power_law_1.2,0.040575999021530154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,64,power_law_1.2,0.041331198811531064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,64,power_law_1.2,0.04642559885978699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,64,power_law_1.2,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,64,power_law_1.2,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,64,power_law_1.2,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,64,power_law_1.2,0.069651198387146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,64,power_law_1.2,0.08162559866905213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,64,power_law_1.2,0.11480319499969482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,64,power_law_1.2,0.1344063997268677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,64,power_law_1.2,0.19366400241851806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,64,power_law_1.2,0.22069759368896485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,64,power_law_1.2,0.38268160820007324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,64,power_law_1.2,0.5578239917755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,64,power_law_1.2,0.915231990814209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,32,balanced,0.04595200220743815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,32,balanced,0.0499946673711141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,32,balanced,0.059114664793014526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,32,balanced,0.09012266993522644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,32,balanced,0.10517332951227824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,32,balanced,0.1074186662832896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,32,balanced,0.10725333293279012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,32,balanced,0.10699733098347981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,32,balanced,0.10975999633471172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,32,balanced,0.10832533240318298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,32,balanced,0.10935466488202412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,32,balanced,0.11219200491905212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,32,balanced,0.11145599683125813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,32,balanced,0.11374400059382121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,32,balanced,0.11859200398127238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,32,balanced,0.11898133158683777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,32,balanced,0.12257599830627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,32,balanced,0.14270933469136557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,32,balanced,0.1414240002632141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,32,balanced,0.18652266263961792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,32,balanced,0.17912532885869345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,32,balanced,0.26358399788538617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,32,balanced,0.21948800484339395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,32,balanced,0.2935520013173421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,32,balanced,0.31725333134333294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,32,balanced,0.43875734011332196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,32,balanced,0.4922720193862915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,8,power_law_1.01,0.052236801385879515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,8,power_law_1.01,0.06483200192451477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,8,power_law_1.01,0.06225280165672302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,8,power_law_1.01,0.0744383990764618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,8,power_law_1.01,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,8,power_law_1.01,0.07756159901618957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,8,power_law_1.01,0.08038399815559387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,8,power_law_1.01,0.08060799837112427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,8,power_law_1.01,0.07996159791946411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,8,power_law_1.01,0.08010240197181702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,8,power_law_1.01,0.08526080250740051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,8,power_law_1.01,0.08600959777832032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,8,power_law_1.01,0.08766080141067505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,8,power_law_1.01,0.09136000275611877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,8,power_law_1.01,0.09713280200958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,8,power_law_1.01,0.10465919971466064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,8,power_law_1.01,0.10758399963378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,8,power_law_1.01,0.11563520431518555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,8,power_law_1.01,0.1282240033149719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,8,power_law_1.01,0.16317440271377565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,8,power_law_1.01,0.17163519859313964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,8,power_law_1.01,0.2250688076019287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,8,power_law_1.01,0.2675839900970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,8,power_law_1.01,0.3569279909133911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,8,power_law_1.01,0.4525119781494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,8,power_law_1.01,0.5997824192047119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,8,power_law_1.01,0.7852735996246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,64,power_law_1.2,1.176473617553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,8,power_law_1.01,0.09494400024414062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,8,power_law_1.01,0.14446719884872436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,8,power_law_1.01,0.23836159706115723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,8,power_law_1.01,0.2234623908996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,8,power_law_1.01,0.26232960224151614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,8,power_law_1.01,0.29795839786529543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,8,power_law_1.01,0.3138816118240356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,8,power_law_1.01,0.3083391904830933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,8,power_law_1.01,0.3120255947113037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,8,power_law_1.01,0.28564479351043703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,8,power_law_1.01,0.33320319652557373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,8,power_law_1.01,0.3303744077682495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,8,power_law_1.01,0.3122944116592407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,8,power_law_1.01,0.31544320583343505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,8,power_law_1.01,0.33115520477294924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,8,power_law_1.01,0.34431359767913816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,8,power_law_1.01,0.3593983888626099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,8,power_law_1.01,0.4004415988922119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,8,power_law_1.01,0.4040192127227783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,8,power_law_1.01,0.4747007846832275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,8,power_law_1.01,0.48899202346801757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,8,power_law_1.01,0.5984640121459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,8,power_law_1.01,0.6894976139068604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,8,power_law_1.01,0.861292839050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,8,power_law_1.01,1.0506048202514648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,8,power_law_1.01,1.3500160217285155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,8,power_law_1.01,1.694905662536621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,32,power_law_1.01,0.02698880136013031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,32,power_law_1.01,0.030188798904418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,32,power_law_1.01,0.027238398790359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,32,power_law_1.01,0.030054399371147157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,32,power_law_1.01,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,32,power_law_1.01,0.02961280047893524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,32,power_law_1.01,0.03663359880447388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,32,power_law_1.01,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,32,power_law_1.01,0.03754239976406097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,32,power_law_1.01,0.03946239948272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,32,power_law_1.01,0.046777600049972536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,32,power_law_1.01,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,32,power_law_1.01,0.05030400156974792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,32,power_law_1.01,0.06999040246009827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,32,power_law_1.01,0.07561600208282471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,32,power_law_1.01,0.08120319843292237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,32,power_law_1.01,0.09371520280838012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,32,power_law_1.01,0.10035840272903443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.01,0.11809920072555542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.01,0.1462399959564209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.01,0.10205440521240235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.01,0.1289023995399475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.01,0.1637503981590271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.01,0.21198720932006837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.01,0.28455679416656493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.01,0.4177152156829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.01,0.5127871990203857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,32,balanced,0.039621333281199135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,32,balanced,0.041706666350364685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,32,balanced,0.04162666698296865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,32,balanced,0.043621331453323364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,32,balanced,0.04574400186538696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,32,balanced,0.062037333846092224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,32,balanced,0.06530133386452992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,32,balanced,0.06433066725730896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,32,balanced,0.06473066906134288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,32,balanced,0.06419200201829274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,32,balanced,0.06610666712125142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,32,balanced,0.06611733138561249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,32,balanced,0.06780266761779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,32,balanced,0.06820799907048543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,32,balanced,0.07455466687679291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,32,balanced,0.07252266506354015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,32,balanced,0.07531733314196269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,32,balanced,0.07909333209196727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,32,balanced,0.08262933293978374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,32,balanced,0.0928000013033549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,32,balanced,0.10000000397364299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,32,balanced,0.1160640021165212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,32,balanced,0.13195199767748514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,32,balanced,0.1691840092341105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,32,balanced,0.20098666350046793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,32,balanced,0.26370133956273395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,32,balanced,0.3323253393173218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,32,balanced,0.04604800045490265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,32,balanced,0.04568533102671305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,32,balanced,0.04567466676235199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,32,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,32,balanced,0.051781331499417625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,32,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,32,balanced,0.05630933245023092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,32,balanced,0.0563679983218511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,32,balanced,0.05782933533191681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,32,balanced,0.058058664202690125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,32,balanced,0.058117335041364036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,32,balanced,0.05991999804973602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,32,balanced,0.061797335743904114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,32,balanced,0.060415998101234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,32,balanced,0.06646933158238728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,32,balanced,0.06818133095900218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,32,balanced,0.07025599976380666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,32,balanced,0.07667199770609538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,32,balanced,0.08082666496435802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,32,balanced,0.094842662413915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,32,balanced,0.10852266351381938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,32,balanced,0.13808000087738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,32,balanced,0.15927466750144958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,32,balanced,0.2087200085322062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,32,balanced,0.2630506753921509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,32,balanced,0.35342931747436523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,32,balanced,0.4474773406982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,32,power_law_1.01,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,32,power_law_1.01,0.04246399998664856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,32,power_law_1.01,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,32,power_law_1.01,0.04162560105323791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,32,power_law_1.01,0.0420415997505188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,32,power_law_1.01,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,32,power_law_1.01,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,32,power_law_1.01,0.04496639966964722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,32,power_law_1.01,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,32,power_law_1.01,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,32,power_law_1.01,0.049369600415229795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,32,power_law_1.01,0.05130879878997803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,32,power_law_1.01,0.05186560153961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,32,power_law_1.01,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,32,power_law_1.01,0.0584447979927063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,32,power_law_1.01,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,32,power_law_1.01,0.05754240155220032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,32,power_law_1.01,0.06416640281677247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,32,power_law_1.01,0.06781439781188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,32,power_law_1.01,0.07806079983711242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,32,power_law_1.01,0.09067519903182983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,8,power_law_1.2,0.08973439931869506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,32,power_law_1.01,0.11223679780960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,32,power_law_1.01,0.13203840255737304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,8,power_law_1.2,0.12851200103759766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,32,power_law_1.01,0.17288960218429567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,8,power_law_1.2,0.11918079853057861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,32,power_law_1.01,0.22001919746398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,8,power_law_1.2,0.1665984034538269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,32,power_law_1.01,0.3002432107925415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,8,power_law_1.2,0.20383360385894775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,32,power_law_1.01,0.3971967935562134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,8,power_law_1.2,0.2111680030822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,8,power_law_1.2,0.19783040285110473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,8,power_law_1.2,0.18615039587020873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,8,power_law_1.2,0.19410560131072999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,8,power_law_1.2,0.19090559482574462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,8,power_law_1.2,0.19594240188598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,8,power_law_1.2,0.20621440410614014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,8,power_law_1.2,0.1995967984199524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,8,power_law_1.2,0.21873281002044678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,8,power_law_1.2,0.2234816074371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,8,power_law_1.2,0.23431038856506348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,8,power_law_1.2,0.24970879554748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,8,power_law_1.2,0.28368000984191893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,8,power_law_1.2,0.2985152006149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,8,power_law_1.2,0.3666687965393066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,8,power_law_1.2,0.3651456117630005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,8,power_law_1.2,0.43955841064453127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,8,power_law_1.2,0.5380159854888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,8,power_law_1.2,0.721727991104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,8,power_law_1.2,0.8919551849365235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,8,power_law_1.2,1.1622207641601563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,8,power_law_1.2,1.5907391548156737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,128,power_law_1.01,0.017555199563503265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,128,power_law_1.01,0.017151999473571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,128,power_law_1.01,0.018118399381637573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,128,power_law_1.01,0.018515199422836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,128,power_law_1.01,0.021568000316619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,128,power_law_1.01,0.03521920144557953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,128,power_law_1.01,0.030527999997138976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,128,power_law_1.01,0.030156800150871278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,128,power_law_1.01,0.030272001028060914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,128,power_law_1.01,0.03049600124359131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,128,power_law_1.01,0.030214399099349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,128,power_law_1.01,0.030899199843406677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,128,power_law_1.01,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,128,power_law_1.01,0.030745598673820495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,128,power_law_1.01,0.0308351993560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,128,power_law_1.01,0.031692799925804135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,128,power_law_1.01,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,128,power_law_1.01,0.03452160060405731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,128,power_law_1.01,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,128,power_law_1.01,0.04249599874019623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,128,power_law_1.01,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,128,power_law_1.01,0.06110079884529114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,128,power_law_1.01,0.07157760262489318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,128,power_law_1.01,0.09627519845962525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,128,power_law_1.01,0.11360000371932984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,128,power_law_1.01,0.16619520187377929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,128,power_law_1.01,0.21045119762420655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,16,balanced,0.022266666094462078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,16,balanced,0.02165333429972331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,16,balanced,0.02207999924818675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,16,balanced,0.023669332265853882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,16,balanced,0.02569599946339925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,16,balanced,0.03325333446264267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,16,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,16,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,16,balanced,0.05011733373006185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,16,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,16,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,16,balanced,0.08224000036716461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,16,balanced,0.08025066554546356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,16,balanced,0.08161066472530365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,16,balanced,0.08172800143559773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,16,balanced,0.08276799817879994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,16,balanced,0.0812906672557195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,16,balanced,0.08948799967765808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,16,balanced,0.09514666597048442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,16,balanced,0.11157332857449849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,16,balanced,0.13261333107948303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,16,balanced,0.16405333081881204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,16,balanced,0.19433067242304483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,16,balanced,0.25338133176167804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,16,balanced,0.31285866101582843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,16,balanced,0.4726133346557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,16,balanced,0.5972906748453776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,4,balanced,0.03986666599909464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,4,balanced,0.03882133215665817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,4,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,4,balanced,0.04354133208592733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,4,balanced,0.04413333535194397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,4,balanced,0.04550399879614512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,4,balanced,0.04408533374468485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,4,balanced,0.046053335070610046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,4,balanced,0.046298667788505554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,4,balanced,0.04631466666857401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,4,balanced,0.04621333380540212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,4,balanced,0.046623999873797096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,4,balanced,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,4,balanced,0.05219733218352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,4,balanced,0.06098666787147522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,4,balanced,0.06417066852251689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,4,balanced,0.06668800115585327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,4,balanced,0.07704533139864604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,4,balanced,0.08237333099047343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,4,balanced,0.09700799981753032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,4,balanced,0.11773866415023804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,4,balanced,0.1395199994246165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,4,balanced,0.15942399700482687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,4,balanced,0.2126986583073934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,4,balanced,0.2653866608937581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,4,balanced,0.37117866675059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,4,balanced,0.47253334522247314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,64,power_law_1.2,0.07416319847106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,64,power_law_1.2,0.07196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,64,power_law_1.2,0.07145599722862243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,64,power_law_1.2,0.06998400092124939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,64,power_law_1.2,0.06796159744262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,64,power_law_1.2,0.06538239717483521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,64,power_law_1.2,0.06688640117645264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,64,power_law_1.2,0.07096959948539734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,64,power_law_1.2,0.0710591971874237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,64,power_law_1.2,0.07278720140457154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,64,power_law_1.2,0.06976640224456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,64,power_law_1.2,0.07219840288162231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,64,power_law_1.2,0.07336320281028748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,64,power_law_1.2,0.07352960109710693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,64,power_law_1.2,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,64,power_law_1.2,0.08095999956130981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,64,power_law_1.2,0.0887167990207672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,64,power_law_1.2,0.1018496036529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.2,0.11230080127716065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.2,0.1314239978790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.2,0.15739519596099855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.2,0.18762240409851075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.2,0.2387392044067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.2,0.334003210067749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.2,0.401632022857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.2,0.5798079967498779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.2,0.8411456108093261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,64,balanced,0.05628266433874766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,64,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,64,balanced,0.04399466514587402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,64,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,64,balanced,0.04784533381462097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,64,balanced,0.05176533261934916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,64,balanced,0.053823997577031456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,64,balanced,0.05297600229581197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,64,balanced,0.05276800195376078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,64,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,64,balanced,0.05454933146635691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,64,balanced,0.05578133463859558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,64,balanced,0.057536001006762184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,64,balanced,0.05985599756240845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,64,balanced,0.06440000236034393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,64,balanced,0.06443733473618825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,64,balanced,0.0695360004901886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,64,balanced,0.07630399862925212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,64,balanced,0.08126399914423625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,64,balanced,0.09210667014122009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,64,balanced,0.10040000081062317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,64,balanced,0.11868266264597575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,64,balanced,0.1378773351510366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,4,power_law_1.2,0.05258240103721619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,4,power_law_1.2,0.055430400371551516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,4,power_law_1.2,0.0738752007484436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,4,power_law_1.2,0.08506240248680115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,4,power_law_1.2,0.0857151985168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,4,power_law_1.2,0.11003520488739013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,4,power_law_1.2,0.11246720552444459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,64,balanced,0.1739573280016581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,64,balanced,0.20932267109553018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,64,balanced,0.27666666110356647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,64,balanced,0.3394720156987508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,4,power_law_1.2,0.11815680265426635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,4,power_law_1.2,0.11939200162887573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,4,power_law_1.2,0.11939200162887573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,4,power_law_1.2,0.1272063970565796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,4,power_law_1.2,0.12679040431976318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,4,power_law_1.2,0.13010560274124144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,4,power_law_1.2,0.13235199451446533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,4,power_law_1.2,0.14652800559997559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,4,power_law_1.2,0.14840960502624512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,128,power_law_1.2,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,4,power_law_1.2,0.16609920263290406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,128,power_law_1.2,0.031052801012992858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,4,power_law_1.2,0.19719680547714233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,128,power_law_1.2,0.03079040050506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,128,power_law_1.2,0.03386240005493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,128,power_law_1.2,0.03416320085525513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,128,power_law_1.2,0.0350271999835968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,128,power_law_1.2,0.03563520014286041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,128,power_law_1.2,0.035692799091339114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,128,power_law_1.2,0.03617280125617981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,128,power_law_1.2,0.03624959886074066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,128,power_law_1.2,0.03692159950733185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,128,power_law_1.2,0.03720960021018982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,128,power_law_1.2,0.037401598691940305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,128,power_law_1.2,0.03864319920539856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,128,power_law_1.2,0.04113920032978058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,128,power_law_1.2,0.04179840087890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,128,power_law_1.2,0.0434688001871109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,128,power_law_1.2,0.044736000895500186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,128,power_law_1.2,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,128,power_law_1.2,0.05208960175514221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,128,power_law_1.2,0.05768960118293762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,128,power_law_1.2,0.07377920150756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,128,power_law_1.2,0.08942720293998718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,128,power_law_1.2,0.12664320468902587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,4,power_law_1.2,0.20626559257507324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,128,power_law_1.2,0.16911360025405883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,128,power_law_1.2,0.22664320468902588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,4,power_law_1.2,0.24965760707855225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,4,power_law_1.2,0.3053951978683472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,128,power_law_1.2,0.3246592044830322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,4,power_law_1.2,0.37871999740600587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,4,power_law_1.2,0.4691328048706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,4,power_law_1.2,0.6256063938140869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,4,power_law_1.2,0.7933440208435059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,4,power_law_1.2,1.150220775604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,32,balanced,0.038959999879201256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,32,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,32,balanced,0.037690666814645134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,32,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,32,balanced,0.041434665520985924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,32,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,32,balanced,0.04740799963474274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,32,balanced,0.04794133206208547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,32,balanced,0.047983999053637184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,32,balanced,0.04796266555786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,32,balanced,0.04782933493455251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,32,balanced,0.05034666756788889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,32,balanced,0.04991999765237173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,32,balanced,0.05609600245952606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,32,balanced,0.060720001657803856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,32,balanced,0.06260799864927928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,32,balanced,0.05611200133959452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,32,balanced,0.06229866544405619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,32,balanced,0.06428800026575725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,32,balanced,0.07437333464622498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,32,balanced,0.08069866895675659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,32,balanced,0.09896533687909444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,32,balanced,0.11990400155385335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,32,balanced,0.15472533305486044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,32,balanced,0.18742932875951132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,32,balanced,0.24923733870188394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,32,balanced,0.31891733407974243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,4,power_law_1.2,1.6437568664550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,2,balanced,0.0458186666170756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,2,balanced,0.05641066531340281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,2,balanced,0.07308266560236613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,2,balanced,0.11351999640464783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,2,balanced,0.1756053368250529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,2,balanced,0.30582932631174725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,2,balanced,0.3077813386917114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,2,balanced,0.30696000655492145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,2,balanced,0.30557332436243695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,2,balanced,0.3090239961942037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,2,balanced,0.3076373338699341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,2,balanced,0.31165866057078045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,2,balanced,0.3147839903831482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,2,balanced,0.31962132453918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,2,balanced,0.324074665705363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,2,balanced,0.3264853358268738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,2,balanced,0.3325600028038025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,2,balanced,0.34810133775075275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,2,balanced,0.3638613224029541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,2,balanced,0.39589866002400714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,2,balanced,0.42814401785532635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,2,balanced,0.47467732429504395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,2,balanced,0.5333653291066488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,2,balanced,0.6681386629740397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,2,balanced,0.7647199630737305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,2,balanced,1.0724053382873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,2,balanced,1.291167974472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,32,power_law_1.01,0.02635520100593567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,32,power_law_1.01,0.024300800263881685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,32,power_law_1.01,0.02446720004081726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,32,power_law_1.01,0.02455040067434311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,32,power_law_1.01,0.026419198513031004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,32,power_law_1.01,0.028070399165153505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,32,power_law_1.01,0.02773759961128235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,32,power_law_1.01,0.028038400411605834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,32,power_law_1.01,0.030559998750686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,32,power_law_1.01,0.03146879971027374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,32,power_law_1.01,0.032620799541473386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,32,power_law_1.01,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,32,power_law_1.01,0.03423359990119934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,2,power_law_1.2,0.16062719821929933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,2,power_law_1.2,0.20307838916778564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,2,power_law_1.2,0.31729280948638916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,2,power_law_1.2,0.3906048059463501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,2,power_law_1.2,0.5394879817962647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,32,power_law_1.01,0.03701759874820709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,32,power_law_1.01,0.04080640077590943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,32,power_law_1.01,0.04113920032978058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,32,power_law_1.01,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,32,power_law_1.01,0.04514560103416443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,32,power_law_1.01,0.0525439977645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,32,power_law_1.01,0.055878400802612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,32,power_law_1.01,0.0657920002937317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,32,power_law_1.01,0.07825279831886292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,2,power_law_1.2,0.8741056442260742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,2,power_law_1.2,0.9028287887573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,2,power_law_1.2,0.9577792167663575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,32,power_law_1.01,0.10083199739456176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,32,power_law_1.01,0.10698239803314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,32,power_law_1.01,0.13215999603271483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,32,power_law_1.01,0.1841215968132019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,32,power_law_1.01,0.2353343963623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,2,power_law_1.2,1.0059519767761231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,2,power_law_1.2,1.0082688331604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,2,power_law_1.2,1.034444808959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,2,power_law_1.2,1.0762175559997558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,2,power_law_1.2,1.065567970275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,2,power_law_1.2,1.1506815910339356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,2,power_law_1.2,1.2113984107971192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,2,power_law_1.2,1.2038656234741212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,2,power_law_1.2,1.2930879592895508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,2,power_law_1.2,1.4184255599975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,2,power_law_1.2,1.4712448120117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,2,power_law_1.2,1.7225919723510743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,2,power_law_1.2,1.5916735649108886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,2,power_law_1.2,1.9121343612670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,2,power_law_1.2,2.257369613647461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,2,power_law_1.2,2.8172479629516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,2,power_law_1.2,3.8236351013183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,2,power_law_1.2,4.943436813354492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,2,power_law_1.2,6.137190246582032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,4,power_law_1.2,0.03253119885921478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,4,power_law_1.2,0.030559998750686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,4,power_law_1.2,0.031302401423454286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,4,power_law_1.2,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,4,power_law_1.2,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,4,power_law_1.2,0.038553598523139956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,4,power_law_1.2,0.03984000086784363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,4,power_law_1.2,0.04067200124263763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,4,power_law_1.2,0.04108799993991852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,4,power_law_1.2,0.04054400026798248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,4,power_law_1.2,0.041657599806785586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,4,power_law_1.2,0.0427839994430542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,4,power_law_1.2,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,4,power_law_1.2,0.04794879853725433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,4,power_law_1.2,0.0526528000831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,4,power_law_1.2,0.05534719824790955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,4,power_law_1.2,0.05907840132713318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,4,power_law_1.2,0.07294080257415772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,4,power_law_1.2,0.08005120158195496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,4,power_law_1.2,0.10219520330429077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,4,power_law_1.2,0.11363199949264527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,4,power_law_1.2,0.15093120336532592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,4,power_law_1.2,0.22120320796966553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,4,power_law_1.2,0.31000959873199463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,32,4,power_law_1.01,0.07259520292282104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,32,4,power_law_1.01,0.08085119724273682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,32,4,power_law_1.01,0.08462719917297364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,32,4,power_law_1.01,0.09774079918861389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,32,4,power_law_1.01,0.10718079805374145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,32,4,power_law_1.01,0.11717760562896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,32,4,power_law_1.01,0.12229759693145752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,32,4,power_law_1.01,0.12554240226745605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,32,4,power_law_1.01,0.1235584020614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,32,4,power_law_1.01,0.12144639492034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,32,4,power_law_1.01,0.125491201877594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,32,4,power_law_1.01,0.1296512007713318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,32,4,power_law_1.01,0.13521920442581176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,32,4,power_law_1.01,0.12534400224685668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,32,4,power_law_1.01,0.13959039449691774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,32,4,power_law_1.01,0.1422976016998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,32,4,power_law_1.01,0.14870400428771974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,32,4,power_law_1.01,0.1665343999862671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,32,4,power_law_1.01,0.17813119888305665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,32,4,power_law_1.01,0.20959360599517823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,32,4,power_law_1.01,0.24476161003112792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,32,4,power_law_1.01,0.31139841079711916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,32,4,power_law_1.01,0.3764800071716309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,32,4,power_law_1.01,0.5169151782989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,4,power_law_1.2,0.3941375970840454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,32,4,power_law_1.01,0.6988351821899415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,32,4,power_law_1.01,0.8956416130065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,32,4,power_law_1.01,1.1459775924682618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,4,power_law_1.2,0.5527167797088623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,2,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,2,balanced,0.05176533261934916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,2,balanced,0.0729973316192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,2,balanced,0.10903466741243999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,2,balanced,0.16146133343378702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,2,balanced,0.1909653345743815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,2,balanced,0.19151467084884644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,2,balanced,0.19430933396021524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,2,balanced,0.19676266113917032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,2,balanced,0.20062933365503946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,2,balanced,0.1991893251736959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,2,balanced,0.20463999112447104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,2,balanced,0.20582934220631918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,2,balanced,0.20938666661580405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,2,balanced,0.21970667441685995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,2,balanced,0.22340800364812216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,2,balanced,0.23306133349736533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,2,balanced,0.2711679935455322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,2,balanced,0.27782400449117023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,2,balanced,0.3510133425394694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,2,balanced,0.3792053461074829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,2,balanced,0.5375466744105021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,2,balanced,0.5936906735102335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,2,balanced,0.8496533234914144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,2,balanced,1.0035359859466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,2,balanced,1.493226687113444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,2,balanced,1.8685812950134277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,2,power_law_1.2,0.01794559955596924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,2,power_law_1.2,0.01841920018196106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,2,power_law_1.2,0.020044800639152528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,2,power_law_1.2,0.023865599930286408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,2,power_law_1.2,0.02797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,2,power_law_1.2,0.031046399474143983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,2,power_law_1.2,0.036057600378990175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,2,power_law_1.2,0.039961600303649904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,2,power_law_1.2,0.03864319920539856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,2,power_law_1.2,0.04252159893512726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,2,power_law_1.2,0.0426367998123169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,2,power_law_1.2,0.04563199877738953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,2,power_law_1.2,0.047116801142692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,2,power_law_1.2,0.04687359929084778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,2,power_law_1.2,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,2,power_law_1.2,0.051660799980163576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,2,power_law_1.2,0.05637120008468628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,2,power_law_1.2,0.0646336019039154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,2,power_law_1.2,0.07169280052185059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,2,power_law_1.2,0.09093760251998902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,2,power_law_1.2,0.10140160322189332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,2,power_law_1.2,0.12943999767303466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,2,power_law_1.2,0.14844160079956054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,2,power_law_1.2,0.1818560004234314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,2,power_law_1.2,0.21542398929595946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,2,power_law_1.2,0.2840384006500244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,2,power_law_1.2,0.36510720252990725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,4,power_law_1.2,0.743174409866333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,4,power_law_1.2,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,4,power_law_1.2,0.031481599807739256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,4,power_law_1.2,0.03612799942493439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,4,power_law_1.2,0.0430400013923645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,4,power_law_1.2,0.04785279929637909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,4,power_law_1.2,0.05891839861869812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,4,power_law_1.2,0.05889279842376709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,4,power_law_1.2,0.05971840023994446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,4,power_law_1.2,0.06193280220031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,4,power_law_1.2,0.06327040195465088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,4,power_law_1.2,0.06555520296096802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,4,power_law_1.2,0.06849920153617858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,4,power_law_1.2,0.06958079934120179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,4,power_law_1.2,0.07200639843940734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,4,power_law_1.2,0.07773439884185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,4,power_law_1.2,0.08378880023956299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,4,power_law_1.2,0.0893119990825653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,4,power_law_1.2,0.10375039577484131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,4,power_law_1.2,0.11571840047836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,4,power_law_1.2,0.14248960018157958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,4,power_law_1.2,0.1662976026535034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,4,power_law_1.2,0.21719679832458497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,4,power_law_1.2,0.2663104057312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,4,power_law_1.2,0.37345280647277834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,4,power_law_1.2,0.4978816032409668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,1,power_law_1.2,0.09708799719810486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,1,power_law_1.2,0.1250048041343689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,1,power_law_1.2,0.18625279664993286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,1,power_law_1.2,0.2573888063430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,1,power_law_1.2,0.3400576114654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,1,power_law_1.2,0.5709248065948487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,1,power_law_1.2,0.584819221496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,1,power_law_1.2,0.6206528186798096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,1,power_law_1.2,0.6620480060577393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,1,power_law_1.2,0.6592959880828857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,1,power_law_1.2,0.6715456008911133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,1,power_law_1.2,0.7029952049255371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,1,power_law_1.2,0.7058112144470214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,1,power_law_1.2,0.7203904151916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,1,power_law_1.2,0.7462528228759766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,1,power_law_1.2,0.7741695880889893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,1,power_law_1.2,0.7965055942535401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,1,power_law_1.2,0.88089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,4,power_law_1.2,0.7072447776794434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.2,0.9753215789794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.2,1.1458687782287598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.2,1.2116479873657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.2,1.5038975715637206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.2,1.881216049194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.2,2.449625587463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.2,3.0230464935302734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.2,4.156492614746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.2,5.365977478027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,4,power_law_1.2,0.954150390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,128,balanced,0.04456533491611481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,128,balanced,0.04594666759173075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,128,balanced,0.04558399816354116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,128,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,128,balanced,0.0476693312327067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,128,balanced,0.053898667295773826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,128,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,128,balanced,0.052058666944503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,128,balanced,0.053472002347310386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,128,balanced,0.053957333167394005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,128,balanced,0.053247998158137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,128,balanced,0.055829331278800964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,128,balanced,0.058058664202690125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,128,balanced,0.055786664287249245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,128,balanced,0.060826669136683144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,128,balanced,0.06098133325576782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,128,balanced,0.062165334820747375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,128,balanced,0.06866133213043213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,128,balanced,0.0728053351243337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,128,balanced,0.08133333424727122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,128,balanced,0.08891200025876363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,128,balanced,0.10730133454004924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,128,balanced,0.1236799955368042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,128,balanced,0.16528532902399698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,128,balanced,0.20729599396387735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,128,balanced,0.2789226571718852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,128,balanced,0.35282667477925617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,1,balanced,0.09705600142478943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,1,balanced,0.10114666819572449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,1,balanced,0.11741333206494649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,1,balanced,0.15430399775505066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,1,balanced,0.21928532918294272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,1,balanced,0.3362666765848796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,1,balanced,0.4471093416213989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,1,balanced,0.44890133539835614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,1,balanced,0.45362667242685956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,1,balanced,0.45127467314402264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,1,balanced,0.456928014755249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,1,balanced,0.46727466583251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,1,balanced,0.47203731536865234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,1,balanced,0.4757546583811442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,1,balanced,0.4846133391062419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,1,balanced,0.48507734139760333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,1,balanced,0.5076320171356201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,1,balanced,0.5511093139648438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,1,balanced,0.5861013333002726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,1,balanced,0.6694453557332357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,1,balanced,0.7433280150095621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,1,balanced,0.909546693166097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,1,balanced,1.083679993947347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,1,balanced,1.4452427228291829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,1,balanced,1.9345547358194988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,1,balanced,2.663424015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,1,balanced,3.5836159388224282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,8,power_law_1.2,0.052179199457168576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,8,power_law_1.2,0.05578879714012146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,8,power_law_1.2,0.0627839982509613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,8,power_law_1.2,0.07609599828720093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,8,power_law_1.2,0.0779263973236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,8,power_law_1.2,0.07912960052490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,8,power_law_1.2,0.07917439937591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,8,power_law_1.2,0.08000640273094177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,8,power_law_1.2,0.08170239925384522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,8,power_law_1.2,0.08089600205421447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,8,power_law_1.2,0.08470399975776673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,8,power_law_1.2,0.08677120208740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,8,power_law_1.2,0.08843520283699036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,8,power_law_1.2,0.0935808002948761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,8,power_law_1.2,0.10210560560226441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,8,power_law_1.2,0.10547200441360474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,8,power_law_1.2,0.11055359840393067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,8,power_law_1.2,0.12163840532302857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,8,power_law_1.2,0.1345023989677429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,8,power_law_1.2,0.17374720573425292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,8,power_law_1.2,0.1912832021713257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,8,power_law_1.2,0.25883519649505615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,8,power_law_1.2,0.2888832092285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,8,power_law_1.2,0.36877439022064207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,8,power_law_1.2,0.5301248073577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,8,power_law_1.2,0.7722432136535644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,8,power_law_1.2,0.9509823799133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,32,power_law_1.2,0.07894399762153625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,32,power_law_1.2,0.07340800166130065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,32,power_law_1.2,0.07061120271682739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,32,power_law_1.2,0.07153279781341552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,32,power_law_1.2,0.06878719925880432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,32,power_law_1.2,0.06583679914474487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,32,power_law_1.2,0.07133439779281617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,32,power_law_1.2,0.07580800056457519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,32,power_law_1.2,0.07264000177383423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,32,power_law_1.2,0.07033600211143494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,32,power_law_1.2,0.07341439723968506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,32,power_law_1.2,0.07371519804000855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,32,power_law_1.2,0.07605760097503662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,32,power_law_1.2,0.07464960217475891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,32,power_law_1.2,0.08199040293693542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,32,power_law_1.2,0.08526080250740051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,32,power_law_1.2,0.08974080085754395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,32,power_law_1.2,0.10308480262756348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.2,0.11267839670181275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.2,0.13614720106124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.2,0.15162240266799926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.2,0.19396480321884155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.2,0.23851521015167237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.2,0.32216320037841795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.2,0.4014143943786621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.2,0.6398784160614014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.2,0.863532829284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,8,balanced,0.04201599955558777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,8,balanced,0.042303999265034996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,8,balanced,0.04560533165931702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,8,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,8,balanced,0.058373332023620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,8,balanced,0.08308266599973042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,8,balanced,0.08303999900817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,8,balanced,0.08438400427500407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,8,balanced,0.0846666693687439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,8,balanced,0.08548800150553386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,8,balanced,0.0863146682580312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,8,balanced,0.08915733297665913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,8,balanced,0.09177600344022115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,8,balanced,0.09126399954160054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,8,balanced,0.1009333332379659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,8,balanced,0.10310399532318115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,8,balanced,0.10948800047238667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,8,balanced,0.12147200107574463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,8,balanced,0.13473066687583923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,8,balanced,0.15686399737993875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,8,balanced,0.17633599042892456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,8,balanced,0.23035732905069986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,8,balanced,0.27874133984247845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,8,balanced,0.37724800904591876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,8,balanced,0.48656535148620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,8,balanced,0.6790880362192789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,8,balanced,0.8774933020273844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,4,power_law_1.01,0.06821119785308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,4,power_law_1.01,0.07331200242042542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,4,power_law_1.01,0.07761920094490052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,4,power_law_1.01,0.09774079918861389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,4,power_law_1.01,0.09793279767036438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,4,power_law_1.01,0.09886080026626587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,4,power_law_1.01,0.09968640208244324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,4,power_law_1.01,0.1003648042678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,4,power_law_1.01,0.10140160322189332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,4,power_law_1.01,0.10074880123138427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,4,power_law_1.01,0.10341119766235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,4,power_law_1.01,0.11942399740219116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,4,power_law_1.01,0.12311680316925049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,4,power_law_1.01,0.12858879566192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,4,power_law_1.01,0.14725760221481324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,4,power_law_1.01,0.1565567970275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,4,power_law_1.01,0.17475199699401855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,4,power_law_1.01,0.24352641105651857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,4,power_law_1.01,0.28316800594329833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,4,power_law_1.01,0.36846721172332764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,4,power_law_1.01,0.4673471927642822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,4,power_law_1.01,0.7504703998565674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,4,power_law_1.01,0.7668543815612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,4,power_law_1.01,1.1237759590148926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,4,power_law_1.01,1.636729621887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,16,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,16,balanced,0.03128000100453695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,16,balanced,0.031397332747777305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,16,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,16,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,16,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,16,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,16,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,16,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,16,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,16,balanced,0.03751466671625773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,16,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,16,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,16,balanced,0.03972266614437103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,16,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,16,balanced,0.044010668992996216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,16,balanced,0.04391466577847799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,16,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,16,balanced,0.0516480008761088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,16,balanced,0.05792533357938131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,16,balanced,0.06244266529877981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,16,balanced,0.08030400176843007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,16,balanced,0.10595732927322388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,16,balanced,0.14658666650454202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,16,balanced,0.18755199511845908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,16,balanced,0.25301865736643475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,16,balanced,0.3181333343187968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,4,power_law_1.01,2.471014404296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,64,power_law_1.2,0.03903999924659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,64,power_law_1.2,0.03307519853115082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,64,power_law_1.2,0.031206399202346802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,64,power_law_1.2,0.030508801341056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,64,power_law_1.2,0.031327998638153075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,64,power_law_1.2,0.03333759903907776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,64,power_law_1.2,0.03258239924907684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,64,power_law_1.2,0.033267199993133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,64,power_law_1.2,0.03386240005493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,64,power_law_1.2,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,64,power_law_1.2,0.03333759903907776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,64,power_law_1.2,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,64,power_law_1.2,0.03560320138931274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,64,power_law_1.2,0.03699840009212494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,64,power_law_1.2,0.040780800580978396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,64,power_law_1.2,0.04230400025844574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,64,power_law_1.2,0.04341759979724884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,64,power_law_1.2,0.0490880012512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,64,power_law_1.2,0.04981119930744171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,64,power_law_1.2,0.05978239774703979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,4,power_law_1.01,3.4989246368408202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,64,power_law_1.2,0.06750079989433289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,64,power_law_1.2,0.0807807981967926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,64,power_law_1.2,0.11994240283966065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,64,power_law_1.2,0.17110400199890136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,64,power_law_1.2,0.1886847972869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,64,power_law_1.2,0.2631808042526245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,64,balanced,0.03019733230272929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,64,balanced,0.027274665733178455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,64,balanced,0.027189334233601887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,64,balanced,0.02916266769170761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,64,power_law_1.2,0.3190335988998413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,64,balanced,0.03130666663249334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,64,balanced,0.03128000100453695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,64,balanced,0.030597334106763203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,64,balanced,0.03165333221356074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,64,balanced,0.03232000023126602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,64,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,64,balanced,0.032698666055997215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,64,balanced,0.031658666829268135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,64,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,64,balanced,0.03533333291610082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,64,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,64,balanced,0.0395359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,64,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,64,balanced,0.03944533318281174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,64,balanced,0.039434666434923805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,64,balanced,0.0439626673857371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,64,balanced,0.044069334864616394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,64,balanced,0.0484799991051356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,64,balanced,0.05179200073083242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,64,balanced,0.06734399994214375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,64,balanced,0.08770133058230083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,64,balanced,0.12159466743469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,64,balanced,0.15044800440470377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,2,power_law_1.01,0.06551679968833923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,2,power_law_1.01,0.08575999736785889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,2,power_law_1.01,0.10736000537872314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,2,power_law_1.01,0.15296640396118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,2,power_law_1.01,0.20723841190338135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,2,power_law_1.01,0.23077120780944824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,2,power_law_1.01,0.31264638900756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,2,power_law_1.01,0.3221503973007202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,2,power_law_1.01,0.31071999073028567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,2,power_law_1.01,0.32750720977783204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,2,power_law_1.01,0.3393856048583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,2,power_law_1.01,0.35164799690246584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,2,power_law_1.01,0.36265599727630615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,2,power_law_1.01,0.37407360076904295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,2,power_law_1.01,0.37649919986724856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,2,power_law_1.01,0.4057024002075195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,2,power_law_1.01,0.41432957649230956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,2,power_law_1.01,0.45435519218444825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,2,power_law_1.01,0.46156158447265627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,2,power_law_1.01,0.5346047878265381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,2,power_law_1.01,0.5681983947753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,2,power_law_1.01,0.6884543895721436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,2,power_law_1.01,0.7796031951904296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,2,power_law_1.01,0.9968640327453613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,2,power_law_1.01,1.223635196685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,2,power_law_1.01,1.6257535934448242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,2,power_law_1.01,2.1215232849121093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,4,power_law_1.01,0.033024001121521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,4,power_law_1.01,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,4,power_law_1.01,0.03226880133152008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,4,power_law_1.01,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,4,power_law_1.01,0.034796801209449765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,4,power_law_1.01,0.03485440015792847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,4,power_law_1.01,0.035641598701477054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,4,power_law_1.01,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,4,power_law_1.01,0.036550399661064145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,4,power_law_1.01,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,4,power_law_1.01,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,4,power_law_1.01,0.039628800749778745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,4,power_law_1.01,0.03943679928779602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,4,power_law_1.01,0.04133760035037994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,4,power_law_1.01,0.050380802154541014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,4,power_law_1.01,0.05129600167274475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,4,power_law_1.01,0.05355520248413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,4,power_law_1.01,0.06025599837303162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.01,0.06533759832382202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.01,0.08375679850578308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.01,0.09587200284004212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.01,0.12710399627685548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.01,0.1600383996963501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.01,0.19360640048980712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.01,0.25194880962371824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.01,0.34620161056518556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,128,power_law_1.2,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,128,power_law_1.2,0.033478400111198424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,128,power_law_1.2,0.03363839983940124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,128,power_law_1.2,0.03550080060958862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,128,power_law_1.2,0.03555839955806732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,128,power_law_1.2,0.03754880130290985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,128,power_law_1.2,0.037049600481987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,128,power_law_1.2,0.03772160112857818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,128,power_law_1.2,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,128,power_law_1.2,0.037862399220466615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,128,power_law_1.2,0.038848000764846805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,128,power_law_1.2,0.03904640078544617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,16,balanced,0.045653333266576133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,128,power_law_1.2,0.03935360014438629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,128,power_law_1.2,0.04041599929332733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,128,power_law_1.2,0.043315199017524716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,128,power_law_1.2,0.04348799884319306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.01,0.5046080112457275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,128,power_law_1.2,0.044972801208496095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,128,power_law_1.2,0.04690560102462769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,128,power_law_1.2,0.05002239942550659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,16,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,128,power_law_1.2,0.05585920214653015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,16,balanced,0.0645546664794286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,16,balanced,0.09197333455085754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,128,power_law_1.2,0.06455039978027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,16,balanced,0.13331733147303262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,16,balanced,0.21728533506393433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,16,balanced,0.22168533007303873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,16,balanced,0.22031466166178384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,16,balanced,0.22030933698018393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,16,balanced,0.22140266497929892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,16,balanced,0.22269866863886514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,16,balanced,0.22194133202234903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,16,balanced,0.22398932774861655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,16,balanced,0.227018674214681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,16,balanced,0.23121599356333414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,16,balanced,0.2326080004374186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,16,balanced,0.23564799626668295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,16,balanced,0.2408213416735331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,16,balanced,0.2485439976056417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,16,balanced,0.26109333833058673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,16,balanced,0.2828106681505839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,16,balanced,0.30803734064102173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,16,balanced,0.33027732372283936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,128,power_law_1.2,0.08351359963417053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,16,balanced,0.4030933380126953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,128,power_law_1.2,0.10335999727249146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,128,power_law_1.2,0.14746880531311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,128,power_law_1.2,0.19404799938201905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,128,power_law_1.2,0.2588864088058472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,16,balanced,0.44961599508921307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,16,balanced,0.5964320103327433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,16,balanced,0.6912960211435953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,128,power_law_1.2,0.3756799936294556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,2,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,2,power_law_1.2,0.03258239924907684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,2,power_law_1.2,0.033888000249862674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,2,power_law_1.2,0.037484800815582274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,2,power_law_1.2,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,2,power_law_1.2,0.05825279951095581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,2,power_law_1.2,0.07537279725074768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,2,power_law_1.2,0.0741312026977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,2,power_law_1.2,0.07827839851379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,2,power_law_1.2,0.08027520179748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,2,power_law_1.2,0.08150399923324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,2,power_law_1.2,0.08233600258827209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,2,power_law_1.2,0.08347520232200623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,2,power_law_1.2,0.08609920144081115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,2,power_law_1.2,0.08646399974822998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,2,power_law_1.2,0.09283199906349182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,2,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,2,power_law_1.2,0.09742720127105713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,2,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,2,balanced,0.0422026664018631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,2,power_law_1.2,0.10124160051345825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,2,balanced,0.043925335009892784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,2,power_law_1.2,0.11384960412979125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,2,power_law_1.2,0.12598400115966796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,2,power_law_1.2,0.14777599573135375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,2,power_law_1.2,0.1671872019767761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,2,power_law_1.2,0.21342079639434813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,2,power_law_1.2,0.24072959423065185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,2,power_law_1.2,0.3175935983657837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,2,power_law_1.2,0.37562239170074463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,2,power_law_1.2,0.6115327835083008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,2,balanced,0.04353600243727366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,2,balanced,0.04526400069395701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,2,balanced,0.04560000201066335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,2,balanced,0.04554666578769684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,2,balanced,0.045567999283472695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,2,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,2,balanced,0.04808533191680908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,2,balanced,0.04796266555786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,2,balanced,0.049498667319615684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,2,balanced,0.060405333836873375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,2,balanced,0.0620959997177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,2,balanced,0.06735466420650482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,2,balanced,0.0848479966322581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,2,balanced,0.08507733543713887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,2,balanced,0.10466667016347249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,4,balanced,0.04205866654713949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,4,balanced,0.04222933451334635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,4,balanced,0.04334400097529093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,4,balanced,0.04806933303674062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,4,balanced,0.05318933228651682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,4,balanced,0.0709386666615804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,2,balanced,0.12600533167521158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,2,balanced,0.16451199849446616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,2,balanced,0.21854400634765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,2,balanced,0.31009600559870404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,2,balanced,0.39661868413289386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,2,balanced,0.5662986834843954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,2,balanced,0.740453322728475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,4,balanced,0.07437333464622498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,4,balanced,0.0768693337837855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,4,balanced,0.07716266810894012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,4,balanced,0.07815999786059062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,4,balanced,0.07500799993673961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,4,balanced,0.07621333499749501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,4,balanced,0.07734400033950806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,4,balanced,0.0795253316561381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,4,balanced,0.08318933347860973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,4,balanced,0.08523199955622356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,4,balanced,0.08821333448092143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,4,balanced,0.0974026620388031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,4,balanced,0.10212266445159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,4,balanced,0.11742933591206868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,4,balanced,0.13268799583117166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,4,balanced,0.16925332943598428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,4,balanced,0.20758932828903198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,4,balanced,0.2837066650390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,4,balanced,0.3791999816894531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,4,balanced,0.5463519891103109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,4,balanced,0.6218933264414469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,2,power_law_1.2,0.6738687992095947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,4,balanced,0.025274666647116344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,4,balanced,0.02743999908367793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,4,balanced,0.036015999813874565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,4,balanced,0.049786667029062905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,4,balanced,0.07229866584142049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,4,balanced,0.07610133290290833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,4,balanced,0.07831466694672902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,4,balanced,0.07978133360544841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,4,balanced,0.09787733356157939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,32,power_law_1.2,0.0634112000465393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,32,power_law_1.2,0.06120960116386413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,32,power_law_1.2,0.059545600414276124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,32,power_law_1.2,0.0616320013999939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,32,power_law_1.2,0.06037759780883789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,32,power_law_1.2,0.0582144021987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,4,balanced,0.09304533402125041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,32,power_law_1.2,0.06190720200538635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,4,balanced,0.09577600161234538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,4,balanced,0.1469066639741262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,32,power_law_1.2,0.06301440000534057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,32,power_law_1.2,0.06162559986114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,32,power_law_1.2,0.06361600160598754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,32,power_law_1.2,0.0638592004776001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,32,power_law_1.2,0.06412799954414368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,32,power_law_1.2,0.0651199996471405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,32,power_law_1.2,0.06647040247917176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,4,balanced,0.1344213287035624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,4,balanced,0.1349120040734609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,4,balanced,0.16310933232307434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,4,balanced,0.16527466972668967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,4,balanced,0.1844373345375061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,4,balanced,0.18966400623321533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,4,balanced,0.19946134090423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,4,balanced,0.21206400791803995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,4,balanced,0.22560532887776694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,4,balanced,0.24044267336527506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,4,balanced,0.2608693242073059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,32,power_law_1.2,0.0721343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,4,balanced,0.4297279914220174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,32,power_law_1.2,0.07268480062484742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,4,balanced,0.4642719825108846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,32,power_law_1.2,0.07553920149803162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,4,balanced,0.6710879802703857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,32,power_law_1.2,0.08279680013656616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,32,power_law_1.2,0.08807039856910706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,32,power_law_1.2,0.10692479610443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,32,power_law_1.2,0.11891839504241944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,32,power_law_1.2,0.1540287971496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,32,power_law_1.2,0.1775871992111206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,32,power_law_1.2,0.25763840675354005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,4,balanced,0.8835679690043131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,32,power_law_1.2,0.31319680213928225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,32,power_law_1.2,0.44225921630859377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,32,power_law_1.2,0.5774208068847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,1,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,1,balanced,0.056602666775385536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,1,balanced,0.08240533371766408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,1,balanced,0.08239999910195668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,1,balanced,0.0823520024617513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,1,balanced,0.08225599924723308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,1,balanced,0.08444266517957051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,1,balanced,0.0860640009244283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,1,balanced,0.08676266670227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,1,balanced,0.08877333005269368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,1,balanced,0.08618133266766866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,1,balanced,0.08980799714724223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,1,balanced,0.0899733304977417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,1,balanced,0.09095999598503113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,1,balanced,0.1090826690196991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,1,balanced,0.10775466759999593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,1,balanced,0.10733866691589355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,1,balanced,0.1295840044816335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,1,balanced,0.1363200048605601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,1,balanced,0.1613866686820984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,1,balanced,0.19176000356674194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,1,balanced,0.24180267254511514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,1,balanced,0.3032960096995036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,1,balanced,0.4120853344599406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,1,balanced,0.5179200172424316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,1,balanced,0.7339519659678141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,1,balanced,0.9518133004506429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,2,balanced,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,2,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,2,balanced,0.05725333094596863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,2,balanced,0.056287998954455055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,2,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,2,balanced,0.05937600135803223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,2,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,2,balanced,0.059749335050582886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,2,balanced,0.06046933432420095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,2,balanced,0.06002666552861532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,2,balanced,0.062314664324124656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,2,balanced,0.061247999469439186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,2,balanced,0.06201600035031637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,2,balanced,0.06615466872851054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,2,balanced,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,2,balanced,0.07646400233109792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,2,balanced,0.07502933343251546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,2,balanced,0.08668266733487447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,2,balanced,0.08819733063379924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,2,balanced,0.1074079970518748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,2,balanced,0.11854933698972066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,2,balanced,0.15408000349998474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,2,balanced,0.18966933091481528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,2,balanced,0.25034666061401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,2,balanced,0.30057066679000854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,2,balanced,0.41384534041086835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,2,balanced,0.5236800114313761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,4,power_law_1.01,0.07219840288162231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,4,power_law_1.01,0.09408640265464782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,4,power_law_1.01,0.14401919841766359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,4,power_law_1.01,0.1506432056427002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,4,power_law_1.01,0.18974080085754394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,4,power_law_1.01,0.22142720222473145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,4,power_law_1.01,0.21736319065093995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,4,power_law_1.01,0.2738048076629639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,4,power_law_1.01,0.26486399173736574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,4,power_law_1.01,0.3029247999191284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,4,power_law_1.01,0.3010175943374634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,4,power_law_1.01,0.27818880081176756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,4,power_law_1.01,0.28707199096679686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,4,power_law_1.01,0.3147072076797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,4,power_law_1.01,0.32505600452423095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,4,power_law_1.01,0.3288127899169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,4,power_law_1.01,0.3474623918533325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,4,power_law_1.01,0.37212159633636477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,4,power_law_1.01,0.4070079803466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,4,power_law_1.01,0.47121281623840333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,4,power_law_1.01,0.49555201530456544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,4,power_law_1.01,0.5921279907226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,4,power_law_1.01,0.6525824069976807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,4,power_law_1.01,0.8698368072509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,4,power_law_1.01,0.9857664108276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,4,power_law_1.01,1.3120832443237305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,4,power_law_1.01,1.6442304611206056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,64,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,64,balanced,0.029631999631722767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,64,balanced,0.029487999776999157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,64,balanced,0.031167998909950256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,64,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,64,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,64,balanced,0.03290133426586787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,64,balanced,0.032245332996050514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,64,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,64,balanced,0.03365333378314972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,64,balanced,0.033717334270477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,64,balanced,0.033520000676314034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,64,balanced,0.033728001018365227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,64,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,64,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,64,balanced,0.03839999934037527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,64,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,64,balanced,0.03979733337958654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,64,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,64,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,64,balanced,0.04385066529115041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,64,balanced,0.05110399921735128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,64,balanced,0.05619733532269796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,64,balanced,0.07577600081761678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,64,balanced,0.09541866183280945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,64,balanced,0.12844266494115195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,64,balanced,0.15447466572125754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,2,power_law_1.2,0.05225600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,2,power_law_1.2,0.05651199817657471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,2,power_law_1.2,0.06281599998474122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,2,power_law_1.2,0.07256960272789001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,2,power_law_1.2,0.08350080251693726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,2,power_law_1.2,0.10223360061645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,2,power_law_1.2,0.10858880281448365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,2,power_law_1.2,0.12821120023727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,2,power_law_1.2,0.13000320196151732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,2,power_law_1.2,0.13639039993286134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,2,power_law_1.2,0.1320575952529907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,2,power_law_1.2,0.1442944049835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,2,power_law_1.2,0.15486719608306884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,2,power_law_1.2,0.16690560579299926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,2,power_law_1.2,0.17919360399246215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,2,power_law_1.2,0.1845247983932495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,2,power_law_1.2,0.20964479446411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,2,power_law_1.2,0.2531840085983276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,2,power_law_1.2,0.2542464017868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,2,power_law_1.2,0.31429119110107423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,2,power_law_1.2,0.37500159740447997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,2,power_law_1.2,0.497651195526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,2,power_law_1.2,0.6318848133087158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,2,power_law_1.2,0.8629440307617188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,2,power_law_1.2,1.1364735603332519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,2,power_law_1.2,1.531174373626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,2,power_law_1.2,2.0188671112060548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,1,power_law_1.2,0.05445759892463684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,1,power_law_1.2,0.07432960271835327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,1,power_law_1.2,0.11100800037384033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,1,power_law_1.2,0.14979840517044068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,1,power_law_1.2,0.18808319568634033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,1,power_law_1.2,0.32259840965270997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,1,power_law_1.2,0.3454528093338013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,1,power_law_1.2,0.34897279739379883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,1,power_law_1.2,0.36953599452972413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,1,power_law_1.2,0.375052809715271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,1,power_law_1.2,0.37580161094665526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,1,power_law_1.2,0.3839679956436157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,1,power_law_1.2,0.39299840927124025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,1,power_law_1.2,0.4045760154724121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,1,power_law_1.2,0.4262400150299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,1,power_law_1.2,0.44846720695495607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,1,power_law_1.2,0.4517824172973633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,1,power_law_1.2,0.506489610671997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,1,power_law_1.2,0.5223104000091553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,1,power_law_1.2,0.6144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,1,power_law_1.2,0.7041408061981201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,1,power_law_1.2,0.893331241607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,1,power_law_1.2,1.0660608291625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,1,power_law_1.2,1.4159296035766602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,2,balanced,0.12173866232236226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,2,balanced,0.19684799512227377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,2,balanced,0.33052267630894977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,2,balanced,0.6114399830500284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,2,balanced,1.1350666681925456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,2,balanced,1.6195039749145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,2,balanced,1.6296693483988445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,2,balanced,1.6287147204081218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,1,power_law_1.2,1.7894527435302734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,2,balanced,1.634229342142741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,2,power_law_1.2,0.0642304003238678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,2,power_law_1.2,0.07998719811439514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,2,balanced,1.6418347358703613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,2,power_law_1.2,0.0887167990207672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,2,balanced,1.6517759958902996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,2,power_law_1.2,0.11823999881744385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,2,power_law_1.2,0.14597760438919066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,2,balanced,1.6593173344930012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,2,power_law_1.2,0.1738752007484436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,2,balanced,1.6660799980163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,2,power_law_1.2,0.23722879886627196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,2,balanced,1.6793920199076335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,2,power_law_1.2,0.253056001663208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,2,power_law_1.2,0.24979839324951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,2,balanced,1.6962347030639648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,2,power_law_1.2,0.26769280433654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,2,balanced,1.7061440149943035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,2,power_law_1.2,0.27457919120788576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,2,power_law_1.2,0.271891188621521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,2,power_law_1.2,0.28407039642333987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,2,power_law_1.2,0.29499518871307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,2,power_law_1.2,0.31196160316467286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,2,power_law_1.2,0.33089919090270997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,2,power_law_1.2,0.34387199878692626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,2,balanced,1.7301653226216633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,2,power_law_1.2,0.37504000663757325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,2,power_law_1.2,0.38239359855651855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,2,balanced,1.9172426859537761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,2,power_law_1.2,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,2,power_law_1.2,0.43683199882507323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,2,power_law_1.2,0.03492479920387268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,2,power_law_1.2,0.4756800174713135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,2,power_law_1.2,0.04175359904766083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,2,power_law_1.2,0.5916351795196533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,2,power_law_1.2,0.05283839702606201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,2,power_law_1.2,0.0592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,2,power_law_1.2,0.6779967784881592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,2,power_law_1.2,0.07429760098457336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,2,power_law_1.2,0.0769919991493225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,2,power_law_1.2,0.8735936164855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,2,balanced,1.826288064320882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,2,power_law_1.2,1.1171839714050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,2,balanced,2.357034683227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,2,power_law_1.2,1.5933695793151856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,2,balanced,2.0256959597269693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,2,power_law_1.2,0.08205440044403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,2,power_law_1.2,0.08321279883384705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,2,balanced,2.9568001429239907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,2,power_law_1.2,0.08454399704933166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,2,power_law_1.2,0.0856000006198883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,2,power_law_1.2,0.08876799941062927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,2,balanced,2.433904012044271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,2,power_law_1.2,0.08980479836463928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,2,power_law_1.2,0.09802240133285522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,2,power_law_1.2,0.10338560342788697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,2,power_law_1.2,0.1107200026512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,2,balanced,4.034842809041341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,2,power_law_1.2,0.1156991958618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,2,power_law_1.2,0.13054720163345337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,2,power_law_1.2,0.14357759952545165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,2,balanced,3.8829867045084634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,2,power_law_1.2,1.8793664932250977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,2,balanced,6.481882731119792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,2,power_law_1.2,0.17895040512084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,1,power_law_1.2,2.4751871109008787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,2,power_law_1.2,0.20193920135498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,2,power_law_1.2,0.2672384023666382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,2,balanced,6.321674982706706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,2,power_law_1.2,0.36548480987548826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,2,power_law_1.2,0.46259198188781736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,2,power_law_1.2,0.5808127880096435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,2,power_law_1.2,0.8088704109191894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,1,power_law_1.2,3.1526079177856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,32,balanced,0.03551999976237615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,32,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,32,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,32,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,32,balanced,0.0432586669921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,32,balanced,0.054842665791511536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,32,balanced,0.06966400146484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,32,balanced,0.06268799801667531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,32,balanced,0.08823999762535095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,32,balanced,0.08687466382980347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,32,balanced,0.08624533812204997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,32,balanced,0.10053333640098572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,32,balanced,0.09891200065612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,32,balanced,0.09852799773216248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,32,balanced,0.11443733175595601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,32,balanced,0.1079360047976176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,32,balanced,0.10161599516868591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,32,balanced,0.121370663245519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,32,balanced,0.1390666663646698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,32,balanced,0.17577600479125977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,32,balanced,0.21194666624069214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,32,balanced,0.27384533484776813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,32,balanced,0.352730671564738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,32,balanced,0.47355735301971436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,32,balanced,0.61517866452535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,32,balanced,0.9251786867777506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,32,balanced,1.2025173505147297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,2,power_law_1.2,1.1386752128601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,128,balanced,0.017871999492247898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,128,balanced,0.017685333887736004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,128,balanced,0.019071999937295914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,128,balanced,0.01905599981546402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,128,balanced,0.02107733239730199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,128,balanced,0.02693866689999898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,128,balanced,0.025519999365011852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,128,balanced,0.027349332968393963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,128,balanced,0.026895999908447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,128,balanced,0.027210667729377747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,128,balanced,0.027280000348885853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,128,balanced,0.027189334233601887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,128,balanced,0.027509334186712902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,128,balanced,0.029205332199732464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,128,balanced,0.029253333806991577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,128,balanced,0.029520000020662945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,128,balanced,0.0312266672650973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,128,balanced,0.03376533339420954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,128,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,128,balanced,0.03915199885765711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,128,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,128,balanced,0.04807466765244802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,128,balanced,0.05417599777380625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,128,balanced,0.0664106657107671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,128,balanced,0.08459200461705525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,128,balanced,0.11349333326021831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,128,balanced,0.14334932963053384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,16,power_law_1.01,0.02232320010662079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,16,power_law_1.01,0.025216001272201537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,16,power_law_1.01,0.025593599677085875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,16,power_law_1.01,0.025760000944137572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,16,power_law_1.01,0.02630400061607361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,16,power_law_1.01,0.028934401273727418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,16,power_law_1.01,0.032153600454330446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,16,power_law_1.01,0.03374080061912536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,16,power_law_1.01,0.04217599928379059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,16,power_law_1.01,0.04227840006351471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,16,power_law_1.01,0.043833601474761966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,16,power_law_1.01,0.06103039979934692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,16,power_law_1.01,0.05971199870109558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,16,power_law_1.01,0.061267197132110596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,16,power_law_1.01,0.07313920259475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,16,power_law_1.01,0.07340160012245178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,16,power_law_1.01,0.07429760098457336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,16,power_law_1.01,0.07872639894485474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,16,power_law_1.01,0.08075519800186157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,16,power_law_1.01,0.08786560297012329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,16,power_law_1.01,0.09570559859275818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,16,power_law_1.01,0.11665279865264892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,16,power_law_1.01,0.1403839945793152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,16,power_law_1.01,0.17712639570236205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,16,power_law_1.01,0.22426240444183348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,16,power_law_1.01,0.30538880825042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,16,power_law_1.01,0.4144896030426025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,64,balanced,0.05097599824269613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,64,balanced,0.050101334849993386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,64,balanced,0.05009600023428599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,64,balanced,0.052058666944503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,64,balanced,0.05232533315817515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,64,balanced,0.060229331254959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,64,balanced,0.06260799864927928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,64,balanced,0.06439466774463654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,64,balanced,0.06447466711203258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,64,balanced,0.06437333424886067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,2,power_law_1.2,0.048256000876426695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,2,power_law_1.2,0.054790401458740236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,2,power_law_1.2,0.06681600213050842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,2,power_law_1.2,0.08528640270233154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,2,power_law_1.2,0.10080640316009522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,2,power_law_1.2,0.11663999557495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,2,power_law_1.2,0.139302396774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,2,power_law_1.2,0.14635519981384276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,2,power_law_1.2,0.1635967969894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,2,power_law_1.2,0.15519360303878785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,2,power_law_1.2,0.16599040031433104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,2,power_law_1.2,0.17687679529190065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,2,power_law_1.2,0.18290560245513915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,2,power_law_1.2,0.19579520225524902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,2,power_law_1.2,0.21043200492858888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,2,power_law_1.2,0.22005119323730468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,2,power_law_1.2,0.23434240818023683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,2,power_law_1.2,0.2633984088897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,2,power_law_1.2,0.2990976095199585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,2,power_law_1.2,0.3615295886993408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,2,power_law_1.2,0.40677762031555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,2,power_law_1.2,0.5341248035430908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,2,power_law_1.2,0.6598591804504395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,2,power_law_1.2,0.8647871971130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,2,power_law_1.2,1.1090944290161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,2,power_law_1.2,1.5729791641235351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,2,power_law_1.2,2.013279914855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,64,balanced,0.06629333396752675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,64,balanced,0.07048533360163371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,64,balanced,0.06845866640408833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,64,balanced,0.07494933406511943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,8,power_law_1.01,0.053324800729751584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,8,power_law_1.01,0.056454402208328244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,8,power_law_1.01,0.05777919888496399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,8,power_law_1.01,0.057631999254226685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,8,power_law_1.01,0.05961599946022034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,8,power_law_1.01,0.06748160123825073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,8,power_law_1.01,0.0673088014125824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,8,power_law_1.01,0.06852480173110961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,8,power_law_1.01,0.06782720088958741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,8,power_law_1.01,0.0694208025932312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,8,power_law_1.01,0.07021440267562866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,8,power_law_1.01,0.0711296021938324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,8,power_law_1.01,0.07390080094337463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,8,power_law_1.01,0.07508479952812194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,8,power_law_1.01,0.07955840229988098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,8,power_law_1.01,0.0827072024345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,8,power_law_1.01,0.08578559756278992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,8,power_law_1.01,0.10421119928359986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,8,power_law_1.01,0.10964479446411132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,8,power_law_1.01,0.13009920120239257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,8,power_law_1.01,0.16346240043640137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,8,power_law_1.01,0.2259200096130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,8,power_law_1.01,0.2662015914916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,8,power_law_1.01,0.33754239082336424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,8,power_law_1.01,0.4936192035675049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,8,power_law_1.01,0.7226111888885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,8,power_law_1.01,1.0102335929870605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,4,power_law_1.01,0.09080320000648498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,4,power_law_1.01,0.10766719579696656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,4,power_law_1.01,0.1263167977333069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,4,power_law_1.01,0.1350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,4,power_law_1.01,0.152019202709198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,4,power_law_1.01,0.17107199430465697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,4,power_law_1.01,0.1668992042541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,4,power_law_1.01,0.18984960317611693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,4,power_law_1.01,0.19191039800643922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,4,power_law_1.01,0.18812799453735352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,4,power_law_1.01,0.1906175971031189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,4,power_law_1.01,0.19122560024261476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,4,power_law_1.01,0.19886720180511475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,4,power_law_1.01,0.20160000324249266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,4,power_law_1.01,0.21469440460205078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,4,power_law_1.01,0.2097856044769287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,4,power_law_1.01,0.21687679290771483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,4,power_law_1.01,0.23714559078216552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,4,power_law_1.01,0.2703295946121216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,4,power_law_1.01,0.3078144073486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,4,power_law_1.01,0.37496318817138674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,4,power_law_1.01,0.4677760124206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,4,power_law_1.01,0.5767168045043946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,4,power_law_1.01,0.7609792232513428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,4,power_law_1.01,0.996793556213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,4,power_law_1.01,1.4072256088256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,4,power_law_1.01,1.798054313659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,64,balanced,0.09139200051625569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,64,balanced,0.09346133470535278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,64,balanced,0.09103467067082723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,64,balanced,0.10776533683141072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,64,balanced,0.11497599879900615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,64,balanced,0.13154133160909018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,64,balanced,0.14910399913787842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,64,balanced,0.1878239909807841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,64,balanced,0.22132267554601034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,64,balanced,0.2834666570027669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,64,balanced,0.35020800431569415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,64,balanced,0.47447999318440753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,64,balanced,0.6039359966913859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,8,power_law_1.01,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,8,power_law_1.01,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,8,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,8,power_law_1.01,0.035872000455856326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,8,power_law_1.01,0.035743999481201175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,8,power_law_1.01,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,8,power_law_1.01,0.03713920116424561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,8,power_law_1.01,0.03795199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,8,power_law_1.01,0.038406398892402646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,8,power_law_1.01,0.03916159868240356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,8,power_law_1.01,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,8,power_law_1.01,0.04434559941291809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,8,power_law_1.01,0.04266240000724793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,8,power_law_1.01,0.04534400105476379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,8,power_law_1.01,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,8,power_law_1.01,0.05308799743652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,8,power_law_1.01,0.05744640231132507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,8,power_law_1.01,0.06625279784202576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.01,0.0691648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.01,0.09084799885749817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.01,0.10028159618377686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,8,balanced,0.06619733572006226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,8,balanced,0.06538666784763336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,8,balanced,0.06614399949709575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,8,balanced,0.0706879993279775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,8,balanced,0.08231466511885326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,8,balanced,0.10077333450317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,8,balanced,0.114138662815094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,8,balanced,0.11426132917404175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,8,balanced,0.11825066804885864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,8,balanced,0.11527466773986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,8,balanced,0.11614400148391724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,8,balanced,0.11938132842381795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,8,balanced,0.11923733353614807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,8,balanced,0.12160533666610718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,8,balanced,0.1322933336098989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,8,balanced,0.13382400075594583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,8,balanced,0.14006933569908142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,8,balanced,0.1495306690533956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,8,balanced,0.15738667050997415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,8,balanced,0.177456001440684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,8,balanced,0.20425599813461304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,8,balanced,0.25201600790023804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,8,balanced,0.30614932378133136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,8,balanced,0.41515199343363446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,8,balanced,0.5563946564992269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,8,balanced,0.7682240009307861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,8,balanced,1.0045973459879558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.01,0.1425984025001526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.01,0.16595200300216675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,32,8,power_law_1.2,0.04567680060863495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,32,8,power_law_1.2,0.05053439736366272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.01,0.23615360260009766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,32,8,power_law_1.2,0.05015040040016174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,32,8,power_law_1.2,0.05125759840011597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,32,8,power_law_1.2,0.05335680246353149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,32,8,power_law_1.2,0.055308800935745236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,32,8,power_law_1.2,0.05725439786911011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,32,8,power_law_1.2,0.058182400465011594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,32,8,power_law_1.2,0.05904639959335327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,32,8,power_law_1.2,0.06050559878349304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,32,8,power_law_1.2,0.06213120222091675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,32,8,power_law_1.2,0.06444159746170045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,32,8,power_law_1.2,0.06549760103225707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,32,8,power_law_1.2,0.07171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,32,8,power_law_1.2,0.078847998380661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,32,8,power_law_1.2,0.0816703975200653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,32,8,power_law_1.2,0.08458880186080933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,32,8,power_law_1.2,0.09806720018386841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,32,8,power_law_1.2,0.11155200004577637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,32,8,power_law_1.2,0.14644479751586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,32,8,power_law_1.2,0.17293440103530883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,32,8,power_law_1.2,0.2385727882385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,32,8,power_law_1.2,0.2897599935531616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,32,8,power_law_1.2,0.40178561210632324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.01,0.28833279609680174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,32,8,power_law_1.2,0.5063424110412598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,32,8,power_law_1.2,0.8268159866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,32,8,power_law_1.2,0.9527359962463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,8,balanced,0.05825066566467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,8,balanced,0.05958933134873708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,8,balanced,0.06412266691525777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,2,power_law_1.01,0.0535103976726532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,2,power_law_1.01,0.06085119843482971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,8,balanced,0.08062933385372162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,8,balanced,0.09483733773231506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,8,balanced,0.10517866412798564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,8,balanced,0.10529067118962605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,8,balanced,0.10500267148017883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,8,balanced,0.10782399773597717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,8,balanced,0.10568533341089885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,2,power_law_1.01,0.07158399820327759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,8,balanced,0.10844266414642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,8,balanced,0.11059733231862386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,8,balanced,0.10957866907119751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,8,balanced,0.1132426659266154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,8,balanced,0.12077866991360982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,8,balanced,0.12255466977755229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,8,balanced,0.12489066521326701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,8,balanced,0.13495999574661255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,2,power_law_1.01,0.08241279721260071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,8,balanced,0.13863999644915262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,8,balanced,0.15307733416557312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,8,balanced,0.17371733983357748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,8,balanced,0.22760534286499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,8,balanced,0.2646346688270569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,8,balanced,0.3716426690419515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,2,power_law_1.01,0.09148799777030944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,8,balanced,0.4771360158920288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,2,power_law_1.01,0.11416959762573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,2,power_law_1.01,0.11701120138168335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,8,balanced,0.6552266677220663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,2,power_law_1.01,0.1244928002357483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,2,power_law_1.01,0.1255295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,8,balanced,0.8544426759084066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,2,power_law_1.01,0.12962559461593628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,2,power_law_1.01,0.13303680419921876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,2,power_law_1.01,0.13704320192337036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,2,power_law_1.01,0.14387840032577515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,2,power_law_1.01,0.14328960180282593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,2,power_law_1.01,0.1573248028755188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,2,power_law_1.01,0.15831040143966674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.01,0.42970881462097166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,2,power_law_1.01,0.1723456025123596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,2,power_law_1.01,0.20151679515838622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,2,power_law_1.01,0.2227776050567627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,2,power_law_1.01,0.27282559871673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,2,power_law_1.01,0.3422719955444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,2,power_law_1.01,0.43119359016418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,2,power_law_1.01,0.547327995300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,2,power_law_1.01,0.685862398147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,2,power_law_1.01,0.9602879524230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,2,power_law_1.01,1.292262363433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,2,power_law_1.01,1.5232959747314454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.01,0.5541312217712402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,1,power_law_1.2,0.09427840113639832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,1,power_law_1.2,0.10583679676055908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,1,power_law_1.2,0.12947200536727904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,1,power_law_1.2,0.17550719976425172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,1,power_law_1.2,0.2077120065689087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,1,power_law_1.2,0.25370240211486816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,1,power_law_1.2,0.34120960235595704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,1,power_law_1.2,0.34904959201812746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,1,power_law_1.2,0.3625983953475952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,1,power_law_1.2,0.3716799974441528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,1,power_law_1.2,0.38765439987182615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,1,power_law_1.2,0.38830080032348635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,1,power_law_1.2,0.3993855953216553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,1,power_law_1.2,0.41868162155151367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,1,power_law_1.2,0.4182720184326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,1,power_law_1.2,0.4332479953765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,1,power_law_1.2,0.45907840728759763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,1,power_law_1.2,0.5075583934783936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,1,power_law_1.2,0.5463808059692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,1,power_law_1.2,0.6424511909484864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,1,power_law_1.2,0.7350080013275146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,1,power_law_1.2,0.9183296203613281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,1,power_law_1.2,1.0967103958129882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,1,power_law_1.2,1.4704895973205567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,1,power_law_1.2,1.8306304931640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,1,power_law_1.2,2.552979278564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,1,power_law_1.2,3.273478317260742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,2,power_law_1.2,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,2,power_law_1.2,0.046963199973106384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,2,power_law_1.2,0.049292799830436704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,2,power_law_1.2,0.053011202812194826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,2,power_law_1.2,0.052025598287582395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,2,power_law_1.2,0.05392000079154968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,2,power_law_1.2,0.05311999917030334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,2,power_law_1.2,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,2,power_law_1.2,0.05223039984703064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,2,power_law_1.2,0.05256320238113403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,2,power_law_1.2,0.05315840244293213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,2,power_law_1.2,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,2,power_law_1.2,0.058208000659942624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,2,power_law_1.2,0.059654402732849124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,2,power_law_1.2,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,2,power_law_1.2,0.06934400200843811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,2,power_law_1.2,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,2,power_law_1.2,0.0803391993045807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,2,power_law_1.2,0.08664320111274719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,2,power_law_1.2,0.10323840379714966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,2,power_law_1.2,0.1246783971786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,2,power_law_1.2,0.16399999856948852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,2,power_law_1.2,0.19360640048980712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,64,power_law_1.01,0.03177599906921387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,64,power_law_1.01,0.028505599498748778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,64,power_law_1.01,0.02863360047340393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,64,power_law_1.01,0.030611199140548707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,64,power_law_1.01,0.030483201146125793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,64,power_law_1.01,0.031948798894882204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,64,power_law_1.01,0.03245440125465393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,64,power_law_1.01,0.032102400064468385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,64,power_law_1.01,0.032179200649261476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,64,power_law_1.01,0.03255679905414581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,64,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,64,power_law_1.01,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,64,power_law_1.01,0.03415679931640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,64,power_law_1.01,0.03585279881954193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,64,power_law_1.01,0.03857280015945434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,64,power_law_1.01,0.0390720009803772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,64,power_law_1.01,0.04056960046291351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,64,power_law_1.01,0.04602240025997162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,64,power_law_1.01,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,64,power_law_1.01,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,64,power_law_1.01,0.05829120278358459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,64,power_law_1.01,0.07357439994812012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,64,power_law_1.01,0.08689919710159302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,64,power_law_1.01,0.11585279703140258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,64,power_law_1.01,0.1435520052909851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,2,power_law_1.2,0.2798847913742065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,64,power_law_1.01,0.21719040870666503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,64,power_law_1.01,0.26124799251556396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,16,2,balanced,0.09547199805577596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,16,2,balanced,0.10065066814422607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,16,2,balanced,0.1053493320941925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,16,2,balanced,0.12517333030700684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,16,2,balanced,0.14415466785430908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,16,2,balanced,0.1906933387120565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,16,2,balanced,0.19109867016474405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,16,2,balanced,0.19165333112080893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,16,2,balanced,0.1965706745783488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,16,2,balanced,0.19463467597961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,16,2,balanced,0.19824532667795816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,16,2,balanced,0.20220800240834555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,16,2,balanced,0.2051573395729065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,16,2,balanced,0.21049600839614868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,16,2,balanced,0.21387734015782675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,16,2,balanced,0.21659199396769205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,16,2,balanced,0.22612800200780234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,16,2,balanced,0.2541760007540385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,16,2,balanced,0.27426666021347046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,16,2,balanced,0.3299199938774109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,16,2,balanced,0.3813920021057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,16,2,balanced,0.4809866746266683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,16,2,balanced,0.5821439822514852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,16,2,balanced,0.851365327835083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,16,2,balanced,1.0558666388193767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,16,2,balanced,1.5340426762898762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,16,2,balanced,2.0022719701131186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,2,power_law_1.2,0.34268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,16,balanced,0.05189866820971171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,16,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,16,balanced,0.05590933561325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,16,balanced,0.06646933158238728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,16,balanced,0.07230400045712788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,16,balanced,0.0813973347345988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,16,balanced,0.08533333738644917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,16,balanced,0.08185066779454549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,16,balanced,0.0844640036424001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,16,balanced,0.08278400202592213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,16,balanced,0.08090133468310039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,16,balanced,0.08249066770076752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,16,balanced,0.08514133095741272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,16,balanced,0.0849226713180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,16,balanced,0.08866666754086812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,16,balanced,0.09057600299517314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,16,balanced,0.09089066584904988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,16,balanced,0.09936533371607463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,16,balanced,0.10418132940928142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,16,balanced,0.11797333757082622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,16,balanced,0.13006933530171713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,16,balanced,0.1721973419189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,16,balanced,0.18991466363271078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,16,balanced,0.2523946762084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,16,balanced,0.32286399602890015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,16,balanced,0.4225920041402181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,16,balanced,0.5538080135981241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,2,power_law_1.2,0.5580543994903564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,8,balanced,0.02004266654451688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,8,balanced,0.019600000232458115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,8,balanced,0.01918399954835574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,8,balanced,0.0191040001809597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,8,balanced,0.020879998803138733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,8,balanced,0.023317334552605946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,8,balanced,0.0407679999868075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,8,balanced,0.04138133426507314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,8,balanced,0.03668266783157984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,8,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,8,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,8,balanced,0.04275733232498169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,8,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,8,balanced,0.035258665680885315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,8,balanced,0.04474133253097534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,8,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,8,balanced,0.045781334241231285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,8,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,8,balanced,0.06633600095907848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,8,balanced,0.08689066767692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,8,balanced,0.10829866925875346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,8,balanced,0.1469013293584188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,4,power_law_1.01,0.0864575982093811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,4,power_law_1.01,0.10224000215530396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,4,power_law_1.01,0.10675840377807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,4,power_law_1.01,0.11644159555435181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,4,power_law_1.01,0.13211519718170167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,4,power_law_1.01,0.16143360137939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,4,power_law_1.01,0.15731199979782104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,8,balanced,0.1813653310139974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,8,balanced,0.25571733713150024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,8,balanced,0.3280479907989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,8,balanced,0.4758400122324626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,8,balanced,0.6221760114034017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,4,power_law_1.01,0.16061439514160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,4,power_law_1.01,0.1579200029373169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,4,power_law_1.01,0.1549504041671753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,4,power_law_1.01,0.16228480339050294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,4,power_law_1.01,0.15960960388183593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,4,power_law_1.01,0.16394879817962646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,4,power_law_1.01,0.16671359539031982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,4,power_law_1.01,0.1750272035598755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,4,power_law_1.01,0.17452800273895264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,4,power_law_1.01,0.1827455997467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,4,power_law_1.01,0.20281600952148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.01,0.22199039459228515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.01,0.2603840112686157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.01,0.3004096031188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.01,0.3854720115661621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.01,0.4739071846008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.01,0.6121471881866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.01,0.7911295890808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.01,1.1148799896240233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.01,1.4717568397521972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,2,power_law_1.2,0.648473596572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,2,balanced,0.046240001916885376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,2,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,2,balanced,0.06769066552321117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,2,balanced,0.0981760025024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,2,balanced,0.1405333379904429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,2,balanced,0.23286932706832886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,2,balanced,0.23326400915781656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,2,balanced,0.2360480030377706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,2,balanced,0.2367466688156128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,2,balanced,0.24037333329518637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,2,balanced,0.2404693365097046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,2,balanced,0.24558399120966592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,2,balanced,0.2494879961013794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,2,balanced,0.2545706629753113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,2,balanced,0.2656906644503276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,2,balanced,0.2669866681098938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,2,balanced,0.2769706646601359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,2,balanced,0.3038346568743388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,2,balanced,0.32473599910736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,2,balanced,0.3685866594314575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,2,balanced,0.41445334752400714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,2,balanced,0.5324853261311849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,2,balanced,0.612122654914856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,2,balanced,0.8337600231170654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,2,balanced,1.0071252981821697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,2,balanced,1.452965259552002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,2,balanced,1.848848025004069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,8,balanced,0.01978133370478948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,8,balanced,0.01932266727089882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,8,balanced,0.019146667172511418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,8,balanced,0.019333332777023315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,8,balanced,0.02102400114138921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,8,balanced,0.023130667706330616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,8,balanced,0.045978665351867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,8,balanced,0.04635733366012573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,8,balanced,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,8,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,8,balanced,0.03473600000143051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,8,balanced,0.035930665830771126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,8,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,8,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,8,balanced,0.03961066653331121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,8,balanced,0.03799466788768768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,8,balanced,0.04257066547870636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,8,balanced,0.04498666524887085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,8,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,8,balanced,0.06772799789905548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,8,balanced,0.07628266513347626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,8,balanced,0.10188266634941101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,8,balanced,0.12612799803415933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,8,balanced,0.17321600516637167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,8,balanced,0.21228265762329102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,8,balanced,0.30770667394002277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,8,balanced,0.39580265680948895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,1,balanced,0.0545066644748052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,1,balanced,0.06862933437029521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,1,balanced,0.09468799829483032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,1,balanced,0.14756266276041666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,1,balanced,0.2320800026257833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,1,balanced,0.4091626803080241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,1,balanced,0.4070880015691121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,1,balanced,0.41805867354075116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,1,balanced,0.420474648475647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,1,balanced,0.4256853262583415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,1,balanced,0.4231413205464681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,1,balanced,0.43616533279418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,1,balanced,0.43857065836588544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,1,balanced,0.4426186482111613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,1,balanced,0.4570666551589966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,1,balanced,0.4617066780726115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,1,balanced,0.4749759833017985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,1,balanced,0.521941343943278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,1,balanced,0.5562133391698202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,1,balanced,0.63645867506663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,1,balanced,0.708021322886149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,1,balanced,0.9222880204518636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,1,balanced,1.0354026953379314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,8,power_law_1.01,0.11662080287933349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,8,power_law_1.01,0.12127360105514526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,1,balanced,1.4603734016418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,8,power_law_1.01,0.11845120191574096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,8,power_law_1.01,0.11809279918670654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,1,balanced,1.7279359499613445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,8,power_law_1.01,0.11856000423431397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,8,power_law_1.01,0.12577919960021972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,8,power_law_1.01,0.162118399143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,1,balanced,2.5009120305379233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,8,power_law_1.01,0.1960576057434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,8,power_law_1.01,0.20609281063079835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,8,power_law_1.01,0.24413440227508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,1,balanced,3.1778507232666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,8,power_law_1.01,0.1515391945838928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,8,power_law_1.01,0.19461120367050172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,8,power_law_1.01,0.1891711950302124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,8,power_law_1.01,0.2273792028427124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,8,power_law_1.01,0.2940416097640991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,8,power_law_1.01,0.28584320545196534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,8,power_law_1.01,0.27654399871826174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,8,power_law_1.01,0.34204161167144775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,8,power_law_1.01,0.41573119163513184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,8,power_law_1.2,0.06802560091018676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,8,power_law_1.2,0.07746559977531434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,8,power_law_1.01,0.5819071769714356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,8,power_law_1.2,0.07635200023651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,8,power_law_1.01,0.5974783897399902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,8,power_law_1.2,0.08119040131568908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,8,power_law_1.2,0.09144960045814514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,8,power_law_1.2,0.09207680225372314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,8,power_law_1.2,0.09306880235671997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,8,power_law_1.2,0.09478399753570557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,8,power_law_1.2,0.09192960262298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,8,power_law_1.2,0.09397760033607483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,8,power_law_1.2,0.0942463994026184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,8,power_law_1.2,0.09452800154685974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,8,power_law_1.2,0.09783679842948914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,8,power_law_1.2,0.10092799663543701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,8,power_law_1.2,0.10620800256729127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,8,power_law_1.2,0.1065600037574768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,8,power_law_1.2,0.11527680158615113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,8,power_law_1.2,0.12683520317077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,8,power_law_1.2,0.14034559726715087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,8,power_law_1.2,0.1670912027359009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,8,power_law_1.2,0.18776320219039916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,8,power_law_1.2,0.2519040107727051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,8,power_law_1.2,0.2859584093093872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,8,power_law_1.01,1.0655743598937988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,8,power_law_1.2,0.38279678821563723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,8,power_law_1.2,0.5155648231506348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,8,power_law_1.2,0.7755072116851807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,8,power_law_1.2,1.010707187652588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,8,power_law_1.01,1.3600447654724122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,8,power_law_1.01,2.1474559783935545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,8,power_law_1.01,2.785113525390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,8,power_law_1.01,0.05736960172653198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,8,power_law_1.01,0.07307519912719726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,8,power_law_1.01,0.08201599717140198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,8,power_law_1.01,0.09717760086059571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,8,power_law_1.01,0.10004479885101318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,8,power_law_1.01,0.09830399751663207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,8,power_law_1.01,0.10590720176696777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,8,power_law_1.01,0.10388480424880982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,8,power_law_1.01,0.10207359790802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,8,power_law_1.01,0.10207359790802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,8,power_law_1.01,0.10329600572586059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,8,power_law_1.01,0.10535039901733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,8,power_law_1.01,0.11201280355453491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,8,power_law_1.01,0.11413760185241699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,8,power_law_1.01,0.11549439430236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,8,power_law_1.01,0.11416959762573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,8,power_law_1.01,0.12210559844970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,8,power_law_1.01,0.1332479953765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,8,power_law_1.01,0.14014079570770263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,8,power_law_1.01,0.16002559661865234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,8,power_law_1.01,0.17788159847259521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,8,power_law_1.01,0.2205183982849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,8,power_law_1.01,0.24305920600891112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,8,power_law_1.01,0.3110912084579468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,8,power_law_1.01,0.39159679412841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,8,power_law_1.01,0.5188352108001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,8,power_law_1.01,0.7039360046386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,8,power_law_1.01,3.985990524291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,8,power_law_1.01,5.276774215698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,8,balanced,0.0390079990029335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,8,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,8,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,8,balanced,0.03992533435424169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,8,balanced,0.04167466859022776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,8,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,8,balanced,0.04144533226887385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,8,balanced,0.04177600145339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,8,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,8,balanced,0.04118400067090988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,8,balanced,0.040474665661652885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,8,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,8,balanced,0.04329599936803182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,8,balanced,0.041573333243529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,8,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,8,balanced,0.0440533310174942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,8,balanced,0.045610666275024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,8,balanced,0.05186666548252106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,8,balanced,0.04980266590913137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,8,balanced,0.06208533545335134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,8,balanced,0.06409066418806712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,8,balanced,0.07801066835721333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,8,balanced,0.09258666634559631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,8,balanced,0.11958932876586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,8,balanced,0.14220266540845236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,8,balanced,0.19137599070866904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,8,balanced,0.24252265691757202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,4,balanced,0.022442666192849476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,4,balanced,0.022090665996074677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,4,balanced,0.021162666380405426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,4,balanced,0.023120000958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,4,balanced,0.02310933421055476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,4,balanced,0.024143998821576435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,4,balanced,0.03148266673088074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,4,balanced,0.027482666075229645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,4,balanced,0.05295999844868978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,4,balanced,0.05242133140563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,4,balanced,0.053616002202034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,4,balanced,0.053455998500188194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,4,balanced,0.05189333359400431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,4,balanced,0.041834667325019836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,4,balanced,0.06343466540177663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,4,balanced,0.053770666321118675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,2,balanced,0.06453866759936015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,4,balanced,0.0605973352988561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,2,balanced,0.06968000034491222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,2,balanced,0.09307733178138733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,2,balanced,0.1329973340034485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,2,balanced,0.19343467553456625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,2,balanced,0.32180800040562946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,2,balanced,0.3212053378423055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,2,balanced,0.32786667346954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,4,balanced,0.07447466750939687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,4,balanced,0.09264000256856282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,2,balanced,0.3272479971249898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,4,balanced,0.12839999794960022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,2,balanced,0.3294666608174642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,4,balanced,0.15501333276430765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,4,balanced,0.21488000949223837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,4,balanced,0.2709386746088664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,4,balanced,0.3866560061772664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,4,balanced,0.5012266635894775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,4,balanced,0.7330719629923502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,2,balanced,0.3383626540501912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,4,balanced,0.9652693271636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,2,balanced,0.33763734499613446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,2,balanced,0.3423253297805786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,2,balanced,0.34893866380055744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,2,balanced,0.3571573495864868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,2,balanced,0.3636106650034587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,2,balanced,0.37273601690928143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,2,balanced,0.40305066108703613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,2,balanced,0.43134931723276776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,2,balanced,0.4843839804331462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,2,balanced,0.5170240004857382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,2,balanced,0.6181439956029257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,2,balanced,0.7101386388142904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,2,balanced,0.9951732953389486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,2,balanced,1.1929226716359456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,2,balanced,1.700757344563802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,2,balanced,2.1599253018697104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,1,power_law_1.01,0.0572160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,1,power_law_1.01,0.07683839797973632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,1,power_law_1.01,0.1003648042678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,1,power_law_1.01,0.1565567970275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,1,power_law_1.01,0.2137216091156006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,1,power_law_1.01,0.26449921131134035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,1,power_law_1.01,0.35664639472961424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,1,power_law_1.01,0.37147519588470457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,1,power_law_1.01,0.3875839948654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,1,power_law_1.01,0.39925758838653563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,1,power_law_1.01,0.4084159851074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,1,power_law_1.01,0.4263296127319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,1,power_law_1.01,0.44009599685668943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,1,power_law_1.01,0.4519807815551758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,1,power_law_1.01,0.4638783931732178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,1,power_law_1.01,0.4823103904724121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,1,power_law_1.01,0.5192319869995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,1,power_law_1.01,0.5870848178863526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,1,power_law_1.01,0.6169727802276611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,1,power_law_1.01,0.7401408195495606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,1,power_law_1.01,0.8386495590209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,1,power_law_1.01,1.057107162475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,1,power_law_1.01,1.2614336013793945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,1,power_law_1.01,1.6464256286621093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,1,power_law_1.01,2.046112060546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,1,power_law_1.01,2.825395202636719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,1,power_law_1.01,3.6219455718994142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,1,power_law_1.2,0.03431679904460907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,1,power_law_1.2,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,1,power_law_1.2,0.061414402723312375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,1,power_law_1.2,0.08721920251846313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,1,power_law_1.2,0.10732799768447876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,1,power_law_1.2,0.16996480226516725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,1,power_law_1.2,0.18268799781799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,1,power_law_1.2,0.18496639728546144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,1,power_law_1.2,0.18969600200653075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,1,power_law_1.2,0.20039680004119872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,1,power_law_1.2,0.20971519947052003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,1,power_law_1.2,0.21086080074310304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,1,power_law_1.2,0.21549439430236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,1,power_law_1.2,0.21624319553375243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,1,power_law_1.2,0.23379199504852294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,1,power_law_1.2,0.24067199230194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,1,power_law_1.2,0.24483199119567872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,1,power_law_1.2,0.28089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,1,power_law_1.2,0.2837631940841675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,1,power_law_1.2,0.3333695888519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,1,power_law_1.2,0.35632638931274413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,1,power_law_1.2,0.42705278396606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,1,power_law_1.2,0.469542407989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,1,power_law_1.2,0.5993599891662598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,1,power_law_1.2,0.7833983898162842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,1,power_law_1.2,1.040940761566162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,1,power_law_1.2,1.2337023735046386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,16,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,16,balanced,0.04607999821503957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,16,balanced,0.04770666857560476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,16,balanced,0.04755199948946635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,16,balanced,0.05421333511670431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,16,balanced,0.07807466884454091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,16,balanced,0.07830933233102162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,16,balanced,0.07931200166543324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,16,balanced,0.08067733546098073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,16,balanced,0.08145066599051158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,16,balanced,0.08099733293056488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,16,balanced,0.08450667063395183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,16,balanced,0.08335466186205547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,16,balanced,0.08659733335177104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,16,balanced,0.09325333436330159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,16,balanced,0.09540266791979472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,16,balanced,0.10097600022951762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,16,balanced,0.11345066626866658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,16,balanced,0.12000532944997151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,16,balanced,0.14017599821090698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,16,balanced,0.1592639982700348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,16,balanced,0.20082134008407593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,16,balanced,0.2412853240966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,16,balanced,0.3185546596844991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,16,balanced,0.40137068430582684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,16,balanced,0.5621920029322306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,16,balanced,0.7220213413238525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,1,power_law_1.2,0.07511680126190186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,1,power_law_1.2,0.09166079759597778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,1,power_law_1.2,0.1221951961517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,1,power_law_1.2,0.1619328022003174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,1,power_law_1.2,0.16234879493713378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,1,power_law_1.2,0.1642303943634033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,1,power_law_1.2,0.16832000017166138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,1,power_law_1.2,0.16183680295944214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,1,power_law_1.2,0.15645439624786378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,1,power_law_1.2,0.16378240585327147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,1,power_law_1.2,0.15866880416870116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,1,power_law_1.2,0.15886720418930053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,1,power_law_1.2,0.17239680290222167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,1,power_law_1.2,0.1751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,1,power_law_1.2,0.18436479568481445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,1,power_law_1.2,0.18616960048675538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,1,power_law_1.2,0.21500160694122314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,1,power_law_1.2,0.25112960338592527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.2,0.2671360015869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.2,0.3251967906951904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.2,0.40474882125854494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.2,0.5377471923828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.2,0.6684288024902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.2,0.9266304016113281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.2,1.1989760398864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.2,1.7188543319702148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,32,balanced,0.06496533254782359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,32,balanced,0.06585599978764851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,32,balanced,0.06425066788991292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,32,balanced,0.07441066702206929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,32,balanced,0.09317333499590556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,32,balanced,0.12017599741617839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,32,balanced,0.16691199938456217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,32,balanced,0.15916267037391663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,32,balanced,0.16180800398190817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,32,balanced,0.15917332967122397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,32,balanced,0.16243732968966165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,32,balanced,0.16552000244458517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,32,balanced,0.16446933150291443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,32,balanced,0.16340800126393637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,32,balanced,0.16684800386428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,32,balanced,0.1697439948717753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,32,balanced,0.17540266116460165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,32,balanced,0.17947733402252197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,32,balanced,0.18791999419530234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,32,balanced,0.2021333376566569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,32,balanced,0.21741332610448202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,32,balanced,0.2458933393160502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,32,balanced,0.2751946647961934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,32,balanced,0.3354453245798747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,32,balanced,0.42583465576171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,32,balanced,0.5334773461023966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,32,balanced,0.6847519874572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,128,power_law_1.2,0.05671679973602295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,128,power_law_1.2,0.05551999807357788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.2,2.231270408630371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,128,power_law_1.2,0.05571200251579285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,128,power_law_1.2,0.05909759998321533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,128,power_law_1.2,0.060864001512527466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,128,power_law_1.2,0.06440960168838501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,128,power_law_1.2,0.06925439834594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,128,power_law_1.2,0.071424001455307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,128,power_law_1.2,0.07648640275001525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,128,power_law_1.2,0.07644799947738648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,128,power_law_1.2,0.07267839908599853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,128,power_law_1.2,0.07495679855346679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,128,power_law_1.2,0.08240000009536744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,128,power_law_1.2,0.07882879972457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,128,power_law_1.2,0.09333760142326356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,128,power_law_1.2,0.0960640013217926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,128,power_law_1.2,0.10088959932327271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,128,power_law_1.2,0.11397119760513305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,128,power_law_1.2,0.1290112018585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,128,power_law_1.2,0.1598847985267639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,128,power_law_1.2,0.1769600033760071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,128,power_law_1.2,0.2593408107757568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,128,power_law_1.2,0.32483839988708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,128,power_law_1.2,0.45630722045898436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,128,power_law_1.2,0.6830399990081787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,128,power_law_1.2,1.1152447700500487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,128,power_law_1.2,1.3910592079162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,2,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,2,balanced,0.049685334165891014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,2,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,2,balanced,0.05387733379999796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,2,balanced,0.0634933312733968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,2,balanced,0.08382399876912434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,2,balanced,0.0844533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,2,balanced,0.08642133076985677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,2,balanced,0.08725866675376892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,2,balanced,0.08896000186602275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,2,balanced,0.08987200260162354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,2,balanced,0.09269866347312927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,2,balanced,0.09417066971460979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,2,balanced,0.09890133142471313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,2,balanced,0.10398933291435242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,2,balanced,0.10593600074450175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,2,balanced,0.11519466837247212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,2,balanced,0.13082666198412576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,2,balanced,0.14321600397427878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,2,balanced,0.17124799887339273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,2,balanced,0.19703465700149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,2,balanced,0.27211199204126996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,2,balanced,0.3210080067316691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,2,balanced,0.4724479913711548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,2,balanced,0.5858346621195475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,2,balanced,0.8675786654154459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,2,balanced,1.1341119607289631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,1,balanced,0.09937066833178203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,1,balanced,0.11802132924397786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,1,balanced,0.1504639983177185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,1,balanced,0.22047466039657593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,1,balanced,0.34644798437754315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,1,balanced,0.6061386664708456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,1,balanced,0.8537226517995199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,1,balanced,0.8601546287536621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,1,balanced,0.863754669825236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,1,balanced,0.870965321858724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,1,balanced,0.8737920125325521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,1,balanced,0.8966933091481527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,1,balanced,0.9110506375630697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,1,balanced,0.9000106652577718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,1,balanced,0.9167253176371256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,1,balanced,0.9354986349741617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,1,balanced,0.9610666433970133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,1,balanced,1.0089813073476155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,1,balanced,1.0669866402943928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,1,balanced,1.1834293206532795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,1,balanced,1.233413298924764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,1,balanced,1.4200372695922852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,1,balanced,1.6098666191101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,1,balanced,1.9794559478759766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,1,balanced,2.595344066619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,1,balanced,3.3384532928466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,1,balanced,4.553338686625163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,2,balanced,0.05629333357016245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,2,balanced,0.0746613343556722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,2,balanced,0.10125866532325745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,2,balanced,0.1493066648642222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,2,balanced,0.2502506573994954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,2,balanced,0.44971199830373126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,2,balanced,0.45019733905792236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,2,balanced,0.4517813523610433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,2,balanced,0.4535946846008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,2,balanced,0.45841066042582196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,2,balanced,0.4611039956410726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,2,balanced,0.466325322786967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,2,balanced,0.4710986614227295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,2,balanced,0.4785439968109131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,2,balanced,0.4898293415705363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,2,balanced,0.4980693260828654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,2,balanced,0.5046026706695557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,2,balanced,0.5356640021006266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,2,balanced,0.5592106580734253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,2,balanced,0.6055999994277954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,2,balanced,0.6722133159637451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,2,balanced,0.7705866495768229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,2,balanced,0.8712960084279379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,2,balanced,1.176357348759969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,2,balanced,1.3326720396677654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,2,balanced,1.9253652890523274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,2,balanced,2.3737600644429526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,128,power_law_1.01,0.06684160232543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,128,power_law_1.01,0.05334399938583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,128,power_law_1.01,0.058348798751831056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,128,power_law_1.01,0.051635199785232545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,128,power_law_1.01,0.052108800411224364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,128,power_law_1.01,0.052198398113250735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,128,power_law_1.01,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,128,power_law_1.01,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,128,power_law_1.01,0.054611200094223024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,128,power_law_1.01,0.05496320128440857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,128,power_law_1.01,0.05486720204353333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,128,power_law_1.01,0.056985598802566526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,128,power_law_1.01,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,128,power_law_1.01,0.05886719822883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,128,power_law_1.01,0.06524800062179566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,128,power_law_1.01,0.06490240097045899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,128,power_law_1.01,0.07125120162963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,128,power_law_1.01,0.08241279721260071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,128,power_law_1.01,0.0894976019859314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,128,power_law_1.01,0.10944000482559205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,128,power_law_1.01,0.12751359939575196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,128,power_law_1.01,0.16883840560913085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,128,power_law_1.01,0.1966976046562195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,128,power_law_1.01,0.2632895946502686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,128,power_law_1.01,0.33777918815612795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,128,power_law_1.01,0.4677248001098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,128,power_law_1.01,0.6157248020172119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,4,power_law_1.2,0.02122880071401596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,4,power_law_1.2,0.021561600267887115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,4,power_law_1.2,0.023449599742889404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,4,power_law_1.2,0.027276799082756042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,4,power_law_1.2,0.02855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,4,power_law_1.2,0.031052801012992858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,4,power_law_1.2,0.034457600116729735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,4,power_law_1.2,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,4,power_law_1.2,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,4,power_law_1.2,0.03697279989719391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,4,power_law_1.2,0.041791999340057374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,4,power_law_1.2,0.04423680007457733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,4,power_law_1.2,0.049292799830436704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,4,power_law_1.2,0.04467200040817261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,4,power_law_1.2,0.053439998626708986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,4,power_law_1.2,0.05611519813537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,4,power_law_1.2,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,4,power_law_1.2,0.07342720031738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,4,power_law_1.2,0.07961599826812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,4,power_law_1.2,0.10097279548645019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,4,power_law_1.2,0.12480640411376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,4,power_law_1.2,0.1497599959373474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,4,power_law_1.2,0.16227200031280517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,4,power_law_1.2,0.21251199245452881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,4,power_law_1.2,0.2673856019973755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,4,power_law_1.2,0.3614464044570923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,4,power_law_1.2,0.47895040512084963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,8,balanced,0.04269866645336151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,8,balanced,0.043866669138272606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,8,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,8,balanced,0.047685335079828896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,8,balanced,0.052015999952952065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,8,balanced,0.05868266522884369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,8,balanced,0.061674664417902626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,8,balanced,0.06265600025653839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,8,balanced,0.06221333146095276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,8,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,8,balanced,0.06624533236026764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,8,balanced,0.06611200173695882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,8,balanced,0.06740800042947133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,8,balanced,0.07082133491834004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,8,balanced,0.07604266703128815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,8,balanced,0.07874133189519246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,8,balanced,0.08381332953770955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,8,balanced,0.09673600395520528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,8,balanced,0.11107732852300008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,8,balanced,0.13220799962679544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,8,balanced,0.15038933356602988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,8,balanced,0.201855997244517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,8,balanced,0.23783467213312784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,8,balanced,0.33166933059692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,8,balanced,0.4285386800765991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,8,balanced,0.6134613355000814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,8,balanced,0.7763840357462565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,2,power_law_1.01,0.018918399512767792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,2,power_law_1.01,0.0204352006316185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,2,power_law_1.01,0.024006399512290954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,2,power_law_1.01,0.031206399202346802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,2,power_law_1.01,0.039628800749778745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,2,power_law_1.01,0.04389120042324066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,2,power_law_1.01,0.05081599950790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,2,power_law_1.01,0.05244799852371216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,2,power_law_1.01,0.05397120118141174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,2,power_law_1.01,0.05580800175666809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,2,power_law_1.01,0.05798400044441223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,2,power_law_1.01,0.058771198987960814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,2,power_law_1.01,0.0599232017993927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,2,power_law_1.01,0.06119040250778198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,2,power_law_1.01,0.06357120275497437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,2,power_law_1.01,0.06603519916534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,2,power_law_1.01,0.06945279836654664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,2,power_law_1.01,0.08046720027923585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,2,power_law_1.01,0.0866752028465271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,2,power_law_1.01,0.11039999723434449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,2,power_law_1.01,0.11010559797286987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,2,power_law_1.01,0.13563519716262817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,2,power_law_1.01,0.17043839693069457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,2,power_law_1.01,0.2076672077178955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,2,power_law_1.01,0.23914239406585694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,2,power_law_1.01,0.32330880165100095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,2,power_law_1.01,0.03934719860553741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,2,power_law_1.01,0.041555199027061465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,2,power_law_1.01,0.04279040098190308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,2,power_law_1.01,0.04647679924964905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,2,power_law_1.01,0.04572800099849701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,2,power_law_1.01,0.04749439954757691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,2,power_law_1.01,0.047654399275779726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,2,power_law_1.01,0.04748159945011139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,2,power_law_1.01,0.04692479968070984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,2,power_law_1.01,0.04655359983444214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,2,power_law_1.01,0.40221438407897947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,2,power_law_1.01,0.046963199973106384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,2,power_law_1.01,0.04864639937877655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,2,power_law_1.01,0.0482367992401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,2,power_law_1.01,0.05013120174407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,2,power_law_1.01,0.05583999752998352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,2,power_law_1.01,0.05804799795150757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,2,power_law_1.01,0.06254079937934875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,2,power_law_1.01,0.07013760209083557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.01,0.08483840227127075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.01,0.10045440196990967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.01,0.10951679944992065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.01,0.15118080377578735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.01,0.19671679735183717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.01,0.23900799751281737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.01,0.35523200035095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,32,power_law_1.2,0.067603200674057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,32,power_law_1.2,0.06531199812889099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.01,0.4316415786743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,32,power_law_1.2,0.06157439947128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,32,power_law_1.2,0.06419199705123901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,32,power_law_1.2,0.06409599781036376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,32,power_law_1.2,0.06299520134925843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,32,power_law_1.2,0.06518399715423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,32,power_law_1.2,0.06714879870414733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,32,power_law_1.2,0.06785280108451844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,32,power_law_1.2,0.06716799736022949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,32,power_law_1.2,0.06754559874534607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,32,power_law_1.2,0.06880639791488648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,32,power_law_1.2,0.06967039704322815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,32,power_law_1.2,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,32,power_law_1.2,0.0772159993648529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,32,power_law_1.2,0.0796671986579895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,32,power_law_1.2,0.0862272024154663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,32,power_law_1.2,0.1007807970046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,32,power_law_1.2,0.11068160533905029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,32,power_law_1.2,0.14436479806900024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,32,power_law_1.2,0.15527039766311646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,32,power_law_1.2,0.2097088098526001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,32,power_law_1.2,0.2643392086029053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,32,power_law_1.2,0.34654080867767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,32,power_law_1.2,0.46520318984985354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,32,power_law_1.2,0.6597375869750977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,32,power_law_1.2,0.8763456344604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.01,0.5431168079376221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,2,power_law_1.2,0.05238400101661682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,2,power_law_1.2,0.06517120003700257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,2,power_law_1.2,0.08183040022850037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,2,power_law_1.2,0.09645439982414246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,2,power_law_1.2,0.10944639444351197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,2,power_law_1.2,0.1515776038169861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,2,power_law_1.2,0.16298880577087402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,2,power_law_1.2,0.17248640060424805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,2,power_law_1.2,0.17445119619369506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,2,power_law_1.2,0.1745471954345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,2,power_law_1.2,0.18778879642486573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,2,power_law_1.2,0.19162880182266234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,2,power_law_1.2,0.1954751968383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,2,power_law_1.2,0.20414719581604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,2,power_law_1.2,0.21857280731201173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,2,power_law_1.2,0.21904640197753905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,2,power_law_1.2,0.2354367971420288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,2,power_law_1.2,0.2630336046218872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,2,power_law_1.2,0.29288320541381835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,2,power_law_1.2,0.3547136068344116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,2,power_law_1.2,0.42636799812316895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,2,power_law_1.2,0.4866879940032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,2,power_law_1.2,0.6639616012573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,2,power_law_1.2,0.8547455787658691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,2,power_law_1.2,1.0311488151550292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,2,power_law_1.2,1.4302656173706054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,2,power_law_1.2,1.923321533203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,16,power_law_1.01,0.044268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,16,power_law_1.01,0.0361407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,16,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,16,power_law_1.01,0.03182080090045929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,16,power_law_1.01,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,16,power_law_1.01,0.033843201398849485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,16,power_law_1.01,0.03368319869041443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,16,power_law_1.01,0.03403519988059998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,16,power_law_1.01,0.034764799475669864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,16,power_law_1.01,0.035545599460601804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,16,power_law_1.01,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,16,power_law_1.01,0.0385919988155365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,16,power_law_1.01,0.03891839981079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,16,power_law_1.01,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,16,power_law_1.01,0.045510399341583255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,16,power_law_1.01,0.04717440009117126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,16,power_law_1.01,0.047628799080848695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,16,power_law_1.01,0.05124480128288269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,16,power_law_1.01,0.05914880037307739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,16,power_law_1.01,0.06969599723815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,16,power_law_1.01,0.07667840123176575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,16,power_law_1.01,0.10533119440078735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,16,power_law_1.01,0.11556479930877686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,16,power_law_1.01,0.16758400201797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,16,power_law_1.01,0.2113408088684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,16,power_law_1.01,0.28282239437103274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,16,power_law_1.01,0.34403200149536134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,2,power_law_1.01,0.06049280166625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,2,power_law_1.01,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,2,power_law_1.01,0.07009919881820678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,2,power_law_1.01,0.08321279883384705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,2,power_law_1.01,0.08298239707946778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,2,power_law_1.01,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,2,power_law_1.01,0.08739839792251587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,2,power_law_1.01,0.08867200016975403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,2,power_law_1.01,0.09277439713478089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,2,power_law_1.01,0.09800320267677307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,2,power_law_1.01,0.09671679735183716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,2,power_law_1.01,0.0998080015182495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,2,power_law_1.01,0.10079360008239746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,2,power_law_1.01,0.11272959709167481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,2,power_law_1.01,0.1258944034576416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,2,power_law_1.01,0.13329919576644897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,2,power_law_1.01,0.13901439905166627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,2,power_law_1.01,0.16389119625091553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,2,power_law_1.01,0.18145920038223268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,2,power_law_1.01,0.22602241039276122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,2,power_law_1.01,0.262662410736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,2,power_law_1.01,0.40972161293029785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,2,power_law_1.01,0.49259519577026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,2,power_law_1.01,0.7070144176483154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,64,balanced,0.05023466547330221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,64,balanced,0.0499946673711141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,64,balanced,0.05132266879081726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,64,balanced,0.05418666700522105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,64,balanced,0.05624000231424967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,64,balanced,0.0625546673933665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,64,balanced,0.0641546646753947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,64,balanced,0.06474666794141133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,64,balanced,0.06301333506902058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,64,balanced,0.06396799782911937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,64,balanced,0.06402133405208588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,64,balanced,0.06592533489068349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,64,balanced,0.06824533144632976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,64,balanced,0.0684853345155716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,64,balanced,0.07256533205509186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,64,balanced,0.07426133255163829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,64,balanced,0.07624533275763194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,64,balanced,0.08785067001978557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,64,balanced,0.09090133508046468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,64,balanced,0.10130133231480916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,64,balanced,0.11558399597803752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,64,balanced,0.14645333091417947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,64,balanced,0.17038400967915854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,2,power_law_1.01,0.8555456161499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,64,balanced,0.22405334313710532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,64,balanced,0.27530133724212646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,64,balanced,0.3782399892807007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,64,balanced,0.4818026622136434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,2,power_law_1.01,1.24717435836792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,8,balanced,0.021274665991465252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,8,balanced,0.022687998910744984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,8,balanced,0.02514133354028066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,8,balanced,0.027600000301996868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,8,balanced,0.043578664461771645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,8,balanced,0.060826669136683144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,8,balanced,0.06574399769306183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,8,balanced,0.06971733272075653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,8,balanced,0.07283199826876323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,8,balanced,0.07833066582679749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,8,balanced,0.07420266668001811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,8,balanced,0.0791786660750707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,8,balanced,0.08304533362388611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,8,balanced,0.08387733499209087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,8,balanced,0.09053333600362141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,8,balanced,0.09542933106422424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,8,balanced,0.11844266454378764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,8,balanced,0.12179199854532878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,8,balanced,0.18979734182357788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,8,balanced,0.23529066642125449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,8,balanced,0.1954560081164042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,8,balanced,0.22010666131973267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,8,balanced,0.24638932943344116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,8,balanced,0.2841813365618388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,8,balanced,0.3335520029067993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,8,balanced,0.5346239805221558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,8,balanced,0.6217546860376993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,4,balanced,0.04597333570321401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,4,balanced,0.04826133449872335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,4,balanced,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,4,balanced,0.06223999957243601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,4,balanced,0.08110933502515157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,4,balanced,0.11691733201344807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,4,balanced,0.1473973294099172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,4,balanced,0.14230933785438538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,4,balanced,0.14429333806037903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,4,balanced,0.14448533455530801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,4,balanced,0.1455359955628713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,4,balanced,0.15178133050600687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,4,balanced,0.15504533052444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,4,balanced,0.15475733081499735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,4,balanced,0.16500799854596457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,4,balanced,0.16660267114639282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,4,balanced,0.17690134048461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,4,balanced,0.19987199703852335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,4,balanced,0.2243679960568746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,4,balanced,0.2563093304634094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,4,balanced,0.2899893323580424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,4,balanced,0.3596746524175008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,4,balanced,0.4567626714706421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,4,balanced,0.5939786831537882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,4,balanced,0.7733439604441324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,4,balanced,1.074613332748413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,4,balanced,1.37281068166097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,8,power_law_1.2,0.018220800161361694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,8,power_law_1.2,0.018310399353504182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,8,power_law_1.2,0.018297599256038667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,8,power_law_1.2,0.018028800189495087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,8,power_law_1.2,0.019891199469566346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,8,power_law_1.2,0.018675200641155243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,8,power_law_1.2,0.019359999895095827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,8,power_law_1.2,0.022252799570560457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,8,power_law_1.2,0.022937600314617158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,8,power_law_1.2,0.022976000607013703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,8,power_law_1.2,0.03313280045986176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,8,power_law_1.2,0.033817601203918454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,8,power_law_1.2,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,8,power_law_1.2,0.034534400701522826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,8,power_law_1.2,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,2,power_law_1.01,1.362700843811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,8,power_law_1.2,0.03712640106678009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,8,power_law_1.2,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,8,power_law_1.2,0.04439040124416351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.2,0.04575999975204468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.2,0.052742397785186766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.2,0.06188160181045532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.2,0.08597760200500489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.2,0.09896960258483886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,8,power_law_1.01,0.02133760005235672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,8,power_law_1.01,0.0208064004778862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,8,power_law_1.01,0.02074880003929138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,8,power_law_1.01,0.02033279985189438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,8,power_law_1.01,0.0208064004778862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,8,power_law_1.01,0.02414720058441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,8,power_law_1.01,0.04805760085582733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,8,power_law_1.01,0.04833920001983642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,8,power_law_1.01,0.04380159974098206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,8,power_law_1.01,0.040057599544525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,8,power_law_1.01,0.03668479919433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,8,power_law_1.01,0.05058559775352478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,8,power_law_1.01,0.049497601389884946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.2,0.1349951982498169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,8,power_law_1.01,0.03922559916973114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,8,power_law_1.01,0.0625216007232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,8,power_law_1.01,0.05268480181694031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,8,power_law_1.01,0.05896959900856018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,8,power_law_1.01,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.01,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.01,0.12472959756851196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.01,0.1500864028930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.01,0.21157760620117189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.01,0.2657984018325806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.2,0.17120640277862548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.01,0.38008320331573486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.2,0.24495360851287842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.01,0.49685120582580566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.01,0.7285888195037842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,2,power_law_1.01,0.07158399820327759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,2,power_law_1.01,0.0969983994960785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,2,power_law_1.01,0.12360320091247559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,2,power_law_1.01,0.16384639739990234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,2,power_law_1.01,0.22994558811187743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,2,power_law_1.01,0.25886080265045164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,2,power_law_1.01,0.3241215944290161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,2,power_law_1.01,0.3608831882476807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,2,power_law_1.01,0.36060800552368166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,2,power_law_1.01,0.37227520942687986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,2,power_law_1.01,0.3966016054153442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,2,power_law_1.01,0.4002687931060791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,2,power_law_1.01,0.4210239887237549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,2,power_law_1.01,0.42961921691894533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,2,power_law_1.01,0.44279680252075193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,2,power_law_1.01,0.4463488101959229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.01,0.9555968284606934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,2,power_law_1.01,0.46572160720825195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,2,power_law_1.01,0.5121088027954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,2,power_law_1.01,0.5351679801940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,2,power_law_1.01,0.609600019454956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,2,power_law_1.01,0.6465216159820557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,2,power_law_1.01,0.7574719905853271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,2,power_law_1.01,0.9004608154296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,2,power_law_1.01,1.1082624435424804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,2,power_law_1.01,1.3488896369934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,2,power_law_1.01,1.8466304779052733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,2,power_law_1.01,2.3799680709838866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.2,0.30599679946899416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,2,power_law_1.2,0.018489600718021394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,2,power_law_1.2,0.01863040030002594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,2,power_law_1.2,0.02021760046482086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,2,power_law_1.2,0.023628799617290495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,2,power_law_1.2,0.026361599564552307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,2,power_law_1.2,0.029977598786354066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,2,power_law_1.2,0.03139840066432953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,8,power_law_1.01,0.08382080197334289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,8,power_law_1.01,0.08538240194320679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,2,power_law_1.2,0.033228799700737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,8,power_law_1.01,0.08455680012702942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,2,power_law_1.2,0.03710080087184906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,8,power_law_1.01,0.0832256019115448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,2,power_law_1.2,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,2,power_law_1.2,0.04023680090904236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,2,power_law_1.2,0.03935360014438629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,8,power_law_1.01,0.08520320057868958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,2,power_law_1.2,0.04162560105323791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,8,power_law_1.01,0.08971520066261292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,2,power_law_1.2,0.0435263991355896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,8,power_law_1.01,0.09510400295257568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,2,power_law_1.2,0.0499455988407135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,2,power_law_1.2,0.050387197732925416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,2,power_law_1.2,0.054745602607727054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,2,power_law_1.2,0.09514880180358887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,2,power_law_1.2,0.10076160430908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,2,power_law_1.2,0.10124800205230713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,2,power_law_1.2,0.11002240180969239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,8,power_law_1.01,0.12138880491256714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,8,power_law_1.01,0.12718080282211303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,8,power_law_1.01,0.14926719665527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,8,power_law_1.01,0.14328960180282593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,8,power_law_1.01,0.14881919622421264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,8,power_law_1.01,0.1769536018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,2,power_law_1.2,0.12578560113906861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,8,power_law_1.01,0.12851200103759766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,8,power_law_1.01,0.15143680572509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,8,power_law_1.01,0.14791040420532225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,8,power_law_1.01,0.19530240297317505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,8,power_law_1.01,0.252729606628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,2,power_law_1.2,0.15450880527496338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,8,power_law_1.01,0.28601601123809817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,8,power_law_1.01,0.42165122032165525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,2,power_law_1.2,0.19020160436630248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,8,power_law_1.01,0.47341442108154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,8,power_law_1.01,0.6662335872650147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,8,power_law_1.2,0.06968320012092591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,2,power_law_1.2,0.24141440391540528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,8,power_law_1.2,0.09578880071640014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,8,power_law_1.2,0.12012799978256225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,8,power_law_1.2,0.1341375946998596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,8,power_law_1.2,0.16362240314483642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,8,power_law_1.2,0.1699455976486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,8,power_law_1.2,0.19852160215377807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,8,power_law_1.2,0.1930816054344177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,8,power_law_1.2,0.17237759828567506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,8,power_law_1.01,0.8455167770385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,8,power_law_1.2,0.19148800373077393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,2,power_law_1.2,0.3250943899154663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,8,power_law_1.2,0.1638208031654358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,8,power_law_1.2,0.17292799949645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,8,power_law_1.2,0.19900799989700318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,8,power_law_1.2,0.18483840227127074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,8,power_law_1.2,0.19535360336303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,8,power_law_1.2,0.204748797416687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,8,power_law_1.2,0.21714560985565184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,8,power_law_1.2,0.24714241027832032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,8,power_law_1.2,0.2704639911651611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,8,power_law_1.2,0.31857919692993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,8,power_law_1.01,1.2114496231079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,8,power_law_1.2,0.3531008005142212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,16,8,balanced,0.053130666414896645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,8,power_law_1.2,0.4596864223480225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,16,8,balanced,0.0539626677831014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,8,power_law_1.01,1.4638336181640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,16,8,balanced,0.055061335364977516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,16,8,balanced,0.06035199761390686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,16,8,balanced,0.06563733518123627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,16,8,balanced,0.08498666683832805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,16,8,balanced,0.08488532900810242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,16,8,balanced,0.0844533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,16,8,balanced,0.08610666791598003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,16,8,balanced,0.08474666873613994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,16,8,balanced,0.08618666728337605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,16,8,balanced,0.0895306666692098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,16,8,balanced,0.08949333429336548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,16,8,balanced,0.0922933320204417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,16,8,balanced,0.09911466638247173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,8,power_law_1.2,0.541267204284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,16,8,balanced,0.09938666224479675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,16,8,balanced,0.1049013336499532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,16,8,balanced,0.12331733107566833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,8,power_law_1.2,0.7012351989746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,16,8,balanced,0.12774399916330972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,8,power_law_1.2,0.8220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,8,power_law_1.2,1.2557439804077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,8,power_law_1.2,1.6426752090454102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,16,8,balanced,0.14192000031471252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,16,8,balanced,0.16459199786186218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,16,8,balanced,0.19948800404866537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,16,8,balanced,0.23968533674875894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,16,8,balanced,0.34165334701538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,16,8,balanced,0.42534399032592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,16,8,balanced,0.616757353146871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,16,8,balanced,0.7975786526997884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,8,power_law_1.01,2.385247993469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,2,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,2,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,2,balanced,0.03972266614437103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,2,balanced,0.062037333846092224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,2,balanced,0.0885653297106425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,2,balanced,0.09073600172996521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,2,balanced,0.09047999978065491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,2,balanced,0.09313066800435384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,2,balanced,0.0950879951318105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,2,balanced,0.09698667128880818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,2,balanced,0.09948800007502238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,2,balanced,0.1032319962978363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,2,balanced,0.10388799508412679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,2,balanced,0.1076746682325999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,2,balanced,0.11313600341478984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,2,balanced,0.11942399541536967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,2,balanced,0.12402666608492534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,2,balanced,0.13868799805641174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,2,balanced,0.14989866813023886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,2,balanced,0.19290133317311606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,2,balanced,0.20643732945124307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,2,balanced,0.29043734073638916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,2,balanced,0.34187201658884686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,2,balanced,0.4946986834208171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,2,balanced,0.6266133387883505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,2,balanced,0.9023679892222086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,2,balanced,1.1736640135447185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,2,power_law_1.2,0.4196159839630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,8,power_law_1.01,3.2371776580810545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,32,power_law_1.2,0.04639999866485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,32,power_law_1.2,0.0438975989818573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,32,power_law_1.2,0.045228800177574156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,32,power_law_1.2,0.04353919923305512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,32,power_law_1.2,0.04412800073623657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,32,power_law_1.2,0.04519680142402649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,32,power_law_1.2,0.047513601183891294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,32,power_law_1.2,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,32,power_law_1.2,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,32,power_law_1.2,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,32,power_law_1.2,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,32,power_law_1.2,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,32,power_law_1.2,0.04852479994297028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,32,power_law_1.2,0.048767998814582825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,32,power_law_1.2,0.052723199129104614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,32,power_law_1.2,0.0538752019405365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,32,power_law_1.2,0.05632640123367309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,32,power_law_1.2,0.06533120274543762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.2,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.2,0.08121600151062011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.2,0.09835519790649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.2,0.12796159982681274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.2,0.14353280067443847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.2,0.2252608060836792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.2,0.2531264066696167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,4,power_law_1.01,0.09604480266571044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,4,power_law_1.01,0.1423359990119934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,4,power_law_1.01,0.17125760316848754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,4,power_law_1.01,0.21967999935150145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,4,power_law_1.01,0.2844543933868408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,4,power_law_1.01,0.2842688083648682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,4,power_law_1.01,0.409932804107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,4,power_law_1.01,0.37097599506378176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,4,power_law_1.01,0.3739711999893188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,4,power_law_1.01,0.3760256052017212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,4,power_law_1.01,0.3910655975341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,4,power_law_1.01,0.40790400505065916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,4,power_law_1.01,0.3913088083267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,4,power_law_1.01,0.41194877624511717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,4,power_law_1.01,0.4189568042755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,4,power_law_1.01,0.4356224060058594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,4,power_law_1.01,0.4339327812194824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,4,power_law_1.01,0.49372158050537107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,4,power_law_1.01,0.5119167804718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,4,power_law_1.01,0.5799295902252197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,4,power_law_1.01,0.5896383762359619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,4,power_law_1.01,0.6915264129638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,4,power_law_1.01,0.8104191780090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,4,power_law_1.01,1.0272192001342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.2,0.3889791965484619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,4,power_law_1.01,1.1888256072998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,4,power_law_1.01,1.4736703872680663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,4,power_law_1.01,1.9284608840942383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.2,0.5140351772308349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,32,8,balanced,0.01739199956258138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,32,8,balanced,0.017269333203633625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,32,8,balanced,0.017279999951521557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,32,8,balanced,0.01720533271630605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,32,8,balanced,0.019023999571800232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,32,8,balanced,0.018901333212852478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,32,8,balanced,0.019029332945744198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,32,8,balanced,0.021344001094500225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,32,8,balanced,0.023013333479563396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,32,8,balanced,0.02201066662867864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,32,8,balanced,0.029520000020662945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,32,8,balanced,0.03148266673088074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,32,8,balanced,0.03148266673088074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,32,8,balanced,0.031136001149813335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,32,8,balanced,0.03169599920511246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,32,8,balanced,0.031717332700888314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,32,8,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,32,8,balanced,0.039306665460268654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,32,8,balanced,0.0418453315893809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,32,8,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,32,8,balanced,0.051872000098228455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,32,8,balanced,0.07088533540566762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,32,8,balanced,0.08107199768225352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,32,8,balanced,0.10977066556612651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,32,8,balanced,0.13589333494504294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,32,8,balanced,0.19129600127538046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,32,8,balanced,0.23842666546503702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,8,power_law_1.2,0.07704960107803345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,8,power_law_1.01,0.0940671980381012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,8,power_law_1.2,0.09841920137405395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,8,power_law_1.01,0.14384640455245973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,8,power_law_1.2,0.09418879747390747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,8,power_law_1.01,0.1532863974571228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,8,power_law_1.2,0.10766719579696656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,8,power_law_1.01,0.1845695972442627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,8,power_law_1.2,0.12832000255584716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,8,power_law_1.01,0.21920640468597413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,8,power_law_1.2,0.12033920288085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,8,power_law_1.01,0.20127360820770263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,32,power_law_1.01,0.031539198756217954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,8,power_law_1.2,0.12750719785690307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,8,power_law_1.2,0.11953279972076417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,32,power_law_1.01,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,8,power_law_1.2,0.12110719680786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,32,power_law_1.01,0.02956799864768982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,8,power_law_1.2,0.12282240390777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,8,power_law_1.2,0.12407040596008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,8,power_law_1.01,0.22055039405822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,8,power_law_1.01,0.2187391996383667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,8,power_law_1.01,0.21147520542144777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,32,power_law_1.01,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,8,power_law_1.01,0.23286399841308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,32,power_law_1.01,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,8,power_law_1.01,0.22218239307403564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,32,power_law_1.01,0.03281919956207276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,32,power_law_1.01,0.03336319923400879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,8,power_law_1.2,0.12319359779357911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,8,power_law_1.2,0.12424960136413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,8,power_law_1.2,0.12481919527053834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,8,power_law_1.2,0.13186559677124024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,8,power_law_1.2,0.13500800132751464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,8,power_law_1.2,0.14280320405960084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,8,power_law_1.01,0.22942719459533692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,32,power_law_1.01,0.03349120020866394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,32,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,32,power_law_1.01,0.03391999900341034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,32,power_law_1.01,0.03500800132751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,32,power_law_1.01,0.0371071994304657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,32,power_law_1.01,0.037145599722862244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,32,power_law_1.01,0.03772799968719483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,32,power_law_1.01,0.04136320054531097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,32,power_law_1.01,0.042982399463653564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,8,power_law_1.01,0.2376960039138794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,8,power_law_1.01,0.2406912088394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,8,power_law_1.01,0.25157120227813723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,8,power_law_1.01,0.25943679809570314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,8,power_law_1.01,0.28170878887176515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,8,power_law_1.2,0.15266560316085814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,8,power_law_1.01,0.31192960739135744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,8,power_law_1.2,0.17082879543304444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,32,power_law_1.01,0.04410240054130554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,8,power_law_1.01,0.31164801120758057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,8,power_law_1.2,0.19998079538345337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,8,power_law_1.01,0.376582407951355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,8,power_law_1.2,0.2223871946334839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,8,power_law_1.01,0.38932480812072756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,8,power_law_1.2,0.2951872110366821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,8,power_law_1.01,0.4702335834503174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,8,power_law_1.2,0.3342144012451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,8,power_law_1.01,0.5384384155273437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,8,power_law_1.2,0.45394558906555177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,8,power_law_1.01,0.711513614654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,8,power_law_1.2,0.562662410736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,8,power_law_1.2,0.7845503807067871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,8,power_law_1.2,1.104543972015381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,32,power_law_1.01,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,32,power_law_1.01,0.0531711995601654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,32,power_law_1.01,0.061849600076675414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,32,power_law_1.01,0.0754751980304718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,32,power_law_1.01,0.10065280199050904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,32,power_law_1.01,0.12817280292510985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,32,power_law_1.01,0.1797376036643982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,32,power_law_1.01,0.24381439685821532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,8,power_law_1.01,0.9052543640136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,32,power_law_1.01,0.3215487957000732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,32,power_law_1.01,0.3798079967498779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,8,power_law_1.01,1.2383935928344727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,8,power_law_1.01,1.5262592315673829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,128,power_law_1.2,0.031769600510597226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,128,power_law_1.2,0.031763198971748355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,128,power_law_1.2,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,128,power_law_1.2,0.03314560055732727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,128,power_law_1.2,0.03425920009613037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,128,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,128,power_law_1.2,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,128,power_law_1.2,0.035174399614334106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,128,power_law_1.2,0.03586559891700745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,128,power_law_1.2,0.035776001214981076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,128,power_law_1.2,0.036575999855995175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,128,power_law_1.2,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,128,power_law_1.2,0.03666560053825378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,128,power_law_1.2,0.03777279853820801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,128,power_law_1.2,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,128,power_law_1.2,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,128,power_law_1.2,0.04156799912452698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,128,power_law_1.2,0.04625279903411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,128,power_law_1.2,0.05050240159034729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,128,power_law_1.2,0.058847999572753905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,4,power_law_1.01,0.088128000497818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,128,power_law_1.2,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,4,power_law_1.01,0.12195839881896972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,4,power_law_1.01,0.12371200323104858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,128,power_law_1.2,0.0770687997341156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,4,power_law_1.01,0.1555392026901245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,4,power_law_1.01,0.18538880348205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,4,power_law_1.01,0.21986560821533202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,4,power_law_1.01,0.22495999336242675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,4,power_law_1.01,0.22469758987426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,4,power_law_1.01,0.22811520099639893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,4,power_law_1.01,0.24294400215148926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,4,power_law_1.01,0.24669439792633058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,4,power_law_1.01,0.24384000301361083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,4,power_law_1.01,0.24741120338439943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,4,power_law_1.01,0.25370240211486816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,128,power_law_1.2,0.08821759819984436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,4,power_law_1.01,0.27124478816986086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,4,power_law_1.01,0.2844288110733032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,4,power_law_1.01,0.29171841144561766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,4,power_law_1.01,0.33544321060180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,4,power_law_1.01,0.3602303981781006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,4,power_law_1.01,0.43208961486816405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,4,power_law_1.01,0.4844480037689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,4,power_law_1.01,0.6111423969268799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,4,power_law_1.01,0.6661248207092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,4,power_law_1.01,0.8333312034606933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,128,power_law_1.2,0.11603200435638428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,4,power_law_1.01,1.1202752113342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,4,power_law_1.01,1.6382400512695312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,4,power_law_1.01,1.8626367568969726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,128,power_law_1.2,0.14677759408950805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,128,power_law_1.2,0.21092479228973388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,4,power_law_1.01,0.020972800254821778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,4,power_law_1.01,0.024166400730609893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,4,power_law_1.01,0.0268095999956131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,4,power_law_1.01,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,4,power_law_1.01,0.04343680143356323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,4,power_law_1.01,0.04705280065536499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,4,power_law_1.01,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,4,power_law_1.01,0.05252479910850525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,4,power_law_1.01,0.05294719934463501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,4,power_law_1.01,0.05491200089454651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,4,power_law_1.01,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,4,power_law_1.01,0.056576001644134524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,4,power_law_1.01,0.057734400033950806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,4,power_law_1.01,0.061363202333450315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,4,power_law_1.01,0.06796799898147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,4,power_law_1.01,0.07057920098304749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,4,power_law_1.01,0.07521920204162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,4,power_law_1.01,0.09304320216178893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.01,0.09760000109672547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.01,0.12061439752578736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.01,0.1492416024208069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.01,0.1758463978767395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.01,0.20135679244995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.01,0.2562432050704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.01,0.32483201026916503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.01,0.4263040065765381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.01,0.5538303852081299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,128,power_law_1.2,0.2746880054473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,4,power_law_1.2,0.08689919710159302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,4,power_law_1.2,0.10506240129470826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,4,power_law_1.2,0.11633919477462769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,4,power_law_1.2,0.15434240102767943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,4,power_law_1.2,0.17330559492111205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,4,power_law_1.2,0.2205183982849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,4,power_law_1.2,0.23283839225769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,4,power_law_1.2,0.24488959312438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,4,power_law_1.2,0.23914239406585694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,4,power_law_1.2,0.23976318836212157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,4,power_law_1.2,0.24549760818481445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,4,power_law_1.2,0.25057919025421144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,4,power_law_1.2,0.25931520462036134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,4,power_law_1.2,0.26751999855041503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,4,power_law_1.2,0.28513920307159424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,4,power_law_1.2,0.2935551881790161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,4,power_law_1.2,0.3070784091949463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,4,power_law_1.2,0.3378432035446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,4,power_law_1.2,0.3646656036376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,4,power_law_1.2,0.4611519813537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,4,power_law_1.2,0.5184319972991943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,4,power_law_1.2,0.6599679946899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,4,power_law_1.2,0.7016831874847412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,4,power_law_1.2,0.9224063873291015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,4,power_law_1.2,1.2005248069763184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,4,power_law_1.2,1.670854377746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,4,power_law_1.2,1.9712127685546874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,64,power_law_1.01,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,64,power_law_1.01,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,16,power_law_1.2,0.05107839703559876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,64,power_law_1.01,0.03118079900741577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,16,power_law_1.2,0.04689919948577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,64,power_law_1.01,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,16,power_law_1.2,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,16,power_law_1.2,0.04730879962444305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,16,power_law_1.2,0.04030719995498657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,16,power_law_1.2,0.05316479802131653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,16,power_law_1.2,0.05441920161247253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,16,power_law_1.2,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,16,power_law_1.2,0.05511040091514587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,16,power_law_1.2,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,64,power_law_1.01,0.03238399922847748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,16,power_law_1.2,0.057196801900863646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,64,power_law_1.01,0.03375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,16,power_law_1.2,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,64,power_law_1.01,0.034041601419448855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,16,power_law_1.2,0.05834239721298218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,64,power_law_1.01,0.03423359990119934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,16,power_law_1.2,0.06090880036354065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,64,power_law_1.01,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,16,power_law_1.2,0.0673792004585266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,64,power_law_1.01,0.035488000512123107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,16,power_law_1.2,0.06991360187530518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,64,power_law_1.01,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,16,power_law_1.2,0.0753920018672943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,64,power_law_1.01,0.038553598523139956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,16,power_law_1.2,0.08364800214767457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,64,power_law_1.01,0.04069119989871979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,16,power_law_1.2,0.08943359851837158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,16,power_law_1.2,0.09856640100479126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,16,power_law_1.2,0.13383040428161622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,16,power_law_1.2,0.173088002204895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,16,power_law_1.2,0.1585088014602661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,64,power_law_1.01,0.041043201088905336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,64,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,64,power_law_1.01,0.048332801461219786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,64,power_law_1.01,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,64,power_law_1.01,0.056569600105285646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,64,power_law_1.01,0.06639999747276307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,64,power_law_1.01,0.09095039963722229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,64,power_law_1.01,0.10158079862594604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,64,power_law_1.01,0.12729599475860595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,64,power_law_1.01,0.15418239831924438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,64,power_law_1.01,0.22044799327850342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,64,power_law_1.01,0.2599423885345459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,64,power_law_1.01,0.38519039154052737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,64,power_law_1.01,0.43980159759521487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,16,power_law_1.2,0.24194560050964356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,16,power_law_1.2,0.29504640102386476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,16,power_law_1.2,0.42816638946533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,16,power_law_1.2,0.7020095825195313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,128,power_law_1.2,0.05015040040016174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,128,power_law_1.2,0.049497601389884946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,128,power_law_1.2,0.04940800070762634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,128,power_law_1.2,0.05208960175514221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,128,power_law_1.2,0.05325440168380737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,128,power_law_1.2,0.05438079833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,128,power_law_1.2,0.05674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,128,power_law_1.2,0.057094401121139525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,128,power_law_1.2,0.058348798751831056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,128,power_law_1.2,0.05839999914169312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,128,power_law_1.2,0.060063999891281125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,128,power_law_1.2,0.0646016001701355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,128,power_law_1.2,0.06787840127944947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,128,power_law_1.2,0.06586880087852479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,128,power_law_1.2,0.07518720030784606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,128,power_law_1.2,0.08201599717140198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,128,power_law_1.2,0.08343679904937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,128,power_law_1.2,0.10055680274963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,128,power_law_1.2,0.11891839504241944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,128,power_law_1.2,0.1440000057220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,16,power_law_1.2,0.021766400337219237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,16,power_law_1.2,0.022495999932289124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,128,power_law_1.2,0.16365439891815187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,16,power_law_1.2,0.022886399924755097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,128,power_law_1.2,0.21069440841674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,16,power_law_1.2,0.02319999933242798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,128,power_law_1.2,0.26160640716552735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,128,power_law_1.2,0.41351680755615233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,128,power_law_1.2,0.5378943920135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,128,power_law_1.2,0.9007743835449219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,16,power_law_1.2,0.024320000410079957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,16,power_law_1.2,0.024774399399757386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,16,power_law_1.2,0.026918399333953857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,16,power_law_1.2,0.02714880108833313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,16,power_law_1.2,0.03281919956207276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,128,power_law_1.2,1.4476415634155273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,16,power_law_1.2,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,16,power_law_1.2,0.03386879861354828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,16,power_law_1.2,0.0578495979309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,16,power_law_1.2,0.05818880200386047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,16,power_law_1.2,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,16,power_law_1.2,0.057817602157592775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,16,power_law_1.2,0.0591808021068573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,16,power_law_1.2,0.022047999501228332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,16,power_law_1.2,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,16,power_law_1.2,0.02192640006542206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,16,power_law_1.2,0.06577919721603394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,16,power_law_1.2,0.02210559993982315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.2,0.06946560144424438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,16,power_law_1.2,0.022707200050354003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.2,0.07784960269927979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,16,power_law_1.2,0.022720000147819518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.2,0.08610560297966004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,16,power_law_1.2,0.024300800263881685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.2,0.11285760402679443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,16,power_law_1.2,0.025388801097869874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.2,0.14142719507217408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,16,power_law_1.2,0.025222399830818178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,16,power_law_1.2,0.02563839852809906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,16,power_law_1.2,0.02617599964141846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,16,power_law_1.2,0.02736639976501465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,16,power_law_1.2,0.02929919958114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,16,power_law_1.2,0.03047040104866028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,16,power_law_1.2,0.03287039995193482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,16,power_law_1.2,0.03583360016345978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,16,power_law_1.2,0.037747201323509214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,16,power_law_1.2,0.05384320020675659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,16,power_law_1.2,0.05822719931602478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.2,0.06593279838562012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.2,0.188646399974823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.2,0.0747327983379364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.2,0.22873599529266359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.2,0.3329024076461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.2,0.438918399810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.2,0.08578559756278992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.2,0.10628479719161987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.2,0.12502399682998658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.2,0.16261119842529298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.2,0.22643840312957764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.2,0.30456318855285647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.2,0.4110976219177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,32,balanced,0.02771199991305669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,32,balanced,0.027317332724730175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,32,balanced,0.02736533433198929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,32,balanced,0.029253333806991577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,32,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,32,balanced,0.03257066756486893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,32,balanced,0.03160000095764796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,32,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,32,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,32,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,32,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,32,balanced,0.03357866654793421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,32,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,32,balanced,0.03638399889071783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,32,balanced,0.03765333443880081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,32,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,32,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,32,balanced,0.041493333876132965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,32,balanced,0.04043200115362803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,32,balanced,0.04558933277924856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,32,balanced,0.04762133459250132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,32,balanced,0.051957334081331887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,32,balanced,0.057674666245778404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,32,balanced,0.07854400078455608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,32,balanced,0.10430933038393657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,32,balanced,0.1440000037352244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,32,balanced,0.17161067326863608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,2,balanced,0.029839999973773956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,2,balanced,0.031285333136717476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,2,balanced,0.03153600047032038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,2,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,2,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,2,balanced,0.03531199942032496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,2,balanced,0.03400533397992452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,2,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,2,balanced,0.036015999813874565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,2,balanced,0.03586666782697042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,2,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,2,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,2,balanced,0.039408000806967415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,2,balanced,0.03973866750796636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,2,balanced,0.045834665497144066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,2,balanced,0.04799999793370565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,2,balanced,0.049882665276527405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,2,balanced,0.061381335059801735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,2,balanced,0.062181333700815834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,2,balanced,0.08551466464996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,2,balanced,0.10218666990598042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,2,balanced,0.13793599605560303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,2,balanced,0.15239466230074564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,2,balanced,0.21073599656422934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,2,balanced,0.26680533091227215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,2,balanced,0.3766026496887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,2,balanced,0.497162659962972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,2,balanced,0.04637333254019419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,2,balanced,0.0517439991235733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,2,balanced,0.05420800050099691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,2,balanced,0.06649066507816315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,2,balanced,0.09602666894594829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,2,balanced,0.11028800408045451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,2,balanced,0.1106719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,2,balanced,0.1076693336168925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,2,balanced,0.1085599958896637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,2,balanced,0.10961600144704182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,2,balanced,0.11249066392580669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,2,balanced,0.11588799953460693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,2,balanced,0.11730133493741353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,2,balanced,0.12589333454767862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,2,balanced,0.13410666584968567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,2,balanced,0.13766933480898538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,2,balanced,0.14710399508476257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,2,balanced,0.1715946594874064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,2,balanced,0.18303465843200684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,2,balanced,0.2545119921366374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,2,balanced,0.2696959972381592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,2,balanced,0.4002399841944377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,2,balanced,0.4694399833679199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,2,balanced,0.7023413181304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,2,balanced,0.8539626598358154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,2,balanced,1.3063039779663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,2,balanced,1.6043999989827473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,16,balanced,0.051813334226608276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,16,balanced,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,16,balanced,0.05381333331267039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,16,balanced,0.05793599784374237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,16,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,16,balanced,0.07430399954319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,16,balanced,0.08252800007661183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,16,balanced,0.08273066580295563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,16,balanced,0.08564800024032593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,16,balanced,0.08412266770998637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,16,balanced,0.08593066533406575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,16,balanced,0.08903466661771138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,16,balanced,0.08850133419036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,16,balanced,0.09081600109736125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,16,balanced,0.10085333387056987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,16,balanced,0.10122133294741313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,16,balanced,0.10913599530855815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,16,balanced,0.11997866630554199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,16,balanced,0.12892799576123556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,16,balanced,0.14217600226402283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,16,balanced,0.16242667039235434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,16,balanced,0.20153599977493286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,16,balanced,0.23804267247517905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,16,balanced,0.31834665934244794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,16,balanced,0.41387200355529785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,16,balanced,0.5823573271433512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,16,balanced,0.7607146898905436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,8,balanced,0.060032000144322716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,8,balanced,0.06006399790445963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,8,balanced,0.06125866870085398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,8,balanced,0.07249600191911061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,8,balanced,0.08676266670227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,8,balanced,0.10756799578666687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,8,balanced,0.10725333293279012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,8,balanced,0.10661866267522176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,8,balanced,0.10922132929166158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,8,balanced,0.10794132947921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,8,balanced,0.10929600397745769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,8,balanced,0.11128532886505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,8,balanced,0.11133866508801778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,8,balanced,0.11528000235557556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,8,balanced,0.12341333429018657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,8,balanced,0.12455999851226807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,8,balanced,0.12853333353996277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,8,balanced,0.13877333203951517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,8,balanced,0.14308266838391623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,8,balanced,0.16561599572499594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,8,balanced,0.18582399686177573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,8,balanced,0.22573333978652954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,8,balanced,0.2735520005226135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,8,balanced,0.39909334977467853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,8,balanced,0.4907413323720296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,8,balanced,0.694159984588623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,8,balanced,0.9059893290201823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,4,power_law_1.2,0.04669440090656281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,4,power_law_1.2,0.054022401571273804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,4,power_law_1.2,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,4,power_law_1.2,0.06094719767570496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,4,power_law_1.2,0.06211199760437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,4,power_law_1.2,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,4,power_law_1.2,0.06299520134925843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,4,power_law_1.2,0.06377599835395813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,4,power_law_1.2,0.06421120166778564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,4,power_law_1.2,0.06562560200691223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,4,power_law_1.2,0.06540160179138184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,4,power_law_1.2,0.07720320224761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,4,power_law_1.2,0.0782912015914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,4,power_law_1.2,0.08161280155181885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,4,power_law_1.2,0.09253759980201721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,4,power_law_1.2,0.11129599809646606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,4,power_law_1.2,0.12846720218658447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,4,power_law_1.2,0.14079999923706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,4,power_law_1.2,0.16725120544433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,4,power_law_1.2,0.24173440933227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,4,power_law_1.2,0.2923775911331177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,4,power_law_1.2,0.024915200471878052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,4,power_law_1.2,0.03763839900493622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,4,power_law_1.2,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,4,power_law_1.2,0.05381760001182556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,4,power_law_1.2,0.06314240097999572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,4,power_law_1.2,0.06729599833488464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,4,power_law_1.2,0.0840511977672577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,4,power_law_1.2,0.08202880024909973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,4,power_law_1.2,0.0807424008846283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,4,power_law_1.2,0.08613119721412658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,4,power_law_1.2,0.08325120210647582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,4,power_law_1.2,0.0964031994342804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,4,power_law_1.2,0.09723520278930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,4,power_law_1.2,0.09456639885902404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,4,power_law_1.2,0.12280960083007812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,4,power_law_1.2,0.11737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,4,power_law_1.2,0.11606400012969971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,4,power_law_1.2,0.19723520278930665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,4,power_law_1.2,0.23118720054626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,4,power_law_1.2,0.20523519515991212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,4,power_law_1.2,0.19225599765777587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,4,power_law_1.2,0.23516159057617186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,4,power_law_1.2,0.23656959533691407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,4,power_law_1.2,0.2959104061126709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,4,power_law_1.2,0.35146241188049315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,4,power_law_1.2,0.47891840934753416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,1,power_law_1.2,0.05306879878044128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,4,power_law_1.2,0.6355648040771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,4,power_law_1.2,0.45639681816101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,1,power_law_1.2,0.0561024010181427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,1,power_law_1.2,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,1,power_law_1.2,0.06689280271530151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,1,power_law_1.2,0.07630720138549804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,1,power_law_1.2,0.0853056013584137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,1,power_law_1.2,0.10289920568466186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,1,power_law_1.2,0.10720000267028809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,1,power_law_1.01,0.061689597368240354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,1,power_law_1.2,0.11165440082550049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,1,power_law_1.01,0.08008319735527039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,1,power_law_1.2,0.11484160423278808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,1,power_law_1.01,0.08643199801445008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,1,power_law_1.2,0.12108800411224366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,1,power_law_1.01,0.09616000056266785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,1,power_law_1.2,0.1283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,1,power_law_1.01,0.10316159725189208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,1,power_law_1.2,0.13388160467147828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,1,power_law_1.01,0.12206720113754273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,1,power_law_1.01,0.12857600450515747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,1,power_law_1.01,0.1306175947189331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,1,power_law_1.01,0.1336832046508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,1,power_law_1.01,0.13766399621963502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,1,power_law_1.01,0.1422144055366516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,1,power_law_1.2,0.1420032024383545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,1,power_law_1.2,0.153056001663208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,1,power_law_1.2,0.16121599674224854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,1,power_law_1.2,0.1661695957183838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,1,power_law_1.2,0.1959231972694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,1,power_law_1.2,0.22454400062561036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,1,power_law_1.01,0.14261120557785034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,1,power_law_1.2,0.28565120697021484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,1,power_law_1.01,0.1461632013320923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,1,power_law_1.2,0.3524863958358765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,1,power_law_1.01,0.15155199766159058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,1,power_law_1.2,0.45607681274414064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,1,power_law_1.01,0.1473855972290039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,1,power_law_1.2,0.5693696022033692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,4,power_law_1.2,0.5547264099121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,1,power_law_1.01,0.15013760328292847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,1,power_law_1.2,0.7878592014312744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,1,power_law_1.01,0.15895040035247804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,1,power_law_1.01,0.17571840286254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,1,power_law_1.2,1.013766384124756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,1,power_law_1.01,0.19503359794616698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,1,power_law_1.2,1.4520895957946778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,1,power_law_1.01,0.23892478942871093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,1,power_law_1.01,0.28524160385131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,1,power_law_1.2,1.8886783599853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,1,power_law_1.01,0.3779583930969238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,1,power_law_1.01,0.4734015941619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,4,power_law_1.2,0.830720043182373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,1,power_law_1.01,0.6689856052398682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,1,power_law_1.01,0.8558527946472168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,1,power_law_1.01,1.2277824401855468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,1,power_law_1.01,1.6158975601196288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,2,power_law_1.2,0.04863359928131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,2,power_law_1.2,0.060089600086212155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,2,power_law_1.2,0.06317440271377564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,2,power_law_1.2,0.08899840116500854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,2,power_law_1.2,0.10269440412521362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,4,power_law_1.2,1.1548800468444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,2,power_law_1.2,0.12133760452270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,2,power_law_1.2,0.15052160024642944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,2,power_law_1.2,0.14950400590896606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,2,power_law_1.2,0.16012799739837646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,2,power_law_1.2,0.16076799631118774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,2,power_law_1.2,0.15978239774703978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,2,power_law_1.2,0.16799999475479127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,2,power_law_1.2,0.18275840282440187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,2,power_law_1.2,0.18844159841537475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,2,power_law_1.2,0.1956480026245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,2,power_law_1.2,0.20427520275115968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,2,power_law_1.2,0.218451189994812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,2,power_law_1.2,0.24668800830841064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,2,power_law_1.2,0.26071040630340575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,2,power_law_1.2,0.31914238929748534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,2,power_law_1.2,0.34219520092010497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,2,power_law_1.2,0.4364031791687012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,2,power_law_1.2,0.5303040027618409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,2,power_law_1.2,0.698745584487915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,2,power_law_1.2,0.892204761505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,2,power_law_1.2,1.226912021636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,2,power_law_1.2,1.5621631622314454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,2,power_law_1.01,0.026214399933815004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,2,power_law_1.01,0.045465600490570066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,2,power_law_1.01,0.05988479852676391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,2,power_law_1.01,0.08819199800491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,2,power_law_1.01,0.11175040006637574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,2,power_law_1.01,0.13427200317382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,2,power_law_1.01,0.16904319524765016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,2,power_law_1.01,0.17517440319061278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,2,power_law_1.01,0.1817728042602539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,2,power_law_1.01,0.18629759550094604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,2,power_law_1.01,0.18896000385284423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,2,power_law_1.01,0.18791680335998534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,2,power_law_1.01,0.19645440578460693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,2,power_law_1.01,0.20809600353240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,2,power_law_1.01,0.2080319881439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,2,power_law_1.01,0.21542398929595946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,2,power_law_1.01,0.21933441162109374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,2,power_law_1.01,0.2404927968978882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,2,power_law_1.01,0.25840001106262206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,2,power_law_1.01,0.551142406463623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,2,power_law_1.01,0.6096384048461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,2,power_law_1.01,0.423635196685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,2,power_law_1.01,0.4654848098754883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,2,power_law_1.01,0.5778367996215821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,2,power_law_1.01,0.6877056121826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,2,power_law_1.01,0.8975040435791015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,2,power_law_1.01,1.141100788116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,4,power_law_1.2,1.5147647857666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,4,power_law_1.2,1.9824895858764648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,128,balanced,0.04850133260091146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,128,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,128,balanced,0.05003733436266581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,128,balanced,0.05287999908129374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,128,balanced,0.05195199946562449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,128,balanced,0.055946667989095054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,128,balanced,0.057904000083605446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,128,balanced,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,128,balanced,0.06011199951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,128,balanced,0.058287998040517174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,128,balanced,0.062037333846092224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,128,balanced,0.06437866886456807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,128,balanced,0.06818666557470958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,128,balanced,0.0644053320089976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,128,balanced,0.07457600037256877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,128,balanced,0.07843733330567677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,128,balanced,0.07653333246707916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,128,balanced,0.08725333213806152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,128,balanced,0.0969493289788564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,128,balanced,0.10942932963371277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,128,balanced,0.12453333536783855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,128,balanced,0.15661866466204324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,128,balanced,0.18663465976715088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,128,balanced,0.23885866006215414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,128,balanced,0.28836266199747723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,128,balanced,0.3915146589279175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,128,balanced,0.48787200450897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,64,power_law_1.01,0.01756799966096878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,64,power_law_1.01,0.017081600427627564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,64,power_law_1.01,0.017209599912166595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,64,power_law_1.01,0.01804800033569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,64,power_law_1.01,0.019065600633621217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,64,power_law_1.01,0.021907199919223786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,64,power_law_1.01,0.03637759983539581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,64,power_law_1.01,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,64,power_law_1.01,0.03057279884815216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,64,power_law_1.01,0.031667199730873105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,64,power_law_1.01,0.03054719865322113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,64,power_law_1.01,0.031219199299812317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,64,power_law_1.01,0.030880001187324525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,64,power_law_1.01,0.030777600407600404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,64,power_law_1.01,0.03159680068492889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,64,power_law_1.01,0.03123840093612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,64,power_law_1.01,0.032767999172210696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,64,power_law_1.01,0.03650560081005096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,64,power_law_1.01,0.0385919988155365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,64,power_law_1.01,0.044198399782180785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,64,power_law_1.01,0.04800640046596527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,64,power_law_1.01,0.0577023983001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,64,power_law_1.01,0.07159680128097534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,64,power_law_1.01,0.09176959991455078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,64,power_law_1.01,0.11075199842453003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,64,power_law_1.01,0.15374079942703248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,64,power_law_1.01,0.20492160320281982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,1,balanced,0.09590400258700053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,1,balanced,0.1462453305721283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,1,balanced,0.2437653342882792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,1,balanced,0.44571733474731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,1,balanced,0.8364853064219157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,1,balanced,1.6184320449829102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,1,balanced,2.4028213818868003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,1,balanced,2.403264045715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,1,balanced,2.408122698465983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,1,balanced,2.4130026499430337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,1,balanced,2.416565259297689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,1,balanced,2.428623994191488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,1,balanced,2.4370346069335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,1,balanced,2.4589920043945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,1,balanced,2.4786292711893716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,1,balanced,2.4921226501464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,1,balanced,2.515711943308512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,1,balanced,2.5644799868265786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,1,balanced,2.602015972137451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,1,balanced,2.726970672607422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,1,balanced,2.816378593444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,1,balanced,2.9666293462117515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,1,balanced,3.1276960372924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,1,balanced,3.4911254247029624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,1,balanced,3.9987573623657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,8,power_law_1.2,0.08425599932670594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,1,balanced,4.597914695739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,8,power_law_1.2,0.09934080243110657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,8,power_law_1.2,0.1099392056465149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,8,power_law_1.2,0.12588800191879274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,1,balanced,6.168373107910156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,8,power_law_1.2,0.1477504014968872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,8,power_law_1.2,0.14968960285186766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,8,power_law_1.2,0.13825279474258423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,8,power_law_1.2,0.14511359930038453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,8,power_law_1.2,0.14366719722747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,8,power_law_1.2,0.1473536014556885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,8,power_law_1.2,0.14862719774246216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,8,power_law_1.2,0.1572543978691101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,8,power_law_1.2,0.16187520027160646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,8,power_law_1.2,0.16346240043640137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,8,power_law_1.2,0.17701120376586915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,8,power_law_1.2,0.18053760528564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,8,power_law_1.2,0.18896640539169313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,8,power_law_1.2,0.22756481170654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,8,power_law_1.2,0.24808320999145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,8,power_law_1.2,0.29051520824432375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,8,power_law_1.2,0.3139647960662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,8,power_law_1.2,0.4158976078033447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,8,power_law_1.2,0.44289278984069824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,8,power_law_1.2,0.5817279815673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,8,power_law_1.2,0.758841609954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,8,power_law_1.2,1.0551872253417969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,8,power_law_1.2,1.2379008293151856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,16,power_law_1.01,0.06625919938087463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,16,power_law_1.01,0.0753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,16,power_law_1.01,0.07441279888153077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,16,power_law_1.01,0.0819263994693756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,16,power_law_1.01,0.08375679850578308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,16,power_law_1.01,0.08310400247573853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,16,power_law_1.01,0.08426240086555481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,16,power_law_1.01,0.07972480058670044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,16,power_law_1.01,0.08459519743919372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,16,power_law_1.01,0.08462719917297364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,16,power_law_1.01,0.08654080033302307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,16,power_law_1.01,0.09108480215072631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,16,power_law_1.01,0.09144319891929627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,16,power_law_1.01,0.0940671980381012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,16,power_law_1.01,0.10250240564346313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,16,power_law_1.01,0.10553599596023559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,16,power_law_1.01,0.1112064003944397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,16,power_law_1.01,0.1289023995399475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,16,power_law_1.01,0.13657599687576294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,16,power_law_1.01,0.17426559925079346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,16,power_law_1.01,0.21445760726928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,16,power_law_1.01,0.2738687992095947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,16,power_law_1.01,0.2927743911743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,16,power_law_1.01,0.36969599723815916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,16,power_law_1.01,0.46416640281677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,16,power_law_1.01,0.6951168060302735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,16,power_law_1.01,0.8168576240539551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,16,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,16,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,16,balanced,0.053898667295773826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,16,balanced,0.06274666885534923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,16,balanced,0.08492799599965413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,16,balanced,0.11685867110888164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,16,balanced,0.14666666587193808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,16,balanced,0.14637333154678345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,16,balanced,0.14826132853825888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,16,balanced,0.14917332927385965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,16,balanced,0.1490666667620341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,16,balanced,0.15053866306940714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,16,balanced,0.15357866883277893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,16,balanced,0.15562132994333902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,16,balanced,0.16080000003178915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,16,balanced,0.16265599926312765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,16,balanced,0.1705333391825358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,16,balanced,0.17938133080800375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,16,balanced,0.18923733631769815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,16,balanced,0.20753065745035806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,16,balanced,0.22765866915384927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,16,balanced,0.2643360098203023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,16,balanced,0.302021324634552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,16,balanced,0.3762400150299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,16,balanced,0.48071467876434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,16,balanced,0.6207360029220581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,16,balanced,0.797205368677775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,8,power_law_1.01,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,8,power_law_1.01,0.0634112000465393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,8,power_law_1.01,0.07467520236968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,8,power_law_1.01,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,8,power_law_1.01,0.08090879917144775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,8,power_law_1.01,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,8,power_law_1.01,0.085343998670578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,8,power_law_1.01,0.08736000061035157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,8,power_law_1.01,0.08944640159606934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,8,power_law_1.01,0.08503680229187012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,8,power_law_1.01,0.08736000061035157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,8,power_law_1.01,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,8,power_law_1.01,0.09142400026321411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,8,power_law_1.01,0.09327359795570374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,8,power_law_1.01,0.09962239861488342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,8,power_law_1.01,0.10470399856567383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,8,power_law_1.01,0.11424000263214111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,8,power_law_1.01,0.13404159545898436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,8,power_law_1.01,0.1488767981529236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,8,power_law_1.01,0.18083839416503905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,8,power_law_1.01,0.18857599496841432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,8,power_law_1.01,0.24382081031799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,8,power_law_1.01,0.32417280673980714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,8,power_law_1.01,0.42437119483947755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,8,power_law_1.01,0.5430975914001465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,8,power_law_1.01,0.7812607765197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,8,power_law_1.01,0.9837311744689942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,2,balanced,0.03620799879233042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,2,balanced,0.05580799778302511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,2,balanced,0.07662400106589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,2,balanced,0.07914666831493378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,2,balanced,0.07898133496443431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,2,balanced,0.080485333998998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,2,balanced,0.08004799981911977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,2,balanced,0.0809386670589447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,2,balanced,0.08035199840863545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,2,balanced,0.08250133196512859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,2,balanced,0.08557333548863728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,2,balanced,0.08447999755541484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,2,balanced,0.08608532945315044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,2,balanced,0.09101866682370503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,2,balanced,0.09995733698209126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,2,balanced,0.0995306670665741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,2,balanced,0.10525332887967427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,2,balanced,0.12757866581281027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,2,balanced,0.1340053379535675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,2,balanced,0.16715200742085776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,2,balanced,0.19365866978963217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,2,balanced,0.2622773249944051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,2,balanced,0.3166293303171794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,2,balanced,0.44595734278361004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,2,balanced,0.5657013257344564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,2,balanced,0.8078933556874593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,2,balanced,1.0481119950612385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,1,power_law_1.01,0.035174399614334106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,1,power_law_1.01,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,1,power_law_1.01,0.03857919871807099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,1,power_law_1.01,0.043859198689460754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,1,power_law_1.01,0.0507968008518219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,1,power_law_1.01,0.06462079882621766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,1,power_law_1.01,0.06688640117645264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,1,power_law_1.01,0.06695039868354798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,1,power_law_1.01,0.06954240202903747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,1,power_law_1.01,0.07198079824447631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,1,power_law_1.01,0.07592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,1,power_law_1.01,0.0809984028339386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,1,power_law_1.01,0.08388479948043823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,1,power_law_1.01,0.08599680066108703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,1,power_law_1.01,0.09448959827423095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,1,power_law_1.01,0.0986240029335022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,1,power_law_1.01,0.10708479881286621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,1,power_law_1.01,0.13011200428009034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,1,power_law_1.01,0.15094399452209473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,1,power_law_1.01,0.19455360174179076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,1,power_law_1.01,0.23212800025939942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,1,power_law_1.01,0.32427520751953126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,1,power_law_1.01,0.4092735767364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,1,power_law_1.01,0.5862912178039551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,1,power_law_1.01,0.758080005645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,1,power_law_1.01,1.1218239784240722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,1,power_law_1.01,1.4657279968261718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,4,power_law_1.01,0.05086719989776611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,4,power_law_1.01,0.07285760045051574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,4,power_law_1.01,0.10209280252456665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,4,power_law_1.01,0.10319360494613647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,4,power_law_1.01,0.13441280126571656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,4,power_law_1.01,0.1593727946281433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,4,power_law_1.01,0.16880639791488647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,4,power_law_1.01,0.17218559980392456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,4,power_law_1.01,0.18347519636154175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,4,power_law_1.01,0.17479679584503174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,4,power_law_1.01,0.18313599824905397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,4,power_law_1.01,0.19353599548339845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,4,power_law_1.01,0.2061824083328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,4,power_law_1.01,0.2018048048019409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,4,power_law_1.01,0.2237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,4,power_law_1.01,0.22269439697265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,4,power_law_1.01,0.2317375898361206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,4,power_law_1.01,0.25340800285339354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,4,power_law_1.01,0.2753727912902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,4,power_law_1.01,0.32994558811187746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,4,power_law_1.01,0.3642560005187988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,4,power_law_1.01,0.46945919990539553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,4,power_law_1.01,0.5930624008178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,4,power_law_1.01,0.7759935855865479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,4,power_law_1.01,0.927462387084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,4,power_law_1.01,1.2870911598205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,4,power_law_1.01,1.6437952041625976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,8,power_law_1.01,0.0172992005944252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,8,power_law_1.01,0.01738879978656769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,8,power_law_1.01,0.017504000663757326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,8,power_law_1.01,0.01820160001516342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,8,power_law_1.01,0.01863040030002594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,8,power_law_1.01,0.019046400487422944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,8,power_law_1.01,0.01990399956703186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,8,power_law_1.01,0.02101760059595108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,8,power_law_1.01,0.023372800648212434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,8,power_law_1.01,0.022310400009155275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,8,power_law_1.01,0.02990719974040985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,8,power_law_1.01,0.034067198634147644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,8,power_law_1.01,0.03338240087032318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,8,power_law_1.01,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,8,power_law_1.01,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,8,power_law_1.01,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,8,power_law_1.01,0.034508800506591795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,8,power_law_1.01,0.04586879909038544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.01,0.0483711987733841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.01,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.01,0.07528319954872131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.01,0.09682559967041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.01,0.12099839448928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.01,0.16485120058059693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,16,power_law_1.01,0.07418879866600037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,16,power_law_1.01,0.08062080144882203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,16,power_law_1.01,0.07514879703521729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,16,power_law_1.01,0.07633280158042907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,16,power_law_1.01,0.0764415979385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,16,power_law_1.01,0.07443199753761291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,16,power_law_1.01,0.0770687997341156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,16,power_law_1.01,0.07670400142669678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,16,power_law_1.01,0.07646080255508422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,16,power_law_1.01,0.07919999957084656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,16,power_law_1.01,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,16,power_law_1.01,0.07888000011444092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,16,power_law_1.01,0.08364160060882568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,16,power_law_1.01,0.08197759985923767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,16,power_law_1.01,0.08651520013809204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,16,power_law_1.01,0.0886080026626587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,16,power_law_1.01,0.09359999895095825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,16,power_law_1.01,0.1054144024848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.01,0.11699199676513672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.01,0.135916805267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.01,0.15402239561080933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.01,0.18919680118560792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.01,0.23239679336547853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.01,0.3205440044403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.01,0.40897278785705565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.01,0.5672319889068603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.01,0.7737664222717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.01,0.20711679458618165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,64,power_law_1.01,0.05948160290718078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,64,power_law_1.01,0.05763840079307556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,64,power_law_1.01,0.05674880146980286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,64,power_law_1.01,0.059628802537918094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,64,power_law_1.01,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,64,power_law_1.01,0.059699201583862306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,64,power_law_1.01,0.062015998363494876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,64,power_law_1.01,0.06355199813842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,64,power_law_1.01,0.062118399143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,64,power_law_1.01,0.0632960021495819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,64,power_law_1.01,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,64,power_law_1.01,0.06538879871368408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,64,power_law_1.01,0.0657151997089386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,64,power_law_1.01,0.067084801197052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,64,power_law_1.01,0.07152000069618225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,64,power_law_1.01,0.07310720086097718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,64,power_law_1.01,0.07681919932365418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,64,power_law_1.01,0.08994560241699219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,64,power_law_1.01,0.1000704050064087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,64,power_law_1.01,0.1284991979598999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,64,power_law_1.01,0.14867839813232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,64,power_law_1.01,0.19323519468307496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,64,power_law_1.01,0.2342463970184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,64,power_law_1.01,0.3265984058380127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,64,power_law_1.01,0.4115903854370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,64,power_law_1.01,0.5988096237182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,16,power_law_1.01,0.1013375997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,64,power_law_1.01,0.7923583984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,16,power_law_1.01,0.1519551992416382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,16,power_law_1.01,0.19869439601898192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,16,power_law_1.01,0.1933184027671814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,16,power_law_1.01,0.19866880178451538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,16,power_law_1.01,0.19406720399856567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,16,power_law_1.01,0.17685120105743407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,16,power_law_1.01,0.18469120264053346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,16,power_law_1.01,0.1605504035949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,16,power_law_1.01,0.18674559593200685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,16,power_law_1.01,0.17104640007019042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,16,power_law_1.01,0.18406399488449096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,16,power_law_1.01,0.19848959445953368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,16,power_law_1.01,0.20074880123138428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,16,power_law_1.01,0.19829119443893434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,16,power_law_1.01,0.19833600521087646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,16,power_law_1.01,0.21660161018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,16,power_law_1.01,0.23413760662078859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,16,power_law_1.01,0.25486719608306885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.01,0.29484798908233645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,16,power_law_1.01,0.30422399044036863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,16,power_law_1.01,0.3168447971343994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,16,power_law_1.01,0.4131775856018066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,16,power_law_1.01,0.4950719833374023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,16,power_law_1.01,0.658681583404541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,16,power_law_1.01,0.7202879905700683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,16,power_law_1.01,1.0112319946289063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,16,power_law_1.01,1.233836841583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.01,0.38136959075927734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,4,balanced,0.0320266659061114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,4,balanced,0.03136533250411352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,4,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,4,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,4,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,4,balanced,0.03265066693226496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,4,balanced,0.03346133232116699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,4,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,4,balanced,0.03540800015131632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,4,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,4,balanced,0.03368533402681351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,4,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,4,balanced,0.03530666728814443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,4,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,4,balanced,0.039813332259655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,4,balanced,0.041877334316571556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,4,balanced,0.041877334316571556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,4,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,4,balanced,0.04996799925963084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,4,balanced,0.057989334066708885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,4,balanced,0.062314664324124656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,4,balanced,0.0827946662902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,4,balanced,0.11210667093594869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,4,balanced,0.15769599874814352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,4,balanced,0.19062399864196777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,4,balanced,0.26604799429575604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,4,balanced,0.3295733332633972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,2,power_law_1.01,0.0470335990190506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,2,power_law_1.01,0.5664703845977783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,2,power_law_1.01,0.07027199864387512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,2,power_law_1.01,0.09241600036621093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,2,power_law_1.01,0.12110719680786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,2,power_law_1.01,0.17719680070877075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,2,power_law_1.01,0.19487359523773193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,2,power_law_1.01,0.18212480545043946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,2,power_law_1.01,0.1935744047164917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,2,power_law_1.01,0.2011967897415161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,2,power_law_1.01,0.19419519901275634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,2,power_law_1.01,0.19627519845962524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,2,power_law_1.01,0.20303359031677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,2,power_law_1.01,0.21358079910278321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,2,power_law_1.01,0.22120959758758546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,2,power_law_1.01,0.23032960891723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,2,power_law_1.01,0.23480958938598634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,2,power_law_1.01,0.2577984094619751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,2,power_law_1.01,0.26749439239501954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,2,power_law_1.01,0.30283520221710203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,2,power_law_1.01,0.35292160511016846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,2,power_law_1.01,0.4268608093261719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,2,power_law_1.01,0.439961576461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,2,power_law_1.01,0.5555071830749512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,2,power_law_1.01,0.6966464042663574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,2,power_law_1.01,0.8820672035217285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,2,power_law_1.01,1.026035213470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,2,power_law_1.2,0.05777919888496399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,2,power_law_1.2,0.06587520241737366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,2,power_law_1.2,0.07299200296401978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,2,power_law_1.2,0.08814079761505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,2,power_law_1.2,0.09365760087966919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,2,power_law_1.2,0.11395200490951538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,2,power_law_1.2,0.11717120409011841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,32,balanced,0.048384000857671104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,32,balanced,0.0480373352766037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,32,balanced,0.04817600051561991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,32,balanced,0.051728000243504844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,32,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,32,balanced,0.057855998476346336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,32,balanced,0.05808533231417338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,32,balanced,0.0565280020236969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,32,balanced,0.06022400160630544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,32,balanced,0.058058664202690125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,32,balanced,0.06007466713587443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,2,power_law_1.2,0.11866240501403809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,2,power_law_1.2,0.11982719898223877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,2,power_law_1.2,0.12103040218353271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,2,power_law_1.2,0.12376320362091064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,2,power_law_1.2,0.12791680097579955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,2,power_law_1.2,0.13080960512161255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,32,balanced,0.06408533453941345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,32,balanced,0.06400000055631001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,2,power_law_1.2,0.1334720015525818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,32,balanced,0.06607466439406078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,2,power_law_1.2,0.1369472026824951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,2,power_law_1.2,0.1407487988471985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,2,power_law_1.2,0.14673919677734376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,2,power_law_1.2,0.17329920530319215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,2,power_law_1.2,0.1940287947654724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,32,balanced,0.07633600135644276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,32,balanced,0.07640533149242401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,2,power_law_1.2,0.24907519817352294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,32,balanced,0.08377599716186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,32,balanced,0.09818666179974873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,32,balanced,0.09706133604049683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,32,balanced,0.12008532881736755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,2,power_law_1.2,0.31305599212646484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,32,balanced,0.1339359978834788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,32,balanced,0.16485333442687988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,2,power_law_1.2,0.3999232053756714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,32,balanced,0.19131199518839517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,32,balanced,0.24894400437672934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,32,balanced,0.30049065748850506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,2,power_law_1.2,0.5045248031616211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,32,balanced,0.4189866781234741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,32,balanced,0.5334986845652262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,2,power_law_1.2,0.7493120193481445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,8,power_law_1.01,0.06047359704971313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,8,power_law_1.01,0.058246397972106935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,8,power_law_1.01,0.06149119734764099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,8,power_law_1.01,0.06290559768676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,8,power_law_1.01,0.061791998147964475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,8,power_law_1.01,0.06128000020980835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,8,power_law_1.01,0.05877760052680969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,8,power_law_1.01,0.056966400146484374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,8,power_law_1.01,0.057222402095794676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,8,power_law_1.01,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,8,power_law_1.01,0.04981760084629059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,8,power_law_1.01,0.06884480118751526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,8,power_law_1.01,0.06478080153465271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,2,power_law_1.2,0.9377023696899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,8,power_law_1.01,0.06503679752349853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,8,power_law_1.01,0.07006719708442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,8,power_law_1.01,0.0785535991191864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,8,power_law_1.01,0.08672639727592468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,8,power_law_1.01,0.10597120523452759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.01,0.1205888032913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.01,0.16223360300064088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.01,0.1795583963394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,2,power_law_1.2,1.4063360214233398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.01,0.24174079895019532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,2,power_law_1.2,1.8860671997070313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.01,0.31402881145477296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.01,0.4967616081237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.01,0.5628096103668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,1,power_law_1.01,0.051046401262283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,1,power_law_1.01,0.05101439952850342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,1,power_law_1.01,0.060678398609161376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,1,power_law_1.01,0.07342720031738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,1,power_law_1.01,0.08687360286712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,1,power_law_1.01,0.11103999614715576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,1,power_law_1.01,0.11679359674453735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,1,power_law_1.01,0.1189568042755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,1,power_law_1.01,0.12005759477615356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,1,power_law_1.01,0.12455040216445923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,1,power_law_1.01,0.12906880378723146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,1,power_law_1.01,0.13521280288696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,1,power_law_1.01,0.1376512050628662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,1,power_law_1.01,0.14260480403900147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,1,power_law_1.01,0.15148160457611085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,1,power_law_1.01,0.15720959901809692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,1,power_law_1.01,0.1712767958641052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,1,power_law_1.01,0.20149118900299073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,1,power_law_1.01,0.21899518966674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,1,power_law_1.01,0.27473280429840086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,1,power_law_1.01,0.3264703989028931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,1,power_law_1.01,0.44211201667785643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,1,power_law_1.01,0.5562623977661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,1,power_law_1.01,0.7852287769317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,1,power_law_1.01,1.0011072158813477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,1,power_law_1.01,1.4397824287414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,1,power_law_1.01,1.8687551498413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.01,0.9646464347839355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.01,1.2142144203186036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,1,power_law_1.2,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,1,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,1,power_law_1.2,0.043647998571395875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,1,power_law_1.2,0.05751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,1,power_law_1.2,0.056601601839065555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,1,power_law_1.2,0.060147202014923094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,1,power_law_1.2,0.06033920049667359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,1,power_law_1.2,0.06034560203552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,1,power_law_1.2,0.06123520135879516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,1,power_law_1.2,0.0627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,1,power_law_1.2,0.06427519917488098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,1,power_law_1.2,0.06704000234603882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,1,power_law_1.2,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,1,power_law_1.2,0.0720192015171051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,1,power_law_1.2,0.08150399923324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,1,power_law_1.2,0.08312320113182067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,1,power_law_1.2,0.09074559807777405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,1,power_law_1.2,0.10395519733428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,1,power_law_1.2,0.11571199893951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,1,power_law_1.2,0.14843519926071166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,1,power_law_1.2,0.17144960165023804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,8,power_law_1.01,0.03368319869041443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,8,power_law_1.01,0.033004799485206605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,8,power_law_1.01,0.03199360072612763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,8,power_law_1.01,0.03162879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,8,power_law_1.01,0.03237119913101196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,8,power_law_1.01,0.03427839875221252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,8,power_law_1.01,0.034796801209449765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,8,power_law_1.01,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,8,power_law_1.01,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,8,power_law_1.01,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,8,power_law_1.01,0.037363201379776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,8,power_law_1.01,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,8,power_law_1.01,0.03948160111904144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,8,power_law_1.01,0.03983359932899475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,1,power_law_1.2,0.2308799982070923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,8,power_law_1.01,0.04302079975605011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,8,power_law_1.01,0.04478079974651337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,8,power_law_1.01,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,8,power_law_1.01,0.05422719717025757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,8,power_law_1.01,0.060812801122665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,8,power_law_1.01,0.07056000232696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,8,power_law_1.01,0.0809664011001587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,8,power_law_1.01,0.09853439927101135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,8,power_law_1.01,0.1218176007270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,8,power_law_1.01,0.17125760316848754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,8,power_law_1.01,0.22294399738311768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,8,power_law_1.01,0.3060992002487183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,8,power_law_1.01,0.4110527992248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,16,power_law_1.01,0.08549759984016418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,16,power_law_1.01,0.12807680368423463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,1,power_law_1.2,0.2790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,16,power_law_1.01,0.1397760033607483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,16,power_law_1.01,0.1443519949913025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,16,power_law_1.01,0.13737599849700927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,16,power_law_1.01,0.14143359661102295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,16,power_law_1.01,0.12373119592666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,16,power_law_1.01,0.13644800186157227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,16,power_law_1.01,0.1310912013053894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,16,power_law_1.01,0.1241152048110962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,16,power_law_1.01,0.1356927990913391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,16,power_law_1.01,0.13571840524673462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,16,power_law_1.01,0.14243199825286865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,16,power_law_1.01,0.14016000032424927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,16,power_law_1.01,0.14977920055389404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,16,power_law_1.01,0.14888960123062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,16,power_law_1.01,0.15375360250473022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,16,power_law_1.01,0.17874560356140137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,16,power_law_1.01,0.19214080572128295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,16,power_law_1.01,0.21524479389190673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,16,power_law_1.01,0.2531327962875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,16,power_law_1.01,0.30379519462585447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,16,power_law_1.01,0.3433151960372925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,16,power_law_1.01,0.4571904182434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,16,power_law_1.01,0.5524288177490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,1,power_law_1.2,0.38703360557556155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,16,power_law_1.01,0.7650239944458008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,16,power_law_1.01,1.0271936416625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,8,power_law_1.01,0.026624000072479247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,8,power_law_1.01,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,8,power_law_1.01,0.03873920142650604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,8,power_law_1.01,0.044038400053977966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,8,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,8,power_law_1.01,0.045747199654579164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,8,power_law_1.01,0.05299839973449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,8,power_law_1.01,0.05247359871864319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,8,power_law_1.01,0.07246080040931702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,8,power_law_1.01,0.07304959893226623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,8,power_law_1.01,0.07050880193710327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,8,power_law_1.01,0.08579840064048767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,8,power_law_1.01,0.08538879752159119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,8,power_law_1.01,0.08572160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,8,power_law_1.01,0.09389439821243287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,8,power_law_1.01,0.10007679462432861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,8,power_law_1.01,0.09416319727897644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,8,power_law_1.01,0.08593279719352723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.01,0.08859519958496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.01,0.10254080295562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.01,0.10995839834213257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.01,0.14525439739227294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.01,0.16992640495300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.01,0.24306559562683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.01,0.3079551935195923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.01,0.4265279769897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.01,0.5212096214294434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,1,power_law_1.2,0.49457921981811526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,1,power_law_1.2,0.7046336174011231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,1,power_law_1.2,0.9108991622924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,16,power_law_1.01,0.05562880039215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,16,power_law_1.01,0.06885759830474854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,16,power_law_1.01,0.06572160124778748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,16,power_law_1.01,0.06938239932060242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,16,power_law_1.01,0.06934400200843811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,16,power_law_1.01,0.07304959893226623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,16,power_law_1.01,0.07056000232696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,16,power_law_1.01,0.06881279945373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,16,power_law_1.01,0.07192959785461425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,16,power_law_1.01,0.0709824025630951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,16,power_law_1.01,0.07262719869613647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,16,power_law_1.01,0.07504640221595764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,16,power_law_1.01,0.0750976026058197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,16,power_law_1.01,0.07742080092430115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,16,power_law_1.01,0.09075199961662292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,16,power_law_1.01,0.0926144003868103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,16,power_law_1.01,0.09438719749450683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,16,power_law_1.01,0.11108479499816895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,16,power_law_1.01,0.11399680376052856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,16,power_law_1.01,0.13887360095977783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,16,power_law_1.01,0.15873279571533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,16,power_law_1.01,0.1979200005531311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,16,power_law_1.01,0.25620479583740235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,32,power_law_1.01,0.04692479968070984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,16,power_law_1.01,0.37351040840148925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,16,power_law_1.01,0.44170241355895995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,16,power_law_1.01,0.6469056129455566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,16,power_law_1.01,0.9218175888061524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,32,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,32,power_law_1.01,0.04378879964351654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,32,power_law_1.01,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,32,power_law_1.01,0.04547840058803558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,32,power_law_1.01,0.04467200040817261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,32,power_law_1.01,0.04674560129642487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,32,power_law_1.01,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,32,power_law_1.01,0.04775680005550385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,32,power_law_1.01,0.04766719937324524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,32,power_law_1.01,0.050297600030899045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,32,power_law_1.01,0.052083200216293334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,32,power_law_1.01,0.053478401899337766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,32,power_law_1.01,0.057011198997497556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,32,power_law_1.01,0.06357120275497437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,32,power_law_1.01,0.06412799954414368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,32,power_law_1.01,0.06654719710350036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,32,power_law_1.01,0.0765887975692749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,32,power_law_1.01,0.08067200183868409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,32,power_law_1.01,0.09919360280036926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,32,power_law_1.01,0.11843839883804322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,32,power_law_1.01,0.14922879934310912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,32,power_law_1.01,0.1720639944076538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,32,power_law_1.01,0.234771203994751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,32,power_law_1.01,0.294265604019165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,32,power_law_1.01,0.41306238174438475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,32,power_law_1.01,0.5494207859039306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,1,balanced,0.020224000016848247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,1,balanced,0.0210506667693456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,1,balanced,0.021104000508785248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,1,balanced,0.021045332153638203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,1,balanced,0.02298133323589961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,1,balanced,0.02407466620206833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,1,balanced,0.024495999018351238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,1,balanced,0.025050667424996693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,1,balanced,0.025360000630219776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,1,balanced,0.02532266577084859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,1,balanced,0.027637332677841187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,1,balanced,0.02924266705910365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,1,balanced,0.028229333460330963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,1,balanced,0.029872000217437744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,1,balanced,0.053786665201187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,1,balanced,0.05445333321889242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,1,balanced,0.05632533133029938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,1,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,1,balanced,0.04833066463470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,1,balanced,0.06924266616503398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,1,balanced,0.07896000146865845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,1,balanced,0.10197333494822185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,1,balanced,0.12683199842770895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,1,balanced,0.17401599884033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,1,balanced,0.21344000101089478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,1,balanced,0.30988266070683795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,1,balanced,0.3973333438237508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,4,power_law_1.2,0.03438720107078552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,4,power_law_1.2,0.045049598813056944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,4,power_law_1.2,0.05326719880104065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,4,power_law_1.2,0.067084801197052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,4,power_law_1.2,0.0706496000289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,4,power_law_1.2,0.08488960266113281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,4,power_law_1.2,0.0887935996055603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,4,power_law_1.2,0.09497600197792053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,4,power_law_1.2,0.09522560238838196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,4,power_law_1.2,0.10232319831848144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,4,power_law_1.2,0.10512640476226806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,4,power_law_1.2,0.10640000104904175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,4,power_law_1.2,0.12565120458602905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,4,power_law_1.2,0.11689599752426147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,4,balanced,0.04444799820582072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,4,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,4,power_law_1.2,0.12015360593795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,4,balanced,0.04390400151411692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,4,balanced,0.053472002347310386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,4,balanced,0.07037866612275441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,4,balanced,0.07177066802978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,4,balanced,0.07108266651630402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,4,balanced,0.07308800021807353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,4,balanced,0.07142400244871776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,4,balanced,0.07073066631952922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,4,balanced,0.07257600128650665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,4,balanced,0.07644266883532207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,4,balanced,0.07653866708278656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,4,balanced,0.07694399853547414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,4,balanced,0.07996266583601634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,4,balanced,0.08051733175913493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,4,balanced,0.08138133088747661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,4,power_law_1.2,0.20599040985107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,4,power_law_1.2,0.2212671995162964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,4,power_law_1.2,0.17957760095596315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,4,balanced,0.08819733063379924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.2,0.19046399593353272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,4,balanced,0.09290132919947307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,4,balanced,0.10754666725794475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.2,0.20670080184936523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,4,balanced,0.11349333326021831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,4,balanced,0.15197867155075073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,4,balanced,0.17097600301106772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,4,balanced,0.23149865865707397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,4,balanced,0.3012320001920064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,4,balanced,0.43424534797668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,4,balanced,0.569978674252828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.2,0.21942400932312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.2,0.25763840675354005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.2,0.3111743927001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.2,0.41055998802185056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.2,0.50032639503479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.2,0.7272960186004639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,1,power_law_1.2,0.031046399474143983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,1,power_law_1.2,0.032595199346542356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,1,power_law_1.2,0.03371520042419433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,1,power_law_1.2,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,1,power_law_1.2,0.03809280097484589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,1,power_law_1.2,0.036908799409866334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,1,power_law_1.2,0.04080640077590943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,1,power_law_1.2,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,1,power_law_1.2,0.046649599075317384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,1,power_law_1.2,0.046828800439834596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,1,power_law_1.2,0.044531199336051944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,1,power_law_1.2,0.046937599778175354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,1,power_law_1.2,0.04911360144615173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,1,power_law_1.2,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,1,power_law_1.2,0.06255999803543091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,1,power_law_1.2,0.06485120058059693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,1,power_law_1.2,0.06493440270423889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,1,power_law_1.2,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,1,power_law_1.2,0.08160640001296997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,1,power_law_1.2,0.10496000051498414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.2,0.9146495819091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,1,power_law_1.2,0.11165440082550049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,1,power_law_1.2,0.15427839756011963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,1,power_law_1.2,0.17057280540466307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,2,power_law_1.2,0.06536960005760192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,2,power_law_1.2,0.07114880084991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,2,power_law_1.2,0.07824000120162963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,2,power_law_1.2,0.09262719750404358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,2,power_law_1.2,0.1044927954673767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,2,power_law_1.2,0.11679999828338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,2,power_law_1.2,0.14417279958724977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,2,power_law_1.2,0.15266560316085814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,2,power_law_1.2,0.15877759456634521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,2,power_law_1.2,0.16223360300064088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,2,power_law_1.2,0.16016000509262085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,2,power_law_1.2,0.16767359972000123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,2,power_law_1.2,0.17166080474853515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,2,power_law_1.2,0.17604479789733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,2,power_law_1.2,0.18931200504302978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,2,power_law_1.2,0.1913599967956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,2,power_law_1.2,0.20344960689544678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,2,power_law_1.2,0.23141119480133057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.2,0.25731840133666994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.2,0.31429119110107423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,1,power_law_1.2,0.23958399295806884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.2,0.36278400421142576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.2,0.48212480545043945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.2,0.587116813659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.2,0.8262847900390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.2,1.0666048049926757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.2,1.408556842803955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.2,1.9274816513061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,1,power_law_1.2,0.30903680324554444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,32,balanced,0.05359466870625814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,32,balanced,0.05268799761931101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,32,balanced,0.05472533404827118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,32,balanced,0.0646666685740153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,32,balanced,0.06910933554172516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,32,balanced,0.09166399637858073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,32,balanced,0.08809600273768108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,32,balanced,0.08373866478602092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,32,balanced,0.08513066172599792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,32,balanced,0.08969066540400188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,32,balanced,0.08844799796740214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,32,balanced,0.08643200000127156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,32,balanced,0.0869599978129069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,32,balanced,0.08482133348782857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,32,balanced,0.0936959981918335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,32,balanced,0.09129066268603007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,32,balanced,0.09886399904886882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,32,balanced,0.10338133573532104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,32,balanced,0.11015466849009196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,32,balanced,0.1221386690934499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,32,balanced,0.13243732849756876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,32,balanced,0.1607200006643931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,32,balanced,0.18367467323939005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,32,balanced,0.24984532594680786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,32,balanced,0.2995733420054118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,32,balanced,0.40997334321339923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,32,balanced,0.5229546626408895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,16,power_law_1.2,0.06673280000686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,16,power_law_1.2,0.07297919988632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,16,power_law_1.2,0.07370240092277527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,16,power_law_1.2,0.08154879808425904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,16,power_law_1.2,0.08295680284500122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,16,power_law_1.2,0.0786624014377594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,16,power_law_1.2,0.08077440261840821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,16,power_law_1.2,0.08281599879264831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,16,power_law_1.2,0.08294399976730346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,16,power_law_1.2,0.08379520177841186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,16,power_law_1.2,0.08689280152320862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,16,power_law_1.2,0.09136000275611877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,16,power_law_1.2,0.09287679791450501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,16,power_law_1.2,0.09670400023460388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,16,power_law_1.2,0.10460799932479858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,16,power_law_1.2,0.11189759969711303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,16,power_law_1.2,0.11016319990158081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,16,power_law_1.2,0.12753280401229858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,16,power_law_1.2,0.14191999435424804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,16,power_law_1.2,0.1806399941444397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,16,power_law_1.2,0.21971840858459474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,16,power_law_1.2,0.2858495950698853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,16,power_law_1.2,0.2991744041442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,16,power_law_1.2,0.41295361518859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,16,power_law_1.2,0.5076863765716553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,16,power_law_1.2,0.7108416080474853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,16,power_law_1.2,0.9275327682495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,1,power_law_1.2,0.42685441970825194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,64,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,64,balanced,0.047093331813812256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,64,balanced,0.04637866715590159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,64,balanced,0.049914668003718056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,64,balanced,0.049728001157442726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,64,balanced,0.05801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,64,balanced,0.056261335810025535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,64,balanced,0.05731200178464254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,64,balanced,0.0561653325955073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,64,balanced,0.05793599784374237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,64,balanced,0.058042665322621666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,64,balanced,0.058330665032068886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,64,balanced,0.06051200131575266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,64,balanced,0.05823466678460439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,64,balanced,0.0644053320089976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,64,balanced,0.06600533425807953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,64,balanced,0.0661599983771642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,64,balanced,0.07260266443093617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,64,balanced,0.07625600198904674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,64,balanced,0.08490133285522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,64,balanced,0.09274666508038838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,64,balanced,0.11157332857449849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,64,balanced,0.1285546620686849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,64,balanced,0.17564266920089722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,64,balanced,0.21628799041112265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,64,balanced,0.2916959921518962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,64,balanced,0.36902399857838947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,1,power_law_1.2,0.5537343978881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,64,power_law_1.2,0.055936002731323244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,64,power_law_1.2,0.056678402423858645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,64,power_law_1.2,0.0554751992225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,64,power_law_1.2,0.05414400100708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,64,power_law_1.2,0.05296639800071716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,64,power_law_1.2,0.05456640124320984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,64,power_law_1.2,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,64,power_law_1.2,0.05592319965362549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,64,power_law_1.2,0.05668479800224304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,64,power_law_1.2,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,64,power_law_1.2,0.058233600854873654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,64,power_law_1.2,0.06236799955368042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,64,power_law_1.2,0.06369919776916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,64,power_law_1.2,0.06449919939041138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,64,power_law_1.2,0.07317119836807251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,64,power_law_1.2,0.07508479952812194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,64,power_law_1.2,0.0803391993045807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,64,power_law_1.2,0.10033279657363892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,64,power_law_1.2,0.10549119710922242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,64,power_law_1.2,0.1385599970817566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,64,power_law_1.2,0.16002559661865234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,64,power_law_1.2,0.21390719413757325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,64,power_law_1.2,0.23800320625305177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,64,power_law_1.2,0.3051584005355835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,64,power_law_1.2,0.40014081001281737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,64,power_law_1.2,0.5638527870178223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,64,power_law_1.2,0.7020351886749268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,32,power_law_1.01,0.01635199934244156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,32,power_law_1.01,0.01595519930124283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,32,power_law_1.01,0.016096000373363496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,32,power_law_1.01,0.017215999960899352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,32,power_law_1.01,0.017932799458503724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,32,power_law_1.01,0.019974400103092194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,32,power_law_1.01,0.030976000428199767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,32,power_law_1.01,0.03115519881248474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,32,power_law_1.01,0.027923199534416198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,32,power_law_1.01,0.028217598795890808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,32,power_law_1.01,0.028224000334739686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,32,power_law_1.01,0.028352001309394838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,32,power_law_1.01,0.028723201155662535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,32,power_law_1.01,0.0282943993806839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,32,power_law_1.01,0.028543999791145323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,32,power_law_1.01,0.0289792001247406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,32,power_law_1.01,0.03025279939174652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,32,power_law_1.01,0.03275519907474518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,32,power_law_1.01,0.035206401348114015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,32,power_law_1.01,0.04087679982185364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,32,power_law_1.01,0.044512000679969785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,32,power_law_1.01,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,32,power_law_1.01,0.06229119896888733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,32,power_law_1.01,0.08150399923324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,32,power_law_1.01,0.09992319941520691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,32,power_law_1.01,0.14220160245895386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,32,power_law_1.01,0.1896448016166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,64,balanced,0.0316746657093366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,64,balanced,0.0315733328461647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,64,balanced,0.031301334500312805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,64,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,64,balanced,0.03588266670703888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,64,balanced,0.036901332437992096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,64,balanced,0.03566399961709976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,64,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,64,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,64,balanced,0.03733866661787033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,64,balanced,0.037471999724706016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,64,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,64,balanced,0.03935466706752777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,64,balanced,0.039674667020638786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,64,balanced,0.04164800047874451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,64,balanced,0.043968002001444496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,64,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,64,balanced,0.04459733267625173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,64,balanced,0.04373333354791006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,64,balanced,0.04796266555786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,64,balanced,0.05000533163547516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,64,balanced,0.05579199890295664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,64,balanced,0.0582826683918635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,64,balanced,0.07451733450094859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,64,balanced,0.08509866396586101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,64,balanced,0.11077866951624553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,64,balanced,0.13395733634630838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,32,power_law_1.2,0.08075519800186157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,32,power_law_1.2,0.09113600254058837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,32,power_law_1.2,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,32,power_law_1.2,0.09045119881629944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,32,power_law_1.2,0.08885120153427124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,32,power_law_1.2,0.09069439768791199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,32,power_law_1.2,0.08366079926490784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,32,power_law_1.2,0.0846783995628357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,32,power_law_1.2,0.089273601770401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,32,power_law_1.2,0.0915391981601715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,32,power_law_1.2,0.09151999950408936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,32,power_law_1.2,0.09182720184326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,32,power_law_1.2,0.09236479997634887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,32,power_law_1.2,0.09336320161819459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,32,power_law_1.2,0.10032000541687011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,32,power_law_1.2,0.09987199902534485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,32,power_law_1.2,0.10554239749908448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,32,power_law_1.2,0.11323519945144653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,32,power_law_1.2,0.12616319656372071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,32,power_law_1.2,0.14501760005950928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,32,power_law_1.2,0.163264000415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,32,power_law_1.2,0.19918080568313598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,32,power_law_1.2,0.23488640785217285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,32,power_law_1.2,0.34339199066162107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,32,power_law_1.2,0.4175551891326904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,32,power_law_1.2,0.5881216049194335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,32,power_law_1.2,0.7274623870849609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,128,balanced,0.03193599979082743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,128,balanced,0.03164266546567281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,128,balanced,0.031397332747777305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,128,balanced,0.033589333295822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,128,balanced,0.033674667278925575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,128,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,128,balanced,0.035232000052928925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,128,balanced,0.03358400116364161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,128,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,128,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,128,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,128,balanced,0.03735466549793879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,128,balanced,0.035605333745479584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,128,balanced,0.03746666759252548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,128,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,128,balanced,0.03809600075085958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,16,balanced,0.07145600020885468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,128,balanced,0.04015466570854187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,16,balanced,0.07112533350785573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,16,balanced,0.08159466584523518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,128,balanced,0.041722665230433144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,128,balanced,0.04346133271853129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,128,balanced,0.04362666606903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,128,balanced,0.04411733150482178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,128,balanced,0.048058668772379555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,128,balanced,0.050010666251182556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,128,balanced,0.057520002126693726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,128,balanced,0.06124266485373179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,128,balanced,0.07659733295440674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,128,balanced,0.09109866619110107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,64,balanced,0.035775999228159584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,64,balanced,0.03595733394225439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,64,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,64,balanced,0.04178666571776072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,64,balanced,0.04180799921353658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,64,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,64,balanced,0.04177600145339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,64,balanced,0.04177600145339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,64,balanced,0.04201066493988037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,64,balanced,0.04173333446184794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,64,balanced,0.0417546679576238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,64,balanced,0.041573333243529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,64,balanced,0.04558399816354116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,64,balanced,0.04364266494909922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,64,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,64,balanced,0.046613335609436035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,64,balanced,0.04785599807898203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,64,balanced,0.04964800179004669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,64,balanced,0.04983466863632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,64,balanced,0.05369600156943003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,64,balanced,0.055733333031336464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,64,balanced,0.06644799808661143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,64,balanced,0.07041066884994507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,64,balanced,0.09354133407274882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,64,balanced,0.10708799958229065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,64,balanced,0.135343998670578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,64,balanced,0.1668213407198588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,2,power_law_1.01,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,2,power_law_1.01,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,2,power_law_1.01,0.0526528000831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,2,power_law_1.01,0.06161919832229614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,2,power_law_1.01,0.07148799896240235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,2,power_law_1.01,0.0790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,2,power_law_1.01,0.09036160111427308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,2,power_law_1.01,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,2,power_law_1.01,0.09230080246925354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,2,power_law_1.01,0.09587200284004212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,2,power_law_1.01,0.09763839840888977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,2,power_law_1.01,0.10277119874954224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,2,power_law_1.01,0.10493439435958862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,2,power_law_1.01,0.11219840049743653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,2,power_law_1.01,0.11976319551467896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,2,power_law_1.01,0.12509440183639525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,2,power_law_1.01,0.12616319656372071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,2,power_law_1.01,0.14699519872665406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,2,power_law_1.01,0.16428159475326537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,2,power_law_1.01,0.2061824083328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,2,power_law_1.01,0.23248639106750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,2,power_law_1.01,0.3112895965576172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,2,power_law_1.01,0.3835455894470215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,2,power_law_1.01,0.5245888233184814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,2,power_law_1.01,0.6857344150543213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,2,power_law_1.01,0.9745856285095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,2,power_law_1.01,1.307481575012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,8,power_law_1.01,0.028467199206352232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,8,power_law_1.01,0.049619200825691226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,8,power_law_1.01,0.05533440113067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,8,power_law_1.01,0.06273279786109924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,8,power_law_1.01,0.06885759830474854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,8,power_law_1.01,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,64,balanced,0.03145066648721695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,8,power_law_1.01,0.0778432011604309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,8,power_law_1.01,0.07351040244102477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,8,power_law_1.01,0.08650239706039428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,64,balanced,0.03267733256022135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,8,power_law_1.01,0.0818880021572113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,64,balanced,0.03197866678237915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,8,power_law_1.01,0.08034560084342957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,64,balanced,0.03172266731659571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,64,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,64,balanced,0.03714133302370707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,64,balanced,0.03585066646337509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,64,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,64,balanced,0.035802667339642845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,64,balanced,0.0360959991812706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,64,balanced,0.03794133414824804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,64,balanced,0.03625066578388214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,64,balanced,0.040074666341145836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,64,balanced,0.04009066770474116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,64,balanced,0.04200533529122671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,64,balanced,0.04400533437728882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,64,balanced,0.04524266719818115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,64,balanced,0.046298667788505554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,64,balanced,0.048437332113583885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,8,power_law_1.01,0.10682879686355591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,8,power_law_1.01,0.10319360494613647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,8,power_law_1.01,0.09541760087013244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,8,power_law_1.01,0.1597440004348755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,8,power_law_1.01,0.1632256031036377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,8,power_law_1.01,0.16595200300216675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,8,power_law_1.01,0.15901440382003784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,64,balanced,0.05500799914201101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.01,0.14955519437789916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,64,balanced,0.06681600213050842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,64,balanced,0.08730133374532063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.01,0.16367360353469848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,64,balanced,0.09807466467221577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,64,balanced,0.12772267063458762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.01,0.17623679637908934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,64,balanced,0.15471466382344565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,64,balanced,0.2039946715037028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.01,0.2037247896194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,64,balanced,0.2555306752522786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.01,0.24741759300231933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.01,0.31814401149749755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.01,0.4042816162109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.01,0.5476736068725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.01,0.7148223876953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,16,power_law_1.2,0.08368639945983887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,16,power_law_1.2,0.08227199912071229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,16,power_law_1.2,0.0782144010066986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,16,power_law_1.2,0.08081279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,16,power_law_1.2,0.07642239928245545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,16,power_law_1.2,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,16,power_law_1.2,0.08181120157241821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,16,power_law_1.2,0.08152959942817688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,16,power_law_1.2,0.0840448021888733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,16,power_law_1.2,0.08332160115242004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,16,power_law_1.2,0.08181759715080261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,16,power_law_1.2,0.08350719809532166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,16,power_law_1.2,0.0865664005279541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,16,power_law_1.2,0.0859008014202118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,16,power_law_1.2,0.09314560294151306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,16,power_law_1.2,0.09192320108413696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,16,power_law_1.2,0.09874560236930847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,16,power_law_1.2,0.11118719577789307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.2,0.12378879785537719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.2,0.14687999486923217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.2,0.1683583974838257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.2,0.21314558982849122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.2,0.27545599937438964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.2,0.37023999691009524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.2,0.5198272228240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.2,0.747660779953003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.2,0.9498751640319825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,1,balanced,0.0699893335501353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,1,balanced,0.08691199620564778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,1,balanced,0.12504532933235168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,1,balanced,0.2028533418973287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,1,balanced,0.33744533856709796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,1,balanced,0.4731786648432414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,1,balanced,0.47278400262196857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,1,balanced,0.4790026744206746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,1,balanced,0.48287999629974365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,1,balanced,0.48234665393829346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,1,balanced,0.48894933859507245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,1,balanced,0.4984320004781087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,1,balanced,0.5007893244425455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,1,balanced,0.5134720007578532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,1,balanced,0.5290666818618774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,1,balanced,0.5383359988530477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,1,balanced,0.5640480120976766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,1,balanced,0.6175466775894165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,1,balanced,0.6614666779836019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,1,balanced,0.9408586819966634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,1,balanced,0.9318933486938477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,1,balanced,1.5110400517781575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,1,balanced,1.5587040583292644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,1,balanced,2.4230079650878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,1,balanced,2.634064038594564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,32,balanced,0.0529120018084844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,32,balanced,0.054485330979029335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,32,balanced,0.05514133473237356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,1,balanced,4.027797381083171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,32,balanced,0.0643093337615331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,32,balanced,0.07434666653474171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,32,balanced,0.07611733178297679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,32,balanced,0.07724266747633617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,32,balanced,0.07733333110809326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,32,balanced,0.07612266639868419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,32,balanced,0.07805333534876506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,32,balanced,0.07969599962234497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,32,balanced,0.08298666775226593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,32,balanced,0.08291199803352356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,32,balanced,0.08286933104197185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,1,balanced,4.994917233784993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,32,balanced,0.09035733342170715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,32,balanced,0.09101866682370503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,32,balanced,0.09482666850090027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,32,balanced,0.11044266819953918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,32,balanced,0.11880532900492351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,4,power_law_1.2,0.04732159972190857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,32,balanced,0.1490293343861898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,4,power_law_1.2,0.045638400316238406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,32,balanced,0.1693920095761617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,4,power_law_1.2,0.051744002103805545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,32,balanced,0.22833067178726196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,4,power_law_1.2,0.05805439949035644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,32,balanced,0.21076800425847372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,4,power_law_1.2,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,32,balanced,0.2828373312950134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,4,power_law_1.2,0.0574400007724762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,32,balanced,0.3227093418439229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,4,power_law_1.2,0.05610880255699158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,4,power_law_1.2,0.05600000023841858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,4,power_law_1.2,0.0554751992225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,4,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,4,power_law_1.2,0.05374720096588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,4,power_law_1.2,0.06095359921455383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,4,power_law_1.2,0.061452800035476686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,32,balanced,0.44087998072306317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,4,power_law_1.2,0.06162559986114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,32,balanced,0.5288319985071818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,4,power_law_1.2,0.07315199971199035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,4,power_law_1.2,0.0691648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,4,power_law_1.2,0.07541120052337646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,4,power_law_1.2,0.08754559755325317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.2,0.09587839841842652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.2,0.13455359935760497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.2,0.1488960027694702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.2,0.18908799886703492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.2,0.2517888069152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.2,0.37370240688323975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.2,0.4437568187713623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.2,0.6233535766601562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,1,power_law_1.2,0.021408000588417055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,1,power_law_1.2,0.020812800526618956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,1,power_law_1.2,0.021196800470352172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,1,power_law_1.2,0.022176000475883483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,1,power_law_1.2,0.02632960081100464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,1,power_law_1.2,0.03139840066432953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,1,power_law_1.2,0.030041599273681642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,1,power_law_1.2,0.032102400064468385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,1,power_law_1.2,0.027833598852157592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,1,power_law_1.2,0.031737598776817325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,1,power_law_1.2,0.033081600069999696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,1,power_law_1.2,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,1,power_law_1.2,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,1,power_law_1.2,0.03656960129737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,1,power_law_1.2,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,1,power_law_1.2,0.07816960215568543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,1,power_law_1.2,0.09279360175132752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,1,power_law_1.2,0.058412802219390866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.2,0.06637439727783204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,4,balanced,0.0869706670443217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,4,balanced,0.0865226686000824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,4,balanced,0.12985600034395853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,4,balanced,0.12129066387812297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,4,balanced,0.12109866738319397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,4,balanced,0.11593600114186604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,4,balanced,0.12486933668454488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,4,balanced,0.11734933654467265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,4,balanced,0.11786666512489319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,4,balanced,0.12141866485277812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,4,balanced,0.11555733283360799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,4,balanced,0.11010666688283284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,4,balanced,0.11822400490442912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,4,balanced,0.10930666327476501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,4,balanced,0.10663466652234395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,4,balanced,0.10050666332244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,4,balanced,0.09879466891288757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,4,balanced,0.1544533371925354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,4,balanced,0.13556266824404398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.2,0.08641279935836792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,4,balanced,0.17832533518473306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,4,balanced,0.21529066562652588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,4,balanced,0.29595200220743817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,4,balanced,0.36901334921518963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,4,balanced,0.5263199806213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,4,balanced,0.6763412952423096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,4,balanced,0.990069309870402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,4,balanced,1.299839973449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.2,0.10804480314254761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.2,0.14557440280914308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.2,0.8148223876953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.2,0.179967999458313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,4,balanced,0.060080001751581825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,4,balanced,0.06206933160622915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.2,0.2543999910354614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,4,balanced,0.06826133529345195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,4,balanced,0.08131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,4,balanced,0.09710933764775594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,4,balanced,0.14567466576894125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,4,balanced,0.14816000064214072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,4,balanced,0.15891200304031372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,4,balanced,0.1465013325214386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,4,balanced,0.15185599525769553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,4,balanced,0.1474133332570394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,4,balanced,0.1513653298219045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,4,balanced,0.1519306699434916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,4,balanced,0.14572800199190775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,4,balanced,0.15768532951672873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,4,balanced,0.1581760048866272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,4,balanced,0.16090133786201477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,4,balanced,0.1776799956957499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,4,balanced,0.18703466653823853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,4,balanced,0.21432000398635864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,4,balanced,0.23480532566706339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,4,balanced,0.2890933354695638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,4,balanced,0.341264009475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,4,balanced,0.4947253465652466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,4,balanced,0.6019573211669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,4,balanced,0.8626453081766764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,4,balanced,1.1197120348612468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.2,0.32600960731506345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.2,0.47409920692443847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,4,balanced,0.0470773329337438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,4,balanced,0.04776533444722494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,4,balanced,0.04983466863632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,4,balanced,0.05991999804973602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,4,balanced,0.0786240001519521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,4,balanced,0.07843733330567677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,4,balanced,0.0782773345708847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,4,balanced,0.08012799918651581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,4,balanced,0.07838400204976399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,4,balanced,0.07933866480986278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,4,balanced,0.08006933331489563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,4,balanced,0.08416000008583069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,4,balanced,0.08387733499209087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,4,balanced,0.08417600393295288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,4,balanced,0.08924800157546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,4,balanced,0.09125333031018575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,4,balanced,0.09269866347312927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,4,balanced,0.10170666376749675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,4,balanced,0.11055999994277954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,4,balanced,0.12573333581288657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,4,balanced,0.13683199882507324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,4,balanced,0.18862932920455933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,4,balanced,0.22632533311843872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,4,balanced,0.3229440053304036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,4,balanced,0.41648534933725995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,4,balanced,0.606549342473348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,4,balanced,0.7812746365865072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.2,0.6193535804748536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,4,power_law_1.01,0.04738560020923614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,4,power_law_1.01,0.05172479748725891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,4,power_law_1.01,0.05126399993896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,4,power_law_1.01,0.060127997398376466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,4,power_law_1.01,0.060550397634506224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,4,power_law_1.01,0.06227200031280518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,4,power_law_1.01,0.06346240043640136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,4,power_law_1.01,0.06277120113372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,4,power_law_1.01,0.06332799792289734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,4,power_law_1.01,0.06467199921607972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,4,power_law_1.01,0.06477439999580384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,4,power_law_1.01,0.07644799947738648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,4,power_law_1.01,0.07703040242195129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,4,power_law_1.01,0.08268160223960877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,4,power_law_1.01,0.09772160053253173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,4,power_law_1.01,0.10574719905853272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,4,power_law_1.01,0.11544959545135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,4,power_law_1.01,0.13923840522766112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,4,power_law_1.01,0.1837823987007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,4,power_law_1.01,0.23025920391082763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,4,power_law_1.01,0.31960320472717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,4,power_law_1.01,0.38468480110168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,4,power_law_1.01,0.5060863971710206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,4,power_law_1.01,0.7856768131256103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,4,power_law_1.01,0.9345664024353028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,4,power_law_1.01,1.5752127647399903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,4,power_law_1.01,1.915398406982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,8,power_law_1.2,0.04930559992790222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,8,power_law_1.2,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,8,power_law_1.2,0.052249598503112796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,8,power_law_1.2,0.0609279990196228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,8,power_law_1.2,0.0651199996471405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,8,power_law_1.2,0.0657151997089386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,2,balanced,0.03427733232577642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,2,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,8,power_law_1.2,0.07169920206069946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,2,balanced,0.035429333647092186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,2,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,2,balanced,0.0397173340121905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,2,balanced,0.039349332451820374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,2,balanced,0.039701332648595176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,2,balanced,0.04165333261092504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,2,balanced,0.039877332746982574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,2,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,2,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,2,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,2,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,2,balanced,0.042597333590189614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,8,power_law_1.2,0.069651198387146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,8,power_law_1.2,0.07589759826660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,8,power_law_1.2,0.07587199807167053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,8,power_law_1.2,0.07643520236015319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,8,power_law_1.2,0.07677440047264099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,8,power_law_1.2,0.07830399870872498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,8,power_law_1.2,0.08166400194168091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,8,power_law_1.2,0.0877888023853302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,8,power_law_1.2,0.0936191976070404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,8,power_law_1.2,0.09964159727096558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,8,power_law_1.2,0.10647679567337036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,8,power_law_1.2,0.11907199621200562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,8,power_law_1.2,0.14112000465393065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,8,power_law_1.2,0.17097599506378175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,8,power_law_1.2,0.20894720554351806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,8,power_law_1.2,0.2638400077819824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,8,power_law_1.2,0.3588543891906738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,8,power_law_1.2,0.44527997970581057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,2,balanced,0.05013866722583771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,2,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,2,balanced,0.055786664287249245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,8,power_law_1.2,0.5837823867797851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,8,power_law_1.2,0.7863359928131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,2,balanced,0.06442666550477345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,2,balanced,0.07233599821726482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,2,balanced,0.09714667002360027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,2,balanced,0.1018293301264445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,2,balanced,0.13781866431236267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,2,balanced,0.1853546698888143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,2,balanced,0.2548639973004659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,2,balanced,0.34090133508046466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,2,balanced,0.4933493137359619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,2,balanced,0.643669327100118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,4,balanced,0.06657599906126659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,4,balanced,0.07248533268769582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,4,balanced,0.08955732981363933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,4,balanced,0.12450666228930156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,4,balanced,0.17323199907938638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,4,balanced,0.21730132897694907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,4,balanced,0.21913599967956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,4,balanced,0.2200053334236145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,4,balanced,0.21997867027918497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,4,balanced,0.22369599342346191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,4,balanced,0.22702932357788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,4,balanced,0.23079466819763184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,4,balanced,0.23354132970174155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,4,balanced,0.23595199982325235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,4,balanced,0.24697067340215048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,4,balanced,0.25115732351938885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,4,balanced,0.266704003016154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,4,balanced,0.29737599690755206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,4,balanced,0.32282666365305585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,4,balanced,0.48928534984588623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,4,balanced,0.4445066849390666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,4,balanced,0.63701331615448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,4,balanced,0.6916053295135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,4,balanced,1.0146559874216716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,4,balanced,1.1595786412556965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,4,balanced,1.727941354115804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,4,balanced,2.144672075907389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,32,power_law_1.2,0.053350400924682614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,32,power_law_1.2,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,32,power_law_1.2,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,32,power_law_1.2,0.04695680141448975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,32,power_law_1.2,0.04913919866085052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,32,power_law_1.2,0.048563200235366824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,32,power_law_1.2,0.05022720098495483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,32,power_law_1.2,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,32,power_law_1.2,0.05182719826698303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,32,power_law_1.2,0.05281919836997986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,32,power_law_1.2,0.05516800284385681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,32,power_law_1.2,0.05607680082321167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,32,power_law_1.2,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,32,power_law_1.2,0.06377599835395813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,32,power_law_1.2,0.06855679750442505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,32,power_law_1.2,0.07199360132217407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,32,power_law_1.2,0.08094080090522766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,32,power_law_1.2,0.09233279824256897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,32,power_law_1.2,0.10602240562438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,32,power_law_1.2,0.13770240545272827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,32,power_law_1.2,0.15820800065994262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,32,power_law_1.2,0.21383678913116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,32,power_law_1.2,0.2544384002685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,32,power_law_1.2,0.3414144039154053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,32,power_law_1.2,0.46709117889404295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,32,power_law_1.2,0.7092607975006103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,32,power_law_1.2,0.8388544082641601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,1,power_law_1.01,0.10213119983673095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,1,power_law_1.01,0.10805120468139648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,1,power_law_1.01,0.1179967999458313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,1,power_law_1.01,0.14326399564743042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,1,power_law_1.01,0.18404480218887329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,1,power_law_1.01,0.21897599697113038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,1,power_law_1.01,0.24186880588531495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,1,power_law_1.01,0.2840384006500244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,1,power_law_1.01,0.2950272083282471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,1,power_law_1.01,0.30062079429626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,1,power_law_1.01,0.3099071979522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,1,power_law_1.01,0.3201280117034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,1,power_law_1.01,0.3248831987380981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,1,power_law_1.01,0.33773438930511473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,1,power_law_1.01,0.358732795715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,1,power_law_1.01,0.3699520111083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,1,power_law_1.01,0.39041919708251954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,1,power_law_1.01,0.4408895969390869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.01,0.480134391784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.01,0.5836671829223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.01,0.6732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.01,0.8744192123413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.01,1.072339153289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.01,1.465664005279541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,1,power_law_1.01,1.859814453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,1,power_law_1.01,2.64833927154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,1,power_law_1.01,3.4371456146240233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,16,power_law_1.2,0.09917439818382263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,32,balanced,0.04916800061861674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,32,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,32,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,32,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,32,balanced,0.06338666876157124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,32,balanced,0.08111999928951263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,32,balanced,0.08166400094827016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,32,balanced,0.07995733122030894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,32,balanced,0.08268799881140391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,32,balanced,0.08160533507664998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,32,balanced,0.08307733138402303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,32,balanced,0.08571733037630717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,32,balanced,0.08591999610265096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,32,balanced,0.08727999528249104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,16,power_law_1.2,0.1133504033088684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,16,power_law_1.2,0.11729279756546021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,16,power_law_1.2,0.1242751955986023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,16,power_law_1.2,0.12471040487289428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,16,power_law_1.2,0.11946239471435546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,32,balanced,0.1002293328444163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,16,power_law_1.2,0.12126719951629639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,32,balanced,0.09974400202433269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,16,power_law_1.2,0.12095359563827515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,16,power_law_1.2,0.12497279644012452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,16,power_law_1.2,0.12348159551620483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,16,power_law_1.2,0.1230463981628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,16,power_law_1.2,0.12872960567474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,16,power_law_1.2,0.1350335955619812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,32,balanced,0.0995786686738332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,32,balanced,0.11103999614715576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,16,power_law_1.2,0.13737599849700927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,32,balanced,0.11649066209793091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,32,balanced,0.1304213305314382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,16,power_law_1.2,0.14198399782180787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,32,balanced,0.14457600315411887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,32,balanced,0.1747786601384481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,16,power_law_1.2,0.1502079963684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,32,balanced,0.20068265994389853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,32,balanced,0.2704533338546753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,16,power_law_1.2,0.1594879984855652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,32,balanced,0.32078399260838825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,16,power_law_1.2,0.1861631989479065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,32,balanced,0.4314719835917155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,16,power_law_1.2,0.21147520542144777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,32,balanced,0.5605599880218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,16,power_law_1.2,0.2565056085586548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,16,power_law_1.2,0.2914367914199829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,16,power_law_1.2,0.39169280529022216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,16,power_law_1.2,0.4433152198791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,16,power_law_1.2,0.5796288013458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,16,power_law_1.2,0.7999231815338135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,16,power_law_1.2,0.9640064239501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,16,power_law_1.2,1.2950912475585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,4,balanced,0.04413333535194397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,4,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,4,balanced,0.05156800150871277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,4,balanced,0.07645333309968312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,4,balanced,0.10383466879526775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,4,balanced,0.15679466724395752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,4,balanced,0.1581439971923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,4,balanced,0.1602720022201538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,4,balanced,0.16200533509254456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,4,balanced,0.16194132963816324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,4,balanced,0.1639786660671234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,4,balanced,0.16847999890645346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,4,balanced,0.17239999771118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,4,balanced,0.17457600434621176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,4,balanced,0.18204800287882486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,4,balanced,0.18784532944361368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,4,balanced,0.19112533330917358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,4,balanced,0.21426133314768472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,4,balanced,0.22953067223230997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,4,balanced,0.262992004553477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,4,balanced,0.2926773428916931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,4,balanced,0.36908801396687824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,4,balanced,0.43666664759318036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,4,balanced,0.5913493235905966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,4,balanced,0.7247839768727621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,4,balanced,1.0210239887237549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,4,balanced,1.29093337059021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,8,power_law_1.2,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,8,power_law_1.2,0.08469120264053345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,8,power_law_1.2,0.09057279825210571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,8,power_law_1.2,0.0995519995689392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,8,power_law_1.2,0.10164480209350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,8,power_law_1.2,0.10551040172576905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,8,power_law_1.2,0.10552959442138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,8,power_law_1.2,0.10613119602203369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,8,power_law_1.2,0.1103551983833313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,8,power_law_1.2,0.11445759534835816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,8,power_law_1.2,0.11535359621047973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,8,power_law_1.2,0.11657600402832032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,8,power_law_1.2,0.1189568042755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,8,power_law_1.2,0.12755839824676513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,8,power_law_1.2,0.13062399625778198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,8,power_law_1.2,0.14408960342407226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,8,power_law_1.2,0.1403712034225464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,8,power_law_1.2,0.15470080375671386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,8,power_law_1.2,0.17191679477691652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,8,power_law_1.2,0.20753920078277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,8,power_law_1.2,0.23998079299926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,8,power_law_1.2,0.2829567909240723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,8,power_law_1.2,0.40972161293029785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,8,power_law_1.2,0.5195648193359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,256,balanced,0.021594665944576263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,32,power_law_1.2,0.034220799803733826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,256,balanced,0.021151999632517498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,256,balanced,0.023120000958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,32,power_law_1.2,0.02815360128879547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,32,power_law_1.2,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,32,power_law_1.2,0.030239999294281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,32,power_law_1.2,0.029760000109672547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,32,power_law_1.2,0.03132160007953644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,32,power_law_1.2,0.030963200330734252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,32,power_law_1.2,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,32,power_law_1.2,0.0308351993560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,256,balanced,0.026736001173655193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,32,power_law_1.2,0.031193599104881287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,256,balanced,0.04836266736189524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,256,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,32,power_law_1.2,0.03188480138778686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,256,balanced,0.04459733267625173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,8,power_law_1.2,0.605836820602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,256,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,32,power_law_1.2,0.03219839930534363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,256,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,256,balanced,0.04295999805132548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,8,power_law_1.2,0.7464960098266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,32,power_law_1.2,0.03288959860801697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,256,balanced,0.04170133173465729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,32,power_law_1.2,0.03383679986000061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,256,balanced,0.0421973317861557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,32,power_law_1.2,0.03751679956912994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,256,balanced,0.04065600037574768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,32,power_law_1.2,0.03834879994392395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,32,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,256,balanced,0.03961066653331121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,32,power_law_1.2,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,256,balanced,0.047237331668535866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,32,power_law_1.2,0.046463999152183535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,32,power_law_1.2,0.05507199764251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,32,power_law_1.2,0.05987200140953064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,32,power_law_1.2,0.0719488024711609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,32,power_law_1.2,0.0753216028213501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,8,power_law_1.2,1.0413951873779297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,32,power_law_1.2,0.11415679454803467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,256,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,256,balanced,0.04348266621430715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,32,power_law_1.2,0.1345728039741516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,256,balanced,0.04626133541266123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,256,balanced,0.05134400228659312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,256,balanced,0.072543998559316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,256,balanced,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,256,balanced,0.11179733276367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,256,balanced,0.13904533783594766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,256,balanced,0.19589867194493613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,256,balanced,0.2519413431485494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,256,balanced,0.36448001861572266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,256,balanced,0.4766720136006673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,32,power_law_1.2,0.18867199420928954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,32,power_law_1.2,0.23199360370635985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,2,power_law_1.01,0.1128000020980835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,2,power_law_1.01,0.11029119491577148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,2,power_law_1.01,0.17192959785461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,2,power_law_1.01,0.21030399799346924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,2,power_law_1.01,0.21047680377960204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,2,power_law_1.01,0.21128320693969727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,2,power_law_1.01,0.21690239906311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,2,power_law_1.01,0.21137919425964355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,2,power_law_1.01,0.23904640674591066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,2,power_law_1.01,0.2422976016998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,2,power_law_1.01,0.2355583906173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,2,power_law_1.01,0.2608576059341431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,2,power_law_1.01,0.25208320617675783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,2,power_law_1.01,0.256057596206665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,2,power_law_1.01,0.28557438850402833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,2,power_law_1.01,0.27883520126342776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,2,power_law_1.01,0.3110719919204712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,2,power_law_1.01,0.3928767919540405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,2,power_law_1.01,0.4017471790313721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,2,power_law_1.01,0.5164735794067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,2,power_law_1.01,0.6601727962493896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,2,power_law_1.01,0.7777279853820801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,2,power_law_1.01,1.0988096237182616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,2,power_law_1.01,1.6552255630493165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,2,power_law_1.01,1.8347967147827149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,2,power_law_1.01,3.1038911819458006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,2,power_law_1.01,3.7246078491210937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,8,power_law_1.2,0.020204800367355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,8,power_law_1.2,0.024991999566555022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,8,power_law_1.2,0.026035198569297792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,8,power_law_1.2,0.028863999247550964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,8,power_law_1.2,0.0342848002910614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,8,power_law_1.2,0.03665919899940491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,8,power_law_1.2,0.041017600893974306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,8,power_law_1.2,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,8,power_law_1.2,0.044377601146698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,8,power_law_1.2,0.0438400000333786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,8,power_law_1.2,0.04251520037651062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,8,power_law_1.2,0.06037120223045349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,8,power_law_1.2,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,8,power_law_1.2,0.0516543984413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,8,power_law_1.2,0.07074559926986694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,8,power_law_1.2,0.07294719815254211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,2,power_law_1.01,0.051795202493667605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,8,power_law_1.2,0.07447680234909057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,2,power_law_1.01,0.06929280161857605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,8,power_law_1.2,0.08057600259780884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,2,power_law_1.01,0.08010879755020142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,8,power_law_1.2,0.08081920146942138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,8,power_law_1.2,0.08700159788131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,2,power_law_1.01,0.10691839456558228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,8,power_law_1.2,0.08744320273399353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,8,power_law_1.2,0.10375679731369018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,2,power_law_1.01,0.13240959644317626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,2,power_law_1.01,0.15745279788970948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,2,power_law_1.01,0.18954880237579347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,2,power_law_1.01,0.20200960636138915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,8,power_law_1.2,0.1176192045211792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,2,power_law_1.01,0.20264959335327148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,8,power_law_1.2,0.14599679708480834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,2,power_law_1.01,0.21107840538024902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,8,power_law_1.2,0.18369280099868773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,8,power_law_1.2,0.2650943994522095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,8,power_law_1.2,0.3293567895889282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,2,power_law_1.01,0.21247360706329346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,2,power_law_1.01,0.22280960083007811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,2,power_law_1.01,0.2234560012817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,2,power_law_1.01,0.23534719944000243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,2,power_law_1.01,0.2538367986679077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,2,power_law_1.01,0.2598335981369019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,2,power_law_1.01,0.2715519905090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,2,power_law_1.01,0.30230400562286375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,2,power_law_1.01,0.3258879899978638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,2,power_law_1.01,0.3900480031967163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,2,power_law_1.01,0.44513278007507323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,2,power_law_1.01,0.5608191967010498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,2,power_law_1.01,0.6743680000305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,2,power_law_1.01,0.89749755859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,2,power_law_1.01,1.102201557159424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,4,power_law_1.2,0.0382207989692688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,4,power_law_1.2,0.03829759955406189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,4,power_law_1.2,0.0398719996213913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,2,power_law_1.01,1.5950336456298828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,4,power_law_1.2,0.04126079976558685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,4,power_law_1.2,0.041945600509643556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,4,power_law_1.2,0.042719998955726625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,4,power_law_1.2,0.04259839951992035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,4,power_law_1.2,0.043084800243377686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,2,power_law_1.01,2.002412796020508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,4,power_law_1.2,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,4,power_law_1.2,0.043750399351119997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,4,power_law_1.2,0.043968001008033754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,4,power_law_1.2,0.04444800019264221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,4,power_law_1.2,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,4,power_law_1.2,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,4,power_law_1.2,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,4,power_law_1.2,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,4,power_law_1.2,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,4,power_law_1.2,0.06256639957427979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,4,power_law_1.2,0.07505279779434204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,4,power_law_1.2,0.09548159837722778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,4,power_law_1.2,0.11902719736099243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,4,power_law_1.2,0.15499520301818848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,4,power_law_1.2,0.18924800157546998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,4,power_law_1.2,0.26460158824920654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,4,power_law_1.2,0.33831679821014404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,4,power_law_1.2,0.4916031837463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,4,power_law_1.2,0.6453824043273926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,256,power_law_1.01,0.06039040088653565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,256,power_law_1.01,0.06102399826049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,256,power_law_1.01,0.0605184018611908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,256,power_law_1.01,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,256,power_law_1.01,0.06395519971847534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,256,power_law_1.01,0.0658623993396759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,256,power_law_1.01,0.06587520241737366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,256,power_law_1.01,0.06694399714469909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,256,power_law_1.01,0.06632320284843445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,256,power_law_1.01,0.06612480282783509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,256,power_law_1.01,0.06652160286903382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,256,power_law_1.01,0.06664959788322448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,256,power_law_1.01,0.0680512011051178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,256,power_law_1.01,0.06871039867401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,256,power_law_1.01,0.07143679857254029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,256,power_law_1.01,0.07128319740295411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,256,power_law_1.01,0.07311360239982605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,256,power_law_1.01,0.08862720131874084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,256,power_law_1.01,0.09593600034713745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,256,power_law_1.01,0.12715519666671754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,256,power_law_1.01,0.15557760000228882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,256,power_law_1.01,0.20140159130096436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,256,power_law_1.01,0.2404927968978882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,256,power_law_1.01,0.32778880596160886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,256,power_law_1.01,0.4284095764160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,256,power_law_1.01,0.6147776126861573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,4,power_law_1.01,0.03429119884967804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,4,power_law_1.01,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,4,power_law_1.01,0.0353408008813858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,4,power_law_1.01,0.03780480027198792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,4,power_law_1.01,0.03861759901046753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,4,power_law_1.01,0.03868159949779511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,4,power_law_1.01,0.03902080059051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,4,power_law_1.01,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,256,power_law_1.01,0.8123904228210449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,4,power_law_1.01,0.03953920006752014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,4,power_law_1.01,0.04031359851360321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,4,power_law_1.01,0.04123519957065582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,16,balanced,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,16,balanced,0.041509332756201424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,16,balanced,0.04383466641108195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,16,balanced,0.042117332418759666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,16,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,16,balanced,0.05003733436266581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,16,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,16,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,16,balanced,0.051856001218159996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,16,balanced,0.051813334226608276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,16,balanced,0.05189333359400431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,16,balanced,0.054117331902186074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,16,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,16,balanced,0.05574933191140493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,16,balanced,0.05991999804973602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,16,balanced,0.06005866825580597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,16,balanced,0.06403733293215434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,16,balanced,0.07045333087444305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,16,balanced,0.07215466598669688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,16,balanced,0.08417066931724548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,16,balanced,0.09317333499590556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,16,balanced,0.12174933155377705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,4,power_law_1.01,0.04416640102863312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,4,power_law_1.01,0.047705599665641786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,4,power_law_1.01,0.05146239995956421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,16,balanced,0.15382933616638184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,16,balanced,0.2166986664136251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,16,balanced,0.26105600595474243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,16,balanced,0.3609120051066081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,16,balanced,0.46625598271687824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,4,power_law_1.01,0.05980160236358643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,4,power_law_1.01,0.06260480284690857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,4,power_law_1.01,0.07822719812393189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,4,power_law_1.01,0.09171199798583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,4,power_law_1.01,0.09402880072593689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,4,power_law_1.01,0.14099199771881105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,4,power_law_1.01,0.15349760055541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,4,power_law_1.01,0.2425407886505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,4,power_law_1.01,0.3329024076461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,4,power_law_1.01,0.5208767890930176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,4,power_law_1.01,0.7119487762451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,4,power_law_1.01,1.060159969329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,4,power_law_1.01,1.2560832023620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,8,power_law_1.01,0.050937598943710326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,8,power_law_1.01,0.05863680243492127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,8,power_law_1.01,0.0762880027294159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,8,power_law_1.01,0.07208960056304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,8,power_law_1.01,0.08278399705886841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,8,power_law_1.01,0.08988159894943237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,8,power_law_1.01,0.08650239706039428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,32,power_law_1.2,0.0205375999212265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,32,power_law_1.2,0.01902720034122467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,32,power_law_1.2,0.019449600577354433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,32,power_law_1.2,0.019487999379634857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,32,power_law_1.2,0.019443200528621675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,32,power_law_1.2,0.0212351992726326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,32,power_law_1.2,0.022521600127220154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,32,power_law_1.2,0.02398719936609268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,32,power_law_1.2,0.02398719936609268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,32,power_law_1.2,0.024377599358558655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,32,power_law_1.2,0.03790720105171204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,32,power_law_1.2,0.03813759982585907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,32,power_law_1.2,0.03868159949779511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,32,power_law_1.2,0.036985599994659425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,32,power_law_1.2,0.03950720131397247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,8,power_law_1.01,0.09418240189552307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,8,power_law_1.01,0.09681280255317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,8,power_law_1.01,0.08399360179901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,8,power_law_1.01,0.09299839735031128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,8,power_law_1.01,0.09438719749450683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,8,power_law_1.01,0.09502080082893372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,32,power_law_1.2,0.03653759956359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,8,power_law_1.01,0.101254403591156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,32,power_law_1.2,0.04084480106830597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,32,power_law_1.2,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.2,0.04631040096282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.2,0.05766400098800659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.2,0.06652160286903382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.2,0.08913279771804809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.2,0.10844800472259522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,8,power_law_1.01,0.10968960523605346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,8,power_law_1.01,0.10970239639282227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,8,power_law_1.01,0.11541119813919068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,8,power_law_1.01,0.1289728045463562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,8,power_law_1.01,0.14473600387573243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.2,0.14325759410858155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,8,power_law_1.01,0.17749119997024537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,8,power_law_1.01,0.19566719532012938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,8,power_law_1.01,0.258022403717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,8,power_law_1.01,0.3088639974594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,8,power_law_1.01,0.4129216194152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,8,power_law_1.01,0.540934419631958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.2,0.17770880460739136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,8,power_law_1.01,0.7466368198394775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,8,power_law_1.01,0.9179072380065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.2,0.2657407999038696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,4,power_law_1.01,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,4,power_law_1.01,0.03806079924106598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,4,power_law_1.01,0.03797760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,4,power_law_1.01,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,4,power_law_1.01,0.041382399201393125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,4,power_law_1.01,0.04158720076084137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,4,power_law_1.01,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,4,power_law_1.01,0.04291839897632599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,4,power_law_1.01,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,4,power_law_1.01,0.0427264004945755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,4,power_law_1.01,0.044268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,4,power_law_1.01,0.04431360065937042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,4,power_law_1.01,0.044896000623703004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,4,power_law_1.01,0.04583039879798889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,4,power_law_1.01,0.05181440114974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,4,power_law_1.01,0.05429120063781738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,4,power_law_1.01,0.05589759945869446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,4,power_law_1.01,0.06302080154418946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,4,power_law_1.01,0.07568640112876893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,4,power_law_1.01,0.09438080191612244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,8,power_law_1.2,0.019808000326156615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,8,power_law_1.2,0.020025600492954255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,8,power_law_1.2,0.01992959976196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,8,power_law_1.2,0.019487999379634857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,8,power_law_1.2,0.02014079988002777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,4,power_law_1.01,0.11574399471282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,8,power_law_1.2,0.022259199619293214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,8,power_law_1.2,0.046623998880386354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,8,power_law_1.2,0.046675199270248414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,8,power_law_1.2,0.03697920143604279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,8,power_law_1.2,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,8,power_law_1.2,0.03492479920387268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,8,power_law_1.2,0.04251520037651062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,8,power_law_1.2,0.04142720103263855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,8,power_law_1.2,0.03665919899940491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,8,power_law_1.2,0.04280959963798523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,4,power_law_1.01,0.1640064001083374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,8,power_law_1.2,0.03865599930286408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,8,power_law_1.2,0.04234879910945892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,8,power_law_1.2,0.05514240264892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.2,0.06024320125579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,4,power_law_1.01,0.2031615972518921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.2,0.08178560137748718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,4,power_law_1.01,0.2552704095840454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.2,0.10140160322189332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.2,0.13544319868087767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,4,power_law_1.01,0.35201919078826904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.2,0.1638592004776001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.2,0.34907519817352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.2,0.23520638942718505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,4,power_law_1.01,0.4765120029449463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,1,power_law_1.01,0.05485439896583557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,1,power_law_1.01,0.06567040085792542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,1,power_law_1.01,0.06971520185470581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,1,power_law_1.01,0.09611520171165466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,1,power_law_1.01,0.09360640048980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,1,power_law_1.01,0.09166079759597778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,1,power_law_1.01,0.0971455991268158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,1,power_law_1.01,0.09563519954681396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.2,0.30076799392700193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,1,power_law_1.01,0.0948736011981964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,1,power_law_1.01,0.09305599927902222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,1,power_law_1.01,0.09430400133132935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,1,power_law_1.01,0.0939840018749237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,1,power_law_1.01,0.10247039794921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,1,power_law_1.01,0.1059648036956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,1,power_law_1.01,0.12175999879837036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,1,power_law_1.01,0.1221951961517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,1,power_law_1.01,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,1,power_law_1.01,0.14706560373306274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,1,power_law_1.01,0.17170560359954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,1,power_law_1.01,0.21661438941955566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,1,power_law_1.01,0.23573760986328124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,1,power_law_1.01,0.3208767890930176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,1,power_law_1.01,0.39308159351348876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,1,power_law_1.01,0.5522431850433349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,4,power_law_1.01,0.6297279834747315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,1,power_law_1.01,0.7090879917144776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,1,power_law_1.01,1.015014362335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.2,0.4351679801940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,1,power_law_1.01,1.3276927947998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,1,power_law_1.01,0.09868159890174866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,1,power_law_1.01,0.14974080324172973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,1,power_law_1.01,0.2523967981338501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,1,power_law_1.01,0.40638079643249514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,1,power_law_1.01,0.5366655826568604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,1,power_law_1.01,0.8856127738952637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,1,power_law_1.01,0.9287232398986817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,1,power_law_1.01,0.9675392150878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,1,power_law_1.01,0.9566271781921387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,1,power_law_1.01,0.996787166595459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,1,power_law_1.01,1.0044544219970704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,1,power_law_1.01,1.0380352020263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,1,power_law_1.01,1.062598419189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,1,power_law_1.01,1.1076800346374511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,1,power_law_1.01,1.1280960083007812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,1,power_law_1.01,1.1698368072509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,1,power_law_1.01,1.2117695808410645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,1,power_law_1.01,1.3305088043212892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,1,power_law_1.01,1.3535872459411622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,1,power_law_1.01,1.5204095840454102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,1,power_law_1.01,1.5277952194213866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,1,power_law_1.01,1.8120704650878907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,1,power_law_1.01,2.1415168762207033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,1,power_law_1.01,2.6296512603759767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,1,power_law_1.01,3.3680255889892576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,1,power_law_1.01,4.327289581298828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,1,power_law_1.01,5.475948715209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,2,power_law_1.2,0.10384000539779663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,2,power_law_1.2,0.13921279907226564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,2,power_law_1.2,0.14965120553970337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,2,power_law_1.2,0.2179647922515869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,2,power_law_1.2,0.214630389213562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,2,power_law_1.2,0.2091968059539795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,2,power_law_1.2,0.21765120029449464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,2,power_law_1.2,0.22814719676971434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,2,power_law_1.2,0.23350400924682618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,2,power_law_1.2,0.24111359119415282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,2,power_law_1.2,0.24651520252227782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,2,power_law_1.2,0.26330881118774413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,2,power_law_1.2,0.2564352035522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,2,power_law_1.2,0.2579967975616455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,2,power_law_1.2,0.28752639293670657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,2,power_law_1.2,0.32487680912017824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,2,power_law_1.2,0.3236736059188843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.2,0.5723648071289062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,2,power_law_1.2,0.4079552173614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,2,power_law_1.2,0.44635519981384275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,2,power_law_1.2,0.6074944019317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,2,power_law_1.2,0.7279744148254395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,2,power_law_1.2,1.0482815742492675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,2,power_law_1.2,1.315340805053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,2,power_law_1.2,1.9482240676879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,8,power_law_1.2,0.058380800485610965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,8,power_law_1.2,0.06316800117492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,8,power_law_1.2,0.06981120109558106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,8,power_law_1.2,0.07153919935226441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,8,power_law_1.2,0.07978879809379577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,8,power_law_1.2,0.0870464026927948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,8,power_law_1.2,0.08420479893684388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,8,power_law_1.2,0.08584960103034973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,8,power_law_1.2,0.08524799942970276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,8,power_law_1.2,0.08609920144081115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,8,power_law_1.2,0.08410239815711976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,8,power_law_1.2,0.08751999735832214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,8,power_law_1.2,0.08972160220146179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,8,power_law_1.2,0.09121919870376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,8,power_law_1.2,0.1018623948097229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,8,power_law_1.2,0.10915839672088623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,8,power_law_1.2,0.11611520051956177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,8,power_law_1.2,0.13509119749069215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,8,power_law_1.2,0.1545024037361145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,8,power_law_1.2,0.18494720458984376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,8,power_law_1.2,0.19382400512695314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,8,power_law_1.2,0.253657603263855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,8,power_law_1.2,0.33060479164123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,8,power_law_1.2,0.4849408149719238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,8,power_law_1.2,0.6328896045684814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,2,power_law_1.2,2.1200511932373045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,8,power_law_1.2,0.8723263740539551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,8,power_law_1.2,1.0916864395141601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,2,power_law_1.2,3.4266815185546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,32,balanced,0.07468800246715546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,32,balanced,0.07499200105667114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,32,balanced,0.07580266892910004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,32,balanced,0.09101866682370503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,32,balanced,0.1123413344224294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,32,balanced,0.18013334274291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,32,balanced,0.1851466695467631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,32,balanced,0.19845332702000937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,32,balanced,0.16371200482050577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,32,balanced,0.1718133290608724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,32,balanced,0.18395199378331503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,32,balanced,0.17018133401870728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,32,balanced,0.17495999733606973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,32,balanced,0.1650986671447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,32,balanced,0.18459200859069824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,32,balanced,0.1793173352877299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,32,balanced,0.18444265921910605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,32,balanced,0.18330667416254678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,32,balanced,0.18554667631785074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,32,balanced,0.19514666001001993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,32,balanced,0.20374933878580728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,32,balanced,0.22266666094462076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,32,balanced,0.24451732635498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,32,balanced,0.34673066933949787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,32,balanced,0.373088002204895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,32,balanced,0.5110453367233276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,32,balanced,0.6313759883244833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,2,power_law_1.2,4.372051239013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,1,power_law_1.2,0.05434240102767944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,1,power_law_1.2,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,1,power_law_1.2,0.07406719923019409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,1,power_law_1.2,0.09958400130271912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,1,power_law_1.2,0.12515840530395508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,1,power_law_1.2,0.1560320019721985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,1,power_law_1.2,0.20571520328521728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,1,power_law_1.2,0.215500807762146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,1,power_law_1.2,0.22250239849090575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,1,power_law_1.2,0.22846720218658448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,1,power_law_1.2,0.23555200099945067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,1,power_law_1.2,0.24413440227508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,1,power_law_1.2,0.2510080099105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,1,power_law_1.2,0.2655168056488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,1,power_law_1.2,0.27603199481964114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,1,power_law_1.2,0.28513278961181643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,1,power_law_1.2,0.3063807964324951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,1,power_law_1.2,0.36172800064086913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,1,power_law_1.2,0.39699199199676516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,1,power_law_1.2,0.48200321197509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,1,power_law_1.2,0.5422207832336425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,1,power_law_1.2,0.690880012512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,1,power_law_1.2,0.8500096321105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,1,power_law_1.2,1.137299156188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,1,power_law_1.2,1.4463680267333985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,1,power_law_1.2,1.99354248046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,1,power_law_1.2,2.5189056396484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,4,power_law_1.2,0.018355199694633485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,4,power_law_1.2,0.02051199972629547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,4,power_law_1.2,0.022412799298763275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,4,power_law_1.2,0.02584959864616394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,4,power_law_1.2,0.028352001309394838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,4,power_law_1.2,0.03000960052013397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,4,power_law_1.2,0.03463680148124695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,4,power_law_1.2,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,4,power_law_1.2,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,4,power_law_1.2,0.03935999870300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,4,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,4,power_law_1.2,0.04376319944858551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,4,power_law_1.2,0.046054399013519286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,4,power_law_1.2,0.045100799202919005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,4,power_law_1.2,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,4,power_law_1.2,0.0515392005443573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,4,power_law_1.2,0.05472000241279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,4,power_law_1.2,0.08112000226974488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,4,power_law_1.2,0.09148799777030944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,4,power_law_1.2,0.09835519790649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,4,power_law_1.2,0.10164480209350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,4,power_law_1.2,0.11219199895858764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,4,power_law_1.2,0.12412159442901612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,4,power_law_1.2,0.1570623993873596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,4,power_law_1.2,0.1904703974723816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,4,power_law_1.2,0.26109440326690675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,4,power_law_1.2,0.3177664041519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,4,power_law_1.01,0.08674560189247131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,4,power_law_1.01,0.09730560183525086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,4,power_law_1.01,0.09640960097312927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,4,power_law_1.01,0.11719679832458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,4,power_law_1.01,0.13653759956359862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,4,power_law_1.01,0.14949120283126832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,4,power_law_1.01,0.15081599950790406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,4,power_law_1.01,0.1503551959991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,4,power_law_1.01,0.1528831958770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,4,power_law_1.01,0.15128320455551147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,4,power_law_1.01,0.15590399503707886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,4,power_law_1.01,0.15633920431137086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,4,power_law_1.01,0.15735039710998536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,4,power_law_1.01,0.16390399932861327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,4,power_law_1.01,0.17227519750595094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,4,power_law_1.01,0.17431679964065552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,4,power_law_1.01,0.18337279558181763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,4,power_law_1.01,0.20736639499664306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,4,power_law_1.01,0.22254719734191894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,4,power_law_1.01,0.2680255889892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,4,power_law_1.01,0.3040127992630005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,4,power_law_1.01,0.4374591827392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,4,power_law_1.01,0.5169343948364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,4,power_law_1.01,0.7463808059692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,4,power_law_1.01,0.9306303977966308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,4,power_law_1.01,1.3742079734802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,4,power_law_1.01,1.7570367813110352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,8,balanced,0.05808533231417338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,8,balanced,0.06202666461467743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,8,balanced,0.06840000053246816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,8,balanced,0.09645866354306538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,8,balanced,0.11902933319409688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,8,balanced,0.13937066992123923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,8,balanced,0.14071466525395712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,8,balanced,0.1409333348274231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,8,balanced,0.14316800236701965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,8,balanced,0.14352533221244812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,8,balanced,0.14615466197331747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,8,balanced,0.14832533399264017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,8,balanced,0.15034133195877075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,8,balanced,0.151936004559199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,8,balanced,0.15988799929618835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,8,balanced,0.16268799702326456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,8,balanced,0.16939733425776163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,8,balanced,0.18966933091481528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,8,balanced,0.2053119937578837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,8,balanced,0.29334932565689087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,8,balanced,0.28515734275182086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,8,balanced,0.39510401089986164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,8,balanced,0.41567468643188477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,8,balanced,0.5859893163045248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,8,balanced,0.6755626996358236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,8,balanced,0.9791893164316813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,8,balanced,1.2104159990946453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,256,balanced,0.07196799914042155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,256,balanced,0.07037333150704701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,256,balanced,0.07167466481526692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,256,balanced,0.07512533167997996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,256,balanced,0.07361066838105519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,256,balanced,0.07679466903209686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,256,balanced,0.07807999849319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,256,balanced,0.07696000238259633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,256,balanced,0.07713066538174947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,256,balanced,0.07799999912579854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,256,balanced,0.07872533301512401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,256,balanced,0.08085333307584126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,256,balanced,0.08179733157157898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,256,balanced,0.08016000191370647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,256,balanced,0.08430400490760803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,256,balanced,0.08482666810353597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,256,balanced,0.08521599570910136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,256,balanced,0.09040533502896626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,256,balanced,0.09517866373062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,256,balanced,0.1032319962978363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,256,balanced,0.11421333750089009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,256,balanced,0.1360106666882833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,256,balanced,0.15657599767049155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,256,balanced,0.19697600603103638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,256,balanced,0.23525333404541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,256,balanced,0.31362666686375934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,256,balanced,0.39213867982228595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,4,power_law_1.2,0.060166400671005246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,4,power_law_1.2,0.06741759777069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,4,power_law_1.2,0.07514240145683289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,4,power_law_1.2,0.08458880186080933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,4,power_law_1.2,0.09038720130920411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,4,power_law_1.2,0.096288001537323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,4,power_law_1.2,0.10620800256729127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,4,power_law_1.2,0.10620800256729127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,4,power_law_1.2,0.10695680379867553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,4,power_law_1.2,0.10885759592056274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,4,power_law_1.2,0.10938880443572999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,4,power_law_1.2,0.11288319826126099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,4,power_law_1.2,0.11390080451965331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,4,power_law_1.2,0.11779199838638306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,4,power_law_1.2,0.11967999935150146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,4,power_law_1.2,0.12284799814224243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,4,power_law_1.2,0.12904959917068481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,4,power_law_1.2,0.14183679819107056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,4,power_law_1.2,0.15475200414657592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,4,power_law_1.2,0.18559999465942384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,4,power_law_1.2,0.21583359241485595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,4,power_law_1.2,0.2798975944519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,4,power_law_1.2,0.3499135971069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,4,power_law_1.2,0.47555198669433596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,4,power_law_1.2,0.6363903999328613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,4,power_law_1.2,0.909984016418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,4,power_law_1.2,1.168678379058838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,4,balanced,0.07043733199437459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,4,balanced,0.07242133220036824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,4,balanced,0.0766186664501826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,4,balanced,0.0906880001227061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,4,balanced,0.10670933127403259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,4,balanced,0.13396799564361572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,4,balanced,0.13340266545613608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,4,balanced,0.13329600294431052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,4,balanced,0.13740266362826029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,4,balanced,0.13571733236312866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,4,balanced,0.13748799761136374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,4,balanced,0.14058132966359457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,4,balanced,0.14214932918548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,4,balanced,0.1467519998550415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,4,balanced,0.1532426675160726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,4,balanced,0.1541866660118103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,4,balanced,0.15894933541615805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,4,balanced,0.17482666174570718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,4,balanced,0.18589866161346436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,4,balanced,0.21787200371424356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,4,balanced,0.2513493299484253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,4,balanced,0.3208213249842326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,4,balanced,0.39130667845408124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,4,balanced,0.5706773201624552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,4,balanced,0.7082293033599854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,4,balanced,1.0243093172709148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,4,balanced,1.342576026916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,16,power_law_1.2,0.060729598999023436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,16,power_law_1.2,0.06896640062332153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,16,power_law_1.2,0.07518079876899719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,16,power_law_1.2,0.08056319952011108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,16,power_law_1.2,0.08182399868965148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,16,power_law_1.2,0.0840448021888733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,16,power_law_1.2,0.07441920042037964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,16,power_law_1.2,0.08247039914131164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,16,power_law_1.2,0.08310400247573853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,8,power_law_1.2,0.0949567973613739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,16,power_law_1.2,0.0837119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,8,power_law_1.2,0.1453760027885437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,16,power_law_1.2,0.08223360180854797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,8,power_law_1.2,0.19044480323791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,16,power_law_1.2,0.08449280261993408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,8,power_law_1.2,0.22024960517883302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,16,power_law_1.2,0.08689280152320862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,8,power_law_1.2,0.28107519149780275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,16,power_law_1.2,0.08734719753265381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,8,power_law_1.2,0.3025023937225342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,16,power_law_1.2,0.09808639883995056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,8,power_law_1.2,0.2855488061904907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,16,power_law_1.2,0.10195200443267823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,8,power_law_1.2,0.3261120080947876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,8,power_law_1.2,0.2876607894897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,8,power_law_1.2,0.2966207981109619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,8,power_law_1.2,0.2655488014221191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,8,power_law_1.2,0.29789440631866454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,8,power_law_1.2,0.28480000495910646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,8,power_law_1.2,0.2835200071334839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,8,power_law_1.2,0.34269440174102783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,8,power_law_1.2,0.33797760009765626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,16,power_law_1.2,0.10625280141830444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,8,power_law_1.2,0.35891199111938477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,8,power_law_1.2,0.42175998687744143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,8,power_law_1.2,0.42709760665893554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,8,power_law_1.2,0.5193600177764892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,8,power_law_1.2,0.5481279850006103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,8,power_law_1.2,0.6704576015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,16,power_law_1.2,0.1189695954322815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,8,power_law_1.2,0.7991424083709717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,16,power_law_1.2,0.13294719457626342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,16,power_law_1.2,0.15163520574569703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,16,power_law_1.2,0.18069759607315064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,16,power_law_1.2,0.2354367971420288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,16,power_law_1.2,0.30024960041046145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,8,power_law_1.2,1.017523193359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,16,power_law_1.2,0.4321023941040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,8,power_law_1.2,1.0632960319519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,16,power_law_1.2,0.5718400001525878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,16,power_law_1.2,0.8189760208129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,16,power_law_1.2,1.0836544036865234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,8,power_law_1.2,1.4752256393432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,8,power_law_1.2,1.8201280593872071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,1,power_law_1.01,0.05626239776611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,1,power_law_1.01,0.07344639897346497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,1,power_law_1.01,0.09422720074653626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,1,power_law_1.01,0.1424831986427307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,1,power_law_1.01,0.18922879695892333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,1,power_law_1.01,0.23943679332733153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,1,power_law_1.01,0.3196671962738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,1,power_law_1.01,0.3261823892593384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,1,power_law_1.01,0.33868160247802737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,1,power_law_1.01,0.348140811920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,1,power_law_1.01,0.3596672058105469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,1,power_law_1.01,0.3737407922744751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,1,power_law_1.01,0.38283519744873046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,1,power_law_1.01,0.39403519630432127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,1,power_law_1.01,0.4103231906890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,1,power_law_1.01,0.4287680149078369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,1,power_law_1.01,0.4555776119232178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,1,power_law_1.01,0.5153855800628662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,1,power_law_1.01,0.5501120090484619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,1,power_law_1.01,0.6527359962463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,1,power_law_1.01,0.749235200881958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,1,power_law_1.01,0.9436287879943848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,1,power_law_1.01,1.1119872093200684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,1,power_law_1.01,1.4795519828796386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,1,power_law_1.01,1.8330368041992187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,1,power_law_1.01,2.5382335662841795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,1,power_law_1.01,3.238969421386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,64,balanced,0.02548266698916753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,64,balanced,0.025311999022960663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,64,balanced,0.025429333249727886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,64,balanced,0.02700799951950709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,64,balanced,0.027066667874654133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,64,balanced,0.027327999472618103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,64,balanced,0.02712533374627431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,64,balanced,0.027082666754722595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,64,balanced,0.02741866558790207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,64,balanced,0.027162666122118633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,64,balanced,0.029167999823888142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,64,balanced,0.02945599953333537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,64,balanced,0.029215998947620392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,64,balanced,0.03141333411137263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,64,balanced,0.03526933242877325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,64,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,64,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,64,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,64,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,64,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,64,balanced,0.03941866755485535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,64,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,64,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,64,balanced,0.04882133503754934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,64,balanced,0.052239999175071716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,64,balanced,0.06391466657320659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,64,balanced,0.07314133147398631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,128,power_law_1.01,0.05829120278358459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,128,power_law_1.01,0.05608320236206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,128,power_law_1.01,0.056006401777267456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,128,power_law_1.01,0.05885440111160278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,128,power_law_1.01,0.06044800281524658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,128,power_law_1.01,0.06207360029220581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,128,power_law_1.01,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,128,power_law_1.01,0.07084159851074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,8,balanced,0.06178133189678192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,8,balanced,0.06229866544405619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,8,balanced,0.06849066913127899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,8,balanced,0.08872532844543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,8,balanced,0.11547199885050456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,8,balanced,0.1644000013669332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,8,balanced,0.16458666324615479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,128,power_law_1.01,0.07661439776420594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,128,power_law_1.01,0.07642880082130432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,128,power_law_1.01,0.07169920206069946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,8,balanced,0.1646346648534139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,8,balanced,0.16717867056528726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,8,balanced,0.16540799538294473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,8,balanced,0.16707199811935425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,8,balanced,0.1710453430811564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,8,balanced,0.170522669951121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,8,balanced,0.17307732502619425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,8,balanced,0.18066666523615518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,8,balanced,0.1839039921760559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,8,balanced,0.18689066171646118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,128,power_law_1.01,0.07663360238075256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,128,power_law_1.01,0.08133119940757752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,8,balanced,0.20735466480255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,128,power_law_1.01,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,8,balanced,0.21335999170939127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,8,balanced,0.23431466023127237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,8,balanced,0.25546665986378986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,8,balanced,0.2943413257598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,8,balanced,0.3335946798324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,8,balanced,0.4627946615219116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,8,balanced,0.5463519891103109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,8,balanced,0.7726666927337646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,128,power_law_1.01,0.09140480160713196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,8,balanced,1.002346674601237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,128,power_law_1.01,0.09403520226478576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,128,power_law_1.01,0.09809280037879944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,128,power_law_1.01,0.11047680377960205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,128,power_law_1.01,0.11995520591735839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,128,power_law_1.01,0.14237439632415771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,128,power_law_1.01,0.1634752035140991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,128,power_law_1.01,0.22531840801239014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,128,power_law_1.01,0.2611327886581421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,128,power_law_1.01,0.3699903964996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,128,power_law_1.01,0.4812032222747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,128,power_law_1.01,0.6984064102172851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,128,power_law_1.01,0.9288640022277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,2,balanced,0.08488532900810242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,2,balanced,0.13275200128555298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,2,balanced,0.21692800521850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,2,balanced,0.21945599714914957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,2,balanced,0.2128480076789856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,2,balanced,0.21594132979710898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,2,balanced,0.2173866629600525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,2,balanced,0.21900266408920288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,2,balanced,0.2228213349978129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,2,balanced,0.2217493255933126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,2,balanced,0.22510933876037598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,2,balanced,0.2241493264834086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,2,balanced,0.2251840035120646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,2,balanced,0.2497546672821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,2,balanced,0.2555359999338786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,2,balanced,0.2579626639684041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,2,balanced,0.2701546748479207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,2,balanced,0.32173866033554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,2,balanced,0.33367466926574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,2,balanced,0.42261866728464764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,2,balanced,0.5069546699523926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,2,balanced,0.6909759839375814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,2,balanced,0.8968853155771891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,8,power_law_1.2,0.044198399782180785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,8,power_law_1.2,0.04529919922351837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,8,power_law_1.2,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,8,power_law_1.2,0.04469119906425476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,2,balanced,1.2678879896799724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,8,power_law_1.2,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,2,balanced,1.6884907086690266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,8,power_law_1.2,0.04648320078849792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,8,power_law_1.2,0.047654399275779726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,8,power_law_1.2,0.04759039878845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,2,balanced,2.465712070465088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,8,power_law_1.2,0.055251199007034305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,8,power_law_1.2,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,2,balanced,3.2428054809570312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,8,power_law_1.2,0.05927039980888367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,8,power_law_1.2,0.06346880197525025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,8,power_law_1.2,0.0656000018119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,8,power_law_1.2,0.07107840180397033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,8,power_law_1.2,0.08119040131568908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,8,power_law_1.2,0.08538240194320679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,8,power_law_1.2,0.0896127998828888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,8,power_law_1.2,0.11183359622955322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.2,0.12671999931335448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.2,0.16032639741897584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.2,0.18153599500656128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.2,0.2642496109008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.2,0.3364288091659546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.2,0.46014719009399413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.2,0.6280831813812255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,8,power_law_1.01,0.06815360188484192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,8,power_law_1.01,0.08620160222053527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,8,power_law_1.01,0.08975359797477722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,8,power_law_1.2,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,8,power_law_1.01,0.08943359851837158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,8,power_law_1.2,0.037452799081802365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,8,power_law_1.01,0.09789440035820007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,8,power_law_1.2,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,8,power_law_1.01,0.10092159509658813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,8,power_law_1.2,0.044121599197387694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,8,power_law_1.01,0.10302720069885254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,8,power_law_1.01,0.10427520275115967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,8,power_law_1.01,0.10549759864807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,8,power_law_1.01,0.10323840379714966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,8,power_law_1.01,0.10691839456558228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,8,power_law_1.2,0.042956799268722534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,8,power_law_1.01,0.10879360437393189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,8,power_law_1.2,0.046105599403381346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.2,0.956390380859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,8,power_law_1.01,0.10711679458618165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,8,power_law_1.2,0.04563199877738953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,8,power_law_1.01,0.11256320476531982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,8,power_law_1.2,0.04578559994697571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,8,power_law_1.01,0.12213759422302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,8,power_law_1.01,0.12348159551620483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,8,power_law_1.01,0.12792320251464845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,8,power_law_1.01,0.13850239515304566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,8,power_law_1.01,0.15016319751739501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,8,power_law_1.01,0.1835584044456482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,8,power_law_1.2,0.04614399969577789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,8,power_law_1.2,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,8,power_law_1.2,0.04776960015296936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,8,power_law_1.2,0.04865280091762543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,8,power_law_1.2,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,8,power_law_1.2,0.05279359817504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,8,power_law_1.01,0.20905599594116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,8,power_law_1.01,0.29134719371795653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,8,power_law_1.01,0.3581439971923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,8,power_law_1.01,0.5073344230651855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,8,power_law_1.01,0.6407680034637451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,8,power_law_1.01,0.9028863906860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,8,power_law_1.01,1.2609087944030761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,8,power_law_1.2,0.057785600423812866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,8,power_law_1.2,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,8,power_law_1.2,0.06566399931907654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,8,power_law_1.2,0.08049280047416688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,8,power_law_1.2,0.08687999844551086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,8,power_law_1.2,0.10712319612503052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,8,power_law_1.2,0.12756479978561402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,8,power_law_1.2,0.16357120275497436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,8,power_law_1.2,0.22279679775238037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,8,power_law_1.2,0.31741440296173096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,8,power_law_1.2,0.42804479598999023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,8,power_law_1.2,0.6306496143341065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,4,balanced,0.04610666632652283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,4,balanced,0.05146666864554087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,4,balanced,0.06505600114663442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,4,balanced,0.09274133046468098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,4,balanced,0.14417599638303122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,4,balanced,0.25127466519673664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,4,balanced,0.2643040021260579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,4,balanced,0.26422399282455444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,4,balanced,0.27213333050409955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,4,balanced,0.28149867057800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,4,balanced,0.2834346691767375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,4,balanced,0.29605333010355633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,4,balanced,0.30878933270772296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,4,balanced,0.3136693239212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,4,balanced,0.33246399958928424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,4,balanced,0.35079999764760333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,4,balanced,0.42469334602355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,4,balanced,0.42422934373219806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,4,balanced,0.42849600315093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,4,balanced,0.45099198818206787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,4,balanced,0.6427040100097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,4,balanced,0.7033226490020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,4,balanced,0.767520030339559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,4,balanced,0.898255983988444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,4,balanced,1.0359413623809814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,4,balanced,1.7112639745076497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,4,balanced,1.9368747075398762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.2,1.2723775863647462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,8,power_law_1.2,0.8406975746154786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,4,balanced,0.027621333797772724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,4,balanced,0.027280000348885853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,4,balanced,0.02515733242034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,4,balanced,0.031221332649389904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,4,balanced,0.03323200096686681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,4,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,4,balanced,0.04607999821503957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,4,balanced,0.03942933430274328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,4,balanced,0.08317333459854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,4,balanced,0.08165333171685536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,4,balanced,0.08162666857242584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,4,balanced,0.084714670976003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,4,balanced,0.07937600215276082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,4,balanced,0.062165334820747375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,4,balanced,0.09317333499590556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,4,balanced,0.07477866609891255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,4,balanced,0.09638933340708415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,4,balanced,0.11929600437482198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,4,balanced,0.14480533202489218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,4,balanced,0.20283200343449911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,4,balanced,0.25810132424036664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,4,balanced,0.3614186843236287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,4,balanced,0.47169065475463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,4,balanced,0.6806293328603109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,4,balanced,0.8878239790598551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,4,balanced,1.3052639961242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,4,balanced,1.7285119692484539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,8,power_law_1.2,0.021663999557495116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,8,power_law_1.2,0.022118400037288665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,8,power_law_1.2,0.02420479953289032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,8,power_law_1.2,0.025868800282478333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,8,power_law_1.2,0.025548800826072693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,8,power_law_1.2,0.02821120023727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,8,power_law_1.2,0.034041601419448855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,8,power_law_1.2,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,8,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,8,power_law_1.2,0.04227199852466583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,8,power_law_1.2,0.042745599150657655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,8,power_law_1.2,0.06435199975967407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,8,power_law_1.2,0.06562560200691223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,8,power_law_1.2,0.06484479904174804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,8,power_law_1.2,0.06856319904327393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,8,power_law_1.2,0.06968960165977478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,8,power_law_1.2,0.06392959952354431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,8,power_law_1.2,0.06773120164871216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.2,0.06878079771995545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.2,0.0804095983505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.2,0.09222400188446045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.2,0.11316479444503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.2,0.12951680421829223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.2,0.16390399932861327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.2,0.22184960842132567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.2,0.31658239364624025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,2,balanced,0.040250666439533234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,2,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,2,balanced,0.06842133402824402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,2,balanced,0.07221866647402446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,2,balanced,0.07429333527882893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,2,balanced,0.07426133255163829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,2,balanced,0.07223999996980031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,2,balanced,0.07424533367156982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,2,balanced,0.07577600081761678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,2,balanced,0.0748586654663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,2,balanced,0.07441066702206929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,2,balanced,0.07656000057856242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,2,balanced,0.07835199932257335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,2,balanced,0.08493866523106892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,2,balanced,0.08823466300964355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,2,balanced,0.09218133489290874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,2,balanced,0.09709333380063374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,2,balanced,0.11129066348075867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,2,balanced,0.11931199828783672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,2,balanced,0.14617600043614706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,2,balanced,0.17223467429478964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,2,balanced,0.22541866699854532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,2,balanced,0.2786986629168193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,2,balanced,0.3782080014546712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,2,balanced,0.4822719891866048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,2,balanced,0.6751093069712321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,2,balanced,0.8830719788869222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.2,0.4131328105926514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,4,power_law_1.01,0.07700480222702026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,4,power_law_1.01,0.09282559752464295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,4,power_law_1.01,0.09931520223617554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,4,power_law_1.01,0.11631360054016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,4,power_law_1.01,0.13087999820709229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,4,power_law_1.01,0.1561728000640869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,4,power_law_1.01,0.16840319633483886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,4,power_law_1.01,0.17162879705429077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,4,power_law_1.01,0.16288000345230103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,4,power_law_1.01,0.17495039701461793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,4,power_law_1.01,0.17933440208435059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,4,power_law_1.01,0.17956479787826538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,4,power_law_1.01,0.184716796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,4,power_law_1.01,0.19061119556427003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,4,power_law_1.01,0.20119040012359618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,4,power_law_1.01,0.20549120903015136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,4,power_law_1.01,0.21515519618988038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,4,power_law_1.01,0.24407680034637452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,4,power_law_1.01,0.2447295904159546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,4,power_law_1.01,0.30494720935821534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,4,power_law_1.01,0.30655360221862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,4,power_law_1.01,0.3906816005706787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,4,power_law_1.01,0.4528639793395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,4,power_law_1.01,0.6228928089141845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,4,power_law_1.01,0.7395840167999268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,4,power_law_1.01,1.0756863594055175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,4,power_law_1.01,1.3879039764404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,2,balanced,0.06353599826494853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,4,power_law_1.01,0.03407360017299652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,2,balanced,0.06610666712125142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,2,balanced,0.07214400172233582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,4,power_law_1.01,0.032953599095344545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,2,balanced,0.08665066957473755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,2,balanced,0.10921600461006165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,4,power_law_1.01,0.03544960021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,2,balanced,0.1609173317750295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,8,power_law_1.01,0.12190079689025879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,2,balanced,0.1620373328526815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,4,power_law_1.01,0.042131200432777405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,2,balanced,0.15918399890263876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,2,balanced,0.15985600153605142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,2,balanced,0.16261333227157593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,2,balanced,0.16300800442695618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,8,power_law_1.01,0.1855039954185486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,2,balanced,0.16771199305852255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,2,balanced,0.17077332735061646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,2,balanced,0.17327467600504556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,2,balanced,0.18039466937383017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,2,balanced,0.1839146614074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,2,balanced,0.19104532400767008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,2,balanced,0.21451733509699503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,2,balanced,0.23166932662328085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,2,balanced,0.2751520077387492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,2,balanced,0.3126186728477478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,2,balanced,0.40695468584696454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,2,balanced,0.49872533480326336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,8,power_law_1.01,0.16513279676437378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,2,balanced,0.7275306383768717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,2,balanced,0.9337333043416342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,8,power_law_1.01,0.19659520387649537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,4,power_law_1.01,0.047366398572921756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,4,power_law_1.01,0.05575680136680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,2,balanced,1.3645973205566406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,4,power_law_1.01,0.057734400033950806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,8,power_law_1.01,0.22525439262390137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,4,power_law_1.01,0.058169597387313844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,2,balanced,1.7670987447102864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,4,power_law_1.01,0.05985919833183288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,8,power_law_1.01,0.22855679988861083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,4,power_law_1.01,0.058329600095748904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,4,power_law_1.01,0.06062080264091492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,8,power_law_1.01,0.22144639492034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,8,power_law_1.01,0.2353343963623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,8,power_law_1.01,0.22985599040985108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,8,power_law_1.01,0.23610880374908447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,4,power_law_1.01,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,4,power_law_1.01,0.06479359865188598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,4,power_law_1.01,0.06722559928894042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,4,power_law_1.01,0.0771776020526886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,4,power_law_1.01,0.08010879755020142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,4,power_law_1.01,0.0790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,4,power_law_1.01,0.09132159948348999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,4,power_law_1.01,0.09797760248184204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,8,power_law_1.01,0.23468799591064454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,4,power_law_1.01,0.11757440567016601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,8,power_law_1.01,0.25306880474090576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,8,power_law_1.01,0.25029759407043456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,8,power_law_1.01,0.2518656015396118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,8,power_law_1.01,0.2723007917404175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,8,power_law_1.01,0.267903995513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,4,power_law_1.01,0.13474559783935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,4,power_law_1.01,0.1769536018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,8,power_law_1.01,0.28764801025390624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,4,power_law_1.01,0.21099519729614258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,8,power_law_1.01,0.3291584014892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,4,power_law_1.01,0.2888767957687378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,8,power_law_1.01,0.3443648099899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,8,power_law_1.01,0.4005568027496338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,8,power_law_1.01,0.47579522132873536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,8,power_law_1.01,0.5906176090240478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,4,power_law_1.01,0.3798655986785889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,8,power_law_1.01,0.5902463912963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,4,power_law_1.01,0.5532032012939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,8,power_law_1.01,0.7012671947479248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,4,power_law_1.01,0.6676991939544678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,8,power_law_1.01,0.9053503990173339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,8,power_law_1.01,1.3598719596862794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,8,power_law_1.01,1.5188544273376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,4,balanced,0.018895999838908512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,4,balanced,0.018992000569899876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,4,balanced,0.018895999838908512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,4,balanced,0.018917333334684372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,4,balanced,0.019088000059127808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,4,balanced,0.01929066702723503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,4,balanced,0.022986667851607006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,4,balanced,0.01953599974513054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,4,balanced,0.02314666658639908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,4,balanced,0.023050665855407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,4,balanced,0.023354666928450268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,4,balanced,0.025397333006064098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,4,balanced,0.025413334369659424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,4,balanced,0.03164800008138021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,4,balanced,0.03341866781314214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,4,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,4,balanced,0.03370666752258936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,4,balanced,0.038912000755469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,4,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,4,balanced,0.05539200206597646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,4,balanced,0.06158933540185293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,4,balanced,0.08084799846013387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,4,balanced,0.09691199660301208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,4,balanced,0.1274666686852773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,4,balanced,0.16065067052841187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,4,balanced,0.22457599639892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,4,balanced,0.28754667441050213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,64,power_law_1.01,0.0216511994600296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,64,power_law_1.01,0.021849599480628968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,64,power_law_1.01,0.02194560021162033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,64,power_law_1.01,0.021907199919223786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,64,power_law_1.01,0.024646399915218352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,64,power_law_1.01,0.03020800054073334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,64,power_law_1.01,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,64,power_law_1.01,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,64,power_law_1.01,0.05073919892311096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,64,power_law_1.01,0.050367999076843264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,64,power_law_1.01,0.05132160186767578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,64,power_law_1.01,0.05151360034942627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,64,power_law_1.01,0.05140479803085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,64,power_law_1.01,0.048793599009513855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,64,power_law_1.01,0.048076799511909483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,64,power_law_1.01,0.04827519953250885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,64,power_law_1.01,0.051500797271728516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,64,power_law_1.01,0.05255680084228516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,64,power_law_1.01,0.05414400100708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,64,power_law_1.01,0.07712640166282654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,64,power_law_1.01,0.08440319895744323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,64,power_law_1.01,0.11240320205688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,64,power_law_1.01,0.13397120237350463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,64,power_law_1.01,0.18078080415725709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,64,power_law_1.01,0.23452160358428956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,64,power_law_1.01,0.33886721134185793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,64,power_law_1.01,0.4468416213989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,1,4,power_law_1.01,0.08794879913330078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,1,4,power_law_1.01,0.10522880554199218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,1,4,power_law_1.01,0.13233280181884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,1,4,power_law_1.01,0.13571200370788575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,1,4,power_law_1.01,0.1355520009994507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,1,4,power_law_1.01,0.1363711953163147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,1,4,power_law_1.01,0.13716479539871215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,1,4,power_law_1.01,0.1389248013496399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,1,4,power_law_1.01,0.1393728017807007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,1,4,power_law_1.01,0.14135040044784547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,1,4,power_law_1.01,0.1422592043876648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,1,4,power_law_1.01,0.16670080423355102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,1,4,power_law_1.01,0.16830719709396363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,1,4,power_law_1.01,0.17408000230789183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,1,4,power_law_1.01,0.18654719591140748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,1,4,power_law_1.01,0.21347200870513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,1,4,power_law_1.01,0.263372802734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,1,4,power_law_1.01,0.323142409324646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,1,4,power_law_1.01,0.40465922355651857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,1,4,power_law_1.01,0.5107711791992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,1,4,power_law_1.01,0.6499968051910401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,1,4,power_law_1.01,1.118617630004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,1,4,power_law_1.01,1.2114496231079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,2,balanced,0.01947733387351036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,2,balanced,0.021013334393501282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,2,balanced,0.023141334454218548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,2,balanced,0.02756800005833308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,2,balanced,0.043493335445721946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,2,balanced,0.06182933350404104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,2,balanced,0.06295999884605408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,2,balanced,0.0634986658891042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,2,balanced,0.06525333225727081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,2,balanced,0.06542933483918507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,2,balanced,0.0699839989344279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,2,balanced,0.0693280001481374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,2,balanced,0.07238399982452393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,2,balanced,0.07386666536331177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,2,balanced,0.07647466659545898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,2,balanced,0.08055999875068665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,2,balanced,0.08368000388145447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,2,balanced,0.09702933828035991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,2,balanced,0.10108799735705058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,2,balanced,0.1325493355592092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,2,balanced,0.12134933471679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,2,balanced,0.1774666706720988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,2,balanced,0.22609599431355795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,2,balanced,0.23497066895167032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,2,balanced,0.2504319945971171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,2,balanced,0.4379306634267171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,2,balanced,0.4593813419342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,1,4,power_law_1.01,1.7507968902587892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,1,4,power_law_1.01,2.336511993408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,1,4,power_law_1.01,3.3468864440917967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,1,4,power_law_1.01,5.119334411621094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,4,power_law_1.2,0.031711998581886294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,4,power_law_1.2,0.027692800760269164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,4,power_law_1.2,0.026636800169944762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,4,power_law_1.2,0.027871999144554137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,4,power_law_1.2,0.029094401001930236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,4,power_law_1.2,0.029420799016952513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,4,power_law_1.2,0.030163198709487915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,4,power_law_1.2,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,4,power_law_1.2,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,4,power_law_1.2,0.032204800844192506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,4,power_law_1.2,0.03261440098285675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,4,power_law_1.2,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,4,power_law_1.2,0.034841600060462954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,4,power_law_1.2,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,4,power_law_1.2,0.03978880047798157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,4,power_law_1.2,0.04166400134563446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,4,power_law_1.2,0.04793600142002106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,4,power_law_1.2,0.05905280113220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,4,power_law_1.2,0.06154879927635193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,4,power_law_1.2,0.07022719979286193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,16,power_law_1.2,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,4,power_law_1.2,0.07495679855346679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,16,power_law_1.2,0.06536319851875305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,16,power_law_1.2,0.0604095995426178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,16,power_law_1.2,0.06783999800682068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,16,power_law_1.2,0.06792960166931153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,16,power_law_1.2,0.06658560037612915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,16,power_law_1.2,0.06676480174064636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,16,power_law_1.2,0.07036160230636597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,16,power_law_1.2,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,16,power_law_1.2,0.07282559871673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,16,power_law_1.2,0.06883839964866638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,4,power_law_1.2,0.10255359411239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,16,power_law_1.2,0.07184000015258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,16,power_law_1.2,0.07425919771194459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,16,power_law_1.2,0.0784704029560089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,16,power_law_1.2,0.08576639890670776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,16,power_law_1.2,0.08942720293998718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,16,power_law_1.2,0.09528959989547729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,16,power_law_1.2,0.11374080181121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.2,0.1259775996208191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.2,0.15383039712905883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.2,0.1722432017326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.2,0.23464319705963135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.2,0.2590015888214111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,4,power_law_1.2,0.11336959600448608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.2,0.35006721019744874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.2,0.45297918319702146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.2,0.6622528076171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.2,0.7913663864135743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,4,power_law_1.2,0.1850111961364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,8,power_law_1.2,0.0794048011302948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,8,power_law_1.2,0.10055680274963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,8,power_law_1.2,0.09263359904289245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,8,power_law_1.2,0.11773439645767211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,8,power_law_1.2,0.12905600070953369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,8,power_law_1.2,0.12987519502639772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,8,power_law_1.2,0.13068799972534179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,8,power_law_1.2,0.12855679988861085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,8,power_law_1.2,0.1282240033149719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,8,power_law_1.2,0.14357119798660278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,8,power_law_1.2,0.13155839443206788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,8,power_law_1.2,0.14950400590896606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,8,power_law_1.2,0.134278404712677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,8,power_law_1.2,0.13741439580917358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,8,power_law_1.2,0.14714879989624025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,8,power_law_1.2,0.15296000242233276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,8,power_law_1.2,0.15791360139846802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,8,power_law_1.2,0.168614399433136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,4,power_law_1.2,0.2238976001739502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,8,power_law_1.2,0.18241920471191406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,8,power_law_1.2,0.20865919589996337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,8,power_law_1.2,0.2372607946395874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,8,power_law_1.2,0.2991552114486694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,8,power_law_1.2,0.3598975896835327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,8,power_law_1.2,0.47080321311950685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,8,power_law_1.2,0.5638207912445068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,8,power_law_1.2,0.763974380493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,8,power_law_1.2,1.0402048110961915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,256,power_law_1.01,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,256,power_law_1.01,0.06306560039520263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,256,power_law_1.01,0.055199998617172244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,256,power_law_1.01,0.05632640123367309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,256,power_law_1.01,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,256,power_law_1.01,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,256,power_law_1.01,0.05572479963302612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,256,power_law_1.01,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,256,power_law_1.01,0.05795199871063232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,256,power_law_1.01,0.05716480016708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,256,power_law_1.01,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,256,power_law_1.01,0.06176000237464905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,256,power_law_1.01,0.06158080101013184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,256,power_law_1.01,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,256,power_law_1.01,0.06915839910507202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,256,power_law_1.01,0.07003520131111145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,256,power_law_1.01,0.0763584017753601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,256,power_law_1.01,0.08878719806671143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,256,power_law_1.01,0.09887999892234803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,256,power_law_1.01,0.12472319602966309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,256,power_law_1.01,0.15107200145721436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,256,power_law_1.01,0.193996798992157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,4,power_law_1.2,0.3226239919662476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,256,power_law_1.01,0.18773119449615477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,256,power_law_1.01,0.23783679008483888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,256,power_law_1.01,0.3066688060760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,256,power_law_1.01,0.4135615825653076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,256,power_law_1.01,0.5228672027587891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,4,balanced,0.023183998962243397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,4,balanced,0.025263999899228413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,4,balanced,0.029306667546431225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,4,balanced,0.04378133515516917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,4,balanced,0.061797335743904114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,4,balanced,0.0958720048268636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,4,balanced,0.10136533776919048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,4,balanced,0.10502400000890096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,4,balanced,0.10523733496665955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,4,balanced,0.10203733046849568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,4,balanced,0.11008532842000325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,4,balanced,0.1362773378690084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,4,balanced,0.13396267096201578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,4,balanced,0.13664000233014426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,4,balanced,0.20227199792861938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,4,balanced,0.19859200716018677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,4,balanced,0.199727992216746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,4,balanced,0.2469600041707357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,4,balanced,0.2754720052083333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,4,balanced,0.30033600330352783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,4,balanced,0.3166293303171794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,4,balanced,0.3327786723772685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,4,balanced,0.3555413484573364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,4,balanced,0.3646133343378703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,4,balanced,0.3932533264160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,4,balanced,0.6835412979125977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,4,balanced,0.7270613511403402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,4,power_law_1.2,0.3530751943588257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,32,balanced,0.018917333334684372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,32,balanced,0.01886933296918869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,32,balanced,0.018992000569899876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,32,balanced,0.018917333334684372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,32,balanced,0.02093333254257838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,32,balanced,0.021216000119845074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,32,balanced,0.021216000119845074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,32,balanced,0.023130667706330616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,32,balanced,0.024058667321999867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,32,balanced,0.023269332945346832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,32,balanced,0.027104000250498455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,32,balanced,0.02720533311367035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,32,balanced,0.027424000203609467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,32,balanced,0.035605333745479584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,32,balanced,0.04164800047874451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,32,balanced,0.041706666350364685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,32,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,32,balanced,0.04418666660785675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,32,balanced,0.045893331368764244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,32,balanced,0.05717333157857259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,32,balanced,0.0661599983771642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,32,balanced,0.08409066994984944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,32,balanced,0.09900266925493877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,32,balanced,0.13452800114949545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,32,balanced,0.16646933555603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,32,balanced,0.2379680077234904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,32,balanced,0.3039039969444275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,32,power_law_1.01,0.07047039866447449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,32,power_law_1.01,0.08686720132827759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,32,power_law_1.01,0.08234239816665649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,32,power_law_1.01,0.08640639781951905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,32,power_law_1.01,0.0870527982711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,32,power_law_1.01,0.07973759770393371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,32,power_law_1.01,0.08722559809684753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,32,power_law_1.01,0.08780800104141236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,32,power_law_1.01,0.08275200128555298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,32,power_law_1.01,0.0881600022315979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,32,power_law_1.01,0.0915391981601715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,32,power_law_1.01,0.09262080192565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,32,power_law_1.01,0.09507200121879578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,32,power_law_1.01,0.09816319942474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,32,power_law_1.01,0.10001920461654663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,32,power_law_1.01,0.1062656044960022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,32,power_law_1.01,0.11431039571762085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,32,power_law_1.01,0.1278656005859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,32,power_law_1.01,0.14158079624176026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,32,power_law_1.01,0.17703039646148683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,32,power_law_1.01,0.189900803565979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,32,power_law_1.01,0.25180160999298096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,32,power_law_1.01,0.30304000377655027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,32,power_law_1.01,0.4092095851898193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,32,power_law_1.01,0.5034495830535889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,32,power_law_1.01,0.7242623805999756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,32,power_law_1.01,0.9242815971374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,4,balanced,0.023317334552605946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,4,balanced,0.023178666830062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,4,balanced,0.025221332907676697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,4,balanced,0.027722666660944622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,4,balanced,0.033674667278925575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,4,balanced,0.03477866699298223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,4,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,4,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,4,balanced,0.03507733345031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,4,balanced,0.03364266703526179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,4,balanced,0.035546667873859406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,4,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,4,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,4,balanced,0.04068800061941147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,4,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,4,balanced,0.04435733457406362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,4,balanced,0.05034666756788889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,4,balanced,0.05172266562779745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,4,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,4,balanced,0.07387733459472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,4,balanced,0.07845866680145264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,4,balanced,0.10098666946093242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,4,balanced,0.11959466338157654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,4,balanced,0.15404799580574036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,4,balanced,0.18524799744288126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,4,balanced,0.260970671971639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,4,balanced,0.3302453358968099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,16,power_law_1.01,0.11181440353393554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,16,power_law_1.01,0.15799039602279663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,16,power_law_1.01,0.15549440383911134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,16,power_law_1.01,0.15898239612579346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,16,power_law_1.01,0.15926400423049927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,16,power_law_1.01,0.1313599944114685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,16,power_law_1.01,0.1566912055015564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,16,power_law_1.01,0.15936000347137452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,16,power_law_1.01,0.15119359493255616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,16,power_law_1.01,0.1508031964302063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,16,power_law_1.01,0.15614720582962036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,16,power_law_1.01,0.1599679946899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,16,power_law_1.01,0.1636288046836853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,16,power_law_1.01,0.163372802734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,16,power_law_1.01,0.17876479625701905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,16,power_law_1.01,0.18750079870223998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,16,power_law_1.01,0.19071359634399415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,16,power_law_1.01,0.2024319887161255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,16,power_law_1.01,0.208787202835083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,16,power_law_1.01,0.2286911964416504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,16,power_law_1.01,0.25813119411468505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,16,power_law_1.01,0.3365632057189941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,16,power_law_1.01,0.40103678703308104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,16,power_law_1.01,0.5399104118347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,16,power_law_1.01,0.7133632183074952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,16,power_law_1.01,1.0453696250915527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,16,power_law_1.01,1.2723135948181152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,64,power_law_1.2,0.04311679899692535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,64,power_law_1.2,0.04193919897079468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,64,power_law_1.2,0.04268159866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,64,power_law_1.2,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,64,power_law_1.2,0.04284160137176514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,64,power_law_1.2,0.04520959854125976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,64,power_law_1.2,0.046137601137161255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,64,power_law_1.2,0.04442239999771118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,64,power_law_1.2,0.045721599459648134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,64,power_law_1.2,0.04568960070610047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,64,power_law_1.2,0.045721599459648134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,64,power_law_1.2,0.04657280147075653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,64,power_law_1.2,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,64,power_law_1.2,0.04720639884471893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,64,power_law_1.2,0.052742397785186766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,64,power_law_1.2,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,64,power_law_1.2,0.05699840188026428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,64,power_law_1.2,0.06414719820022582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.2,0.07468799948692321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.2,0.07992320060729981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.2,0.10383360385894776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,32,power_law_1.2,0.15568000078201294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,32,power_law_1.2,0.15334399938583373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.2,0.14349440336227418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,32,power_law_1.2,0.15182080268859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,32,power_law_1.2,0.1537279963493347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,32,power_law_1.2,0.13628799915313722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,32,power_law_1.2,0.14290560483932496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,32,power_law_1.2,0.14761600494384766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,32,power_law_1.2,0.14821759462356568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,32,power_law_1.2,0.1474560022354126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,32,power_law_1.2,0.15088000297546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,32,power_law_1.2,0.15543680191040038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,32,power_law_1.2,0.1573632001876831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,32,power_law_1.2,0.15603840351104736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,32,power_law_1.2,0.16799360513687134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,32,power_law_1.2,0.17701760530471802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,32,power_law_1.2,0.18480000495910645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,32,power_law_1.2,0.19863040447235109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,32,power_law_1.2,0.24170238971710206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,32,power_law_1.2,0.24293119907379152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.2,0.17074559926986693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,32,power_law_1.2,0.3041856050491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,32,power_law_1.2,0.37249279022216797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,32,power_law_1.2,0.4346367835998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,4,power_law_1.01,0.03271040022373199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,4,power_law_1.01,0.031200000643730165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,4,power_law_1.01,0.029900801181793214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,4,power_law_1.01,0.0320576012134552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,4,power_law_1.01,0.03210879862308502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,4,power_law_1.01,0.03240959942340851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,4,power_law_1.01,0.03361279964447021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,4,power_law_1.01,0.03278720080852508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,4,power_law_1.01,0.033228799700737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,32,power_law_1.2,0.421292781829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,4,power_law_1.01,0.03553279936313629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,4,power_law_1.01,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,32,power_law_1.2,0.606169605255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,4,power_law_1.01,0.03688960075378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,32,power_law_1.2,0.7221824169158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,4,power_law_1.01,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,4,power_law_1.01,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,4,power_law_1.01,0.0488319993019104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,4,power_law_1.01,0.050918400287628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.2,0.2231935977935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,4,power_law_1.01,0.0511680006980896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,32,power_law_1.2,0.9726271629333496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,4,power_law_1.01,0.05985280275344849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,32,power_law_1.2,1.4034367561340333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,4,power_law_1.01,0.06679040193557739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,4,power_law_1.01,0.08283519744873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,4,power_law_1.01,0.09168000221252441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.2,0.28373119831085203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,4,power_law_1.01,0.10242559909820556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,4,power_law_1.01,0.13325439691543578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,4,power_law_1.01,0.18819199800491332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.2,0.41077117919921874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,8,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,8,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,8,balanced,0.035546667873859406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,8,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,8,balanced,0.03621333340803782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,8,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,8,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,8,balanced,0.035802667339642845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,8,balanced,0.038618666430314384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,8,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,8,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,8,balanced,0.03757333258787791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,8,balanced,0.037733333806196846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,8,balanced,0.04161066561937332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,8,balanced,0.05013866722583771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,8,balanced,0.04974933465321859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,8,balanced,0.04660266637802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,8,balanced,0.06268799801667531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,8,balanced,0.05972266693909963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,8,balanced,0.07502933343251546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,8,balanced,0.07567999760309856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,8,balanced,0.0986293355623881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,8,balanced,0.10713600118954976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,8,balanced,0.13609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,8,balanced,0.16340266664822897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,8,balanced,0.21733866135279337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,8,balanced,0.27663999795913696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,4,power_law_1.01,0.2227008104324341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.2,0.5759615898132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,4,power_law_1.01,0.33598079681396487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,4,power_law_1.01,0.4151103973388672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,1,8,power_law_1.01,0.08791679739952088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,1,8,power_law_1.01,0.08846719861030579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,1,8,power_law_1.01,0.08648319840431214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,1,8,power_law_1.01,0.09032319784164429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,1,8,power_law_1.01,0.09052799940109253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,1,8,power_law_1.01,0.09357439875602722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,1,8,power_law_1.01,0.09264000058174134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,1,8,power_law_1.01,0.09378560185432434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,1,8,power_law_1.01,0.09604480266571044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,1,8,power_law_1.01,0.09564800262451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,1,8,power_law_1.01,0.0974399983882904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,1,8,power_law_1.01,0.11432960033416747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,1,8,power_law_1.01,0.12152320146560669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,1,8,power_law_1.01,0.14041600227355958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,1,8,power_law_1.01,0.16792320013046264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,1,8,power_law_1.01,0.1476032018661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,1,8,power_law_1.01,0.19234559535980225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,1,8,power_law_1.01,0.2733504056930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,1,8,power_law_1.01,0.3270591974258423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,2,power_law_1.01,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,2,power_law_1.01,0.07279360294342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,1,8,power_law_1.01,0.4262400150299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,1,8,power_law_1.01,0.5811840057373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,2,power_law_1.01,0.09932799935340882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,2,power_law_1.01,0.12291840314865113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,2,power_law_1.01,0.16447360515594484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,2,power_law_1.01,0.2020672082901001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,2,power_law_1.01,0.2235647916793823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,1,8,power_law_1.01,0.7777599811553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,2,power_law_1.01,0.2784064054489136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,2,power_law_1.01,0.27777919769287107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,2,power_law_1.01,0.2881472110748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,2,power_law_1.01,0.2980736017227173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,2,power_law_1.01,0.32181758880615235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,2,power_law_1.01,0.3021696090698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,1,8,power_law_1.01,1.081491184234619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,2,power_law_1.01,0.3081599950790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,2,power_law_1.01,0.33857920169830324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,2,power_law_1.01,0.3416640043258667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,2,power_law_1.01,0.36277759075164795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,2,power_law_1.01,0.41283202171325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,2,power_law_1.01,0.44033279418945315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,2,power_law_1.01,0.5115583896636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,2,power_law_1.01,0.5589695930480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,2,power_law_1.01,0.6916543960571289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,2,power_law_1.01,0.825376033782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,2,power_law_1.01,1.0764287948608398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,2,power_law_1.01,1.2909760475158691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,2,power_law_1.01,1.6936128616333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,2,power_law_1.01,2.169657516479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,1,8,power_law_1.01,1.7395519256591796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,1,8,power_law_1.01,2.2356288909912108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,1,8,power_law_1.01,3.220275115966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,32,power_law_1.01,0.07708160281181335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,32,power_law_1.01,0.07415680289268493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,32,power_law_1.01,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,32,power_law_1.01,0.07671040296554565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,32,power_law_1.01,0.07676159739494323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,32,power_law_1.01,0.07533439993858337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,32,power_law_1.01,0.07985919713973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,32,power_law_1.01,0.08042240142822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,32,power_law_1.01,0.08065279722213745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,32,power_law_1.01,0.08089600205421447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,32,power_law_1.01,0.08144639730453491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,32,power_law_1.01,0.08435840010643006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,32,power_law_1.01,0.08329600095748901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,32,power_law_1.01,0.08606079816818238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,32,power_law_1.01,0.09875199794769288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,32,power_law_1.01,0.09957119822502136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,32,power_law_1.01,0.09964159727096558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,32,power_law_1.01,0.11357439756393432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,32,power_law_1.01,0.12165759801864624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,32,power_law_1.01,0.14049919843673705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,32,power_law_1.01,0.16654720306396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,32,power_law_1.01,0.20695040225982667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,32,power_law_1.01,0.244268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,32,power_law_1.01,0.35861759185791015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,32,power_law_1.01,0.49185919761657715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,8,power_law_1.2,0.0324864000082016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,32,power_law_1.01,0.6719552040100097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,8,power_law_1.2,0.0320576012134552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,8,power_law_1.2,0.03455359935760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,8,power_law_1.2,0.04242559969425201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,32,power_law_1.01,0.9010432243347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,8,power_law_1.2,0.04101119935512543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,8,power_law_1.2,0.04750719964504242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,8,power_law_1.2,0.049004799127578734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,8,power_law_1.2,0.05121279954910278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,8,power_law_1.2,0.05181440114974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,8,power_law_1.2,0.053299200534820554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,8,power_law_1.2,0.0511680006980896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,8,power_law_1.2,0.0536575973033905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,8,power_law_1.2,0.0573311984539032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,8,power_law_1.2,0.057555198669433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,8,power_law_1.2,0.06334720253944397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,8,power_law_1.2,0.06510720252990723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,8,power_law_1.2,0.0680895984172821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,8,power_law_1.2,0.07622399926185608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,8,power_law_1.2,0.08885759711265565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,8,power_law_1.2,0.10175360441207885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,8,power_law_1.2,0.10470399856567383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,8,power_law_1.2,0.13382400274276735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,8,power_law_1.2,0.15884799957275392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,8,power_law_1.2,0.22037119865417482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,8,power_law_1.2,0.25502719879150393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,1,8,power_law_1.01,4.352454376220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,8,power_law_1.2,0.3495296001434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,8,power_law_1.2,0.4571839809417725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,4,power_law_1.01,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,4,power_law_1.01,0.0471231997013092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,4,power_law_1.01,0.04640640020370483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,4,power_law_1.01,0.04811519980430603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,4,power_law_1.01,0.047654399275779726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,4,power_law_1.01,0.05072640180587769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,4,power_law_1.01,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,4,power_law_1.01,0.05398399829864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,4,power_law_1.01,0.05506560206413269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,4,power_law_1.01,0.055769598484039305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,4,power_law_1.01,0.057817602157592775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,4,power_law_1.01,0.058208000659942624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,4,power_law_1.01,0.06302080154418946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,4,power_law_1.01,0.06354560256004334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,4,power_law_1.01,0.07141759991645813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,4,power_law_1.01,0.07162879705429077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,4,power_law_1.01,0.07444480061531067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,4,power_law_1.01,0.0892799973487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,4,power_law_1.01,0.11250560283660889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,4,power_law_1.01,0.1308608055114746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,4,power_law_1.01,0.1363711953163147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,4,power_law_1.01,0.1724544048309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,4,power_law_1.01,0.2310336112976074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,4,power_law_1.01,0.30639359951019285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,4,power_law_1.01,0.40748162269592286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,4,power_law_1.01,0.5544256210327149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,4,power_law_1.01,0.6888192176818848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,1,balanced,0.14435733358065286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,1,balanced,0.15065067013104758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,1,balanced,0.15917866428693137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,1,balanced,0.17685333887736002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,1,balanced,0.21030932664871216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,1,balanced,0.2849760055541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,1,balanced,0.28913066784540814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,1,balanced,0.2882080078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,1,balanced,0.29362666606903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,1,balanced,0.2919680078824361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,1,balanced,0.2974133292833964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,1,balanced,0.30427199602127075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,1,balanced,0.3052159945170085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,1,balanced,0.3124693234761556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,1,balanced,0.31093867619832355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,1,balanced,0.3166240056355794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,1,balanced,0.32637866338094074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,1,balanced,0.3556693394978841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,1,balanced,0.3815786838531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,1,balanced,0.44280000527699787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,1,balanced,0.5035413503646851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,4,power_law_1.2,0.10190720558166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,4,power_law_1.2,0.1404031991958618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,4,power_law_1.2,0.14758399724960328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,4,power_law_1.2,0.1915071964263916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,4,power_law_1.2,0.2136768102645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,4,power_law_1.2,0.2130431890487671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,4,power_law_1.2,0.30305280685424807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,4,power_law_1.2,0.2950272083282471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,4,power_law_1.2,0.3036736011505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,4,power_law_1.2,0.2831295967102051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,4,power_law_1.2,0.2906431913375854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,4,power_law_1.2,0.3028671979904175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,4,power_law_1.2,0.2968640089035034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,4,power_law_1.2,0.29294719696044924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,4,power_law_1.2,0.31570560932159425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,4,power_law_1.2,0.3205631971359253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,4,power_law_1.2,0.3392960071563721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,4,power_law_1.2,0.3616640090942383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.2,0.384883189201355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.2,0.44693760871887206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.2,0.5006464004516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.2,0.602348804473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.2,0.6947968006134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.2,0.9456128120422364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.2,1.128870391845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.2,1.5268927574157716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.2,1.9556671142578126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,16,balanced,0.04576000074545542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,16,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,16,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,16,balanced,0.04595733185609182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,16,balanced,0.05320533116658529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,16,balanced,0.07266666491826375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,16,balanced,0.07428800066312154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,16,balanced,0.07524799803892772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,16,balanced,0.07632000247637431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,16,balanced,0.07700799902280171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,16,balanced,0.07701866825421651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,16,balanced,0.08069866895675659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,16,balanced,0.08005333443482716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,16,balanced,0.08244266609350841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,16,balanced,0.08933867017428081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,16,balanced,0.08915199836095174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,16,balanced,0.09481066465377808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,16,balanced,0.10492799679438274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,16,balanced,0.11078932881355286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,16,balanced,0.12851199507713318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,16,balanced,0.14632532993952432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,16,balanced,0.18653333187103271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,16,balanced,0.2200053334236145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,16,balanced,0.2916319966316223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,16,balanced,0.37014933427174884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,16,balanced,0.5192533334096273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,16,balanced,0.6616693337758383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,8,balanced,0.01844266677896182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,8,balanced,0.017946666727463405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,8,balanced,0.01730666682124138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,8,balanced,0.019071999937295914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,8,balanced,0.019098666807015736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,8,balanced,0.019333332777023315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,8,balanced,0.019120000302791595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,8,balanced,0.021370666722456615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,8,balanced,0.023285334308942158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,8,balanced,0.02309333284695943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,8,balanced,0.03089066594839096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,8,balanced,0.03250666707754135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,8,balanced,0.03281066566705704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,8,balanced,0.03164266546567281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,8,balanced,0.03272533416748047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,8,balanced,0.03350933392842611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,8,balanced,0.034287999073664345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,8,balanced,0.03921066721280416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,8,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,8,balanced,0.05585599939028422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,8,balanced,0.061759998401006065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,8,balanced,0.08111999928951263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,8,balanced,0.09779733419418335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,8,balanced,0.12756266196568808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,8,balanced,0.16149866580963135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,8,balanced,0.2246560057004293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,8,balanced,0.2884213328361511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,8,power_law_1.01,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,8,power_law_1.01,0.039443200826644896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,8,power_law_1.01,0.038335999846458434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,8,power_law_1.01,0.0403328001499176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,8,power_law_1.01,0.04056960046291351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,8,power_law_1.01,0.041465601325035094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,8,power_law_1.01,0.04119040071964264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,8,power_law_1.01,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,8,power_law_1.01,0.04158720076084137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,8,power_law_1.01,0.04186240136623383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,8,power_law_1.01,0.042284798622131345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,8,power_law_1.01,0.04359680116176605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,8,power_law_1.01,0.043398401141166686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,8,power_law_1.01,0.04442879855632782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,8,power_law_1.01,0.049983999133110045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,8,power_law_1.01,0.05148800015449524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,8,power_law_1.01,0.0545087993144989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,8,power_law_1.01,0.06219519972801209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,8,power_law_1.01,0.06999040246009827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,8,power_law_1.01,0.09301120042800903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,8,power_law_1.01,0.1210368037223816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,8,power_law_1.01,0.1613312005996704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,1,power_law_1.2,0.1235584020614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,1,power_law_1.2,0.15611519813537597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,1,power_law_1.2,0.2058624029159546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,1,power_law_1.2,0.33019518852233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,8,balanced,0.031701333820819855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,8,balanced,0.037418665985266365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,1,power_law_1.2,0.4824575901031494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,8,balanced,0.05388266841570536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,8,balanced,0.07857066889603932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,1,power_law_1.2,0.6653056144714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,8,balanced,0.12563199798266092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,8,balanced,0.12103466192881267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,8,balanced,0.12264000376065572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,8,balanced,0.12461866935094197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,8,balanced,0.12324800093968709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,8,balanced,0.12168533603350322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,8,balanced,0.12291199962298076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,8,balanced,0.12456533312797546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,8,balanced,0.1243893305460612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,8,balanced,0.12643200159072876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,8,balanced,0.13012267152468363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,8,balanced,0.13145066301027933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,8,balanced,0.1323360006014506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,8,balanced,0.13917866349220276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,8,balanced,0.1476746698220571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,8,balanced,0.15465066830317178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,8,balanced,0.1623146633307139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,8,balanced,0.19638399283091226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,1,power_law_1.2,0.7689727783203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,8,balanced,0.2134666641553243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,8,balanced,0.2640213370323181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,8,balanced,0.30819199482599896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,8,balanced,0.3970506588617961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,1,power_law_1.2,1.0237183570861816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,8,balanced,0.5059839884440104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,1,power_law_1.2,1.0758399963378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,8,power_law_1.01,0.20730879306793212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,1,power_law_1.2,1.0935104370117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,1,power_law_1.2,1.1815808296203614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,1,power_law_1.2,1.2043392181396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,1,power_law_1.2,1.2592576026916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,1,power_law_1.2,1.2891648292541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,1,power_law_1.2,1.3312000274658202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,1,power_law_1.2,1.3884480476379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,1,power_law_1.2,1.4193663597106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,1,power_law_1.2,1.526585578918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.2,1.6134143829345704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.2,1.8054527282714843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.2,2.0243263244628906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,8,power_law_1.01,0.2850624084472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.2,2.4064895629882814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.2,2.5477951049804686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.2,3.151865577697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,1,power_law_1.2,3.972051239013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,1,power_law_1.2,5.151718521118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,1,power_law_1.2,6.170316696166992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,8,power_law_1.01,0.37639679908752444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,8,power_law_1.01,0.5360767841339111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,8,balanced,0.03053866575161616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,8,balanced,0.029359998802344005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,8,balanced,0.031210665901501972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,8,balanced,0.034287999073664345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,8,balanced,0.03871466716130575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,8,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,8,balanced,0.039434666434923805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,8,balanced,0.03976000100374222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,8,balanced,0.04152533411979675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,8,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,8,balanced,0.0413973331451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,8,balanced,0.041477332512537636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,8,balanced,0.04364266494909922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,8,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,8,balanced,0.04757866760094961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,8,balanced,0.050160000721613564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,8,balanced,0.050106664498647056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,8,balanced,0.054117331902186074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,8,balanced,0.058058664202690125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,8,balanced,0.07077333331108093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,8,balanced,0.07677866518497467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,8,balanced,0.11212266484896342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,8,balanced,0.1425386667251587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,8,balanced,0.19226133823394775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,8,balanced,0.236570676167806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,8,balanced,0.3354133367538452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,8,balanced,0.42766400178273517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,8,power_law_1.01,0.7234623908996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,32,balanced,0.0580266664425532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,32,balanced,0.057914664347966514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,32,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,32,balanced,0.06408533453941345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,32,balanced,0.06464000046253204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,32,balanced,0.0783786674340566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,32,balanced,0.09103467067082723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,32,balanced,0.090229332447052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,32,balanced,0.0885813335577647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,32,balanced,0.09201600154240926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,32,balanced,0.09121066331863403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,32,balanced,0.09091200431187947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,32,balanced,0.09227200349171956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,32,balanced,0.09436800082524617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,32,balanced,0.09797333677609761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,32,balanced,0.10082133611043294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,32,balanced,0.10467200477917989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,32,balanced,0.11134399970372517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,32,balanced,0.11780266960461934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,32,balanced,0.13503999511400858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,32,balanced,0.1482080022493998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,32,balanced,0.17971734205881754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,32,balanced,0.20772800842920938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,32,balanced,0.2663573424021403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,32,balanced,0.32992533842722577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,32,balanced,0.4469066858291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,32,balanced,0.5648906628290812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,4,power_law_1.2,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,4,power_law_1.2,0.051148802042007446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,4,power_law_1.2,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,4,power_law_1.2,0.06160640120506287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,4,power_law_1.2,0.06515840291976929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,4,power_law_1.2,0.07070720195770264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,4,power_law_1.2,0.07713919878005981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,4,power_law_1.2,0.08426880240440368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,4,power_law_1.2,0.08252800107002259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,4,power_law_1.2,0.08659200072288513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,4,power_law_1.2,0.08415359854698182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,4,power_law_1.2,0.08909440040588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,4,power_law_1.2,0.09223039746284485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,4,power_law_1.2,0.09861119985580444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,4,power_law_1.2,0.10407040119171143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,4,power_law_1.2,0.11043839454650879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,4,power_law_1.2,0.11393280029296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,4,power_law_1.2,0.1304128050804138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,4,power_law_1.2,0.14433280229568482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,4,power_law_1.2,0.17404799461364745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,4,power_law_1.2,0.19434239864349365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,4,power_law_1.2,0.25410559177398684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,4,power_law_1.2,0.3097791910171509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,4,power_law_1.2,0.41631360054016114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,4,power_law_1.2,0.512556791305542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,4,power_law_1.2,0.7848063945770264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,4,power_law_1.2,0.9804096221923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,8,power_law_1.2,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,8,power_law_1.2,0.04776319861412048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,8,power_law_1.2,0.0525056004524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,8,power_law_1.2,0.052671998739242554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,8,power_law_1.2,0.057734400033950806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,8,power_law_1.2,0.060313600301742556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,8,power_law_1.2,0.060601598024368285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,8,power_law_1.2,0.06218879818916321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,8,power_law_1.2,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,8,power_law_1.2,0.06358399987220764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,8,power_law_1.2,0.06162559986114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,8,power_law_1.2,0.06353279948234558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,8,power_law_1.2,0.06312320232391358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,8,power_law_1.2,0.0661184012889862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,8,power_law_1.2,0.07207040190696716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,8,power_law_1.2,0.0722495973110199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,8,power_law_1.2,0.074726402759552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,8,power_law_1.2,0.08432639837265014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.2,0.09456639885902404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.2,0.11281919479370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.2,0.13719040155410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.2,0.17278079986572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.2,0.20528640747070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.2,0.28679039478302004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,16,power_law_1.2,0.05689600110054016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,16,power_law_1.2,0.07292799949645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,16,power_law_1.2,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,16,power_law_1.2,0.06935039758682252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,16,power_law_1.2,0.06832640171051026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,16,power_law_1.2,0.06407039761543273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,16,power_law_1.2,0.06816639900207519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,16,power_law_1.2,0.07131519913673401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,16,power_law_1.2,0.06988160014152527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,16,power_law_1.2,0.0719488024711609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,16,power_law_1.2,0.07306879758834839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,16,power_law_1.2,0.07713919878005981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,16,power_law_1.2,0.08199040293693542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,16,power_law_1.2,0.0796671986579895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,16,power_law_1.2,0.09072639942169189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,16,power_law_1.2,0.09648000001907349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,16,power_law_1.2,0.09981439709663391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,16,power_law_1.2,0.11566720008850098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,16,power_law_1.2,0.12227200269699097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,16,power_law_1.2,0.1477120041847229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,16,power_law_1.2,0.17258880138397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,16,power_law_1.2,0.21779201030731202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,16,power_law_1.2,0.2836416006088257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,16,power_law_1.2,0.40081281661987306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,16,power_law_1.2,0.531379222869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,16,power_law_1.2,0.7386752128601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,16,power_law_1.2,0.9312064170837402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,2,power_law_1.2,0.04614399969577789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,2,power_law_1.2,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,2,power_law_1.2,0.057081598043441775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,2,power_law_1.2,0.06467840075492859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,2,power_law_1.2,0.06754559874534607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,2,power_law_1.2,0.07954559922218322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,2,power_law_1.2,0.0815168023109436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,2,power_law_1.2,0.08440319895744323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,2,power_law_1.2,0.08456959724426269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,2,power_law_1.2,0.08479999899864196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,2,power_law_1.2,0.08849279880523682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,2,power_law_1.2,0.09015679955482483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.2,0.3988800048828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,2,power_law_1.2,0.09126399755477906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,2,power_law_1.2,0.09522560238838196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,2,power_law_1.2,0.09561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,2,power_law_1.2,0.09713919758796692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,2,power_law_1.2,0.1019327998161316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,2,power_law_1.2,0.11345280408859253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,2,power_law_1.2,0.126803195476532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,2,power_law_1.2,0.15535999536514283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,2,power_law_1.2,0.182425594329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,2,power_law_1.2,0.264902400970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.2,0.5628543853759765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,2,power_law_1.2,0.322271990776062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,2,power_law_1.2,0.4538559913635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,2,power_law_1.2,0.6400191783905029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,1,power_law_1.01,0.05090559720993042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,1,power_law_1.01,0.054150402545928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,1,power_law_1.01,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,1,power_law_1.01,0.06785280108451844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,1,power_law_1.01,0.07891839742660522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,1,power_law_1.01,0.08814079761505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,1,power_law_1.01,0.10393600463867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,1,power_law_1.01,0.10796159505844116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,1,power_law_1.01,0.11014399528503419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,1,power_law_1.01,0.11354240179061889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,1,power_law_1.01,0.11868799924850464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,1,power_law_1.01,0.1270784020423889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,1,power_law_1.01,0.13301119804382325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,1,power_law_1.01,0.13852159976959227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,1,power_law_1.01,0.15253119468688964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,1,power_law_1.01,0.16047359704971315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,2,power_law_1.2,0.9003264427185058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,1,power_law_1.01,0.161190402507782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,1,power_law_1.01,0.19266560077667236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,1,power_law_1.01,0.2215424060821533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,1,power_law_1.01,0.28044800758361815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,1,power_law_1.01,0.34607999324798583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,1,power_law_1.01,0.45243520736694337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,1,power_law_1.01,0.5609536170959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,1,power_law_1.01,0.7843776226043702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,1,power_law_1.01,1.0129407882690429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,1,power_law_1.01,1.453382396697998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,1,power_law_1.01,1.8981632232666015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,32,balanced,0.06005866825580597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,32,balanced,0.06032533446947733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,32,balanced,0.06006933252016703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,32,balanced,0.07601599891980489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,32,balanced,0.09684800108273824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,32,balanced,0.10717866818110149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,32,balanced,0.10455999771753947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,32,balanced,0.10599999626477559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,32,balanced,0.1086346705754598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,32,balanced,0.10531199971834819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,32,balanced,0.10703466335932414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,32,balanced,0.11581866939862569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,32,balanced,0.10502933462460835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,32,balanced,0.10599999626477559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,32,balanced,0.11211199561754863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,32,balanced,0.11009599765141805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,32,balanced,0.12507200241088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,32,balanced,0.11975466211636861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,32,balanced,0.1206773320833842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,32,balanced,0.13404800494511923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,32,balanced,0.13423466682434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,32,balanced,0.20044799645741782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,32,balanced,0.20356800158818564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,32,balanced,0.2767573396364848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,32,balanced,0.35706667105356854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,32,balanced,0.42852266629536945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,32,balanced,0.5735626618067423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,32,power_law_1.2,0.0791808009147644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,32,power_law_1.2,0.08046720027923585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,32,power_law_1.2,0.07553920149803162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,32,power_law_1.2,0.07554559707641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,32,power_law_1.2,0.07180160284042358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,32,power_law_1.2,0.07388799786567687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,32,power_law_1.2,0.07656319737434387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,32,power_law_1.2,0.07647359967231751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,32,power_law_1.2,0.07786239981651306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,32,power_law_1.2,0.07555840015411378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,32,power_law_1.2,0.07523199915885925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,32,power_law_1.2,0.08050559759140015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,32,power_law_1.2,0.0816703975200653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,32,power_law_1.2,0.081740802526474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,32,power_law_1.2,0.08588799834251404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,32,power_law_1.2,0.09100160002708435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,32,power_law_1.2,0.09575679898262024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,32,power_law_1.2,0.10648959875106812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.2,0.11864960193634033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.2,0.14335360527038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.2,0.1668352007865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.2,0.215500807762146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.2,0.24746880531311036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.2,0.35916800498962403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,2,power_law_1.2,1.2024895668029785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.2,0.4629568099975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.2,0.6591616153717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.2,0.8038528442382813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.2,0.9371071815490722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,64,power_law_1.2,0.018246400356292724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,64,power_law_1.2,0.016947199404239655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,64,power_law_1.2,0.016947199404239655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,64,power_law_1.2,0.017401599884033205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,64,power_law_1.2,0.017977599799633027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,64,power_law_1.2,0.020019200444221497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,64,power_law_1.2,0.020601600408554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,64,power_law_1.2,0.027065598964691163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,64,power_law_1.2,0.02800000011920929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,64,power_law_1.2,0.028044798970222475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,64,power_law_1.2,0.027878400683403016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,64,power_law_1.2,0.027667200565338133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,64,power_law_1.2,0.02861439883708954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,64,power_law_1.2,0.028377598524093627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,64,power_law_1.2,0.028812798857688903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,64,power_law_1.2,0.029926401376724244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,64,power_law_1.2,0.031244799494743347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,64,power_law_1.2,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,64,power_law_1.2,0.03455359935760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,64,power_law_1.2,0.04030719995498657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,64,power_law_1.2,0.045587199926376346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,64,power_law_1.2,0.0542527973651886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,64,power_law_1.2,0.06703360080718994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,64,power_law_1.2,0.08945919871330262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,64,power_law_1.2,0.11554559469223022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,64,balanced,0.039034667114416756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,64,balanced,0.03942933430274328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,64,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,64,balanced,0.04181866844495138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,64,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,64,balanced,0.045994664231936135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,64,balanced,0.04624533156553904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,64,balanced,0.04602666695912679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,64,balanced,0.05028266708056132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,64,balanced,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,64,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,64,balanced,0.05012799799442291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,64,balanced,0.0487413356701533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,64,balanced,0.05204799771308899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,64,balanced,0.05820266902446747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,64,balanced,0.05834133426348368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,64,balanced,0.06225599845250448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,64,balanced,0.06817066669464111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,64,balanced,0.07258666555086772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,64,balanced,0.08277866741021474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,64,balanced,0.09311466415723164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,64,balanced,0.11564800143241882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,64,balanced,0.13622400164604187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,64,balanced,0.18282665808995566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,64,balanced,0.2220426599184672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,64,balanced,0.31007466713587445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,64,balanced,0.3922239939371745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,2,power_law_1.2,0.023366400599479677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,2,power_law_1.2,0.02210559993982315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,2,power_law_1.2,0.02170879989862442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,2,power_law_1.2,0.026163199543952943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,2,power_law_1.2,0.03133440017700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,2,power_law_1.2,0.031327998638153075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,2,power_law_1.2,0.030214399099349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,2,power_law_1.2,0.03215999901294708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,2,power_law_1.2,0.03900800049304962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,2,power_law_1.2,0.033318400382995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,2,power_law_1.2,0.03632639944553375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,64,power_law_1.2,0.15696640014648439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,2,power_law_1.2,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,2,power_law_1.2,0.07562879920005798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,2,power_law_1.2,0.09201279878616334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,2,power_law_1.2,0.06694399714469909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,2,power_law_1.2,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,2,power_law_1.2,0.06307839751243591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,2,power_law_1.2,0.08301439881324768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.2,0.09905920028686524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.2,0.13848960399627686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.2,0.17182079553604127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.2,0.23980801105499266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.2,0.30005760192871095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,64,power_law_1.2,0.19809919595718384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,2,balanced,0.04828266799449921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,2,balanced,0.050293331344922386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,2,balanced,0.0543093333641688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,2,balanced,0.06663999954859416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,2,balanced,0.09490133325258891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,2,balanced,0.1393066644668579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,2,balanced,0.1434293289979299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,2,balanced,0.13921067118644714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,2,balanced,0.14243732889493307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,2,balanced,0.14293866356213888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,2,balanced,0.14306132992108664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,2,balanced,0.1493226687113444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.2,0.4361663818359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,2,balanced,0.15266133348147073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,2,balanced,0.15631999572118124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,2,balanced,0.16591999928156534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,2,balanced,0.1706399917602539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,2,balanced,0.18083733320236206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,2,balanced,0.20949333906173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,2,balanced,0.22774400313695273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,2,balanced,0.2762239972750346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,2,balanced,0.3142133355140686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,2,balanced,0.440559983253479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,2,balanced,0.5184853474299113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,2,balanced,0.7546133200327555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,2,balanced,0.9357120196024576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,2,balanced,1.3656320571899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,4,power_law_1.2,0.0964031994342804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,2,balanced,1.7759466171264648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,4,power_law_1.2,0.147052800655365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,4,power_law_1.2,0.2063999891281128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,4,power_law_1.2,0.27023360729217527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,4,power_law_1.2,0.3282560110092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,4,power_law_1.2,0.38462080955505373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,4,power_law_1.2,0.4284736156463623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,4,power_law_1.2,0.4899903774261475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,4,power_law_1.2,0.49936637878417967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,4,power_law_1.2,0.47587199211120607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,4,power_law_1.2,0.5040832042694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,4,power_law_1.2,0.5311423778533936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,4,power_law_1.2,0.5597631931304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,4,power_law_1.2,0.5103551864624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,4,power_law_1.2,0.5492544174194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,4,power_law_1.2,0.5885503768920899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.2,0.5625728130340576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,4,power_law_1.2,0.661568021774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,4,power_law_1.2,0.6661824226379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,4,power_law_1.2,0.750809621810913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,4,power_law_1.2,0.9029184341430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,4,power_law_1.2,0.8672384262084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,4,power_law_1.2,1.0491968154907227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,4,power_law_1.2,1.0716608047485352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,4,power_law_1.2,1.3562751770019532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,4,power_law_1.2,1.6582080841064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,4,power_law_1.2,2.2155967712402345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,4,power_law_1.2,2.799212837219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,16,balanced,0.05425066749254862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,16,balanced,0.05412800113360087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,16,balanced,0.05598400036493937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,16,balanced,0.06274666885534923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,16,balanced,0.07164266705513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,16,balanced,0.09398933251698811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,16,balanced,0.09443199634552002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,16,balanced,0.09339200456937154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,16,balanced,0.09707199533780415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,16,balanced,0.09471999605496724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,16,balanced,0.09634666641553243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,16,balanced,0.09819199641545613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,16,balanced,0.09815466403961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,16,balanced,0.09963732957839966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,16,balanced,0.10946133732795715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,16,balanced,0.10983999570210774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,16,balanced,0.11293333768844604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,16,balanced,0.12773866454760233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,16,balanced,0.13206400473912558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,16,balanced,0.14531733592351279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,16,balanced,0.16310933232307434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,16,balanced,0.1956160068511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,16,balanced,0.22475733359654745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,16,balanced,0.31414933999379474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,16,balanced,0.38208532333374023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,16,balanced,0.5359146595001221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,16,balanced,0.6998933156331381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,1,power_law_1.01,0.10120960474014282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,1,power_law_1.01,0.11249920129776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,1,power_law_1.01,0.12826240062713623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,1,power_law_1.01,0.17993600368499757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,1,power_law_1.01,0.2201472043991089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,1,power_law_1.01,0.25770881175994875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,1,power_law_1.01,0.32651519775390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,1,power_law_1.01,0.3332479953765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,1,power_law_1.01,0.3591871976852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,1,power_law_1.01,0.33980159759521483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,1,power_law_1.01,0.357478404045105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,1,power_law_1.01,0.35538558959960936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,1,power_law_1.01,0.3655359983444214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,1,power_law_1.01,0.3798975944519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,1,power_law_1.01,0.3844160079956055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,1,power_law_1.01,0.39657599925994874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,1,power_law_1.01,0.4149184226989746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,1,power_law_1.01,0.46744961738586427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.01,0.5127359867095947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.01,0.6119872093200683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.01,0.7040256023406982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.01,0.9003583908081054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.01,1.0910783767700196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.01,1.4793343544006348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.01,1.878611183166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.2,0.8296768188476562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,1,power_law_1.01,2.6541120529174806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,1,power_law_1.01,3.422252655029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.2,1.0912704467773438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,1,balanced,0.05046933392683665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,1,balanced,0.058373332023620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,1,balanced,0.07993066807587941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,1,balanced,0.11771200100580852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,1,balanced,0.17739200592041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,1,balanced,0.2118826707204183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,1,balanced,0.21811199188232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,1,balanced,0.2158506711324056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,1,balanced,0.22016000747680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,1,balanced,0.2238346735636393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,1,balanced,0.22380266586939493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,1,balanced,0.2323733369509379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,1,balanced,0.238864004611969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,1,balanced,0.24359466632207236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,1,balanced,0.2531999945640564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,1,balanced,0.265664001305898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,1,balanced,0.28144532442092896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,1,balanced,0.3179519971211751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,1,balanced,0.3513173262278239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,1,balanced,0.47177600860595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,1,balanced,0.5028479894002279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,1,balanced,0.7453813552856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,1,balanced,0.8445226351420084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,1,balanced,1.3265066941579182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,1,balanced,1.541445255279541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,1,balanced,2.3338987032572427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,1,balanced,2.9261226654052734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,4,power_law_1.2,0.03692159950733185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,4,power_law_1.2,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,4,power_law_1.2,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,4,power_law_1.2,0.041657599806785586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,4,power_law_1.2,0.0435263991355896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,4,power_law_1.2,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,4,power_law_1.2,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,4,power_law_1.2,0.056857597827911374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,4,power_law_1.2,0.058316802978515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,4,power_law_1.2,0.059494400024414064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,4,power_law_1.2,0.06112639904022217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,4,power_law_1.2,0.06399359703063964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,4,power_law_1.2,0.06631039977073669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,4,power_law_1.2,0.06787840127944947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,4,power_law_1.2,0.07491840124130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,4,power_law_1.2,0.07697920203208923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,4,power_law_1.2,0.08033279776573181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,128,balanced,0.05299200117588043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,2,power_law_1.01,0.07529600262641907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,128,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,128,balanced,0.052442664901415505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,128,balanced,0.05619733532269796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,128,balanced,0.05436799923578898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,128,balanced,0.06080533564090729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,128,balanced,0.059936001896858215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,128,balanced,0.06163733204205831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,128,balanced,0.060191998879114784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,128,balanced,0.06216000020503998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,128,balanced,0.060133333007494606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,128,balanced,0.06222933530807495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,128,balanced,0.06500266492366791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,128,balanced,0.06438933312892914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,128,balanced,0.07021866738796234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,128,balanced,0.07222400108973186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,128,balanced,0.07646933197975159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,4,power_law_1.2,0.08981760144233704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,4,power_law_1.2,0.11020799875259399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,2,power_law_1.01,0.08227840065956116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,4,power_law_1.2,0.13082239627838135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,2,power_law_1.01,0.10012799501419067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,4,power_law_1.2,0.13611520528793336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,2,power_law_1.01,0.1161344051361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,4,power_law_1.2,0.18803199529647827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,2,power_law_1.01,0.12428159713745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,128,balanced,0.08044800162315369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,2,power_law_1.01,0.13761919736862183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,2,power_law_1.01,0.15204479694366455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,2,power_law_1.01,0.1602687954902649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,2,power_law_1.01,0.16462719440460205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,2,power_law_1.01,0.16341760158538818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,128,balanced,0.08619733651479085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,128,balanced,0.1032480001449585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,128,balanced,0.12262399991353352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,128,balanced,0.1644426683584849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,128,balanced,0.19530133406321207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,128,balanced,0.2625653346379598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,128,balanced,0.3262346585591634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,4,power_law_1.2,0.2332672119140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,128,balanced,0.4522720177968343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,128,balanced,0.5781333446502686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,4,power_law_1.2,0.3319616079330444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,4,power_law_1.2,0.4049344062805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,4,power_law_1.2,0.5685887813568116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,8,power_law_1.2,0.04573439955711365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,8,power_law_1.2,0.05087360143661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,8,power_law_1.2,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,8,power_law_1.2,0.06645119786262513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,8,power_law_1.2,0.07535359859466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,8,power_law_1.2,0.07838079929351807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,8,power_law_1.2,0.07936639785766601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,8,power_law_1.2,0.08037760257720947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,8,power_law_1.2,0.07836800217628478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,8,power_law_1.2,0.08042240142822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,8,power_law_1.2,0.08204159736633301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,8,power_law_1.2,0.08539519906044006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,8,power_law_1.2,0.08883839845657349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,8,power_law_1.2,0.09646080136299133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,8,power_law_1.2,0.10627199411392212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,8,power_law_1.2,0.10940799713134766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,8,power_law_1.2,0.11393920183181763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,8,power_law_1.2,0.13201279640197755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,8,power_law_1.2,0.1449728012084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,8,power_law_1.2,0.18407679796218873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,8,power_law_1.2,0.213702392578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,8,power_law_1.2,0.28399360179901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,8,power_law_1.2,0.34101119041442873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,8,power_law_1.2,0.4700607776641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,8,power_law_1.2,0.6635263919830322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,8,power_law_1.2,0.9087424278259277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,8,power_law_1.2,1.195206356048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,4,power_law_1.2,0.8456319808959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,2,power_law_1.2,0.06464639902114869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,2,power_law_1.2,0.11155200004577637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,2,power_law_1.2,0.09838719964027405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,2,power_law_1.2,0.1328703999519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,2,power_law_1.2,0.13697919845581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,2,power_law_1.2,0.14058879613876343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,2,power_law_1.2,0.14430079460144044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,2,power_law_1.2,0.14675840139389038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,2,power_law_1.2,0.14910720586776732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,2,power_law_1.2,0.15419520139694215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,2,power_law_1.2,0.15739519596099855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,2,power_law_1.2,0.16357120275497436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,2,power_law_1.2,0.16806399822235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,2,power_law_1.2,0.17498879432678222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,2,power_law_1.2,0.18792320489883424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,2,power_law_1.2,0.19966720342636107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,2,power_law_1.2,0.22605440616607667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,2,power_law_1.2,0.253222393989563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,2,power_law_1.2,0.2769984006881714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,2,power_law_1.2,0.3554176092147827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,2,power_law_1.2,0.4065536022186279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,2,power_law_1.2,0.543775987625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,2,power_law_1.2,0.7516223907470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,2,power_law_1.2,1.0194175720214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,1,power_law_1.2,0.06230400204658508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,1,power_law_1.2,0.08170239925384522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,1,power_law_1.2,0.0929535984992981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,1,power_law_1.2,0.14023679494857788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,1,power_law_1.2,0.1418496012687683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,1,power_law_1.2,0.14344320297241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,1,power_law_1.2,0.14398080110549927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,1,power_law_1.2,0.145414400100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,1,power_law_1.2,0.14722559452056885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,16,balanced,0.021141332884629566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,16,balanced,0.02103466788927714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,1,power_law_1.2,0.14837119579315186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,1,power_law_1.2,0.15152640342712403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,1,power_law_1.2,0.16027519702911378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,1,power_law_1.2,0.1610751986503601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,1,power_law_1.2,0.16552319526672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,1,power_law_1.2,0.1835904002189636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,16,balanced,0.02120000123977661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,16,balanced,0.021290667355060577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,16,balanced,0.023423999547958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,1,power_law_1.2,0.18625919818878173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,16,balanced,0.026181332767009735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,16,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,16,balanced,0.029898665845394135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,16,balanced,0.04460800190766653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,16,balanced,0.04364799956480662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,1,power_law_1.2,0.2064255952835083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,1,power_law_1.2,0.23678081035614013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,16,balanced,0.044069334864616394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,16,balanced,0.04766400158405304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,16,balanced,0.046384001771608986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,16,balanced,0.0439626673857371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,16,balanced,0.056234667698542275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,16,balanced,0.055498664577802025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,16,balanced,0.05172266562779745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,16,balanced,0.05992533266544342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,16,balanced,0.06879466772079468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,16,balanced,0.08456533153851827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,16,balanced,0.0974026620388031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,16,balanced,0.11965866883595784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,16,balanced,0.1418346663316091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,16,balanced,0.2050559918085734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,16,balanced,0.25146667162577313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,16,balanced,0.36022400856018066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.2,0.27550079822540285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,16,balanced,0.4707466761271159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.2,0.3413439989089966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.2,0.409497594833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.2,0.5513919830322266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,2,power_law_1.2,1.5656064033508301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.2,0.6987775802612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.2,0.9872127532958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,16,balanced,0.031557333966096245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,16,balanced,0.03156266609827677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,16,balanced,0.0315733328461647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,16,balanced,0.033615998923778534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,16,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,16,balanced,0.03562666724125544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,16,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,16,balanced,0.03796799977620443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,16,balanced,0.03782933453718821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,16,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,16,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,16,balanced,0.04181866844495138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,16,balanced,0.03987200061480204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,16,balanced,0.040005333721637726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,16,balanced,0.04379733403523763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,16,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,16,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,16,balanced,0.0487306664387385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,2,power_law_1.2,1.685843276977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,16,balanced,0.05293866495291392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,16,balanced,0.058261334896087646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,16,balanced,0.06032533446947733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,16,balanced,0.07871466875076294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,16,balanced,0.09074667096138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,16,balanced,0.12356799840927124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,16,balanced,0.1474720040957133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,16,balanced,0.19966399669647217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,16,balanced,0.2531733314196269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.2,1.278707218170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,2,power_law_1.2,2.518374443054199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.2,1.851251220703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,16,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,16,balanced,0.030037333567937214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,16,balanced,0.03166399896144867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,16,balanced,0.03368533402681351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,16,balanced,0.035743998984495796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,16,balanced,0.03757333258787791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,16,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,16,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,16,balanced,0.040021332601706185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,16,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,16,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,16,balanced,0.042863999803860985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,16,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,16,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,16,balanced,0.045834665497144066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,16,balanced,0.049770668148994446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,16,balanced,0.050010666251182556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,16,balanced,0.05383466680844625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,16,balanced,0.07039466500282288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,16,balanced,0.08329600095748901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,16,balanced,0.08623466889063518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,16,balanced,0.11186133821805318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,16,balanced,0.1285920043786367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,16,balanced,0.16357333461443582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,16,balanced,0.2083466649055481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,16,balanced,0.2831626733144124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,16,balanced,0.35275201002756756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,16,power_law_1.01,0.08876799941062927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,16,power_law_1.01,0.10536960363388062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,16,power_law_1.01,0.10065920352935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,16,power_law_1.01,0.1069375991821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,16,power_law_1.01,0.10602879524230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,16,power_law_1.01,0.10214400291442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,16,power_law_1.01,0.10236159563064576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,16,power_law_1.01,0.10483839511871337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,16,power_law_1.01,0.10318080186843873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,16,power_law_1.01,0.1032639980316162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,16,power_law_1.01,0.1072767972946167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,16,power_law_1.01,0.11190400123596192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,16,power_law_1.01,0.11381759643554687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,16,power_law_1.01,0.11752320528030395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,16,power_law_1.01,0.12289279699325562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,16,power_law_1.01,0.12828160524368287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,16,power_law_1.01,0.13232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,16,power_law_1.01,0.14999680519104003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,16,power_law_1.01,0.16437120437622071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,16,power_law_1.01,0.20764799118041993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,16,power_law_1.01,0.23759360313415528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,16,power_law_1.01,0.3080384016036987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.2,2.4400768280029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,16,power_law_1.01,0.3307136058807373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,16,power_law_1.01,0.4274303913116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,16,power_law_1.01,0.5638463973999024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,16,power_law_1.01,0.778553581237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,16,power_law_1.01,0.9426943778991699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,1,power_law_1.2,0.036294400691986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,1,power_law_1.2,0.05351679921150208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,1,power_law_1.2,0.07141759991645813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,1,power_law_1.2,0.09425920248031616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,1,power_law_1.2,0.11493120193481446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,1,power_law_1.2,0.18511359691619872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,1,power_law_1.2,0.19382400512695314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,1,power_law_1.2,0.2021183967590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,1,power_law_1.2,0.20353279113769532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,1,power_law_1.2,0.20851199626922606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,1,power_law_1.2,0.2132863998413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,1,power_law_1.2,0.21709439754486085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,1,power_law_1.2,0.22006399631500245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,1,power_law_1.2,0.22711679935455323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,1,power_law_1.2,0.2396672010421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,1,power_law_1.2,0.24864640235900878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,1,power_law_1.2,0.2631360054016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,1,power_law_1.2,0.29443840980529784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,1,power_law_1.2,0.3295936107635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,1,power_law_1.2,0.397273588180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,1,power_law_1.2,0.4402304172515869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,1,power_law_1.2,0.5573184013366699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,1,power_law_1.2,0.6732607841491699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,32,power_law_1.2,0.08751999735832214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,32,power_law_1.2,0.08580480217933655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,32,power_law_1.2,0.08492799997329711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,32,power_law_1.2,0.08490880131721497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,32,power_law_1.2,0.08265600204467774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,32,power_law_1.2,0.08999680280685425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,32,power_law_1.2,0.09145600199699402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,32,power_law_1.2,0.09452800154685974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,32,power_law_1.2,0.09343360066413879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,32,power_law_1.2,0.0953279972076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,32,power_law_1.2,0.09381759762763978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,32,power_law_1.2,0.09661440253257751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,1,power_law_1.2,0.896025562286377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,32,power_law_1.2,0.09830399751663207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,32,power_law_1.2,0.09968640208244324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,32,power_law_1.2,0.10694400072097779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,32,power_law_1.2,0.11471999883651733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,32,power_law_1.2,0.1217087984085083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,32,power_law_1.2,0.14252159595489503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,32,power_law_1.2,0.15734399557113649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,32,power_law_1.2,0.19344639778137207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,32,power_law_1.2,0.23018240928649902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,32,power_law_1.2,0.3095232009887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,32,power_law_1.2,0.42447361946105955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,1,power_law_1.2,1.1042688369750977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,32,power_law_1.2,0.7346432209014893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,32,power_law_1.2,0.8625984191894531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,32,power_law_1.2,1.111788845062256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,32,power_law_1.2,1.6452159881591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,1,power_law_1.2,1.5450112342834472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,2,balanced,0.04613333443800608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,2,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,2,balanced,0.050901333491007485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,2,balanced,0.06394133468468984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,2,balanced,0.08892800410588582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,2,balanced,0.12708266576131186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,2,balanced,0.1288746694723765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,2,balanced,0.1292586624622345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,2,balanced,0.13153599699338278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,2,balanced,0.1327946682771047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,2,balanced,0.13090133666992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,2,balanced,0.13563199838002524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,2,balanced,0.13715733091036478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,2,balanced,0.13980266451835632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,2,balanced,0.1479200025399526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,2,balanced,0.15011200308799744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,2,balanced,0.16197333733240762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,2,balanced,0.18198400735855103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,2,balanced,0.1928960084915161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,2,balanced,0.22657066583633423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,2,balanced,0.2568959991137187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,2,balanced,0.3359786669413249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,2,balanced,0.39175466696421307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,2,balanced,0.5462453365325928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,2,balanced,0.673413356145223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,2,balanced,0.9695466359456381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,2,balanced,1.266698678334554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,1,power_law_1.2,1.9929536819458007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,4,power_law_1.2,0.08647680282592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,4,power_law_1.2,0.10263680219650269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,4,power_law_1.2,0.11784960031509399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,4,power_law_1.2,0.15240319967269897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,4,power_law_1.2,0.17848960161209107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,4,power_law_1.2,0.22717440128326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,4,power_law_1.2,0.2260672092437744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,4,power_law_1.2,0.22231678962707518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,4,power_law_1.2,0.23708159923553468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,4,power_law_1.2,0.239737606048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,4,power_law_1.2,0.23742079734802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,4,power_law_1.2,0.2541887998580933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,4,power_law_1.2,0.25384960174560545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,4,power_law_1.2,0.2612096071243286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,4,power_law_1.2,0.2831104040145874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,4,power_law_1.2,0.2926784038543701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,4,power_law_1.2,0.3106559991836548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,32,power_law_1.01,0.052153599262237546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,4,power_law_1.2,0.3510272026062012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,32,power_law_1.01,0.05511040091514587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,4,power_law_1.2,0.36985599994659424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,32,power_law_1.01,0.055852800607681274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,4,power_law_1.2,0.4624447822570801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,4,power_law_1.2,0.5180736064910889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,4,power_law_1.2,0.6678080081939697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,4,power_law_1.2,0.6849408149719238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,4,power_law_1.2,0.8881792068481446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,32,power_law_1.01,0.05818880200386047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,4,power_law_1.2,1.1749888420104981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,32,power_law_1.01,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,32,power_law_1.01,0.060678398609161376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,32,power_law_1.01,0.06001920104026794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,32,power_law_1.01,0.06128000020980835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,32,power_law_1.01,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,32,power_law_1.01,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,4,power_law_1.2,1.7225984573364257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,32,power_law_1.01,0.06436480283737182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,32,power_law_1.01,0.06552960276603699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,128,balanced,0.05287466446558634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,128,balanced,0.06043733159701029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,128,balanced,0.042949333786964417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,32,power_law_1.01,0.06686080098152161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,128,balanced,0.045408000548680626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,128,balanced,0.0462719996770223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,128,balanced,0.04602666695912679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,32,power_law_1.01,0.06824319958686828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,128,balanced,0.05017066498597463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,128,balanced,0.048309331138928734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,128,balanced,0.05023466547330221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,128,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,128,balanced,0.05213333169619242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,128,balanced,0.05203733344872793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,4,power_law_1.2,1.9039552688598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,128,balanced,0.050757333636283875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,128,balanced,0.052144000927607216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,128,balanced,0.056602666775385536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,128,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,32,power_law_1.01,0.07730559706687927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,32,power_law_1.01,0.07649279832839966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,32,power_law_1.01,0.08340479731559754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,32,power_law_1.01,0.09780480265617371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,128,balanced,0.056362668673197426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,128,balanced,0.0614026685555776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,128,balanced,0.0624533345301946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,128,balanced,0.0743146687746048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,128,balanced,0.07745600243409474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,128,balanced,0.09435199697812398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,128,balanced,0.11247467001279195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,128,balanced,0.14401599764823914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,128,balanced,0.17495999733606973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,128,balanced,0.22754132747650146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,32,power_law_1.01,0.10686719417572021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,32,power_law_1.01,0.135206401348114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,32,power_law_1.01,0.1537727952003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,32,power_law_1.01,0.1926975965499878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,32,power_law_1.01,0.2335360050201416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,128,balanced,0.2879413366317749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,32,power_law_1.01,0.3041984081268311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,32,power_law_1.01,0.3860863924026489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,32,power_law_1.01,0.5430399894714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,32,power_law_1.01,0.7113664150238037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,32,power_law_1.2,0.049491199851036075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,32,power_law_1.2,0.05196800231933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,32,power_law_1.2,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,32,power_law_1.2,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,32,power_law_1.2,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,32,power_law_1.2,0.04880639910697937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,32,power_law_1.2,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,32,power_law_1.2,0.05397760272026062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,32,power_law_1.2,0.05550079941749573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,32,power_law_1.2,0.05875840187072754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,32,power_law_1.2,0.06254720091819763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,32,power_law_1.2,0.06641280055046081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,32,power_law_1.2,0.06631680130958557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,32,power_law_1.2,0.10654720067977905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,32,power_law_1.2,0.1087615966796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,32,power_law_1.2,0.11892479658126831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,32,power_law_1.2,0.13411200046539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,32,power_law_1.2,0.1657472014427185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.2,0.20079998970031737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.2,0.24293119907379152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.2,0.13735040426254272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.2,0.19812480211257935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.2,0.23331201076507568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.2,0.37229440212249754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.2,0.4484288215637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.2,0.6165056228637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.2,0.8801664352416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,1,power_law_1.2,0.05452160239219665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,1,power_law_1.2,0.05925760269165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,1,power_law_1.2,0.07986559867858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,1,power_law_1.2,0.09594240188598632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,1,power_law_1.2,0.09751679897308349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,1,power_law_1.2,0.09946879744529724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,1,power_law_1.2,0.09935359954833985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,1,power_law_1.2,0.10043519735336304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,1,power_law_1.2,0.09960960149765015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,1,power_law_1.2,0.10028799772262573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,1,power_law_1.2,0.10101759433746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,1,power_law_1.2,0.10542080402374268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,1,power_law_1.2,0.1090175986289978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,1,power_law_1.2,0.11720319986343383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,1,power_law_1.2,0.12997119426727294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,1,power_law_1.2,0.1313472032546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,1,power_law_1.2,0.1397312045097351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,1,power_law_1.2,0.16109440326690674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.2,0.19209599494934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,4,power_law_1.01,0.06941440105438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,4,power_law_1.01,0.0827455997467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,4,power_law_1.01,0.08380799889564514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,4,power_law_1.01,0.10056320428848267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,4,power_law_1.01,0.10669440031051636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,4,power_law_1.01,0.13502720594406128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.2,0.25564160346984866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,4,power_law_1.01,0.12791039943695068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,4,power_law_1.01,0.13006080389022828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,4,power_law_1.01,0.134278404712677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,4,power_law_1.01,0.13749760389328003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,4,power_law_1.01,0.13717759847640992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,4,power_law_1.01,0.14407039880752565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,4,power_law_1.01,0.14647040367126465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,4,power_law_1.01,0.15096319913864137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,4,power_law_1.01,0.16123520135879515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,4,power_law_1.01,0.16481280326843262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,4,power_law_1.01,0.17713279724121095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,4,power_law_1.01,0.1995967984199524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.01,0.22419838905334472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.2,0.30799360275268556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.01,0.2697088003158569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.01,0.2959167957305908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.01,0.3884927988052368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.01,0.46901760101318357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.01,0.6112063884735107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.01,0.7334144115447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.01,1.0449024200439454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.01,1.3707967758178712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.2,0.4073023796081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.2,0.5233983993530273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.2,0.7562367916107178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.2,0.9810111999511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,64,balanced,0.01923199991385142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,64,balanced,0.01894933357834816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,64,balanced,0.020256000260512035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,64,balanced,0.01905599981546402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,64,balanced,0.021375998854637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,64,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,64,balanced,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,64,balanced,0.03127466638882955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,64,balanced,0.031311998764673867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,64,balanced,0.0313226655125618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,64,balanced,0.0313226655125618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,64,balanced,0.0330079992612203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,64,balanced,0.03259733319282532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,64,balanced,0.03346133232116699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,64,balanced,0.0353973334034284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,64,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,64,balanced,0.03667200108369192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,64,balanced,0.04181866844495138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,64,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,64,balanced,0.050373335679372154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,64,balanced,0.05594133337338766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,64,balanced,0.06990933418273926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,64,balanced,0.08958933750788371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,64,balanced,0.11353600025177002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,64,balanced,0.1455839971701304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,64,balanced,0.19960532585779825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.2,1.437286376953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,64,balanced,0.2585120002428691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.2,1.856185531616211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,4,power_law_1.2,0.017734399437904357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,4,power_law_1.2,0.017535999417304993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,4,power_law_1.2,0.01770240068435669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,4,power_law_1.2,0.018335999548435213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,4,power_law_1.2,0.018700799345970152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,4,power_law_1.2,0.01870719939470291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,4,power_law_1.2,0.022694399952888487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,4,power_law_1.2,0.021939200162887574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,4,power_law_1.2,0.03953279852867127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,4,power_law_1.2,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,4,power_law_1.2,0.03943040072917938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,4,power_law_1.2,0.03463680148124695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,4,power_law_1.2,0.03370240032672882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,4,power_law_1.2,0.03162879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,4,power_law_1.2,0.03463039994239807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,4,power_law_1.2,0.033062401413917544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,4,power_law_1.2,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,4,power_law_1.2,0.0458624005317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.2,0.04811519980430603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.2,0.0630400002002716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.2,0.07624319791793824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.2,0.09701120257377624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.2,0.12055679559707641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.2,0.16388479471206666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,2,power_law_1.2,0.10065920352935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,2,power_law_1.2,0.11230080127716065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,2,power_law_1.2,0.1278272032737732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,2,power_law_1.2,0.14881279468536376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,2,power_law_1.2,0.17136640548706056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,2,power_law_1.2,0.18140799999237062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,2,power_law_1.2,0.21969280242919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,2,power_law_1.2,0.22251520156860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,2,power_law_1.2,0.2343616008758545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,2,power_law_1.2,0.2421504020690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,2,power_law_1.2,0.24728960990905763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,2,power_law_1.2,0.25301759243011473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,2,power_law_1.2,0.25656321048736574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,2,power_law_1.2,0.2686847925186157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,2,power_law_1.2,0.2667327880859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,2,power_law_1.2,0.2705919981002808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,2,power_law_1.2,0.28803839683532717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,2,power_law_1.2,0.32274560928344725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,2,power_law_1.2,0.3700031995773315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,2,power_law_1.2,0.457919979095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,2,power_law_1.2,0.5387775897979736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,2,power_law_1.2,0.6844992160797119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,2,power_law_1.2,0.8538432121276855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,2,power_law_1.2,1.1786751747131348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,2,power_law_1.2,1.4742783546447753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,2,power_law_1.2,2.140550422668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,2,power_law_1.2,2.938150405883789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,8,balanced,0.025733334322770435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,8,balanced,0.02309333284695943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,8,balanced,0.02345066765944163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,8,balanced,0.02513066679239273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,8,balanced,0.03316266586383184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,8,balanced,0.05231466889381409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,8,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,8,balanced,0.054560000697771706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,8,balanced,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,8,balanced,0.05640000104904175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,8,balanced,0.06454400221506755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,8,balanced,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,8,balanced,0.07442133128643036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.2,0.20699520111083985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,8,balanced,0.06930666665236156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,8,balanced,0.07462933162848155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,8,balanced,0.08086933195590973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,8,balanced,0.08665066957473755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,8,balanced,0.09814932942390442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,8,balanced,0.15889066457748413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,8,balanced,0.19853333632151285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,8,balanced,0.17176532745361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,8,balanced,0.19693867365519205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,8,balanced,0.22842667500178018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,8,balanced,0.28307199478149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,8,balanced,0.34068799018859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,8,balanced,0.524394671122233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,8,balanced,0.6492053270339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.2,0.2935296058654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,8,balanced,0.02380266785621643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,8,balanced,0.023893333971500397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,8,balanced,0.025807999074459076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,8,balanced,0.027477333943049114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,8,balanced,0.03999999910593033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,8,balanced,0.05605866511662801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,8,balanced,0.05919999877611796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,8,balanced,0.060080001751581825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,8,balanced,0.06480533381303151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,8,balanced,0.06385600070158641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,8,balanced,0.06593066453933716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,8,balanced,0.08363733688990276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,8,balanced,0.08266133566697438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,8,balanced,0.08436266581217448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,8,balanced,0.12764267126719156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,8,balanced,0.12596266468365988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,8,balanced,0.13221866885821024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,8,balanced,0.14033599694569907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,8,balanced,0.14966932932535806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,8,balanced,0.1601706643899282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,8,balanced,0.1869866649309794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,8,balanced,0.2172266642252604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,8,balanced,0.25353066126505536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,8,balanced,0.3171573281288147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,8,balanced,0.38606401284535724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,8,balanced,0.5994240045547485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,8,balanced,0.7530612945556641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.2,0.3824575901031494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,8,balanced,0.03384000062942505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,8,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,8,balanced,0.03329599897066752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,8,balanced,0.03559466699759165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,8,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,8,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,8,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,8,balanced,0.0394400010506312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,8,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,8,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,8,balanced,0.037733333806196846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,8,balanced,0.039861333866914116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,8,balanced,0.03818666686614355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,8,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,8,balanced,0.04580800235271454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,8,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,8,balanced,0.04966933528582255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,8,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,8,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,8,balanced,0.07258666555086772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,8,balanced,0.07435200115044911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,8,balanced,0.09899733463923137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,8,balanced,0.1183093289534251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,8,balanced,0.15956800182660422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,8,balanced,0.19407999515533447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,8,balanced,0.26293333371480304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,8,balanced,0.34409598509470624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,16,power_law_1.2,0.07365760207176208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,16,power_law_1.2,0.06855679750442505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,16,power_law_1.2,0.06627839803695679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,16,power_law_1.2,0.07219200134277344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,16,power_law_1.2,0.06804479956626892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,16,power_law_1.2,0.07061759829521179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,16,power_law_1.2,0.0691648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,16,power_law_1.2,0.07489920258522034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,16,power_law_1.2,0.07584000229835511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,16,power_law_1.2,0.07679359912872315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,16,power_law_1.2,0.07807999849319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,16,power_law_1.2,0.07922559976577759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,16,power_law_1.2,0.07966079711914062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,16,power_law_1.2,0.08056319952011108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,16,power_law_1.2,0.08935040235519409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,16,power_law_1.2,0.09444479942321778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,16,power_law_1.2,0.09778559803962708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,16,power_law_1.2,0.1090880036354065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.2,0.12101759910583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.2,0.15123840570449829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.2,0.1776128053665161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.2,0.2091968059539795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.2,0.28360960483551023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.2,0.35632638931274413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.2,0.44579200744628905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.2,0.6718463897705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,2,power_law_1.01,0.15970560312271118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,2,power_law_1.01,0.23779840469360353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,2,power_law_1.01,0.324019193649292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,2,power_law_1.01,0.4798592090606689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,2,power_law_1.01,0.5515391826629639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,2,power_law_1.01,0.8845760345458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,2,power_law_1.01,0.9654463768005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,2,power_law_1.01,0.9991168022155762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,2,power_law_1.01,0.990783977508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,2,power_law_1.01,0.9955327987670899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,2,power_law_1.01,1.039027214050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,2,power_law_1.01,1.0500224113464356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,2,power_law_1.01,1.058784008026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,2,power_law_1.01,1.0887871742248536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,2,power_law_1.01,1.1340352058410645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.2,0.8743167877197265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,2,power_law_1.01,1.1842240333557128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,2,power_law_1.01,1.205299186706543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,2,power_law_1.01,1.4418623924255372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,2,power_law_1.01,0.07474560141563416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,2,power_law_1.01,0.09593600034713745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,2,power_law_1.01,0.10435199737548828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,2,power_law_1.01,0.13238400220870972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,2,power_law_1.01,0.160697603225708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,2,power_law_1.01,1.387731170654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,2,power_law_1.01,0.22008960247039794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,2,power_law_1.01,1.6429376602172852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,2,power_law_1.01,0.22768640518188477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,2,power_law_1.01,0.23109118938446044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,2,power_law_1.01,1.5368255615234374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,2,power_law_1.01,0.23958399295806884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,2,power_law_1.01,0.24335360527038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,2,power_law_1.01,1.763987159729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,2,power_law_1.01,0.24732160568237305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,2,power_law_1.01,0.2578687906265259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,2,power_law_1.01,0.2595328092575073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,2,power_law_1.01,2.1694015502929687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,2,power_law_1.01,0.2744256019592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,2,power_law_1.01,0.2933759927749634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,2,power_law_1.01,0.2961535930633545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,2,power_law_1.01,2.599648094177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,2,power_law_1.01,0.322489595413208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,2,power_law_1.01,0.3723200082778931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,2,power_law_1.01,0.40844159126281737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,2,power_law_1.01,0.5135295867919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,2,power_law_1.01,3.8241729736328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,2,power_law_1.01,0.639737606048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,2,power_law_1.01,0.810489559173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,2,power_law_1.01,0.9422975540161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,2,power_law_1.01,4.717324829101562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,2,power_law_1.01,1.2592448234558105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,2,power_law_1.01,1.5531392097473145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,2,power_law_1.01,5.73570556640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,2,power_law_1.01,2.2525503158569338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,2,power_law_1.01,2.8870271682739257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,4,balanced,0.11848533153533936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,4,balanced,0.11915733416875203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,4,balanced,0.18226667245229086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,4,balanced,0.1809013287226359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,4,balanced,0.1794933279355367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,4,balanced,0.17597333590189615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,4,balanced,0.1776426633199056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,4,balanced,0.18807466824849448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,4,balanced,0.1880106727282206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,4,balanced,0.18263467152913412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,4,balanced,0.17145599921544394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,4,balanced,0.1711946725845337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,4,balanced,0.16702399651209512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,4,balanced,0.16274666786193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,4,balanced,0.1572533349196116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,4,balanced,0.13873599966367087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,4,balanced,0.13446399569511414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,4,balanced,0.23945599794387817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,4,balanced,0.19388800859451294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,4,balanced,0.2656053304672241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,4,balanced,0.335968017578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,4,balanced,0.4613279898961385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,4,balanced,0.5891786813735962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,4,balanced,0.8423413435618082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,4,balanced,1.0891520182291667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,4,balanced,1.5871466000874836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,4,balanced,2.0848213831583657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,16,balanced,0.05034666756788889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,16,balanced,0.050373335679372154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,16,balanced,0.054458667834599815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,16,balanced,0.06099733213583628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,16,balanced,0.06464000046253204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,16,balanced,0.07266133526961009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,16,balanced,0.07469866673151652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,16,balanced,0.07461866736412048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,16,balanced,0.07284800211588542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,16,balanced,0.07271466652552287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,16,balanced,0.07195200026035309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,16,balanced,0.07462933162848155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,16,balanced,0.07453866799672444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,16,balanced,0.07397333284219106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,16,balanced,0.08066666622956593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,16,balanced,0.07859733204046886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,16,balanced,0.07947733501593272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,16,balanced,0.08680533369382222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,16,balanced,0.0936853289604187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,16,balanced,0.10705066720644633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,16,balanced,0.12135466933250427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,16,balanced,0.1600213348865509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,16,balanced,0.1812959909439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,16,balanced,0.23522132635116577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,16,balanced,0.2970293362935384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,16,balanced,0.39717864990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,16,balanced,0.5112640062967936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,64,power_law_1.2,0.03216640055179596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,64,power_law_1.2,0.03185920119285583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,64,power_law_1.2,0.03234559893608093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,64,power_law_1.2,0.03478400111198425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,64,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,64,power_law_1.2,0.03653759956359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,64,power_law_1.2,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,64,power_law_1.2,0.038022398948669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,64,power_law_1.2,0.04023039937019348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,64,power_law_1.2,0.04017919898033142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,64,power_law_1.2,0.04207360148429871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,64,power_law_1.2,0.04466559886932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,64,power_law_1.2,0.044198399782180785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,64,power_law_1.2,0.04747520089149475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,64,power_law_1.2,0.05503360033035278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,64,power_law_1.2,0.05563520193099976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,64,power_law_1.2,0.06109439730644226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,64,power_law_1.2,0.07432960271835327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,64,power_law_1.2,0.08892160058021545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,64,power_law_1.2,0.11187200546264649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,64,power_law_1.2,0.16206079721450806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,64,power_law_1.2,0.2301248073577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,64,power_law_1.2,0.261190390586853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,32,balanced,0.04806933303674062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,32,balanced,0.048010667165120445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,32,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,32,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,32,balanced,0.05197866757710775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,32,balanced,0.05579199890295664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,32,balanced,0.05579199890295664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,32,balanced,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,32,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,32,balanced,0.058117335041364036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,32,balanced,0.05949866771697998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,32,balanced,0.06078400214513143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,64,power_law_1.2,0.49370880126953126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,32,balanced,0.0639626681804657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,64,power_law_1.2,0.5966335773468018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,32,balanced,0.06846400101979573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,32,balanced,0.07833600044250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,32,balanced,0.07427200178305308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,32,balanced,0.07752533257007599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,32,balanced,0.08683733145395915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,32,balanced,0.0957493285338084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,32,balanced,0.1072746713956197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,32,balanced,0.12169599533081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,32,balanced,0.15359999736150107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,32,balanced,0.17484267552693686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,32,balanced,0.21863466501235962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,32,balanced,0.26469866434733075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,32,balanced,0.3693759838740031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,32,balanced,0.47162131468455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,64,power_law_1.2,1.032902431488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,32,power_law_1.2,0.09837440252304078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,32,power_law_1.2,0.08751999735832214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,32,power_law_1.2,0.08750079870223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,32,power_law_1.2,0.09352319836616516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,32,power_law_1.2,0.08617600202560424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,32,power_law_1.2,0.08981760144233704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,8,power_law_1.2,0.059539198875427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,32,power_law_1.2,0.09266560077667237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,8,power_law_1.2,0.07227519750595093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,32,power_law_1.2,0.0885312020778656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,8,power_law_1.2,0.08188160061836243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,32,power_law_1.2,0.0897599995136261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,8,power_law_1.2,0.08800640106201171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,8,power_law_1.2,0.1026304006576538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,8,power_law_1.2,0.11223679780960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,8,power_law_1.2,0.11910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,8,power_law_1.2,0.11879040002822876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,32,power_law_1.2,0.09177600145339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,8,power_law_1.2,0.11342719793319703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,32,power_law_1.2,0.09470080137252808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,8,power_law_1.2,0.11117440462112427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,32,power_law_1.2,0.097324800491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,8,power_law_1.2,0.11586560010910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,32,power_law_1.2,0.09720320105552674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,8,power_law_1.2,0.11095679998397827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,8,power_law_1.2,0.11818879842758179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,8,power_law_1.2,0.11921919584274292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,8,power_law_1.2,0.1317952036857605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,8,power_law_1.2,0.13770240545272827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,8,power_law_1.2,0.14018559455871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,64,power_law_1.2,1.3517375946044923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,8,power_law_1.2,0.15274239778518678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,8,power_law_1.2,0.17562240362167358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,8,power_law_1.2,0.20191359519958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,8,power_law_1.2,0.22922880649566652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,8,power_law_1.2,0.3064448118209839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,32,power_law_1.2,0.10323840379714966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,32,power_law_1.2,0.10919040441513062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,32,power_law_1.2,0.112716805934906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,32,power_law_1.2,0.1193343997001648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,32,power_law_1.2,0.14072959423065184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,32,power_law_1.2,0.15802240371704102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,8,power_law_1.2,0.35866880416870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,32,power_law_1.2,0.1951807975769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,8,power_law_1.2,0.4878528118133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,32,power_law_1.2,0.22259840965270997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,8,power_law_1.2,0.6249855995178223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,8,power_law_1.2,0.9099136352539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,8,power_law_1.2,1.1370816230773926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,32,power_law_1.2,0.2953216075897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,32,power_law_1.2,0.34724481105804444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,32,power_law_1.2,0.48837761878967284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,32,power_law_1.2,0.5292543888092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,32,power_law_1.2,0.8189503669738769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,32,power_law_1.2,1.017414379119873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,2,balanced,0.047194664676984154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,2,balanced,0.05421866476535797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,2,balanced,0.07425599793593089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,2,balanced,0.07459199925263722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,2,balanced,0.07495466868082683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,2,balanced,0.07328000168005626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,2,balanced,0.08760000268618266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,128,balanced,0.020453333854675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,128,balanced,0.019066666563351948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,2,balanced,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,2,balanced,0.07429333527882893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,2,balanced,0.07977066437403361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,2,balanced,0.07251733541488647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,2,balanced,0.07825600107510884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,2,balanced,0.07690133154392242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,2,balanced,0.07019199927647908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,2,balanced,0.07356266677379608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,2,balanced,0.07593066493670146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,2,balanced,0.07589866717656453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,2,balanced,0.1074133316675822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,2,balanced,0.09990400075912476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,128,balanced,0.019013332823912304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,2,balanced,0.12388267119725545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,128,balanced,0.02117866774400075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,2,balanced,0.14641599853833517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,128,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,128,balanced,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,128,balanced,0.03123733401298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,128,balanced,0.03136533250411352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,128,balanced,0.03142933299144109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,128,balanced,0.032831999162832894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,128,balanced,0.03197333216667175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,128,balanced,0.03331733246644338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,128,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,128,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,128,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,128,balanced,0.03643200049797694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,128,balanced,0.03756266583998998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,128,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,128,balanced,0.04153066625197729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,2,balanced,0.19274133443832397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,128,balanced,0.04836800197760264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,2,balanced,0.23804267247517905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,2,balanced,0.32873066266377765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,2,balanced,0.42110931873321533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,2,balanced,0.6082293192545573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,2,balanced,0.7978826363881429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,128,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,128,balanced,0.06379733482996623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,128,balanced,0.08082133531570435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,128,balanced,0.10494400064150493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,128,balanced,0.13170133034388223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,128,balanced,0.18157333135604858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,128,balanced,0.2330933411916097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,16,power_law_1.01,0.06167680025100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,16,power_law_1.01,0.07725440263748169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,16,power_law_1.01,0.07138559818267823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,16,power_law_1.01,0.07617279887199402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,16,power_law_1.01,0.07793279886245727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,16,power_law_1.01,0.07429760098457336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,16,power_law_1.01,0.07675520181655884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,16,power_law_1.01,0.07787520289421082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,16,power_law_1.01,0.07854719758033753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,16,power_law_1.01,0.07769600152969361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,16,power_law_1.01,0.07865599989891052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,16,power_law_1.01,0.081740802526474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,16,power_law_1.01,0.08365439772605895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,16,power_law_1.01,0.08357759714126586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,16,power_law_1.01,0.09673600196838379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,16,power_law_1.01,0.09839360117912292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,16,power_law_1.01,0.10226559638977051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,16,power_law_1.01,0.11827199459075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,16,power_law_1.01,0.12101759910583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,16,power_law_1.01,0.14789119958877564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,16,power_law_1.01,0.17066240310668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,16,power_law_1.01,0.2166912078857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,16,power_law_1.01,0.2659519910812378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,16,power_law_1.01,0.3924992084503174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,16,power_law_1.01,0.5210239887237549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,16,power_law_1.01,0.7761407852172851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,16,power_law_1.01,0.9369471549987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,4,balanced,0.0506933331489563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,4,balanced,0.05413866539796194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,4,balanced,0.06857066849867503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,4,balanced,0.09712533156077068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,4,balanced,0.13359466195106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,4,balanced,0.1649333337942759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,4,balanced,0.16471466422080994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,4,balanced,0.16476800044377646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,4,balanced,0.16833599408467612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,4,balanced,0.1665600041548411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,4,balanced,0.16914133230845133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,4,balanced,0.17332265774408975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,4,balanced,0.17473600308100382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,4,balanced,0.1774079998334249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,4,balanced,0.18638400236765543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,4,balanced,0.18890132506688437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,4,balanced,0.19991467396418253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,4,balanced,0.22825600703557333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,4,balanced,0.23715200026830038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,4,balanced,0.3163253267606099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,4,balanced,0.335258682568868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,4,balanced,0.5084693431854248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,4,balanced,0.4975200096766154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,4,balanced,0.7474880218505859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,4,balanced,0.8097173372904459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,4,balanced,1.1947840054829915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,4,balanced,1.4665600458780925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,16,power_law_1.2,0.050783997774124144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,16,power_law_1.2,0.04014720022678375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,16,power_law_1.2,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,16,power_law_1.2,0.04062080085277557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,16,power_law_1.2,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,16,power_law_1.2,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,16,power_law_1.2,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,16,power_law_1.2,0.04105600118637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,16,power_law_1.2,0.04142079949378967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,16,power_law_1.2,0.041868799924850465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,16,power_law_1.2,0.04174720048904419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,16,power_law_1.2,0.04261119961738587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,16,power_law_1.2,0.04322560131549835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,16,power_law_1.2,0.044710400700569156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,16,power_law_1.2,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,16,power_law_1.2,0.05334399938583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,16,power_law_1.2,0.056992000341415404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,16,power_law_1.2,0.060761600732803345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,16,power_law_1.2,0.06425600051879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,16,power_law_1.2,0.08236799836158752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,16,power_law_1.2,0.08584319949150085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,16,power_law_1.2,0.10002559423446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,16,power_law_1.2,0.1424064040184021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,16,power_law_1.2,0.1696320056915283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,64,power_law_1.01,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,64,power_law_1.01,0.046105599403381346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,64,power_law_1.01,0.045286399126052854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,16,power_law_1.2,0.19957120418548585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,64,power_law_1.01,0.04813440144062042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,64,power_law_1.01,0.049830400943756105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,16,power_law_1.2,0.2998208045959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,64,power_law_1.01,0.051641601324081424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,64,power_law_1.01,0.05267840027809143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,64,power_law_1.01,0.05167359709739685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,64,power_law_1.01,0.057843202352523805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,64,power_law_1.01,0.05869439840316772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,64,power_law_1.01,0.057766401767730714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,64,power_law_1.01,0.055430400371551516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,64,power_law_1.01,0.05581439733505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,64,power_law_1.01,0.06069759726524353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,64,power_law_1.01,0.06533759832382202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,64,power_law_1.01,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,64,power_law_1.01,0.06725119948387145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,64,power_law_1.01,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,64,power_law_1.01,0.08394880294799804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,64,power_law_1.01,0.0998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,64,power_law_1.01,0.11918720006942748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,64,power_law_1.01,0.15745919942855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,16,power_law_1.2,0.36764800548553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,64,power_law_1.01,0.196288001537323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,64,power_law_1.01,0.25409278869628904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,64,power_law_1.01,0.31256320476531985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,64,power_law_1.01,0.4816448211669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,64,power_law_1.01,0.6159615993499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,1,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,1,balanced,0.08166400094827016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,1,balanced,0.1362986663977305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,1,balanced,0.1386560002962748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,1,balanced,0.13411200046539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,1,balanced,0.13586666186650595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,1,balanced,0.1359946628411611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,1,balanced,0.13564800222714743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,1,balanced,0.1386613349119822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,1,balanced,0.1388266682624817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,1,balanced,0.14014933506647745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,1,balanced,0.1416320006052653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,1,balanced,0.14226667086283365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,1,balanced,0.1495626668135325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,1,balanced,0.1718133290608724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,1,balanced,0.17118932803471884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,1,balanced,0.17119999726613364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,1,balanced,0.20756266514460245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,1,balanced,0.21576533714930216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,1,balanced,0.2816320061683655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,1,balanced,0.30637866258621216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,1,balanced,0.405898650487264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,1,balanced,0.49454931418100995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,1,balanced,0.7000586986541748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,1,balanced,0.896117369333903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,1,balanced,1.3104106585184734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,1,balanced,1.7086933453877766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,32,power_law_1.01,0.10306559801101685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,32,power_law_1.01,0.0964352011680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,32,power_law_1.01,0.08974720239639282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,32,power_law_1.01,0.09393280148506164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,32,power_law_1.01,0.09312000274658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,32,power_law_1.01,0.0846015989780426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,32,power_law_1.01,0.09488639831542969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,32,power_law_1.01,0.09472640156745911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,32,power_law_1.01,0.09401599764823913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,32,power_law_1.01,0.09290879964828491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,32,power_law_1.01,0.09720320105552674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,32,power_law_1.01,0.09538559913635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,32,power_law_1.01,0.10184320211410522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,32,power_law_1.01,0.10158079862594604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,32,power_law_1.01,0.109990394115448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,32,power_law_1.01,0.11017600297927857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,32,power_law_1.01,0.11438720226287842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,32,power_law_1.01,0.13178240060806273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,32,power_law_1.01,0.1352895975112915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,32,power_law_1.01,0.1535871982574463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,32,power_law_1.01,0.1827072024345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,32,power_law_1.01,0.23678719997406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,32,power_law_1.01,0.24849278926849366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,32,power_law_1.01,0.3207871913909912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,32,power_law_1.01,0.4062784194946289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,32,power_law_1.01,0.5743423938751221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,32,power_law_1.01,0.6639296054840088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,32,balanced,0.05253866811593374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,32,balanced,0.05220800141493479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,32,balanced,0.05248000224431356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,32,balanced,0.058143998185793556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,32,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,32,balanced,0.06573333342870076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,32,balanced,0.07049599786599477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,32,balanced,0.07208533088366191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,32,balanced,0.07052266597747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,32,balanced,0.0726506660381953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,32,balanced,0.0711893339951833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,32,balanced,0.07258133093516032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,32,balanced,0.07670933504899342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,32,balanced,0.0765173335870107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,32,balanced,0.0835093359152476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,32,balanced,0.08517866333325703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,32,balanced,0.09113599856694539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,32,balanced,0.10541333754857381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,32,balanced,0.11901866396268208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,32,balanced,0.14526399970054626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,32,balanced,0.1643946667512258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,32,balanced,0.21013865868250528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,32,balanced,0.25249600410461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,32,balanced,0.34151466687520343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,32,balanced,0.43538665771484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,32,balanced,0.6063146591186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,32,balanced,0.784554640452067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,8,power_law_1.2,0.07612159848213196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,8,power_law_1.2,0.07578880190849305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,8,power_law_1.2,0.08268799781799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,8,power_law_1.2,0.0877568006515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,8,power_law_1.2,0.10610560178756714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,8,power_law_1.2,0.1041599988937378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,8,power_law_1.2,0.11456639766693115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,8,power_law_1.2,0.1162559986114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,8,power_law_1.2,0.11798399686813354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,8,power_law_1.2,0.12054400444030762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,8,power_law_1.2,0.12328959703445434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,8,power_law_1.2,0.12383359670639038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,8,power_law_1.2,0.12654720544815062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,8,power_law_1.2,0.12947200536727904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,8,power_law_1.2,0.14190080165863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,8,power_law_1.2,0.15163520574569703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,8,power_law_1.2,0.14754559993743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,8,power_law_1.2,0.17281919717788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,8,power_law_1.2,0.17842559814453124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,8,power_law_1.2,0.20453119277954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,8,power_law_1.2,0.2506239891052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,8,power_law_1.2,0.2760832071304321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,8,power_law_1.2,0.3092223882675171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,8,power_law_1.2,0.43903360366821287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,8,power_law_1.2,0.5623807907104492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,8,power_law_1.2,0.770304012298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,8,power_law_1.2,0.04748159945011139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,8,power_law_1.2,0.044460800290107724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,8,power_law_1.2,0.04659200012683869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,8,power_law_1.2,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,8,power_law_1.2,0.049439999461174014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,8,power_law_1.2,0.05486720204353333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,8,power_law_1.2,0.05324800014495849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,8,power_law_1.2,0.053887999057769774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,8,power_law_1.2,0.0536191999912262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,8,power_law_1.2,0.053497600555419925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,2,power_law_1.01,0.019731199741363524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,8,power_law_1.2,0.05512319803237915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,8,power_law_1.2,0.056435197591781616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,8,power_law_1.2,0.05715199708938599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,8,power_law_1.2,0.059564799070358276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,8,power_law_1.2,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,8,power_law_1.2,0.06369280219078063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,8,power_law_1.2,0.06695680022239685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,8,power_law_1.2,0.0748416006565094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,8,power_law_1.2,0.08369920253753663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,2,power_law_1.01,0.01963520050048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,8,power_law_1.2,0.09736319780349731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,2,power_law_1.01,0.019654400646686554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,8,power_law_1.2,0.11198079586029053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,8,power_law_1.2,0.14025599956512452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,8,power_law_1.2,0.18071680068969725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,2,power_law_1.01,0.01982080042362213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,2,power_law_1.01,0.02064639925956726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,2,power_law_1.01,0.020377600193023683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,2,power_law_1.01,0.02181120067834854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,2,power_law_1.01,0.02184319943189621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,2,power_law_1.01,0.02364159971475601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,2,power_law_1.01,0.023596799373626708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,2,power_law_1.01,0.026111999154090883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,2,power_law_1.01,0.04857600033283234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,8,power_law_1.2,0.21227519512176513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,2,power_law_1.01,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,2,power_law_1.01,0.05095679759979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,2,power_law_1.01,0.03787519931793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,2,power_law_1.01,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,2,power_law_1.01,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,2,power_law_1.01,0.0440064013004303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.01,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,8,power_law_1.2,0.3052864074707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.01,0.06640639901161194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.01,0.07549440264701843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.01,0.09953920245170593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.01,0.12461440563201905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,8,power_law_1.2,0.41546878814697263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,8,power_law_1.2,0.9393024444580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.01,0.1717568039894104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.01,0.21013760566711426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,2,power_law_1.2,0.07170559763908387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,2,power_law_1.2,0.08262400031089782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,2,power_law_1.2,0.08588799834251404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,2,power_law_1.2,0.10616960525512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,2,power_law_1.2,0.12108800411224366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,2,power_law_1.2,0.13601919412612914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,2,power_law_1.2,0.16219520568847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,2,power_law_1.2,0.1704767942428589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,2,power_law_1.2,0.1693567991256714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,2,power_law_1.2,0.17034239768981935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,2,power_law_1.2,0.16799999475479127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,2,power_law_1.2,0.17007360458374024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,8,balanced,0.04821866750717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,2,power_law_1.2,0.177183997631073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,8,balanced,0.04552533229192098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,2,power_law_1.2,0.18250240087509156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,8,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,8,balanced,0.05171733101209005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,2,power_law_1.2,0.1941248059272766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,8,balanced,0.05925333499908447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,8,balanced,0.08106133341789246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,8,balanced,0.10043199857076009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,2,power_law_1.2,0.19871360063552856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,8,balanced,0.10149866342544556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,8,balanced,0.1011253297328949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,2,power_law_1.2,0.20892798900604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,2,power_law_1.2,0.22580480575561523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,2,power_law_1.2,0.2553728103637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,2,power_law_1.2,0.29869439601898196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,8,balanced,0.10311466455459595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,8,balanced,0.1032373309135437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,8,balanced,0.1053013304869334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,8,balanced,0.10660800337791443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,8,balanced,0.10922132929166158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,8,balanced,0.1127306620279948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,8,balanced,0.11582400401433308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,8,balanced,0.12339199582735698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,8,balanced,0.13340800007184347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,8,balanced,0.14065066973368326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,8,balanced,0.16157333056131998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,8,balanced,0.17819732427597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,8,balanced,0.2153973380724589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,2,power_law_1.2,0.343507194519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,8,balanced,0.26763200759887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,8,balanced,0.3415786822636922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,8,balanced,0.4363466501235962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,8,balanced,0.5973866780598959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,8,balanced,0.756282647450765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.01,0.30629119873046873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,2,power_law_1.2,0.4357503890991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,2,power_law_1.2,0.5278656005859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,2,power_law_1.2,0.7584959983825683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,2,power_law_1.2,0.9232064247131347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,2,power_law_1.2,1.2863167762756347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,2,power_law_1.2,1.57456636428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,8,power_law_1.2,0.5828159809112549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.01,0.39271039962768556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,32,power_law_1.2,0.13237760066986085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,32,power_law_1.2,0.15843839645385743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,32,power_law_1.2,0.15791360139846802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,32,power_law_1.2,0.1562432050704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,16,balanced,0.04595200220743815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,16,balanced,0.04584000011285146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,16,balanced,0.049738665421803795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,16,balanced,0.0727893312772115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,16,balanced,0.09708266456921895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,16,balanced,0.14845333496729532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,16,balanced,0.20019733905792236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,16,balanced,0.20189867417017618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,16,balanced,0.20057600736618042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,16,balanced,0.2015786568323771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,16,balanced,0.2050559918085734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,16,balanced,0.20577067136764526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,16,balanced,0.20633065700531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,16,balanced,0.20839999119440714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,16,balanced,0.2121866742769877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,32,power_law_1.2,0.16120320558547974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,16,balanced,0.21630932887395224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,16,balanced,0.21820267041524252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,16,balanced,0.2257279952367147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,32,power_law_1.2,0.15070079565048217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,16,balanced,0.23596266905466715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,16,balanced,0.252949337164561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,32,power_law_1.2,0.1323904037475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,16,balanced,0.26945600907007855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,16,balanced,0.301530659198761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,32,power_law_1.2,0.14890880584716798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,32,power_law_1.2,0.1450495958328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,32,power_law_1.2,0.15000959634780883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,32,power_law_1.2,0.15695359706878662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,32,power_law_1.2,0.16197119951248168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,16,balanced,0.3345706860224406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,16,balanced,0.39318398634592694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,16,balanced,0.4787626663843791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,16,balanced,0.5959200064341227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,16,balanced,0.7672533194224039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,32,power_law_1.2,0.14892799854278566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,32,power_law_1.2,0.1661247968673706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,32,power_law_1.2,0.16361600160598755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,32,power_law_1.2,0.17870719432830812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,32,power_law_1.2,0.1801408052444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,32,power_law_1.2,0.1942080020904541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,32,power_law_1.2,0.20381441116333007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,32,power_law_1.2,0.2536767959594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,2,balanced,0.02363733450571696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,32,power_law_1.2,0.2898751974105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,2,balanced,0.02333866556485494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,2,balanced,0.023061332603295643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,64,power_law_1.01,0.02643199861049652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,2,balanced,0.027445333699385326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,2,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,32,power_law_1.2,0.35778560638427737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,2,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,64,power_law_1.01,0.026943999528884887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,32,power_law_1.2,0.44563841819763184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,64,power_law_1.01,0.02593280076980591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,64,power_law_1.01,0.026867198944091796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,32,power_law_1.2,0.6005184173583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,64,power_law_1.01,0.031667199730873105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,32,power_law_1.2,0.782316780090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,2,balanced,0.03160000095764796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,2,balanced,0.03373866776625315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,2,balanced,0.04121600091457367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,32,power_law_1.2,1.0105792045593263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,32,power_law_1.2,1.525817584991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,64,power_law_1.01,0.04227840006351471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,64,power_law_1.01,0.05672320127487183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,2,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,64,power_law_1.01,0.05860480070114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,2,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,2,balanced,0.07645866771539052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,2,balanced,0.07668266693751018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,2,balanced,0.09242133299509685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,2,balanced,0.06593599915504456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,2,balanced,0.05765866736570994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,2,balanced,0.0649599979321162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,2,balanced,0.0848640004793803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,2,balanced,0.10107733805974324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,2,balanced,0.14032000303268433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,2,balanced,0.17459199825922647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,2,balanced,0.24316267172495523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,64,power_law_1.01,0.07313920259475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,2,balanced,0.303274671236674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,2,balanced,0.43906132380167645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,2,balanced,0.5657973289489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,2,balanced,0.8308533032735189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,2,balanced,1.092901309331258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,64,power_law_1.01,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,64,power_law_1.01,0.07137920260429383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,64,power_law_1.01,0.06779519915580749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,64,power_law_1.01,0.0689791977405548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,64,power_law_1.01,0.06656640172004699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,64,power_law_1.01,0.06217600107192993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,64,power_law_1.01,0.05996800065040588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,64,power_law_1.01,0.07148799896240235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,64,power_law_1.01,0.06907520294189454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,64,power_law_1.01,0.08170239925384522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,64,power_law_1.01,0.09018880128860474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,64,power_law_1.01,0.108787202835083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,64,power_law_1.01,0.13014400005340576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,64,power_law_1.01,0.18042880296707153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,64,power_law_1.01,0.2510143995285034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,64,power_law_1.01,0.3195519924163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,64,power_law_1.01,0.4679103851318359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,64,power_law_1.01,0.5997759819030761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,4,balanced,0.05253866811593374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,4,balanced,0.05825066566467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,4,balanced,0.07115733126799266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,4,balanced,0.10008000334103902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,4,balanced,0.1546026666959127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,4,balanced,0.15466666221618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,4,balanced,0.15621333320935568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,4,balanced,0.15618667006492615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,4,balanced,0.15611199537913004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,4,balanced,0.15621333320935568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,4,balanced,0.15643733739852905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,4,balanced,0.16135467092196146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,4,balanced,0.16154133280118307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,4,balanced,0.16292267044385275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,4,balanced,0.16802666584650675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,4,balanced,0.16988267501195273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,4,balanced,0.17038933436075845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,4,balanced,0.1835199991861979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,4,balanced,0.19183466831843057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,4,balanced,0.2150719960530599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,4,balanced,0.2278826634089152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,4,balanced,0.30551467339197796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,4,balanced,0.3447893460591634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,4,balanced,0.47197333971659344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,4,balanced,0.6157866716384888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,4,balanced,0.8780053456624349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,4,balanced,1.1530240376790364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,32,8,power_law_1.01,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,32,8,power_law_1.01,0.0748799979686737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,32,8,power_law_1.01,0.07542399764060974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,32,8,power_law_1.01,0.07955200076103211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,32,8,power_law_1.01,0.08423680067062378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,32,8,power_law_1.01,0.08528000116348267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,32,8,power_law_1.01,0.088646399974823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,32,8,power_law_1.01,0.08897280097007751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,32,8,power_law_1.01,0.0899071991443634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,32,8,power_law_1.01,0.09114239811897278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,32,8,power_law_1.01,0.08959360122680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,32,8,power_law_1.01,0.09333119988441467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,32,8,power_law_1.01,0.09679359793663025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,32,8,power_law_1.01,0.09386879801750184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,32,8,power_law_1.01,0.10017919540405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,32,8,power_law_1.01,0.10005760192871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,32,8,power_law_1.01,0.1077888011932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,32,8,power_law_1.01,0.11941759586334229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,32,8,power_law_1.01,0.13363200426101685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,32,8,power_law_1.01,0.15349119901657104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,32,8,power_law_1.01,0.18053760528564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,32,8,power_law_1.01,0.22013440132141113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,16,power_law_1.2,0.07896320223808288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,32,8,power_law_1.01,0.276857590675354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,16,power_law_1.2,0.10297600030899048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,32,8,power_law_1.01,0.36075520515441895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,32,8,power_law_1.01,0.4624447822570801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,32,8,power_law_1.01,0.6568448066711425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,32,8,power_law_1.01,0.8225343704223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,16,power_law_1.2,0.09386240243911743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,16,power_law_1.2,0.098854398727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,16,power_law_1.2,0.10366719961166382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,16,power_law_1.2,0.08268799781799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,16,power_law_1.2,0.10176000595092774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,16,power_law_1.2,0.09578239917755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,16,power_law_1.2,0.09459840059280396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,16,power_law_1.2,0.09420160055160523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,16,power_law_1.2,0.09793919920921326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,16,power_law_1.2,0.10095360279083251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,16,power_law_1.2,0.09623680114746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,16,power_law_1.2,0.09979519844055176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,16,power_law_1.2,0.1081663966178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,16,power_law_1.2,0.10858240127563476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,16,power_law_1.2,0.11349120140075683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,16,power_law_1.2,0.12570879459381104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,16,power_law_1.2,0.13100800514221192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,16,power_law_1.2,0.1557055950164795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,16,power_law_1.2,0.1801151990890503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,16,power_law_1.2,0.21056001186370848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,1,power_law_1.01,0.09466239809989929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,16,power_law_1.2,0.2567487955093384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,1,power_law_1.01,0.10651520490646363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,16,power_law_1.2,0.34371840953826904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,1,power_law_1.01,0.12997759580612184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,16,power_law_1.2,0.4367743968963623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,1,power_law_1.01,0.18732160329818726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,16,power_law_1.2,0.6218111991882325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,1,power_law_1.01,0.22770559787750244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,1,power_law_1.01,0.26404480934143065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,16,power_law_1.2,0.7713664054870606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,1,power_law_1.01,0.3434880018234253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,8,balanced,0.03156266609827677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,8,balanced,0.03141333411137263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,1,power_law_1.01,0.35331199169158933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,8,balanced,0.033615998923778534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,8,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,8,balanced,0.0782293329636256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,1,power_law_1.01,0.36463360786437987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,1,power_law_1.01,0.3630784034729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,1,power_law_1.01,0.3777087926864624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,1,power_law_1.01,0.38559999465942385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,1,power_law_1.01,0.39983999729156494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,1,power_law_1.01,0.40033278465270994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,8,balanced,0.07945600152015686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,8,balanced,0.0787360022465388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,8,balanced,0.07818666597207387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,8,balanced,0.07964266836643219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,8,balanced,0.08031466603279114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,8,balanced,0.08028266827265422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,8,balanced,0.08269866804281871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,8,balanced,0.08212266862392426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,8,balanced,0.08386133114496867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,8,balanced,0.08850666880607605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,8,balanced,0.08869866530100505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,8,balanced,0.08858666817347209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,8,balanced,0.09362133344014485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,8,balanced,0.10044800241788228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,8,balanced,0.10692800084749858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,8,balanced,0.11828800042470296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,8,balanced,0.14006400108337402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,1,power_law_1.01,0.4139967918395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,8,balanced,0.1530400017897288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,8,balanced,0.19470399618148804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,1,power_law_1.01,0.41918721199035647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,8,balanced,0.2442880074183146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,8,balanced,0.3200533390045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,1,power_law_1.01,0.44021759033203123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,8,balanced,0.3875519831975301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,1,power_law_1.01,0.4928895950317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,1,power_law_1.01,0.5323328018188477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,1,power_law_1.01,0.6241151809692382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,1,power_law_1.01,0.7049791812896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,1,power_law_1.01,0.9029439926147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,1,power_law_1.01,1.0750016212463378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,1,power_law_1.01,1.4461119651794434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,1,power_law_1.01,1.8168256759643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,1,power_law_1.01,2.5410688400268553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,1,power_law_1.01,3.2732990264892576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,8,power_law_1.2,0.142739200592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,8,power_law_1.2,0.23168001174926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,8,power_law_1.2,0.2008512020111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,8,power_law_1.2,0.29171199798583985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,8,power_law_1.2,0.38255360126495364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,8,power_law_1.2,0.3962176084518433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,8,power_law_1.2,0.3795775890350342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,8,power_law_1.2,0.35362560749053956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,8,power_law_1.2,0.34816000461578367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,8,power_law_1.2,0.39612159729003904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,8,power_law_1.2,0.39032320976257323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,8,power_law_1.2,0.3879744052886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,8,power_law_1.2,0.4184000015258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,8,power_law_1.2,0.4239232063293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,8,power_law_1.2,0.44683518409729006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,8,power_law_1.2,0.4452159881591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,8,power_law_1.2,0.477459192276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,8,power_law_1.2,0.5322624206542969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,8,power_law_1.2,0.5753471851348877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,8,power_law_1.2,0.6303872108459473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,16,power_law_1.2,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,16,power_law_1.2,0.050963199138641356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,8,power_law_1.2,0.655679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,16,power_law_1.2,0.05002239942550659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,16,power_law_1.2,0.04302720129489899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,16,power_law_1.2,0.044684800505638125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,16,power_law_1.2,0.05443199872970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,16,power_law_1.2,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,16,power_law_1.2,0.05692160129547119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,16,power_law_1.2,0.05598719716072083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,16,power_law_1.2,0.05822719931602478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,16,power_law_1.2,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,8,power_law_1.2,0.716864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,8,power_law_1.2,0.8307647705078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,8,power_law_1.2,1.107801628112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,8,power_law_1.2,1.2704575538635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,16,power_law_1.2,0.060575997829437254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,16,power_law_1.2,0.061715197563171384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,16,power_law_1.2,0.06599040031433105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,8,power_law_1.2,1.8060672760009766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,16,power_law_1.2,0.07204480171203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,16,power_law_1.2,0.07664639949798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,16,power_law_1.2,0.08111360073089599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,16,power_law_1.2,0.10050560235977173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,8,power_law_1.2,2.103647994995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,16,power_law_1.2,0.10087679624557495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,16,power_law_1.2,0.12505600452423096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,16,power_law_1.2,0.14339840412139893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,16,power_law_1.2,0.18707200288772582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,16,power_law_1.2,0.23938560485839844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,16,power_law_1.2,0.3174015998840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,16,power_law_1.2,0.41010560989379885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,16,power_law_1.2,0.5897664070129395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,128,balanced,0.029663999875386555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,128,balanced,0.02717866748571396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,128,balanced,0.027077332139015198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,128,balanced,0.02752000093460083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,128,balanced,0.029498666524887085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,128,balanced,0.030074665943781536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,128,balanced,0.02994133283694585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,128,balanced,0.031258667508761086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,128,balanced,0.03123733401298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,128,balanced,0.0308693324526151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,128,balanced,0.03162133445342382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,128,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,128,balanced,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,128,balanced,0.03355200091997782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,128,balanced,0.033557333052158356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,128,balanced,0.0364479993780454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,128,balanced,0.036015999813874565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,128,balanced,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,128,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,128,balanced,0.043418665726979576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,128,balanced,0.04368533194065094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,128,balanced,0.04618666569391886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,128,balanced,0.04820266862710317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,128,balanced,0.06358933448791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,128,balanced,0.08066133161385854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,128,balanced,0.11337066690127055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,8,balanced,0.04388799766699473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,128,balanced,0.13864533106486002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,8,balanced,0.04350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,8,balanced,0.0436106671889623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,8,balanced,0.049642667174339294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,8,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,8,balanced,0.05646933118502299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,8,balanced,0.0581279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,8,balanced,0.059258664647738137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,8,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,8,balanced,0.059861332178115845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,8,balanced,0.06018133461475372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,8,balanced,0.06033066908518473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,8,balanced,0.061941335598627724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,8,balanced,0.06400000055631001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,8,balanced,0.06651199857393901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,8,balanced,0.06837333242098491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,8,balanced,0.07017066578070323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,8,balanced,0.08243200182914734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,8,balanced,0.08218133449554443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,8,balanced,0.11451733112335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,8,balanced,0.12541866302490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,8,balanced,0.16871466239293417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,8,balanced,0.1818079948425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,8,balanced,0.2604106664657593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,8,balanced,0.2958933313687642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,8,balanced,0.40706666310628253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,8,balanced,0.5049226681391398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,4,power_law_1.2,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,4,power_law_1.2,0.07838079929351807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,4,power_law_1.2,0.0698303997516632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,4,power_law_1.2,0.07745280265808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,4,power_law_1.2,0.07761920094490052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,4,power_law_1.2,0.07800959944725036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,4,power_law_1.2,0.0794431984424591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,4,power_law_1.2,0.08639360070228577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,4,power_law_1.2,0.08650239706039428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,4,power_law_1.2,0.09077119827270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,4,power_law_1.2,0.08768640160560608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,4,power_law_1.2,0.09740800261497498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,4,power_law_1.2,0.09667840003967285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,4,power_law_1.2,0.0963648021221161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,4,power_law_1.2,0.06880639791488648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,4,power_law_1.2,0.12015360593795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,4,power_law_1.2,0.07758079767227173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,4,power_law_1.2,0.1195904016494751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,4,power_law_1.2,0.09050880074501037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,4,power_law_1.2,0.13969919681549073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,4,power_law_1.2,0.10777599811553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,4,power_law_1.2,0.11976319551467896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,4,power_law_1.2,0.153056001663208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,4,power_law_1.2,0.16026240587234497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,4,power_law_1.2,0.1511615991592407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,4,power_law_1.2,0.1506880044937134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,4,power_law_1.2,0.19051519632339478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,4,power_law_1.2,0.1586624026298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,4,power_law_1.2,0.1560960054397583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,4,power_law_1.2,0.15858559608459472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,4,power_law_1.2,0.16378240585327147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,4,power_law_1.2,0.17133439779281617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,4,power_law_1.2,0.17820160388946532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,4,power_law_1.2,0.1875264048576355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,4,power_law_1.2,0.19418879747390747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,4,power_law_1.2,0.23998079299926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,4,power_law_1.2,0.20364799499511718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,4,power_law_1.2,0.23962879180908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,4,power_law_1.2,0.2602112054824829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,4,power_law_1.2,0.3289151906967163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,4,power_law_1.2,0.4010303974151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,4,power_law_1.2,0.5157631874084473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,4,power_law_1.2,0.26080000400543213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,4,power_law_1.2,0.5917632102966308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,4,power_law_1.2,0.7765183925628663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,4,power_law_1.2,0.9489472389221192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,16,power_law_1.2,0.7094848155975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,4,power_law_1.2,1.3394047737121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,4,power_law_1.2,0.3894592046737671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,4,power_law_1.2,0.4260608196258545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,4,power_law_1.2,1.8181760787963868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,4,power_law_1.2,0.6068607807159424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,4,power_law_1.2,0.9527935981750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,4,power_law_1.2,1.2628735542297362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,128,power_law_1.01,0.019705599546432494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,128,power_law_1.01,0.02229759991168976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,128,power_law_1.01,0.019494399428367615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,128,power_law_1.01,0.022252799570560457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,128,power_law_1.01,0.021004800498485566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,128,power_law_1.01,0.026438400149345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,128,power_law_1.01,0.02558720111846924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,128,power_law_1.01,0.041126400232315063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,128,power_law_1.01,0.04081920087337494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,128,power_law_1.01,0.04032639861106872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,128,power_law_1.01,0.04140160083770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,128,power_law_1.01,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,128,power_law_1.01,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,128,power_law_1.01,0.04110719859600067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,128,power_law_1.01,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,128,power_law_1.01,0.03930239975452423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,128,power_law_1.01,0.04490880072116852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,128,power_law_1.01,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,128,power_law_1.01,0.0515392005443573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,128,power_law_1.01,0.06088320016860962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,128,power_law_1.01,0.078438401222229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,128,power_law_1.01,0.10037120580673217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,4,power_law_1.2,1.4872511863708495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,128,power_law_1.01,0.12706559896469116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,128,power_law_1.01,0.1631168007850647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,128,power_law_1.01,0.21338241100311278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,128,power_law_1.01,0.2923264026641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,128,power_law_1.01,0.38738560676574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,1,balanced,0.09750399986902873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,1,balanced,0.14698666334152222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,1,balanced,0.24583999315897623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,1,balanced,0.45079998175303143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,1,balanced,0.8439573446909586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,1,balanced,1.6328746477762859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,1,balanced,2.4168532689412436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,1,balanced,2.4420533180236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,1,balanced,2.448197364807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,1,balanced,2.4539519945780435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,1,balanced,2.443023999532064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,1,balanced,2.44595734278361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,1,balanced,2.4606293042500815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,1,balanced,2.4723946253458657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,1,balanced,2.493562698364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,1,balanced,2.505839983622233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,1,balanced,2.535168011983236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,1,balanced,2.589200019836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,1,balanced,2.6395626068115234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,64,balanced,0.059263999263445534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,64,balanced,0.059978668888409935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,64,balanced,0.05991999804973602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,64,balanced,0.06233066817124685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,64,balanced,0.06435200075308482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,64,balanced,0.07640000184377034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,1,balanced,2.77510929107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,64,balanced,0.08815466364224751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,64,balanced,0.0885759989420573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,64,balanced,0.08599467078844707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,64,balanced,0.0879306693871816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,64,balanced,0.08896000186602275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,64,balanced,0.09021866321563721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,64,balanced,0.09301333626111348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,1,balanced,2.8510986963907876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,64,balanced,0.09072533249855042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,64,balanced,0.09532266855239868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,64,balanced,0.09702400366465251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,64,balanced,0.10249599814414978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,64,balanced,0.10947733124097188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,64,balanced,0.11515200138092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,64,balanced,0.13119999567667642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,1,balanced,3.062704086303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,64,balanced,0.14386666814486185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,64,balanced,0.17274133364359537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,64,balanced,0.20016000668207803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,64,balanced,0.2573759953180949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,64,balanced,0.31627732515335083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,1,balanced,3.2494398752848306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,64,balanced,0.423306663831075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,64,balanced,0.5353493293126425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,1,balanced,3.62334410349528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,128,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,128,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,128,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,128,balanced,0.037776000797748566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,128,balanced,0.03868266691764196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,128,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,128,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,128,balanced,0.039306665460268654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,128,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,128,balanced,0.039834665755430855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,128,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,128,balanced,0.040048000713189445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,128,balanced,0.03967999915281931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,128,balanced,0.04168533285458883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,128,balanced,0.04462933540344238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,128,balanced,0.043807998299598694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,128,balanced,0.04497066636880239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,128,balanced,0.045978665351867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,128,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,128,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,128,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,128,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,128,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,1,balanced,4.194181442260742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,1,balanced,4.95414384206136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,128,balanced,0.0643146683772405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,128,balanced,0.0707946668068568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,128,balanced,0.08760533730189006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,128,balanced,0.10377599795659383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,1,balanced,6.70904541015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,8,balanced,0.048453330993652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,8,balanced,0.05316799879074097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,8,balanced,0.06817066669464111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,8,balanced,0.09846933682759602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,8,balanced,0.1320266624291738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,8,balanced,0.15782933433850607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,8,balanced,0.1600213348865509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,8,balanced,0.1625279982884725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,8,balanced,0.16432000199953714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,8,balanced,0.1623199979464213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,8,balanced,0.16155200203259787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,8,balanced,0.16548800468444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,8,balanced,0.1644319991270701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,8,balanced,0.16781866550445557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,8,balanced,0.17573332786560059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,8,balanced,0.17523199319839478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,8,balanced,0.18247467279434204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,8,balanced,0.20939733584721884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,8,balanced,0.21107733249664307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,8,balanced,0.28892799218495685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,1,power_law_1.01,0.08414720296859741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,8,balanced,0.2765866716702779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,8,balanced,0.3708639939626058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,8,balanced,0.36500799655914307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,8,balanced,0.508458654085795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,8,balanced,0.5657120148340861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,8,balanced,0.8098506927490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,8,balanced,0.9468533198038737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,1,power_law_1.01,0.11149439811706544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,1,power_law_1.01,0.1608896017074585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,1,power_law_1.01,0.24832639694213868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,1,power_law_1.01,0.41096959114074705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,1,power_law_1.01,0.5581888198852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,1,power_law_1.01,0.6504767894744873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,1,power_law_1.01,0.8413824081420899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,1,power_law_1.01,0.8736384391784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,1,power_law_1.01,0.9073599815368653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,1,power_law_1.01,0.9351936340332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,1,power_law_1.01,0.9845696449279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,1,power_law_1.01,0.9992639541625976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,1,power_law_1.01,1.036736011505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,1,power_law_1.01,1.0700480461120605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,4,balanced,0.02345066765944163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,4,balanced,0.023157333334287006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,4,balanced,0.025221332907676697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,4,balanced,0.029322666426499683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,4,balanced,0.0320266659061114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,4,balanced,0.04423466821511587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,4,balanced,0.04726399978001913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,4,balanced,0.049914668003718056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,4,balanced,0.04889066517353058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,4,balanced,0.05126399795214335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,4,balanced,0.05306133131186167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,4,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,4,balanced,0.057578667998313904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,4,balanced,0.06058666606744131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,4,balanced,0.0739573339621226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,4,balanced,0.07634666562080383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,4,balanced,0.08075733482837677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,4,balanced,0.12133866548538208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,4,balanced,0.1278986632823944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,4,balanced,0.14406399925549826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,1,power_law_1.01,1.0949888229370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,4,balanced,0.1616320013999939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,4,balanced,0.19226133823394775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,4,balanced,0.22068800528844199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,1,power_law_1.01,1.1456000328063964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,4,balanced,0.2781706651051839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,4,balanced,0.3201706608136495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,4,balanced,0.49958399931589764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,1,power_law_1.01,1.229759979248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,4,balanced,0.6088266770044962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,1,power_law_1.01,1.305388832092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,1,power_law_1.01,1.5114432334899903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,1,power_law_1.01,1.5395903587341309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,1,power_law_1.01,1.8420927047729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,1,power_law_1.01,2.0837888717651367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,1,power_law_1.01,2.6000768661499025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,1,power_law_1.01,2.9096832275390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,1,power_law_1.01,3.728460693359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,1,power_law_1.01,5.019449615478516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,32,balanced,0.019274666905403137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,32,balanced,0.01893866683046023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,32,balanced,0.018922666708628338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,32,balanced,0.019023999571800232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,32,balanced,0.019018666197856266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,32,balanced,0.023205332458019257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,32,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,32,balanced,0.039306665460268654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,32,balanced,0.032842665910720825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,32,balanced,0.03179733455181122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,32,balanced,0.03170666595300039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,32,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,32,balanced,0.03327466547489166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,32,balanced,0.033557333052158356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,32,balanced,0.03804266701141993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,32,balanced,0.040048000713189445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,32,balanced,0.03956799954175949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,32,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,32,balanced,0.04625066618124644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,32,balanced,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,32,balanced,0.06422399977842967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,32,balanced,0.08709866801897685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,32,balanced,0.10248532891273499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,32,balanced,0.13988799850145975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,32,balanced,0.17705599466959634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,32,balanced,0.25083200136820477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,32,balanced,0.3188213308652242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,2,power_law_1.2,0.01703680008649826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,2,power_law_1.2,0.017164799571037292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,2,power_law_1.2,0.016518400609493257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,16,balanced,0.01738133281469345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,2,power_law_1.2,0.01759359985589981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,16,balanced,0.01718933383623759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,16,balanced,0.01728533332546552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,16,balanced,0.017136000096797943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,2,power_law_1.2,0.018156799674034118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,16,balanced,0.0191040001809597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,16,balanced,0.018976000448067982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,16,balanced,0.021183999876181286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,16,balanced,0.0210506667693456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,16,balanced,0.02319466571013133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,16,balanced,0.021984001000722248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,16,balanced,0.02367999901374181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,16,balanced,0.025279998779296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,16,balanced,0.02526933451493581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,16,balanced,0.03295466552178065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,16,balanced,0.03693866729736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,16,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,16,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,16,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,16,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,16,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,16,balanced,0.05548266569773356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,16,balanced,0.07038400073846181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,16,balanced,0.08158933122952779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,16,balanced,0.11227200428644817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,2,power_law_1.2,0.01834239959716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,16,balanced,0.13759467005729675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,2,power_law_1.2,0.019449600577354433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,16,balanced,0.19221333662668863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,2,power_law_1.2,0.019116799533367156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,2,power_law_1.2,0.020153599977493285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,2,power_law_1.2,0.020351999998092653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,2,power_law_1.2,0.021529600024223328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,2,power_law_1.2,0.023129600286483764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,2,power_law_1.2,0.02712959945201874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,2,power_law_1.2,0.02539519965648651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,16,balanced,0.24702399969100952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,2,power_law_1.2,0.026649600267410277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,2,power_law_1.2,0.028780800104141236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,2,power_law_1.2,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,2,power_law_1.2,0.03894400000572205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.2,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.2,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.2,0.05985280275344849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.2,0.07861120104789734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.2,0.09583359956741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.2,0.12572159767150878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.2,0.15916160345077515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.2,0.22225921154022216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,32,power_law_1.2,0.06538239717483521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,32,power_law_1.2,0.06235520243644714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,32,power_law_1.2,0.0596671998500824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,32,power_law_1.2,0.061299198865890504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,32,power_law_1.2,0.060703998804092406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,32,power_law_1.2,0.05987200140953064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,32,power_law_1.2,0.06480640172958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,32,power_law_1.2,0.06391680240631104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,32,power_law_1.2,0.0640064001083374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,32,power_law_1.2,0.06433280110359192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,32,power_law_1.2,0.06543999910354614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,32,power_law_1.2,0.06603519916534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,32,power_law_1.2,0.06776959896087646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,32,power_law_1.2,0.06813439726829529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,32,power_law_1.2,0.07356799840927124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,32,power_law_1.2,0.07260159850120544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,32,power_law_1.2,0.07869439721107482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.2,0.28458240032196047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,32,power_law_1.2,0.08718079924583436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,32,power_law_1.2,0.09552000164985656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,32,power_law_1.2,0.1168768048286438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,32,power_law_1.2,0.13727999925613404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,32,power_law_1.2,0.17277439832687377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,32,power_law_1.2,0.22423040866851807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,32,power_law_1.2,0.28049280643463137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,32,power_law_1.2,0.36487040519714353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,32,power_law_1.2,0.5391168117523193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,32,power_law_1.2,0.6901760101318359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,4,power_law_1.01,0.047865599393844604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,4,power_law_1.01,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,4,power_law_1.01,0.05025920271873474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,4,power_law_1.01,0.06044800281524658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,4,power_law_1.01,0.056729602813720706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,4,power_law_1.01,0.05602560043334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,4,power_law_1.01,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,4,power_law_1.01,0.05637120008468628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,4,power_law_1.01,0.05563520193099976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,4,power_law_1.01,0.055615997314453124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,4,power_law_1.01,0.05554559826850891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,4,power_law_1.01,0.059724801778793336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,4,power_law_1.01,0.06121600270271301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,4,power_law_1.01,0.06195840239524841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,4,power_law_1.01,0.06862080097198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,4,power_law_1.01,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,4,power_law_1.01,0.07262719869613647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,4,power_law_1.01,0.08906880021095276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.01,0.10173439979553223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.01,0.12715519666671754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,8,balanced,0.047914668917655945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,8,balanced,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,8,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,8,balanced,0.0562720000743866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,8,balanced,0.062128002444903054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,8,balanced,0.07620800038178761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,8,balanced,0.07640000184377034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,8,balanced,0.07640533149242401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.01,0.1459328055381775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,8,balanced,0.07633066674073537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,8,balanced,0.07824000219504039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,8,balanced,0.07768533130486806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,8,balanced,0.0806826651096344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,8,balanced,0.08071466783682506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,8,balanced,0.08057599763075511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,8,balanced,0.08872000376383464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,8,balanced,0.09081066648165385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,8,balanced,0.09471467137336731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,8,balanced,0.10534399747848511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,8,balanced,0.11520000298817952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,8,balanced,0.13804266850153604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,8,balanced,0.15365333358446756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,8,balanced,0.1919893423716227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,8,balanced,0.23413334290186563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,8,balanced,0.3223466674486796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,8,balanced,0.3990186850229899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.01,0.21021440029144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,8,balanced,0.5640373229980469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,8,balanced,0.7214986483256022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.01,0.23991680145263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.01,0.33600640296936035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,2,power_law_1.01,0.04633600115776062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,2,power_law_1.01,0.05082240104675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,2,power_law_1.01,0.05366399884223938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.01,0.4592895984649658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,2,power_law_1.01,0.06684160232543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,2,power_law_1.01,0.07429119944572449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,2,power_law_1.01,0.09567999839782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,2,power_law_1.01,0.10006400346755981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,2,power_law_1.01,0.10216959714889526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,2,power_law_1.01,0.10572799444198608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,2,power_law_1.01,0.1098688006401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,2,power_law_1.01,0.1125823974609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,2,power_law_1.01,0.11757440567016601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,2,power_law_1.01,0.12366080284118652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,2,power_law_1.01,0.1246783971786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,2,power_law_1.01,0.13519359827041627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,2,power_law_1.01,0.14295680522918702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,2,power_law_1.01,0.15456639528274535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,2,power_law_1.01,0.18315520286560058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,2,power_law_1.01,0.21109120845794677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,2,power_law_1.01,0.2581376075744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,2,power_law_1.01,0.3130111932754517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,2,power_law_1.01,0.4168704032897949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,2,power_law_1.01,0.5319551944732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,2,power_law_1.01,0.7224703788757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,2,power_law_1.01,1.0238143920898437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,2,power_law_1.01,1.3525440216064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,2,power_law_1.01,1.7109312057495116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.01,0.6856639862060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.01,0.8773056030273437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,2,power_law_1.2,0.1054144024848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,2,power_law_1.2,0.13171199560165406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,2,power_law_1.2,0.17057280540466307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,2,power_law_1.2,0.23826560974121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,2,power_law_1.2,0.2781440019607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,2,power_law_1.2,0.4825727939605713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,2,power_law_1.2,0.5099584102630615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,2,power_law_1.2,0.5283584117889404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,2,power_law_1.2,0.5375167846679687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,2,power_law_1.2,0.5213376045227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,2,power_law_1.2,0.5649983882904053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,2,power_law_1.2,0.5857215881347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,2,power_law_1.2,0.5853824138641357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,2,power_law_1.2,0.5808320045471191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,2,power_law_1.2,0.627891206741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,2,power_law_1.2,0.6471615791320801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,64,power_law_1.01,0.16152960062026978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,2,power_law_1.2,0.7019775867462158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,64,power_law_1.01,0.15852799415588378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,2,power_law_1.2,0.7808320045471191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,64,power_law_1.01,0.156550395488739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,2,power_law_1.2,0.7588287830352783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,64,power_law_1.01,0.15990400314331055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,2,power_law_1.2,0.916256046295166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,64,power_law_1.01,0.1542847990989685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,2,power_law_1.2,0.9559167861938477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,64,power_law_1.01,0.14087040424346925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,64,power_law_1.01,0.1341375946998596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,2,power_law_1.2,1.1596799850463868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,64,power_law_1.01,0.16588799953460692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,2,power_law_1.2,1.3473024368286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,64,power_law_1.01,0.15263999700546266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,64,power_law_1.01,0.15111680030822755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,64,power_law_1.01,0.16799360513687134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,64,power_law_1.01,0.16429439783096314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,64,power_law_1.01,0.15767040252685546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,64,power_law_1.01,0.16486400365829468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,2,power_law_1.2,1.714841651916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,64,power_law_1.01,0.15942399501800536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,64,power_law_1.01,0.16985599994659423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,64,power_law_1.01,0.17854080200195313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,64,power_law_1.01,0.18931200504302978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,64,power_law_1.01,0.19048320055007933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,64,power_law_1.01,0.21420800685882568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,64,power_law_1.01,0.24646399021148682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,64,power_law_1.01,0.30492799282073973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,64,power_law_1.01,0.36296958923339845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,64,power_law_1.01,0.49763197898864747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,2,power_law_1.2,2.4949056625366213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,128,balanced,0.02757333219051361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,64,power_law_1.01,0.6141119956970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,64,power_law_1.01,0.8945343971252442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,64,power_law_1.01,1.0761024475097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,128,balanced,0.027237333357334137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,128,balanced,0.027221334477265675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,128,balanced,0.02945599953333537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,128,balanced,0.029994666576385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,2,power_law_1.2,3.28287353515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,128,balanced,0.029658667743206024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,128,balanced,0.030965333183606465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,128,balanced,0.03170666595300039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,128,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,128,balanced,0.03143466760714849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,2,power_law_1.2,3.5535167694091796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,128,balanced,0.031717332700888314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,128,balanced,0.033786666889985405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,128,balanced,0.03365333378314972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,128,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,128,balanced,0.03991466760635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,128,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,128,balanced,0.04014399896065394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,128,balanced,0.04364266494909922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,128,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,128,balanced,0.05188799897829691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,128,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,128,balanced,0.06637866795063019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,128,balanced,0.07444799939791362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,128,balanced,0.09513599673906963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,128,balanced,0.11344533165295918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,128,balanced,0.15016000469525656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,128,balanced,0.1890453298886617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,64,balanced,0.043706665436426796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,64,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,64,balanced,0.03781333317359289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,64,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,64,balanced,0.041493333876132965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,64,balanced,0.04384533564249674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,64,balanced,0.04390933116277059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,64,balanced,0.04562666515509287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,64,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,64,balanced,0.04684266448020935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,64,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,64,balanced,0.047637333472569786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,64,balanced,0.049882665276527405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,64,balanced,0.04987733562787374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,64,balanced,0.05416533350944519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,64,balanced,0.05561066667238871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,64,balanced,0.057850668827692665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,64,balanced,0.06196799874305725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,64,balanced,0.0664160003264745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,64,balanced,0.07471466561158498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,64,balanced,0.08140799899895985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,64,balanced,0.1029813289642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,64,balanced,0.12038933237393697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,64,balanced,0.17867734034856161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,64,balanced,0.22236265738805136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,64,balanced,0.30684266487757367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,64,balanced,0.3877013524373372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,64,power_law_1.2,0.017254400253295898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,64,power_law_1.2,0.016889600455760954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,64,power_law_1.2,0.017407999932765962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,64,power_law_1.2,0.018035200238227845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,64,power_law_1.2,0.018918399512767792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,64,power_law_1.2,0.021766400337219237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,64,power_law_1.2,0.03596160113811493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,64,power_law_1.2,0.035436800122261046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,64,power_law_1.2,0.0304639995098114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,64,power_law_1.2,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,64,power_law_1.2,0.030675199627876282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,64,power_law_1.2,0.031097599864006044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,64,power_law_1.2,0.030713599920272828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,64,power_law_1.2,0.030476799607276915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,64,power_law_1.2,0.03128319978713989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,64,power_law_1.2,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,64,power_law_1.2,0.03245440125465393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,64,power_law_1.2,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,64,power_law_1.2,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,64,power_law_1.2,0.044275200366973876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,64,power_law_1.2,0.047577598690986635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,64,power_law_1.2,0.05985280275344849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,64,power_law_1.2,0.07548800110816956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,64,power_law_1.2,0.09975680112838745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,64,power_law_1.2,0.12268160581588745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,64,power_law_1.2,0.1619328022003174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,64,power_law_1.2,0.21969280242919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,2,power_law_1.2,0.03813759982585907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,2,power_law_1.2,0.03956480026245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,2,power_law_1.2,0.039878401160240176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,2,power_law_1.2,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,2,power_law_1.2,0.04391680061817169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,2,power_law_1.2,0.04485119879245758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,2,power_law_1.2,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,2,power_law_1.2,0.04552960097789764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,2,power_law_1.2,0.04560000002384186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,2,power_law_1.2,0.046419200301170346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,2,power_law_1.2,0.046515199542045596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,2,power_law_1.2,0.04798719882965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,2,power_law_1.2,0.048902401328086854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,2,power_law_1.2,0.05000960230827332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,2,power_law_1.2,0.05591679811477661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,2,power_law_1.2,0.055718398094177245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,2,power_law_1.2,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,2,power_law_1.2,0.07443199753761291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,2,power_law_1.2,0.08589439988136291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,2,power_law_1.2,0.10563199520111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,2,power_law_1.2,0.13278080224990846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,2,power_law_1.2,0.18152960538864135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,2,power_law_1.2,0.23562240600585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,2,power_law_1.2,0.32261760234832765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,2,power_law_1.2,0.3837631940841675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,2,power_law_1.2,0.5894847869873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,2,power_law_1.2,0.7447807788848877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,16,power_law_1.01,0.09827839732170104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,16,power_law_1.01,0.11785600185394288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,16,power_law_1.01,0.11730560064315795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,16,power_law_1.01,0.12388479709625244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,16,power_law_1.01,0.1198848009109497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,16,power_law_1.01,0.1185215950012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,16,power_law_1.01,0.12229759693145752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,16,power_law_1.01,0.11743359565734864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,16,power_law_1.01,0.12436480522155761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,16,power_law_1.01,0.12457599639892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,16,power_law_1.01,0.12295039892196655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,16,power_law_1.01,0.13050880432128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,16,power_law_1.01,0.1309183955192566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,16,power_law_1.01,0.13578879833221436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,16,power_law_1.01,0.14419200420379638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,16,power_law_1.01,0.14546560049057006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,16,power_law_1.01,0.1538879990577698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,16,power_law_1.01,0.17579519748687744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,16,power_law_1.01,0.1983423948287964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,16,power_law_1.01,0.2399616003036499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,16,power_law_1.01,0.26737279891967775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,16,power_law_1.01,0.35041279792785646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,16,power_law_1.01,0.3976448059082031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,16,power_law_1.01,0.5157887935638428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,16,power_law_1.01,0.6572351932525635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,16,power_law_1.01,0.974124813079834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,16,power_law_1.01,1.1799039840698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,8,balanced,0.06026133398214976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,8,balanced,0.06373866895834605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,8,balanced,0.06015466650327047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,8,balanced,0.05983466903368632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,8,balanced,0.059802666306495667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,8,balanced,0.061936000982920326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,8,balanced,0.06206933160622915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,8,balanced,0.06260266900062561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,8,balanced,0.06628266473611195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,8,balanced,0.06522666911284129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,8,balanced,0.0628959983587265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,8,balanced,0.06596266726652782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,8,balanced,0.06597333153088887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,8,balanced,0.06809066732724507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,8,balanced,0.0718560020128886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,8,balanced,0.07145600020885468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,8,balanced,0.07692799965540568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,8,balanced,0.08662399649620056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,8,balanced,0.08690667152404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,8,balanced,0.11729066570599873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,8,balanced,0.11641599734624226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,8,balanced,0.1520746648311615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,8,balanced,0.17661333084106445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,8,balanced,0.2330026626586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,8,balanced,0.2807146708170573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,8,balanced,0.3953813314437866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,8,balanced,0.500485340754191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,8,balanced,0.041519999504089355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,8,balanced,0.03956799954175949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,8,balanced,0.04140799989302953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,8,balanced,0.041893333196640015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,8,balanced,0.043706665436426796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,8,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,8,balanced,0.04192533095677694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,8,balanced,0.043920000394185386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,8,balanced,0.04363733530044556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,8,balanced,0.04246933261553446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,8,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,8,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,8,balanced,0.04641599953174591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,8,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,8,balanced,0.04786133269468943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,8,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,8,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,8,balanced,0.05607999861240387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,8,balanced,0.0581279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,8,balanced,0.06910400092601776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,8,balanced,0.08478933572769165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,8,balanced,0.11556266744931538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,8,balanced,0.1376053293546041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,8,balanced,0.18928533792495728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,8,balanced,0.23197867472966513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,8,balanced,0.3240480025609334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,8,balanced,0.4166773160298665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,8,power_law_1.2,0.04862079918384552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,2,power_law_1.01,0.09514880180358887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,8,power_law_1.2,0.05233280062675476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,2,power_law_1.01,0.147052800655365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,8,power_law_1.2,0.054067200422286986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,8,power_law_1.2,0.05594239830970764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,8,power_law_1.2,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,8,power_law_1.2,0.06450560092926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,8,power_law_1.2,0.06528639793395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,8,power_law_1.2,0.06633599996566772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,8,power_law_1.2,0.06846719980239868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,8,power_law_1.2,0.06773759722709656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,2,power_law_1.01,0.20030720233917237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,8,power_law_1.2,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,2,power_law_1.01,0.3121216058731079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,2,power_law_1.01,0.39448320865631104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,2,power_law_1.01,0.461030387878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,8,power_law_1.2,0.07802240252494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,2,power_law_1.01,0.6498432159423828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,8,power_law_1.2,0.08039680123329163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,8,power_law_1.2,0.08314239978790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,2,power_law_1.01,0.6782783985137939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,8,power_law_1.2,0.09304959774017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,8,power_law_1.2,0.09530879855155945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,8,power_law_1.2,0.10076160430908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,8,power_law_1.2,0.11354880332946778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,8,power_law_1.2,0.13215359449386596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,8,power_law_1.2,0.15665919780731202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,8,power_law_1.2,0.20385921001434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,8,power_law_1.2,0.26112000942230223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,2,power_law_1.01,0.6723392009735107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,8,power_law_1.2,0.32748799324035643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,2,power_law_1.01,0.7271743774414062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,2,power_law_1.01,0.7239168167114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,2,power_law_1.01,0.7583104133605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,8,power_law_1.2,0.47830400466918943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,2,power_law_1.01,0.7259647846221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,2,power_law_1.01,0.7671679973602294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,2,power_law_1.01,0.8139712333679199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,2,power_law_1.01,0.8088447570800781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,2,power_law_1.01,0.8448960304260253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,2,power_law_1.01,0.9412032127380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,2,power_law_1.01,0.9589056015014649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,2,power_law_1.01,1.0499711990356446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,2,power_law_1.01,1.1105600357055665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,2,power_law_1.01,1.278592014312744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,2,power_law_1.01,1.504032039642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,2,power_law_1.01,1.8839168548583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,2,power_law_1.01,2.209407997131348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,2,power_law_1.01,2.9838592529296877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,8,power_law_1.2,0.572217607498169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,2,power_law_1.01,3.6210880279541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,8,power_law_1.2,0.8277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,2,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,2,balanced,0.06021333237489065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,2,balanced,0.0841493308544159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,2,balanced,0.0832426647345225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,2,balanced,0.0848586658636729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,2,balanced,0.08455999692281087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,2,balanced,0.08407466610272725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,2,balanced,0.08707732955614726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,2,balanced,0.08890666564305623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,2,balanced,0.08913600444793701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,2,balanced,0.08919466535250346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,2,balanced,0.08954667051633199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,2,balanced,0.08912533521652222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,2,balanced,0.09236799677213033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,2,balanced,0.11071466406186421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,2,balanced,0.10814399520556132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,2,balanced,0.10481599966684978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,2,balanced,0.12780800461769104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,2,balanced,0.13352533181508383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,2,balanced,0.15524799625078836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,2,balanced,0.18398932615915933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,2,balanced,0.2379306753476461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,2,balanced,0.2879040042559306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,2,balanced,0.39324267705281574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,2,balanced,0.48633599281311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,2,balanced,0.7057279745737711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,2,balanced,0.8967573642730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,8,power_law_1.2,1.1241727828979493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,128,power_law_1.01,0.01713919937610626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,128,power_law_1.01,0.01785600036382675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,128,power_law_1.01,0.018988800048828126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,128,power_law_1.01,0.02215680032968521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,16,power_law_1.2,0.041791999340057374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,16,power_law_1.2,0.03464959859848023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,16,power_law_1.2,0.03307519853115082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,16,power_law_1.2,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,16,power_law_1.2,0.031488001346588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,16,power_law_1.2,0.03317759931087494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,16,power_law_1.2,0.03343999981880188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,16,power_law_1.2,0.03412480056285858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,16,power_law_1.2,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,16,power_law_1.2,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,16,power_law_1.2,0.03806079924106598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,16,power_law_1.2,0.039110401272773744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,16,power_law_1.2,0.039603200554847715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,16,power_law_1.2,0.04087679982185364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,16,power_law_1.2,0.04538240134716034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,16,power_law_1.2,0.04764159917831421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,16,power_law_1.2,0.04864639937877655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,16,power_law_1.2,0.05355520248413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,16,power_law_1.2,0.05959039926528931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,16,power_law_1.2,0.0685375988483429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,16,power_law_1.2,0.08407040238380432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,16,power_law_1.2,0.10872960090637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,16,power_law_1.2,0.12911360263824462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,16,power_law_1.2,0.19946880340576173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,8,power_law_1.01,0.020262399315834047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,8,power_law_1.01,0.01984640061855316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,8,power_law_1.01,0.019551999866962433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,8,power_law_1.01,0.019699199497699736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,8,power_law_1.01,0.020448000729084016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,8,power_law_1.01,0.022035199403762817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,8,power_law_1.01,0.04586879909038544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,8,power_law_1.01,0.04589439928531647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,8,power_law_1.01,0.03692800104618073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,8,power_law_1.01,0.03443840146064758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,8,power_law_1.01,0.035155200958251955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,8,power_law_1.01,0.04095360040664673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,8,power_law_1.01,0.039289599657058714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,8,power_law_1.01,0.03672960102558136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,8,power_law_1.01,0.04301440119743347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,8,power_law_1.01,0.03852800130844116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,8,power_law_1.01,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,8,power_law_1.01,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.01,0.06039680242538452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.01,0.08221439719200134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,16,power_law_1.2,0.2740799903869629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.01,0.10236159563064576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.01,0.1373952031135559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.01,0.16593919992446898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.01,0.2368959903717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.01,0.2999039888381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,8,balanced,0.021551998953024547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,8,balanced,0.022426667312781017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,8,balanced,0.02314666658639908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,8,balanced,0.023178666830062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,8,balanced,0.030591999491055805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,8,balanced,0.03270400067170461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,8,balanced,0.04119999955097834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,8,balanced,0.039813332259655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,8,balanced,0.05583466589450836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,8,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,8,balanced,0.04930666585763296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,8,balanced,0.068122665087382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,8,balanced,0.06846933563550313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,8,balanced,0.0664106657107671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,8,balanced,0.08201066652933757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,8,balanced,0.07971733311812083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,8,balanced,0.07355199754238129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,8,balanced,0.0794239987929662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,8,balanced,0.09130666653315227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,8,balanced,0.10083733002344768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,8,balanced,0.11495467027028401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,8,balanced,0.13901866475741068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,8,balanced,0.15454933047294617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,8,balanced,0.23161600033442178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,8,balanced,0.26904000838597614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,8,balanced,0.38969067732493085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,8,balanced,0.5065866708755493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,16,power_law_1.2,0.3704063892364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,16,power_law_1.01,0.0536191999912262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,16,power_law_1.01,0.05064319968223572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,16,power_law_1.01,0.050918400287628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,16,power_law_1.01,0.05557760000228882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,16,power_law_1.01,0.04938240051269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,16,power_law_1.01,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,16,power_law_1.01,0.055852800607681274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,16,power_law_1.01,0.057606399059295654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,16,power_law_1.01,0.05826560258865356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,16,power_law_1.01,0.05797119736671448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,16,power_law_1.01,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,16,power_law_1.01,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,16,power_law_1.01,0.061887997388839724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,16,power_law_1.01,0.06273279786109924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,16,power_law_1.01,0.06843519806861878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,16,power_law_1.01,0.0728384017944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,16,power_law_1.01,0.07839360237121581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,16,power_law_1.01,0.0908352017402649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,16,power_law_1.01,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,16,power_law_1.01,0.12019840478897095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,16,power_law_1.01,0.1325824022293091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,16,power_law_1.01,0.1669119954109192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,16,power_law_1.01,0.21375999450683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,16,power_law_1.01,0.2710975885391235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,16,power_law_1.01,0.3605247974395752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,16,power_law_1.01,0.5210752010345459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,16,power_law_1.01,0.6883456230163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.01,0.4364031791687012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,16,power_law_1.2,0.4867712020874023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.01,0.5719744205474854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,2,balanced,0.017504000415404636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,2,balanced,0.019023999571800232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,2,balanced,0.02046400060256322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,2,balanced,0.023418667415777843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,2,balanced,0.02914133419593175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,2,balanced,0.02985599885384242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,2,balanced,0.031498665610949196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,2,balanced,0.0313226655125618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,2,balanced,0.0332640012105306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,2,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,2,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,2,balanced,0.03566399961709976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,2,balanced,0.03734933336575826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,2,balanced,0.04180799921353658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,2,balanced,0.04282666742801666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,2,balanced,0.042453333735466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,2,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,2,balanced,0.0539626677831014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,2,balanced,0.055871998270352684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,2,balanced,0.0800853321949641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,2,balanced,0.10358933607737224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,2,balanced,0.10470933715502422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,2,balanced,0.11356266339619954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,2,balanced,0.1842026710510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,32,power_law_1.01,0.07767040133476258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,2,balanced,0.19454934199651083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,32,power_law_1.01,0.07827200293540955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,2,balanced,0.2736426591873169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,2,balanced,0.35573867956797284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,32,power_law_1.01,0.07807360291481018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,32,power_law_1.01,0.0804032027721405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,32,power_law_1.01,0.08047999739646912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,32,power_law_1.01,0.07545599937438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,32,power_law_1.01,0.0809216022491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,32,power_law_1.01,0.07875199913978577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,32,power_law_1.01,0.07859839797019959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,32,power_law_1.01,0.07969279885292054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,32,power_law_1.01,0.08172799944877625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,32,power_law_1.01,0.08538879752159119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,32,power_law_1.01,0.08579199910163879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,32,power_law_1.01,0.08807039856910706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,32,power_law_1.01,0.09315199851989746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,32,power_law_1.01,0.09319040179252625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,32,power_law_1.01,0.10047359466552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,32,power_law_1.01,0.11850240230560302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,32,power_law_1.01,0.12641919851303102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,32,power_law_1.01,0.15193599462509155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,32,power_law_1.01,0.1829375982284546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,32,power_law_1.01,0.2320703983306885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,32,power_law_1.01,0.2360703945159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,32,power_law_1.01,0.3051136016845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,32,power_law_1.01,0.3809216022491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,32,power_law_1.01,0.5177663803100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,32,power_law_1.01,0.6441472053527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,2,power_law_1.01,0.05649920105934143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,2,power_law_1.01,0.06872959733009339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,2,power_law_1.01,0.0833791971206665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,2,power_law_1.01,0.10824960470199585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,2,power_law_1.01,0.12989439964294433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,2,power_law_1.01,0.15456639528274535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,2,power_law_1.01,0.1922943949699402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,2,power_law_1.01,0.19553279876708984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,2,power_law_1.01,0.19898879528045654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,2,power_law_1.01,0.19535360336303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,2,power_law_1.01,0.21013760566711426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,2,power_law_1.01,0.21689600944519044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,2,power_law_1.01,0.2147520065307617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,2,power_law_1.01,0.2361664056777954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,2,power_law_1.01,0.2409343957901001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,2,power_law_1.01,0.2479423999786377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,2,power_law_1.01,0.25464320182800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,2,power_law_1.01,0.28923521041870115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,2,power_law_1.01,0.303436803817749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,2,power_law_1.01,0.3546112060546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,2,power_law_1.01,0.41205759048461915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,2,power_law_1.01,0.5244736194610595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,2,power_law_1.01,0.6094783782958985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,2,power_law_1.01,0.7968319892883301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,2,power_law_1.01,0.9615743637084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,2,power_law_1.01,1.3395071983337403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,2,power_law_1.01,1.7571775436401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,128,power_law_1.01,0.11560959815979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,128,power_law_1.01,0.039647999405860904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,128,power_law_1.01,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,128,power_law_1.01,0.04222080111503601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,128,power_law_1.01,0.04357759952545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,128,power_law_1.01,0.043961599469184875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,128,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,128,power_law_1.01,0.045535999536514285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,128,power_law_1.01,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,128,power_law_1.01,0.04627839922904968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,128,power_law_1.01,0.04673919975757599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,128,power_law_1.01,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,128,power_law_1.01,0.04856959879398346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,128,power_law_1.01,0.05125759840011597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,128,power_law_1.01,0.05637120008468628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,128,power_law_1.01,0.057081598043441775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,128,power_law_1.01,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,128,power_law_1.01,0.06865919828414917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,128,power_law_1.01,0.07089920043945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,128,power_law_1.01,0.08526080250740051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,128,power_law_1.01,0.09675520062446594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,128,power_law_1.01,0.13793280124664306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,128,power_law_1.01,0.15530879497528077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,128,power_law_1.01,0.2138495922088623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,128,power_law_1.01,0.2748863935470581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,128,power_law_1.01,0.3828160047531128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,128,power_law_1.01,0.4701695919036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,4,power_law_1.01,0.07508479952812194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,4,power_law_1.01,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,4,power_law_1.01,0.08637440204620361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,4,power_law_1.01,0.09990400075912476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,4,power_law_1.01,0.10635520219802856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,4,power_law_1.01,0.12395520210266113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,4,power_law_1.01,0.1388159990310669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,4,power_law_1.01,0.14311039447784424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,4,power_law_1.01,0.13416320085525513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,4,power_law_1.01,0.1378175973892212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,4,power_law_1.01,0.1346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,4,power_law_1.01,0.1437376022338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,4,power_law_1.01,0.14469120502471924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,4,power_law_1.01,0.14725120067596437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,4,power_law_1.01,0.14965759515762328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,4,power_law_1.01,0.15015039443969727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,4,power_law_1.01,0.1614848017692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,4,power_law_1.01,0.17657599449157715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,4,power_law_1.01,0.19274879693984986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,4,power_law_1.01,0.23002879619598388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,4,power_law_1.01,0.2609663963317871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,2,power_law_1.01,0.025824001431465148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,2,power_law_1.01,0.024857600033283234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,2,power_law_1.01,0.029548799991607665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,2,power_law_1.01,0.036134400963783266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,4,power_law_1.01,0.347321605682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,2,power_law_1.01,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,2,power_law_1.01,0.0496832013130188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,4,power_law_1.01,0.405241584777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,2,power_law_1.01,0.052671998739242554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,4,power_law_1.01,0.5639743804931641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,4,power_law_1.01,0.7040639877319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,4,power_law_1.01,0.9972864151000976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,2,power_law_1.01,0.05494400262832642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,2,power_law_1.01,0.055827200412750244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,4,power_law_1.01,1.3557184219360352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,2,power_law_1.01,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,2,power_law_1.01,0.05802879929542541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,2,power_law_1.01,0.06032639741897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,2,power_law_1.01,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,2,power_law_1.01,0.0690496027469635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,2,power_law_1.01,0.07415680289268493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,2,power_law_1.01,0.07616000175476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,2,power_law_1.01,0.08069120049476623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,2,power_law_1.01,0.09292160272598267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,2,power_law_1.01,0.09660159945487976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,2,power_law_1.01,0.11074559688568116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,2,power_law_1.01,0.13141759634017944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,2,power_law_1.01,0.1743872046470642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,2,power_law_1.01,0.20359039306640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,2,power_law_1.01,0.284499192237854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,2,power_law_1.01,0.36840319633483887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,2,power_law_1.01,0.5609471797943115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,2,power_law_1.01,0.6934271812438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,4,power_law_1.2,0.14016640186309814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,4,power_law_1.2,0.21715199947357178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,4,power_law_1.2,0.22367360591888427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,4,power_law_1.2,0.3211967945098877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,4,power_law_1.2,0.4272831916809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,4,power_law_1.2,0.4860032081604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,4,power_law_1.2,0.5580351829528809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,4,power_law_1.2,0.6625535964965821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,4,power_law_1.2,0.57358717918396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,4,power_law_1.2,0.6409023761749267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,4,power_law_1.2,0.6242239952087403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,4,power_law_1.2,0.6232575893402099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,4,power_law_1.2,0.6385536193847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,4,power_law_1.2,0.6838143825531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,4,power_law_1.2,0.6822783946990967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,4,power_law_1.2,0.6726592063903809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,4,power_law_1.2,0.7480127811431885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,4,power_law_1.2,0.8347519874572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,4,power_law_1.2,0.8125375747680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,4,power_law_1.2,0.9869312286376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,4,power_law_1.2,0.9915200233459472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,4,power_law_1.2,1.122828769683838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,4,power_law_1.2,1.233510398864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,4,power_law_1.2,1.5126784324645997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,4,power_law_1.2,1.8451072692871093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,4,power_law_1.2,2.3315008163452147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,4,power_law_1.2,2.795846366882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,2,power_law_1.01,0.026752001047134398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,2,power_law_1.01,0.03868159949779511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,2,power_law_1.01,0.05484799742698669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,2,power_law_1.01,0.06413440108299255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,2,power_law_1.01,0.07823359966278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,2,power_law_1.01,0.10334080457687378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,2,power_law_1.01,0.10955519676208496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,2,power_law_1.01,0.11230080127716065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,2,power_law_1.01,0.1141055941581726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,2,power_law_1.01,0.11716480255126953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,2,power_law_1.01,0.11882239580154419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,2,power_law_1.01,0.11996159553527833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,2,power_law_1.01,0.12476799488067628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,2,power_law_1.01,0.12339199781417846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,2,power_law_1.01,0.14461439847946167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,2,power_law_1.01,0.14869120121002197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,2,power_law_1.01,0.1434175968170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,2,power_law_1.01,0.28626561164855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.01,0.32214400768280027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.01,0.23210880756378174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.01,0.24377601146697997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.01,0.282425594329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.01,0.32488958835601806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.01,0.4204095840454102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.01,0.5455808162689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.01,0.7351808071136474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.01,0.9426112174987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,32,power_law_1.01,0.055078399181365964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,32,power_law_1.01,0.05256959795951843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,32,power_law_1.01,0.05367040038108826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,32,power_law_1.01,0.05141760110855102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,32,power_law_1.01,0.04924800097942352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,32,power_law_1.01,0.05497599840164184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,32,power_law_1.01,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,32,power_law_1.01,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,32,power_law_1.01,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,32,power_law_1.01,0.05422080159187317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,32,power_law_1.01,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,32,power_law_1.01,0.05548800230026245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,32,power_law_1.01,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,32,power_law_1.01,0.056620800495147706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,32,power_law_1.01,0.056966400146484374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,32,power_law_1.01,0.05989120006561279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,32,power_law_1.01,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,32,power_law_1.01,0.07382400035858154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.01,0.07895680069923401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.01,0.09410560131072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.01,0.10574079751968384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.01,0.14428160190582276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.01,0.14856319427490233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.01,0.22929279804229735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.01,0.2776832103729248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.01,0.39336318969726564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.01,0.4907711982727051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,1,power_law_1.01,0.0414463996887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,1,power_law_1.01,0.043654400110244754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,1,power_law_1.01,0.04769920110702515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,1,power_law_1.01,0.05423359870910645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,1,power_law_1.01,0.05756160020828247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,1,power_law_1.01,0.055641597509384154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,1,power_law_1.01,0.05626239776611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,1,power_law_1.01,0.05633280277252197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,1,power_law_1.01,0.05608959794044495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,1,power_law_1.01,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,1,power_law_1.01,0.055801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,1,power_law_1.01,0.057196801900863646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,1,power_law_1.01,0.05743359923362732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,1,power_law_1.01,0.06279680132865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,1,power_law_1.01,0.0689791977405548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,1,power_law_1.01,0.07049599885940552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,1,power_law_1.01,0.07724159955978394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,1,power_law_1.01,0.09010559916496277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.01,0.10341119766235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.01,0.12460800409317016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,4,power_law_1.2,0.03772799968719483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,4,power_law_1.2,0.04022400081157684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,4,power_law_1.2,0.0429504007101059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,4,power_law_1.2,0.04775039851665497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,4,power_law_1.2,0.04999040067195892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,4,power_law_1.2,0.053651201725006106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,4,power_law_1.2,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.01,0.14434560537338256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,4,power_law_1.2,0.05578240156173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,4,power_law_1.2,0.05511680245399475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,4,power_law_1.2,0.05616000294685364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,4,power_law_1.2,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,4,power_law_1.2,0.05988479852676391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,4,power_law_1.2,0.06220160126686096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,4,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.01,0.19297280311584472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,4,power_law_1.2,0.06729599833488464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,4,power_law_1.2,0.07029119729995728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,4,power_law_1.2,0.07491199970245362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.01,0.24320640563964843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,4,power_law_1.2,0.08453119993209839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,4,power_law_1.2,0.09408640265464782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,4,power_law_1.2,0.1123263955116272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,4,power_law_1.2,0.1358464002609253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.01,0.3331648111343384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,1,balanced,0.020448000480731327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,1,balanced,0.021317332983016968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,1,balanced,0.023141334454218548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,1,balanced,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,1,balanced,0.0516533354918162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,1,balanced,0.051781331499417625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,1,balanced,0.05578133463859558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,1,balanced,0.057071998715400696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,1,balanced,0.05831466615200043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,1,balanced,0.06172266602516174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,1,balanced,0.0660159985224406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,4,power_law_1.2,0.17952640056610109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,1,balanced,0.0637600024541219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,1,balanced,0.06602666775385539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,1,balanced,0.0721919983625412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,4,power_law_1.2,0.2269376039505005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,1,balanced,0.06796800096829732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,1,balanced,0.07045333087444305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,1,balanced,0.07717866698900859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,1,balanced,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,1,balanced,0.09054399530092876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,1,balanced,0.1560373306274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,1,balanced,0.18339200814565024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,1,balanced,0.1677280068397522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,1,balanced,0.1786880095799764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,1,balanced,0.28914666175842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.01,0.4292543888092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,1,balanced,0.30661867062250775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,1,balanced,0.4370453357696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,1,balanced,0.571674664815267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,8,balanced,0.03156266609827677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,8,balanced,0.029487999776999157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,8,balanced,0.02924266705910365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,8,balanced,0.0323840007185936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,8,balanced,0.033359999457995095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,8,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,8,balanced,0.03850133220354716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,8,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,8,balanced,0.03955733279387156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,8,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,8,balanced,0.03922666609287262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,8,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,8,balanced,0.03855466594298681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,8,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,8,balanced,0.04398933549722036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,8,balanced,0.043552001317342125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,8,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,8,balanced,0.05678399900595347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,8,balanced,0.04974400003751119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,8,balanced,0.06242666641871134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,8,balanced,0.06169599791367849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,8,balanced,0.07899199922879536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,8,balanced,0.08285333216190338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,8,balanced,0.10941867033640544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,8,balanced,0.12345066666603088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,8,balanced,0.16482133666674295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,8,balanced,0.18772266308466592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,4,power_law_1.2,0.31215999126434324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,4,power_law_1.2,0.45119361877441405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,4,power_law_1.2,0.6851391792297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.01,0.6137343883514405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.01,0.7917568206787109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,4,power_law_1.2,0.9534720420837403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,16,balanced,0.02345066765944163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,16,balanced,0.02309866746266683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,16,balanced,0.023082666099071503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,16,balanced,0.023386667172114056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,16,balanced,0.025199999411900837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,16,balanced,0.025146665672461193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,16,balanced,0.02535466601451238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,16,balanced,0.028586665789286297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,16,balanced,0.029215998947620392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,16,balanced,0.02736533433198929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,16,balanced,0.028517333169778187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,16,balanced,0.02914133419593175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,16,balanced,0.028581333657105763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,16,balanced,0.03149333347876867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,16,balanced,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,16,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,16,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,16,balanced,0.03737066686153412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,16,balanced,0.03772266705830892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,16,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,16,balanced,0.045653333266576133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,16,balanced,0.04866666595141093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,16,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,16,balanced,0.06672533353169759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,16,balanced,0.08538132905960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,16,balanced,0.10601600011189778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,16,balanced,0.12948800126711527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,32,balanced,0.044768000642458596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,32,balanced,0.04444799820582072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,32,balanced,0.044677332043647766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,32,balanced,0.0499893327554067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,32,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,32,balanced,0.07632000247637431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,32,balanced,0.09214400251706441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,32,balanced,0.09143466750780742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,32,balanced,0.09237866600354512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,32,balanced,0.09310932954152425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,32,balanced,0.09278933207194011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,32,balanced,0.09681600332260132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,32,balanced,0.09917333722114563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,32,balanced,0.09876799583435059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,32,balanced,0.1069546639919281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,32,balanced,0.10774399836858113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,32,balanced,0.11431466539700826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,32,balanced,0.12478933731714885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,32,balanced,0.13572266697883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,32,balanced,0.15893866618474325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,32,balanced,0.17359467347462973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,32,balanced,0.2170506715774536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,32,balanced,0.2619946599006653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,32,balanced,0.3441866636276245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,32,balanced,0.43161598841349286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,32,balanced,0.5923839807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,32,balanced,0.7635573546091715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,4,power_law_1.2,0.12392319440841675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,4,power_law_1.2,0.15180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,4,power_law_1.2,0.18354560136795045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,4,power_law_1.2,0.23975040912628173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,4,power_law_1.2,0.27952640056610106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,4,power_law_1.2,0.4355455875396729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,4,power_law_1.2,0.40665597915649415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,4,power_law_1.2,0.3905344009399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,4,power_law_1.2,0.4030144214630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,4,power_law_1.2,0.39290239810943606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,4,power_law_1.2,0.4259200096130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,4,power_law_1.2,0.4327104091644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,4,power_law_1.2,0.4421504020690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,4,power_law_1.2,0.4617663860321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,4,power_law_1.2,0.4730559825897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,4,power_law_1.2,0.47716479301452636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,4,power_law_1.2,0.5261375904083252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,4,power_law_1.2,0.5841792106628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,4,power_law_1.2,0.6015488147735596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,4,power_law_1.2,0.7275008201599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,4,power_law_1.2,0.8405695915222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,4,power_law_1.2,0.9913087844848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,4,power_law_1.2,1.0137215614318849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,4,power_law_1.2,1.2918975830078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,4,power_law_1.2,1.6979839324951171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,4,power_law_1.2,2.188447952270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,4,power_law_1.2,0.07971839904785157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,4,power_law_1.2,2.7916799545288087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,4,power_law_1.2,0.0989247977733612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,4,power_law_1.2,0.10017919540405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,4,power_law_1.2,0.12310400009155273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,4,power_law_1.2,0.14057600498199463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,4,power_law_1.2,0.13627519607543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,4,power_law_1.2,0.1671872019767761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,4,power_law_1.2,0.16751359701156615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,4,power_law_1.2,0.1949504017829895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,4,power_law_1.2,0.17066240310668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,4,power_law_1.2,0.18931200504302978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,4,power_law_1.2,0.1825279951095581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,4,power_law_1.2,0.18410880565643312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,4,power_law_1.2,0.1837823987007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,4,power_law_1.2,0.19263360500335694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,4,power_law_1.2,0.20426878929138184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,4,power_law_1.2,0.2105407953262329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,4,power_law_1.2,0.23086080551147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,4,power_law_1.2,0.24519040584564208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,4,power_law_1.2,0.29887359142303466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,4,power_law_1.2,0.32654719352722167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,4,power_law_1.2,0.401804780960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,4,power_law_1.2,0.4928127765655518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,4,power_law_1.2,0.6735487937927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,4,power_law_1.2,0.8217535972595215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,4,power_law_1.2,1.170905590057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,4,power_law_1.2,1.4734848022460938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,256,balanced,0.06039466460545858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,256,balanced,0.0603413333495458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,256,balanced,0.06006399790445963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,256,balanced,0.06381333371003468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,256,balanced,0.06211199859778086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,256,balanced,0.06559999783833821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,256,balanced,0.06311999758084615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,256,balanced,0.0643039991458257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,256,balanced,0.0670773337284724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,256,balanced,0.06667733192443848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,256,balanced,0.06471999982992808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,256,balanced,0.06840533514817555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,256,balanced,0.0705973356962204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,256,balanced,0.06842666864395142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,256,balanced,0.07540800174077351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,256,balanced,0.07602666815121968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,256,balanced,0.07666666805744171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,256,balanced,0.08481066425641377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,256,balanced,0.08834667007128398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,256,balanced,0.10493866602579753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,256,balanced,0.12377066413561504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,256,balanced,0.16476800044377646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,256,balanced,0.19662400086720785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,256,balanced,0.25989866256713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,256,balanced,0.32440000772476196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,256,balanced,0.4492959976196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,256,balanced,0.5764106512069702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,8,balanced,0.033813332517941795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,8,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,8,balanced,0.03306133300065994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,8,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,8,balanced,0.036848001182079315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,8,balanced,0.03580799947182337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,8,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,8,balanced,0.03772266705830892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,8,balanced,0.03920533259709676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,8,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,8,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,8,balanced,0.04177600145339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,8,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,8,balanced,0.04481600224971771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,8,balanced,0.04775466521581014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,32,power_law_1.01,0.024665600061416625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,32,power_law_1.01,0.0244159996509552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,32,power_law_1.01,0.025900799036026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,8,balanced,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,8,balanced,0.054197331269582115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,32,power_law_1.01,0.024006399512290954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,8,balanced,0.06035733222961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,8,balanced,0.062261333068211876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,32,power_law_1.01,0.024057599902153014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,8,balanced,0.07261866827805837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,32,power_law_1.01,0.027641600370407103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,32,power_law_1.01,0.02685439884662628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,32,power_law_1.01,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,32,power_law_1.01,0.03771519958972931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,32,power_law_1.01,0.03749760091304779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,32,power_law_1.01,0.05378559827804565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,32,power_law_1.01,0.053286397457122804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,32,power_law_1.01,0.05459200143814087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,8,balanced,0.08201600114504497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,8,balanced,0.11356799801190694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,8,balanced,0.13788800438245138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,8,balanced,0.18532266219456991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,8,balanced,0.23033599058787027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,8,balanced,0.3340959946314494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,8,balanced,0.4458719889322917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,32,power_law_1.01,0.0596671998500824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,32,power_law_1.01,0.060134398937225345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,32,power_law_1.01,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,32,power_law_1.01,0.08391680121421814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,32,power_law_1.01,0.08919680118560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.01,0.11357439756393432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.01,0.14940799474716188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.01,0.09548159837722778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.01,0.1168063998222351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.01,0.14977279901504517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.01,0.20785279273986818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.01,0.26845440864562986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.01,0.39225599765777586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.01,0.4568192005157471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,16,balanced,0.03815466662247976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,16,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,16,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,16,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,16,balanced,0.04574400186538696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,16,balanced,0.0651039977868398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,16,balanced,0.06709866722424825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,16,balanced,0.06543999910354614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,16,balanced,0.06442133088906606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,16,balanced,0.06426133215427399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,16,balanced,0.0658186674118042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,16,balanced,0.06664533416430156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,16,balanced,0.06824533144632976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,16,balanced,0.06849599877993266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,16,balanced,0.0743093341588974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,16,balanced,0.07261333366235097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,16,balanced,0.07628266513347626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,16,balanced,0.08072533210118611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,16,balanced,0.08251733581225078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,16,balanced,0.09304533402125041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,16,balanced,0.10404266913731892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,16,balanced,0.12247467041015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,16,balanced,0.14820266763369241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,16,balanced,0.18981866041819254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,16,balanced,0.22288000583648682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,16,balanced,0.29470400015513104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,16,balanced,0.3778239885965983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,8,power_law_1.2,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,8,power_law_1.2,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,8,power_law_1.2,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,8,power_law_1.2,0.04709759950637817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,8,power_law_1.2,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,8,power_law_1.2,0.05050879716873169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,8,power_law_1.2,0.051769602298736575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,8,power_law_1.2,0.05120639801025391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,8,power_law_1.2,0.052198398113250735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,8,power_law_1.2,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,8,power_law_1.2,0.05629439949989319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,8,power_law_1.2,0.05658239722251892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,8,power_law_1.2,0.060038399696350095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,8,power_law_1.2,0.06480640172958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,8,power_law_1.2,0.07221760153770447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,8,power_law_1.2,0.07412480115890503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,8,power_law_1.2,0.0756608009338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,8,power_law_1.2,0.08900480270385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,8,power_law_1.2,0.0938368022441864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,8,power_law_1.2,0.115065598487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,8,power_law_1.2,0.1411967992782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,8,power_law_1.2,0.18200960159301757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,8,power_law_1.2,0.2212735891342163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,8,power_law_1.2,0.3125760078430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,8,power_law_1.2,0.3857088088989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,8,power_law_1.2,0.5856768131256104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,8,power_law_1.2,0.7417984008789062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,2,balanced,0.07102400064468384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,2,balanced,0.09949866930643718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,2,balanced,0.15135467052459717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,2,balanced,0.2594826618830363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,2,balanced,0.4700213273366292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,2,balanced,0.5758399963378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,2,balanced,0.5803360144297282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,4,balanced,0.019706666469573975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,2,balanced,0.5782026847203573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,2,balanced,0.5821653207143148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,2,balanced,0.5840906699498495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,2,balanced,0.582693338394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,2,balanced,0.5921866496404012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,2,balanced,0.5960213343302408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,2,balanced,0.6011146704355875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,2,balanced,0.6083573500315348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,4,balanced,0.019189332922299702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,2,balanced,0.610426664352417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,4,balanced,0.021136000752449036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,4,balanced,0.02312533309062322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,4,balanced,0.027653334041436512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,4,balanced,0.0414986660083135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,4,balanced,0.04348800083001455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,4,balanced,0.04578666885693868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,4,balanced,0.04760533571243286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,4,balanced,0.04897599915663401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,4,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,4,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,4,balanced,0.056277334690093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,4,balanced,0.058373332023620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,4,balanced,0.07067733506361644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,4,balanced,0.07020266850789388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,4,balanced,0.07402666906515758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,4,balanced,0.09044800202051799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,4,balanced,0.09781333804130554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,4,balanced,0.11809066931406657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,2,balanced,0.6272853215535482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,4,balanced,0.12601600090662637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,4,balanced,0.1416213313738505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,4,balanced,0.15825066963831583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,4,balanced,0.1805386741956075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,4,balanced,0.20347734292348227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,4,balanced,0.33188800017038983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,4,balanced,0.3769386609395345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,2,balanced,0.6479039986928304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,2,balanced,0.6706186930338541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,2,balanced,0.8135786851247152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,2,balanced,0.7426559925079346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,2,balanced,1.1967199643452961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,2,balanced,0.9210720062255859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,2,balanced,1.420197327931722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,2,balanced,1.4440587361653645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,2,balanced,2.506458600362142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,2,balanced,2.658143997192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,8,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,8,balanced,0.04702933132648468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,8,balanced,0.04800533254941305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,8,balanced,0.049045334259668984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,8,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,8,balanced,0.04971200227737427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,8,balanced,0.04801600178082784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,8,balanced,0.05016533533732096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,8,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,8,balanced,0.04931733508904775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,8,balanced,0.050240000089009605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,8,balanced,0.04632000128428141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,8,balanced,0.04836800197760264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,8,balanced,0.04826133449872335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,8,balanced,0.05013333261013031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,8,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,8,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,8,balanced,0.060559997955958046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,8,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,8,balanced,0.0811413327852885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,8,balanced,0.09909866253534953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,8,balanced,0.12991467118263245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,8,balanced,0.15228799978892008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,8,balanced,0.20709866285324097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,8,balanced,0.2619146704673767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,8,balanced,0.367301344871521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,8,balanced,0.4662453333536784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,8,balanced,0.04101333270470301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,8,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,8,balanced,0.045253331462542214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,8,balanced,0.04985066751639048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,8,balanced,0.05189333359400431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,8,balanced,0.06426133215427399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,8,balanced,0.0642080008983612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,8,balanced,0.06527466575304668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,8,balanced,0.0681279997030894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,8,balanced,0.06788266698519389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,8,balanced,0.06817600131034851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,8,balanced,0.07025599976380666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,8,balanced,0.0729013333717982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,8,balanced,0.07667733232180278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,8,balanced,0.08663466572761536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,8,balanced,0.08927466471989949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,8,balanced,0.09504000345865886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,8,balanced,0.10913599530855815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,8,balanced,0.11924800276756287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,8,balanced,0.14293866356213888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,8,balanced,0.16158933440844217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,8,balanced,0.21715732415517172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,16,power_law_1.01,0.03192960023880005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,16,power_law_1.01,0.039417600631713866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,16,power_law_1.01,0.039052799344062805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,16,power_law_1.01,0.03941119909286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,16,power_law_1.01,0.039852800965309146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,16,power_law_1.01,0.045414400100708005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,16,power_law_1.01,0.05825279951095581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,8,balanced,0.26072533925374347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,8,balanced,0.36137068271636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,8,balanced,0.4615093469619751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,8,balanced,0.6605173349380493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,8,balanced,0.8447253704071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,16,power_law_1.01,0.054150402545928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,16,power_law_1.01,0.05631999969482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,16,power_law_1.01,0.08392320275306701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,16,power_law_1.01,0.08454399704933166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,16,power_law_1.01,0.08599680066108703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,16,power_law_1.01,0.09308800101280212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,16,power_law_1.01,0.09058560132980346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,16,power_law_1.01,0.08623999953269959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,16,power_law_1.01,0.08333439826965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,16,power_law_1.01,0.08751999735832214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,16,power_law_1.01,0.08022400140762329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.01,0.08339840173721313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.01,0.10308480262756348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.01,0.11126400232315063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.01,0.13879040479660035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.01,0.1696895956993103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.01,0.23198719024658204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.01,0.306713604927063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.01,0.4654655933380127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.01,0.5893695831298829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,8,balanced,0.049839998284975685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,8,balanced,0.05205333232879639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,8,balanced,0.05377600093682607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,8,balanced,0.06407999992370605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,8,balanced,0.08669867118199666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,8,balanced,0.12200533350308736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,8,balanced,0.12293866276741028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,8,balanced,0.12341866890589397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,8,balanced,0.12482133507728577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,8,balanced,0.12665599584579468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,8,balanced,0.12637866536776224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,8,balanced,0.12824533383051553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,8,balanced,0.13061867157618204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,8,balanced,0.1309386690457662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,8,balanced,0.13939199844996134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,8,balanced,0.141077329715093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,8,balanced,0.14377599954605103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,8,balanced,0.15643200278282166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,8,balanced,0.16842132806777954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,8,balanced,0.18685332934061685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,8,balanced,0.208079993724823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,8,balanced,0.2504533330599467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,8,balanced,0.2895786762237549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,8,balanced,0.38338665167490643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,8,balanced,0.4567466576894124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,8,balanced,0.6410400072733561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,8,balanced,0.8198613325754801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,64,power_law_1.2,0.06520959734916687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,64,power_law_1.2,0.06561279892921448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,64,power_law_1.2,0.0665727972984314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,64,power_law_1.2,0.0698303997516632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,64,power_law_1.2,0.06564480066299438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,64,power_law_1.2,0.06999040246009827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,64,power_law_1.2,0.0719488024711609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,64,power_law_1.2,0.06989439725875854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,64,power_law_1.2,0.07167999744415283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,64,power_law_1.2,0.07831680178642272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,64,power_law_1.2,0.07138559818267823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,64,power_law_1.2,0.0756991982460022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,64,power_law_1.2,0.07745280265808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,64,power_law_1.2,0.08087040185928344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,64,power_law_1.2,0.08920320272445678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,64,power_law_1.2,0.09107199907302857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,64,power_law_1.2,0.10002559423446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,64,power_law_1.2,0.11563520431518555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,64,power_law_1.2,0.12876800298690796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,64,power_law_1.2,0.16526080369949342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,64,power_law_1.2,0.19382400512695314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,64,power_law_1.2,0.25496959686279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,64,power_law_1.2,0.28206720352172854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,64,power_law_1.2,0.36131839752197265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,64,power_law_1.2,0.42768001556396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,8,power_law_1.01,0.04428159892559051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,8,power_law_1.01,0.042777600884437564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,8,power_law_1.01,0.042412799596786496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,8,power_law_1.01,0.044614401459693906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,64,power_law_1.2,0.6377664089202881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,8,power_law_1.01,0.045177599787712096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,8,power_law_1.01,0.04678399860858917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,8,power_law_1.01,0.04676479995250702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,8,power_law_1.01,0.04835839867591858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,8,power_law_1.01,0.05548160076141358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,8,power_law_1.01,0.05496320128440857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,8,power_law_1.01,0.05663359761238098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,8,power_law_1.01,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,8,power_law_1.01,0.06247680187225342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,8,balanced,0.018954666952292126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,8,balanced,0.017322666943073273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,8,balanced,0.020314666132132213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,8,balanced,0.01878400022784869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,8,balanced,0.021130666136741638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,8,balanced,0.02346133440732956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,8,balanced,0.025498665869235992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,8,balanced,0.025386666258176167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,8,balanced,0.027248000105222065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,8,balanced,0.02720533311367035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,8,balanced,0.02739199995994568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,8,balanced,0.030960001051425934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,8,power_law_1.01,0.06760960221290588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,64,power_law_1.2,0.8563199996948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,8,balanced,0.03123733401298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,8,balanced,0.029296000798543293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,8,balanced,0.0306986669699351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,8,balanced,0.03194133440653483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,8,balanced,0.045509333411852516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,8,balanced,0.04971733192602793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,8,balanced,0.051728000243504844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,8,balanced,0.06751466790835063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,8,balanced,0.06343466540177663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,8,balanced,0.07321600119272868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,8,balanced,0.08871466914812724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,8,balanced,0.12395733594894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,8,power_law_1.01,0.0863103985786438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,8,balanced,0.14647466937700906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,8,balanced,0.2038559913635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,8,balanced,0.26128000020980835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,8,power_law_1.01,0.09758080244064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,8,power_law_1.01,0.09247360229492188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,8,power_law_1.01,0.1227455973625183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,8,power_law_1.01,0.1544000029563904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,8,power_law_1.01,0.20040318965911866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,8,power_law_1.01,0.2773632049560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,8,power_law_1.01,0.3380863904953003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,8,power_law_1.2,0.04906879961490631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,8,power_law_1.2,0.049472001194953916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,8,power_law_1.2,0.05048959851264954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,8,power_law_1.2,0.061766397953033444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,8,power_law_1.2,0.06596480011940002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,8,power_law_1.2,0.06666880249977111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,8,power_law_1.2,0.0723583996295929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,8,power_law_1.2,0.073990398645401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,8,power_law_1.2,0.07191680073738098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,8,power_law_1.2,0.07533439993858337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,8,power_law_1.2,0.07884160280227662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,8,power_law_1.2,0.07986559867858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,8,power_law_1.2,0.08485760092735291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,8,power_law_1.2,0.09142400026321411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,8,power_law_1.2,0.10117759704589843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,8,power_law_1.2,0.10222079753875732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,8,power_law_1.2,0.11020159721374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,8,power_law_1.2,0.12706559896469116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,8,power_law_1.2,0.13611520528793336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,8,power_law_1.2,0.1659775972366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,8,power_law_1.2,0.19511040449142455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,8,power_law_1.2,0.2597631931304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,8,power_law_1.2,0.32085120677948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,8,power_law_1.2,0.47718400955200196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,8,power_law_1.2,0.5701375961303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,8,power_law_1.2,0.8778752326965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,8,power_law_1.01,0.5379776000976563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,8,power_law_1.2,1.0536512374877929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,8,power_law_1.01,0.7491456031799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,8,power_law_1.01,0.9769344329833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,8,power_law_1.01,0.07240319848060608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,8,power_law_1.01,0.08764799833297729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,8,power_law_1.01,0.07761920094490052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,8,power_law_1.01,0.08099200129508972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,8,power_law_1.01,0.07847679853439331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,8,power_law_1.01,0.09322239756584168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,8,power_law_1.01,0.09082880020141601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,8,power_law_1.01,0.0906495988368988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,8,power_law_1.01,0.09331200122833253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,8,power_law_1.01,0.095551997423172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,8,power_law_1.01,0.091430401802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,8,power_law_1.01,0.09237120151519776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,8,power_law_1.01,0.09610880017280579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,8,power_law_1.01,0.0964031994342804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,8,power_law_1.01,0.10421119928359986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,8,power_law_1.01,0.10631040334701539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,8,power_law_1.01,0.10900479555130005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,32,1,power_law_1.2,0.08307200074195861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,8,power_law_1.01,0.1239743947982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,32,1,power_law_1.2,0.08873599767684937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,8,power_law_1.01,0.13541760444641113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,32,1,power_law_1.2,0.09633920192718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,8,power_law_1.01,0.15964159965515137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,32,1,power_law_1.2,0.11733119487762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,8,power_law_1.01,0.18173439502716066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,32,1,power_law_1.2,0.13946880102157594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,8,power_law_1.01,0.2346496105194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,32,1,power_law_1.2,0.15796480178833008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,32,1,power_law_1.2,0.19827840328216553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,32,1,power_law_1.2,0.20650880336761473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,32,1,power_law_1.2,0.20706560611724853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,32,1,power_law_1.2,0.2164031982421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,32,1,power_law_1.2,0.22449920177459717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,32,1,power_law_1.2,0.23145599365234376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,8,power_law_1.01,0.2839807987213135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,32,1,power_law_1.2,0.2349247932434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,8,power_law_1.01,0.3856192111968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,32,1,power_law_1.2,0.24591999053955077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,8,power_law_1.01,0.4442304134368896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,32,1,power_law_1.2,0.2557055950164795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,32,1,power_law_1.2,0.2638015985488892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,8,power_law_1.01,0.6603328227996826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,32,1,power_law_1.2,0.28533120155334474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,32,1,power_law_1.2,0.33470079898834226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,32,1,power_law_1.2,0.3728832006454468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,32,1,power_law_1.2,0.4600512027740479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,32,1,power_law_1.2,0.5387584209442139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,8,power_law_1.01,1.51145601272583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,32,1,power_law_1.2,0.7146560192108155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,8,power_law_1.01,0.8752511978149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,32,1,power_law_1.2,0.8817791938781738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,32,1,power_law_1.2,1.216825580596924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,32,1,power_law_1.2,1.5585920333862304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,32,1,power_law_1.2,2.22540168762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,32,1,power_law_1.2,2.8936511993408205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,8,power_law_1.01,1.8152511596679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,32,power_law_1.2,0.09541119933128357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,32,power_law_1.2,0.1135103940963745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,32,power_law_1.2,0.10929280519485474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,32,power_law_1.2,0.1145408034324646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,32,power_law_1.2,0.11493760347366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,32,power_law_1.2,0.10895999670028686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,32,power_law_1.2,0.09598079919815064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,32,power_law_1.2,0.1087488055229187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,32,power_law_1.2,0.10974080562591552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,32,power_law_1.2,0.1108288049697876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,32,power_law_1.2,0.10886399745941162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,32,power_law_1.2,0.10839040279388427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,32,power_law_1.2,0.11268479824066162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,32,power_law_1.2,0.10736000537872314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,32,power_law_1.2,0.11754879951477051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,32,power_law_1.2,0.12318719625473022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,32,power_law_1.2,0.12599680423736573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,32,power_law_1.2,0.1361407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,32,power_law_1.2,0.14644479751586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,32,power_law_1.2,0.17878400087356566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,32,power_law_1.2,0.19583359956741334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,32,power_law_1.2,0.2479680061340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,32,power_law_1.2,0.30354559421539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,32,power_law_1.2,0.4229248046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,32,power_law_1.2,0.5205120086669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,32,power_law_1.2,0.7328447818756103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,32,power_law_1.2,0.9532608032226563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,1,power_law_1.2,0.054054397344589236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,1,power_law_1.2,0.07062399983406067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,1,power_law_1.2,0.08895360231399536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,1,power_law_1.2,0.13197439908981323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,1,power_law_1.2,0.16758400201797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,1,power_law_1.2,0.21587839126586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,1,power_law_1.2,0.299564790725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,1,power_law_1.2,0.3129280090332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,1,power_law_1.2,0.3297343969345093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,1,power_law_1.2,0.3371072053909302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,1,power_law_1.2,0.35119359493255614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,1,power_law_1.2,0.3712768077850342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,1,power_law_1.2,0.3732991933822632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,1,power_law_1.2,0.3857728004455566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,1,power_law_1.2,0.4133120059967041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,1,power_law_1.2,0.4228672027587891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,1,power_law_1.2,0.4580992221832275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,1,power_law_1.2,0.5169216156005859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,1,power_law_1.2,0.5503551959991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,1,power_law_1.2,0.6424960136413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,1,power_law_1.2,0.7331840038299561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,1,power_law_1.2,0.9294015884399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,1,power_law_1.2,1.0546303749084474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,1,power_law_1.2,1.3511232376098632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,1,power_law_1.2,1.6445184707641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,1,power_law_1.2,2.231411170959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,1,power_law_1.2,2.839289665222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,32,balanced,0.04855999847253164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,32,balanced,0.0480373352766037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,32,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,32,balanced,0.0562720000743866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,32,balanced,0.062074666221936546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,32,balanced,0.07226666808128357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,32,balanced,0.07115200161933899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,32,balanced,0.07223999996980031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,32,balanced,0.07044800122578938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,32,balanced,0.07052266597747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,32,balanced,0.07097066442171733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,32,balanced,0.0705813318490982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,32,balanced,0.07257600128650665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,32,balanced,0.07249600191911061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,32,balanced,0.07665599882602692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,32,balanced,0.0775626649459203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,32,balanced,0.08045866588751475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,32,balanced,0.08685333530108134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,32,balanced,0.09333333373069763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,32,balanced,0.10718400279680888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,32,balanced,0.12170132994651794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,32,balanced,0.150218665599823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,32,balanced,0.1741973360379537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,32,balanced,0.23333332935969034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,32,balanced,0.28461867570877075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,32,balanced,0.39049065113067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,32,balanced,0.49927465120951336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,1,power_law_1.01,0.036051198840141296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,1,power_law_1.01,0.049465599656105044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,1,power_law_1.01,0.06638079881668091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,1,power_law_1.01,0.09462400078773499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,1,power_law_1.01,0.11725440025329589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,1,power_law_1.01,0.17605119943618774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,1,power_law_1.01,0.18273279666900635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,1,power_law_1.01,0.19367040395736695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,1,power_law_1.01,0.19686399698257445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,1,power_law_1.01,0.19941760301589967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,1,power_law_1.01,0.20763518810272216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,1,power_law_1.01,0.2076416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,1,power_law_1.01,0.21271679401397706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,1,power_law_1.01,0.2220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,1,power_law_1.01,0.2232127904891968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,1,power_law_1.01,0.23218560218811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,1,power_law_1.01,0.24296960830688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,1,power_law_1.01,0.266214394569397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,1,power_law_1.01,0.2744960069656372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,1,power_law_1.01,0.3175679922103882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,1,power_law_1.01,0.3545408010482788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,1,power_law_1.01,0.4304192066192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,1,power_law_1.01,0.4742591857910156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,1,power_law_1.01,0.605951976776123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,1,power_law_1.01,0.7475903987884521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,1,power_law_1.01,1.020524787902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,1,power_law_1.01,1.2665023803710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,2,power_law_1.2,0.04501760005950928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,2,power_law_1.2,0.052275198698043826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,2,power_law_1.2,0.06416000127792358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,2,power_law_1.2,0.07534719705581665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,2,power_law_1.2,0.076665598154068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,2,power_law_1.2,0.07875840067863464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,2,power_law_1.2,0.08310400247573853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,2,power_law_1.2,0.08042240142822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,16,power_law_1.01,0.06782720088958741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,2,power_law_1.2,0.08559359908103943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,2,power_law_1.2,0.08465920090675354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,16,power_law_1.01,0.07555199861526489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,16,power_law_1.01,0.0754688024520874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,16,power_law_1.01,0.08155519962310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,16,power_law_1.01,0.08142079710960388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,2,power_law_1.2,0.08842880129814149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,16,power_law_1.01,0.07901440262794494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,16,power_law_1.01,0.07650560140609741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,16,power_law_1.01,0.07996159791946411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,16,power_law_1.01,0.08162559866905213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,16,power_law_1.01,0.081632000207901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,16,power_law_1.01,0.0825215995311737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,16,power_law_1.01,0.08549119830131531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,16,power_law_1.01,0.08854399919509888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,16,power_law_1.01,0.09014400243759155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,16,power_law_1.01,0.09812480211257935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,2,power_law_1.2,0.09142400026321411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,16,power_law_1.01,0.09953920245170593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,16,power_law_1.01,0.10945919752120972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,16,power_law_1.01,0.12280960083007812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,16,power_law_1.01,0.13445760011672975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,16,power_law_1.01,0.16934399604797362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,16,power_law_1.01,0.19563519954681396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,16,power_law_1.01,0.2575743913650513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,2,power_law_1.2,0.09110400080680847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,16,power_law_1.01,0.2668800115585327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,2,power_law_1.2,0.10038399696350098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,16,power_law_1.01,0.3751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,2,power_law_1.2,0.1082111954689026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,16,power_law_1.01,0.43018240928649903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,2,power_law_1.2,0.11504640579223632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,16,power_law_1.01,0.5994175910949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,2,power_law_1.2,0.12312959432601929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,16,power_law_1.01,0.732096004486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,2,power_law_1.2,0.14420479536056519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,2,power_law_1.2,0.15987839698791503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,2,power_law_1.2,0.2257920026779175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,2,power_law_1.2,0.23489279747009278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,2,power_law_1.2,0.32660479545593263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,2,power_law_1.2,0.40316162109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,2,power_law_1.2,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,2,power_law_1.2,0.04540160000324249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,2,power_law_1.2,0.047193598747253415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,2,power_law_1.2,0.055180799961090085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,2,power_law_1.2,0.05480319857597351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,2,power_law_1.2,0.05360640287399292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,2,power_law_1.2,0.05480960011482239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,2,power_law_1.2,0.054662400484085084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,2,power_law_1.2,0.05330560207366943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,2,power_law_1.2,0.05322239995002746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,2,power_law_1.2,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,2,power_law_1.2,0.05509759783744812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,2,power_law_1.2,0.056729602813720706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,2,power_law_1.2,0.05912320017814636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,2,power_law_1.2,0.06416640281677247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,2,power_law_1.2,0.06850559711456299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,2,power_law_1.2,0.07420799732208253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,2,power_law_1.2,0.08720639944076539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.2,0.09556480050086975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.2,0.11107840538024902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,2,power_law_1.2,0.5095295906066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.2,0.1343168020248413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.2,0.16586240530014038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,2,power_law_1.2,0.7832448005676269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.2,0.22465920448303223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,2,power_law_1.2,1.0901568412780762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.2,0.2817984104156494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.2,0.4253119945526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,2,power_law_1.2,1.3516608238220216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,64,power_law_1.01,0.030329599976539612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,64,power_law_1.01,0.0282368004322052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,64,power_law_1.01,0.026899200677871705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,64,power_law_1.01,0.02874239981174469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,64,power_law_1.01,0.029369598627090453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,64,power_law_1.01,0.030476799607276915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,64,power_law_1.01,0.030681601166725157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,64,power_law_1.01,0.03028480112552643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,64,power_law_1.01,0.03091840147972107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,64,power_law_1.01,0.0313728004693985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,64,power_law_1.01,0.03162240087985992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,64,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,64,power_law_1.01,0.0331712007522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,64,power_law_1.01,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,64,power_law_1.01,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,64,power_law_1.01,0.03866879940032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,64,power_law_1.01,0.03991680145263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,64,power_law_1.01,0.04237439930438995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,64,power_law_1.01,0.04527359902858734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,64,power_law_1.01,0.0497408002614975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,64,power_law_1.01,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,64,power_law_1.01,0.07252479791641235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,64,power_law_1.01,0.07934079766273498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,64,power_law_1.01,0.10718719959259033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,64,power_law_1.01,0.13729920387268066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,64,power_law_1.01,0.18498560190200805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,64,power_law_1.01,0.22568318843841553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,16,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,16,balanced,0.051882664362589516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,16,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,16,balanced,0.058378666639328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,16,balanced,0.0663679987192154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,16,balanced,0.0855466624101003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,16,balanced,0.08668266733487447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,16,balanced,0.0851039985815684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,16,balanced,0.0885706643263499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,16,balanced,0.08725333213806152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,16,balanced,0.08889066179593404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,16,balanced,0.09065066774686177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,16,balanced,0.09058133761088054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,16,balanced,0.09242133299509685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,16,balanced,0.09950400392214458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,16,balanced,0.09995733698209126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,16,balanced,0.10326932867368062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,16,balanced,0.11746666828791301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,16,balanced,0.12333333492279053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,16,balanced,0.13612266381581625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,16,balanced,0.15215466419855753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,16,balanced,0.18171733617782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,16,balanced,0.21146132548650107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,16,balanced,0.2916693290074666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,16,balanced,0.3545386791229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,16,balanced,0.4933706521987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.2,0.5808000087738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,16,balanced,0.6463946501413981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,8,power_law_1.2,0.03463680148124695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,8,power_law_1.2,0.030847999453544616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,8,power_law_1.2,0.030099201202392577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,8,power_law_1.2,0.032332798838615416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,8,power_law_1.2,0.03246079981327057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,8,power_law_1.2,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,8,power_law_1.2,0.03529599905014038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,8,power_law_1.2,0.035385599732398985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,8,power_law_1.2,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,8,power_law_1.2,0.03570559918880463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,8,power_law_1.2,0.03653759956359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,8,power_law_1.2,0.03648000061511993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,8,power_law_1.2,0.0379967987537384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,8,power_law_1.2,0.03951359987258911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,8,power_law_1.2,0.04394879937171936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,8,power_law_1.2,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,8,power_law_1.2,0.04799999892711639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,8,power_law_1.2,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,8,power_law_1.2,0.06270080208778381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,8,power_law_1.2,0.06857600212097167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,8,power_law_1.2,0.08689280152320862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,8,power_law_1.2,0.11373440027236939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,8,power_law_1.2,0.14758399724960328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,8,power_law_1.2,0.19975680112838745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,8,power_law_1.2,0.24817919731140137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,8,power_law_1.2,0.3524224042892456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.2,0.6674431800842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,2,balanced,0.08905599514643352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,2,balanced,0.1156160036722819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,2,balanced,0.18280533949534097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,2,balanced,0.18652266263961792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,2,balanced,0.19268266359965006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,2,balanced,0.18690667549769083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,2,balanced,0.18695465723673502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,2,balanced,0.19722133874893188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,2,balanced,0.18282665808995566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,2,balanced,0.1791093349456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,2,balanced,0.18291199207305908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,2,balanced,0.1827626625696818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,2,balanced,0.1800853411356608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,2,balanced,0.16852800051371256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,2,balanced,0.15625600020090738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,2,balanced,0.15795733531316122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,2,balanced,0.14430399735768637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,2,balanced,0.2638240059216817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,2,balanced,0.2202933430671692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,2,balanced,0.2972799936930339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,2,balanced,0.364464004834493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,2,balanced,0.5198986530303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,2,balanced,0.6647040049235026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,2,balanced,0.964250644048055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,8,power_law_1.2,0.5334015846252441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,2,balanced,1.2630613644917805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,2,balanced,1.868783950805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,2,balanced,2.468501408894857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,32,power_law_1.01,0.09915519952774048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,32,power_law_1.01,0.030124801397323608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,32,power_law_1.01,0.028832000494003297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,32,power_law_1.01,0.03000960052013397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,32,power_law_1.01,0.02993920147418976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,32,power_law_1.01,0.031206399202346802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,32,power_law_1.01,0.031091201305389404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,32,power_law_1.01,0.031667199730873105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,32,power_law_1.01,0.03221760094165802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,32,power_law_1.01,0.03283840119838714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,32,power_law_1.01,0.03364480137825012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,32,power_law_1.01,0.03402239978313446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,32,power_law_1.01,0.034790399670600894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,32,power_law_1.01,0.03563520014286041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,32,power_law_1.01,0.03976959884166718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,32,power_law_1.01,0.04103040099143982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,32,power_law_1.01,0.0416703999042511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,32,power_law_1.01,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,32,power_law_1.01,0.04845440089702606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,32,power_law_1.01,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,32,power_law_1.01,0.05791360139846802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,32,power_law_1.01,0.07160959839820862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,32,power_law_1.01,0.08970879912376403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,32,power_law_1.01,0.12475520372390747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,32,power_law_1.01,0.14419840574264525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,32,power_law_1.01,0.2016511917114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,32,power_law_1.01,0.2648576021194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,2,power_law_1.2,0.08141440153121948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,2,power_law_1.2,0.09868800044059753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,2,power_law_1.2,0.1080191969871521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,2,power_law_1.2,0.14062080383300782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,2,power_law_1.2,0.13879040479660035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,2,power_law_1.2,0.14547200202941896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,2,power_law_1.2,0.15024640560150146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,2,power_law_1.2,0.1503551959991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,2,power_law_1.2,0.15410560369491577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,2,power_law_1.2,0.1648576021194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,2,power_law_1.2,0.15752320289611815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,2,power_law_1.2,0.16707839965820312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,2,power_law_1.2,0.1766144037246704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,2,power_law_1.2,0.18978559970855713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,2,power_law_1.2,0.19733760356903077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,2,power_law_1.2,0.206060791015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,2,power_law_1.2,0.24439680576324463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,2,power_law_1.2,0.27035520076751707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,2,power_law_1.2,0.33758718967437745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,2,power_law_1.2,0.39731199741363527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,2,power_law_1.2,0.48083839416503904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,2,power_law_1.2,0.6741568088531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,2,power_law_1.2,0.7703231811523438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,1,power_law_1.01,0.058259201049804685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,1,power_law_1.01,0.0763264000415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,1,power_law_1.01,0.12248320579528808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,1,power_law_1.01,0.16060800552368165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,1,power_law_1.01,0.20986878871917725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,1,power_law_1.01,0.3154623985290527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,1,power_law_1.01,0.33680000305175783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,1,power_law_1.01,0.3470016002655029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,1,power_law_1.01,0.34790399074554446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,1,power_law_1.01,0.3590912103652954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,1,power_law_1.01,0.35898239612579347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,1,power_law_1.01,0.3724224090576172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,1,power_law_1.01,0.3771647930145264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,1,power_law_1.01,0.38629119396209716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,1,power_law_1.01,0.40355839729309084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,1,power_law_1.01,0.40885119438171386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,1,power_law_1.01,0.4255360126495361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,1,power_law_1.01,0.47343997955322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,1,power_law_1.01,0.5139455795288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,1,power_law_1.01,0.5864831924438476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,1,power_law_1.01,0.620959997177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,1,power_law_1.01,0.7657983779907227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,1,power_law_1.01,0.9111295700073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,1,power_law_1.01,1.181107234954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,1,power_law_1.01,1.486412811279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,1,power_law_1.01,2.043673515319824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,1,power_law_1.01,2.5879999160766602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,8,balanced,0.03682666768630346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,8,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,2,power_law_1.2,1.2444095611572266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,8,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,8,balanced,0.04085866610209147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,8,balanced,0.045935998360315956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,8,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,8,balanced,0.046906664967536926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,8,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,8,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,8,balanced,0.04764266808827718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,8,balanced,0.0481333335240682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,8,balanced,0.047882666190465294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,8,balanced,0.05004799862702688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,8,balanced,0.05177066723505656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,8,balanced,0.053210665782292686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,8,balanced,0.05610666672388712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,8,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,8,balanced,0.06293866535027821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,8,balanced,0.0661599983771642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,8,balanced,0.07660800218582153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,8,balanced,0.08251200119654338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,8,balanced,0.10880000392595927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,8,balanced,0.12970133622487387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,8,balanced,0.17485866943995157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,8,balanced,0.2140480081240336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,8,balanced,0.2940853238105774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,8,balanced,0.3760480086008708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,2,power_law_1.2,1.6116735458374023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,8,power_law_1.2,0.07292799949645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,8,power_law_1.2,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,8,power_law_1.2,0.07957760095596314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,8,power_law_1.2,0.0906112015247345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,8,power_law_1.2,0.08563200235366822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,8,power_law_1.2,0.09306880235671997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,8,power_law_1.2,0.09176959991455078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,8,power_law_1.2,0.09212160110473633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,8,power_law_1.2,0.09189760088920593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,8,power_law_1.2,0.096806401014328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,8,power_law_1.2,0.09404799938201905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,8,power_law_1.2,0.09461119771003723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,8,power_law_1.2,0.09809280037879944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,8,power_law_1.2,0.09931520223617554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,8,power_law_1.2,0.10414719581604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,8,power_law_1.2,0.10703359842300415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,8,power_law_1.2,0.11244159936904907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,8,power_law_1.2,0.12842240333557128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,8,power_law_1.2,0.14307199716567992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,8,power_law_1.2,0.16547839641571044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,8,power_law_1.2,0.18735359907150267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,8,power_law_1.2,0.24860799312591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,8,power_law_1.2,0.2926464080810547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,8,power_law_1.2,0.43442559242248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,8,power_law_1.2,0.5595583915710449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,8,power_law_1.2,0.7941760063171387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,8,power_law_1.2,1.0163840293884276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,2,power_law_1.2,2.3953535079956056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,128,power_law_1.2,0.057920002937316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,128,power_law_1.2,0.05677440166473389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,128,power_law_1.2,0.05676159858703613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,128,power_law_1.2,0.0599232017993927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,128,power_law_1.2,0.056543999910354616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,128,power_law_1.2,0.05914239883422852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,128,power_law_1.2,0.061504000425338747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,128,power_law_1.2,0.05996800065040588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,128,power_law_1.2,0.060083198547363284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,128,power_law_1.2,0.05983359813690185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,128,power_law_1.2,0.06247680187225342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,128,power_law_1.2,0.061401599645614625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,128,power_law_1.2,0.06279680132865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,128,power_law_1.2,0.06354560256004334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,128,power_law_1.2,0.06770560145378113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,128,power_law_1.2,0.06769919991493226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,128,power_law_1.2,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,128,power_law_1.2,0.08168960213661194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,128,power_law_1.2,0.09088640213012696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,128,power_law_1.2,0.11240960359573364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,128,power_law_1.2,0.12671999931335448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,128,power_law_1.2,0.17116160392761232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,128,power_law_1.2,0.19587839841842652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,128,power_law_1.2,0.27616000175476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,2,power_law_1.2,2.6630016326904298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,128,power_law_1.2,0.372544002532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,128,power_law_1.2,0.542790412902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,128,power_law_1.2,0.6492032051086426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,16,power_law_1.2,0.05400320291519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,16,power_law_1.2,0.046803200244903566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,16,power_law_1.2,0.04536960124969482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,16,power_law_1.2,0.047705599665641786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,16,power_law_1.2,0.04714879989624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,16,power_law_1.2,0.04872959852218628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,16,power_law_1.2,0.050483202934265135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,16,power_law_1.2,0.05089920163154602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,16,power_law_1.2,0.05156480073928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,16,power_law_1.2,0.05134080052375793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,16,power_law_1.2,0.05364480018615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,16,power_law_1.2,0.05367040038108826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,16,power_law_1.2,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,16,power_law_1.2,0.05902079939842224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,16,power_law_1.2,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,16,power_law_1.2,0.0698751986026764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,16,power_law_1.2,0.07415040135383606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,16,power_law_1.2,0.0858560025691986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,16,power_law_1.2,0.09435520172119141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,16,power_law_1.2,0.11840640306472779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,16,power_law_1.2,0.13565440177917482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,16,power_law_1.2,0.16671359539031982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,16,power_law_1.2,0.21710081100463868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,16,power_law_1.2,0.2925055980682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,16,power_law_1.2,0.36687359809875486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,16,power_law_1.2,0.5355072021484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,1,power_law_1.2,0.0359935998916626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,1,power_law_1.2,0.04156160056591034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,1,power_law_1.2,0.05371519923210144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,1,power_law_1.2,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,1,power_law_1.2,0.08481280207633972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,1,power_law_1.2,0.1223296046257019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,1,power_law_1.2,0.12930560111999512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,1,power_law_1.2,0.13621120452880858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,1,power_law_1.2,0.14251519441604615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,1,power_law_1.2,0.14266239404678344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,1,power_law_1.2,0.1512063980102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,1,power_law_1.2,0.15549440383911134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,1,power_law_1.2,0.16157439947128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,1,power_law_1.2,0.16440320014953613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,1,power_law_1.2,0.17330559492111205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,1,power_law_1.2,0.18328959941864015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,1,power_law_1.2,0.19881600141525269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,1,power_law_1.2,0.23311998844146728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,1,power_law_1.2,0.2651968002319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,1,power_law_1.2,0.33597440719604493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,1,power_law_1.2,0.3867775917053223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,1,power_law_1.2,0.5153664112091064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,1,power_law_1.2,0.6351935863494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,1,power_law_1.2,0.8818112373352051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,16,power_law_1.2,0.6945663928985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,1,power_law_1.2,1.1184000015258788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,1,power_law_1.2,1.594438362121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,16,power_law_1.2,0.055257600545883176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,16,power_law_1.2,0.06911360025405884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,16,power_law_1.2,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,16,power_law_1.2,0.07145599722862243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,1,power_law_1.2,2.113113594055176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,16,power_law_1.2,0.07046399712562561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,16,power_law_1.2,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,16,power_law_1.2,0.06865280270576476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,16,power_law_1.2,0.07015039920806884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,16,power_law_1.2,0.07322880029678344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,16,power_law_1.2,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,16,power_law_1.2,0.0710528016090393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,16,power_law_1.2,0.07464320063591004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,16,power_law_1.2,0.07662720084190369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,16,power_law_1.2,0.07712640166282654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,16,power_law_1.2,0.09000959992408752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,16,power_law_1.2,0.0902079999446869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,16,power_law_1.2,0.0965503990650177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,16,power_law_1.2,0.11312639713287354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,16,power_law_1.2,0.11896320581436157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,16,power_law_1.2,0.15000959634780883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,16,power_law_1.2,0.1747712016105652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,16,power_law_1.2,0.23248000144958497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,16,power_law_1.2,0.28556160926818847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,16,power_law_1.2,0.4124095916748047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,16,power_law_1.2,0.5537024021148682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,16,power_law_1.2,0.8170111656188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,16,power_law_1.2,1.1095616340637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,16,power_law_1.2,0.10641280412673951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,16,power_law_1.2,0.15077120065689087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,16,power_law_1.2,0.13345919847488402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,16,power_law_1.2,0.14425599575042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,16,power_law_1.2,0.14232959747314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,16,power_law_1.2,0.12408319711685181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,16,power_law_1.2,0.13288960456848145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,16,power_law_1.2,0.13672959804534912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,16,power_law_1.2,0.1385472059249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,16,power_law_1.2,0.14372479915618896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,16,power_law_1.2,0.14338560104370118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,16,power_law_1.2,0.14341119527816773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,16,power_law_1.2,0.15041919946670532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,16,power_law_1.2,0.15217920541763305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,16,power_law_1.2,0.157478404045105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,16,power_law_1.2,0.1723647952079773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,16,power_law_1.2,0.18437119722366332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,16,power_law_1.2,0.21308159828186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,16,power_law_1.2,0.21998720169067382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,16,power_law_1.2,0.2877824068069458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,16,power_law_1.2,0.2888511896133423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,16,power_law_1.2,0.3655296087265015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,16,power_law_1.2,0.46222720146179197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,16,power_law_1.2,0.5653503894805908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,16,power_law_1.2,0.7532224178314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,16,power_law_1.2,1.1163711547851562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,16,power_law_1.2,1.4318911552429199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,64,balanced,0.032144000132878624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,64,balanced,0.029893333713213604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,64,balanced,0.02811199923356374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,64,balanced,0.03181333343187968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,64,balanced,0.03398933261632919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,64,balanced,0.033999999364217125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,64,balanced,0.03226666649182638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,64,balanced,0.03401600072781245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,64,balanced,0.033930666744709015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,64,balanced,0.0340693344672521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,64,balanced,0.03405333310365677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,64,balanced,0.03422933320204417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,64,balanced,0.03589333345492681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,64,balanced,0.03615466753641764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,64,balanced,0.03812800099452337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,64,balanced,0.04009066770474116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,64,balanced,0.039861333866914116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,64,balanced,0.04165333261092504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,64,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,64,balanced,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,64,balanced,0.047781333327293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,64,balanced,0.05726933479309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,64,balanced,0.06228266656398773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,64,balanced,0.09446932872136433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,64,balanced,0.10666132966677348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,64,balanced,0.14261333147684732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,64,balanced,0.17258665959040323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,4,power_law_1.2,0.031699201464653014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,4,power_law_1.2,0.03159680068492889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,4,power_law_1.2,0.03208959996700287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,4,power_law_1.2,0.035046398639678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,4,power_law_1.2,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,4,power_law_1.2,0.036550399661064145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,4,power_law_1.2,0.037299200892448425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,4,power_law_1.2,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,4,power_law_1.2,0.0381056010723114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,4,power_law_1.2,0.03891200125217438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,4,power_law_1.2,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,4,power_law_1.2,0.0425024002790451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,4,power_law_1.2,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,4,power_law_1.2,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,4,power_law_1.2,0.05490559935569763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,4,power_law_1.2,0.05916159749031067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,4,power_law_1.2,0.06672639846801758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,4,power_law_1.2,0.07741439938545228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,128,balanced,0.12588799993197122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,128,balanced,0.1418293317159017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,128,balanced,0.04757333298524221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,4,power_law_1.2,0.09745280146598816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,128,balanced,0.048351998130480446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,128,balanced,0.04758933186531067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,128,balanced,0.053743998209635414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,128,balanced,0.07042133311430614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,128,balanced,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,128,balanced,0.07082133491834004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,128,balanced,0.07214400172233582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,128,balanced,0.07254933317502339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,128,balanced,0.07453333338101704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,128,balanced,0.07789333164691925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,128,balanced,0.07829866806666057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,128,balanced,0.08518399794896443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,128,balanced,0.08643200000127156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,128,balanced,0.09263466795285542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,128,balanced,0.10222400228182475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,128,balanced,0.11169067025184631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,128,balanced,0.13133866588274637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,128,balanced,0.14763733744621277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,128,balanced,0.18769599994023642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,128,balanced,0.22503999869028726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,128,balanced,0.30028265714645386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,128,balanced,0.3776906728744507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,128,balanced,0.5201760133107504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,128,balanced,0.6727786858876547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,4,power_law_1.2,0.13040000200271606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,4,power_law_1.2,0.14926079511642457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,4,power_law_1.2,0.23558399677276612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,4,power_law_1.2,0.2849152088165283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,4,power_law_1.2,0.4218751907348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,4,power_law_1.2,0.5336192131042481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,16,power_law_1.2,0.06975359916687011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,16,power_law_1.2,0.0745024025440216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,16,power_law_1.2,0.07546240091323853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,16,power_law_1.2,0.08218240141868591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,16,power_law_1.2,0.08341119885444641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,16,power_law_1.2,0.08328319787979126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,16,power_law_1.2,0.07953280210494995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,16,power_law_1.2,0.08106880187988282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,16,power_law_1.2,0.08469120264053345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,16,power_law_1.2,0.08518400192260742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,16,power_law_1.2,0.08593279719352723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,16,power_law_1.2,0.08895999789237977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,16,power_law_1.2,0.08915200233459472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,16,power_law_1.2,0.09159039855003356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,16,power_law_1.2,0.10001920461654663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,16,power_law_1.2,0.10368000268936158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,16,power_law_1.2,0.11290240287780762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,16,power_law_1.2,0.12930560111999512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,16,power_law_1.2,0.13668479919433593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,16,power_law_1.2,0.1754688024520874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,16,power_law_1.2,0.2063744068145752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,16,power_law_1.2,0.2757055997848511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,16,power_law_1.2,0.29252479076385496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,16,power_law_1.2,0.3884991884231567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,16,power_law_1.2,0.4926464080810547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,4,power_law_1.2,0.8680447578430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,16,power_law_1.2,0.6746240139007569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,128,balanced,0.05401599903901418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,128,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,128,balanced,0.05392533540725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,128,balanced,0.05859733124574026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,128,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,128,balanced,0.06517333288987477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,128,balanced,0.06427200138568878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,128,balanced,0.06551999847094218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,128,balanced,0.06414933502674103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,128,balanced,0.06463466584682465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,128,balanced,0.06432533264160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,128,balanced,0.06634666522343953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,128,balanced,0.06771199901898702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,128,balanced,0.06636266907056172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,128,balanced,0.07431999842325847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,128,balanced,0.07326399783293407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,16,power_law_1.2,0.8552127838134765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,128,balanced,0.07587199906508128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,128,balanced,0.0831573357184728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,128,balanced,0.08896533648173015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,128,balanced,0.1011306643486023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,128,balanced,0.11154133081436157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,128,balanced,0.1439466675122579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,128,balanced,0.16857065757115683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,128,balanced,0.21981332699457803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,128,balanced,0.2680533329645793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,128,balanced,0.3696906566619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,128,balanced,0.46767465273539227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,4,power_law_1.2,1.1316160202026366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,64,power_law_1.2,0.0632960021495819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,64,power_law_1.2,0.043161600828170776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,64,power_law_1.2,0.042259201407432556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,64,power_law_1.2,0.044940799474716187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,64,power_law_1.2,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,64,power_law_1.2,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,64,power_law_1.2,0.04622719883918762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,64,power_law_1.2,0.04689919948577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,64,power_law_1.2,0.04769920110702515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,64,power_law_1.2,0.047353601455688475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,64,power_law_1.2,0.048492801189422605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,64,power_law_1.2,0.05000960230827332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,64,power_law_1.2,0.05115519762039185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,64,power_law_1.2,0.053414398431777955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,64,power_law_1.2,0.057920002937316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,64,power_law_1.2,0.06004480123519897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,64,power_law_1.2,0.0629696011543274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,64,power_law_1.2,0.07145599722862243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,64,power_law_1.2,0.07714560031890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,64,power_law_1.2,0.0915391981601715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,64,power_law_1.2,0.10423680543899536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,64,power_law_1.2,0.13831039667129516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,64,power_law_1.2,0.16760319471359253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,64,power_law_1.2,0.23153278827667237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,64,power_law_1.2,0.30210559368133544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,64,power_law_1.2,0.45223679542541506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,64,power_law_1.2,0.5608128070831299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,32,balanced,0.054005334774653115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,32,balanced,0.055258666475613914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,32,balanced,0.055248002211252846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,32,balanced,0.06525866687297821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,32,balanced,0.06980266670385997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,32,balanced,0.09273599584897359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,32,balanced,0.1174720029036204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,32,balanced,0.11434132854143779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,32,balanced,0.10936533411343892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,32,balanced,0.11615467071533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,32,balanced,0.11372799674669902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,32,balanced,0.11411733428637187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,32,balanced,0.11364799737930298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,32,balanced,0.1179200013478597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,32,balanced,0.12632532914479574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,32,balanced,0.12421866257985432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,32,balanced,0.1267680029074351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,32,balanced,0.12796800335248312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,32,balanced,0.1451466679573059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,32,balanced,0.15996266404787698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,32,balanced,0.16899732748667398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,32,balanced,0.2046239972114563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,32,balanced,0.23537067572275797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,32,balanced,0.2995199958483378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,32,balanced,0.40531734625498456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,32,balanced,0.5249760150909424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,32,balanced,0.6831626892089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,32,balanced,0.05665066838264465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,32,balanced,0.04781333108743032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,32,balanced,0.047541335225105286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,32,balanced,0.0480373352766037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,32,balanced,0.052469333012898765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,32,balanced,0.07421866556008656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,32,balanced,0.0735093355178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,32,balanced,0.07665599882602692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,32,balanced,0.07618666688601176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,32,balanced,0.07640533149242401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,32,balanced,0.07621333499749501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,32,balanced,0.07857066889603932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,32,balanced,0.07650133470694225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,32,balanced,0.0790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,32,balanced,0.0844586690266927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,8,balanced,0.019445333629846573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,32,balanced,0.0848586658636729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,8,balanced,0.019082666685183842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,32,balanced,0.08691199620564778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,32,balanced,0.09501333038012187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,32,balanced,0.09610133369763692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,32,balanced,0.10706133643786113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,32,balanced,0.11858666936556499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,32,balanced,0.1402506629625956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,32,balanced,0.16220800081888834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,32,balanced,0.200490673383077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,32,balanced,0.2409706711769104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,32,balanced,0.3243146737416585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,32,balanced,0.4073280096054077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,8,balanced,0.018853332847356796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,8,balanced,0.019018666197856266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,8,balanced,0.019808000574509304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,8,balanced,0.020949333906173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,8,balanced,0.02126399924357732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,8,balanced,0.023157333334287006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,8,balanced,0.02514133354028066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,8,balanced,0.023061332603295643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,8,balanced,0.0313226655125618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,8,balanced,0.035232000052928925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,8,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,8,balanced,0.031727999448776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,8,balanced,0.0355679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,8,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,8,balanced,0.03548266738653183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,8,balanced,0.04736533264319102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,8,balanced,0.049626668294270836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,8,balanced,0.06432533264160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,8,balanced,0.07638933261235555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,8,balanced,0.09877333045005798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,8,balanced,0.12202133735020955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,8,balanced,0.1666826605796814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,8,balanced,0.20875734090805054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,8,balanced,0.2966880003611247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,8,balanced,0.38228801886240643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,1,power_law_1.01,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,1,power_law_1.01,0.03728640079498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,1,power_law_1.01,0.038764798641204835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,1,power_law_1.01,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,1,power_law_1.01,0.0585919976234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,1,power_law_1.01,0.05985280275344849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,1,power_law_1.01,0.06102399826049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,1,power_law_1.01,0.06103039979934692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,1,power_law_1.01,0.061952000856399535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,1,power_law_1.01,0.061561602354049685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,1,power_law_1.01,0.06509439945220948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,1,power_law_1.01,0.06778240203857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,1,power_law_1.01,0.06962559819221496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,1,power_law_1.01,0.07086719870567322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,1,power_law_1.01,0.08156160116195679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,1,power_law_1.01,0.08389760255813598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,1,power_law_1.01,0.08963840007781983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,1,power_law_1.01,0.10620160102844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,1,power_law_1.01,0.11594239473342896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,1,power_law_1.01,0.14636160135269166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,1,power_law_1.01,0.1717311978340149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,1,power_law_1.01,0.2302464008331299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,1,power_law_1.01,0.2764863967895508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,2,balanced,0.09319466352462769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,2,balanced,0.09504533807436626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,2,balanced,0.09954667091369629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,2,balanced,0.11545067032178243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,2,balanced,0.13291733463605246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,2,balanced,0.17463467518488565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,2,balanced,0.17509865760803223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,1,balanced,0.11218133568763733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,1,balanced,0.15077867110570273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,1,balanced,0.22272533178329468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,1,balanced,0.32468799750010174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,1,balanced,0.5286933183670044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,1,balanced,0.9486986796061198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,2,balanced,0.17489065726598105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,1,balanced,0.8982079823811849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,2,balanced,0.18017599980036417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,2,balanced,0.1786880095799764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,1,balanced,0.8700106938680013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,2,balanced,0.18035733699798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,2,balanced,0.18473066886266074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,2,balanced,0.1853760083516439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,2,balanced,0.18959999084472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,2,balanced,0.18997333447138467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,2,balanced,0.19185600678126016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,2,balanced,0.1990506649017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,2,balanced,0.21451733509699503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,2,balanced,0.22639467318852743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,2,balanced,0.2542240023612976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,2,balanced,0.28732266028722125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,1,balanced,0.8702293237050375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,2,balanced,0.34652264912923175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,1,balanced,0.8841919898986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,2,balanced,0.4161813259124756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,1,balanced,0.8862880071004232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,1,power_law_1.01,0.38844799995422363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,2,balanced,0.5875253280003866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,1,balanced,0.8345386981964111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,2,balanced,0.7141653696695963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,1,balanced,0.8432426452636719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,1,balanced,0.883957306543986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,2,balanced,1.0293707052866619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,1,balanced,0.8504906495412191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,2,balanced,1.3418134053548176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,1,balanced,0.8726239999135336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,1,balanced,0.9302240212758383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,1,balanced,0.9156800111134847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,1,balanced,0.9363839626312256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,1,balanced,1.0091573397318523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,1,balanced,1.0487093130747478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,1,balanced,1.2111732959747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,1,balanced,1.3708640734354656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,1,balanced,2.198970635732015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,1,balanced,2.4083305994669595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,1,power_law_1.01,0.4937280178070068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,1,balanced,3.3780800501505532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,1,balanced,4.410223960876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,4,balanced,0.029498666524887085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,4,balanced,0.029663999875386555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,4,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,4,balanced,0.035904000202814736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,4,balanced,0.05193600058555603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,4,balanced,0.060319999853769936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,4,balanced,0.06004266440868378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,4,balanced,0.06100266675154368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,4,balanced,0.062021334966023765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,4,balanced,0.0606826643149058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,4,balanced,0.0621013343334198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,4,balanced,0.06208533545335134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,4,balanced,0.06376533210277557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,4,balanced,0.06464000046253204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,4,balanced,0.06974400083223979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,4,balanced,0.07084799806276958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,4,balanced,0.07119999825954437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,4,balanced,0.07558399935563405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,4,balanced,0.07978133360544841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,4,balanced,0.09480533003807068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,4,balanced,0.10361599922180176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,4,balanced,0.12796266873677573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,4,balanced,0.1418719987074534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,4,balanced,0.19719467560450235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,4,balanced,0.2232746680577596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,4,balanced,0.3070506652196248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,4,balanced,0.37536001205444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,1,power_law_1.01,0.6981632232666015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,1,power_law_1.01,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,1,power_law_1.01,0.03959679901599884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,1,power_law_1.01,0.05389440059661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,1,power_law_1.01,0.07131519913673401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,1,power_law_1.01,0.0815168023109436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,1,power_law_1.01,0.11362559795379638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,1,power_law_1.01,0.12031359672546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,1,power_law_1.01,0.12346880435943604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,1,power_law_1.01,0.12626559734344484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,1,power_law_1.01,0.12700159549713136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,1,power_law_1.01,0.13435519933700563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,1,power_law_1.01,0.13915519714355468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,1,power_law_1.01,0.1434880018234253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,1,power_law_1.01,0.14873600006103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,1,power_law_1.01,0.1587007999420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,1,power_law_1.01,0.1669312000274658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,1,power_law_1.01,0.17697919607162477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,1,power_law_1.01,0.20281600952148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,1,power_law_1.01,0.22536959648132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,1,power_law_1.01,0.2760960102081299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,1,power_law_1.01,0.32718079090118407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,1,power_law_1.01,0.43408641815185545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,1,power_law_1.01,0.5455935955047607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,1,power_law_1.01,0.7473279953002929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,1,power_law_1.01,0.9561280250549317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,1,power_law_1.01,1.3740032196044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,1,power_law_1.01,1.7908031463623046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,1,power_law_1.01,0.9078975677490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,4,power_law_1.2,0.019782400131225585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,4,power_law_1.2,0.019289599359035493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,4,power_law_1.2,0.018751999735832213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,4,power_law_1.2,0.019276799261569978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,4,power_law_1.2,0.019814400374889372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,4,power_law_1.2,0.020211200416088104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,4,power_law_1.2,0.02536959946155548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,4,power_law_1.2,0.023391999304294586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,4,power_law_1.2,0.02815360128879547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,4,power_law_1.2,0.026399999856948853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,4,power_law_1.2,0.02733440101146698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,4,power_law_1.2,0.029657599329948426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,4,power_law_1.2,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,4,power_law_1.2,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,4,power_law_1.2,0.04394879937171936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,4,power_law_1.2,0.04232319891452789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,4,power_law_1.2,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,4,power_law_1.2,0.05689600110054016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.2,0.06510720252990723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,2,power_law_1.2,0.06215680241584778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,2,power_law_1.2,0.06392959952354431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,2,power_law_1.2,0.07164160013198853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,2,power_law_1.2,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.2,0.08617600202560424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,1,power_law_1.2,0.03597440123558045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,2,power_law_1.2,0.08867200016975403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,1,power_law_1.2,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,2,power_law_1.2,0.09018880128860474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,1,power_law_1.2,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,2,power_law_1.2,0.0898688018321991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,1,power_law_1.2,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,1,power_law_1.2,0.05854079723358154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,2,power_law_1.2,0.09148160219192505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,1,power_law_1.2,0.0781823992729187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,2,power_law_1.2,0.09126399755477906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,1,power_law_1.2,0.08120319843292237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,1,power_law_1.2,0.08403840065002441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,1,power_law_1.2,0.08659200072288513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,1,power_law_1.2,0.08869119882583618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,1,power_law_1.2,0.09601280093193054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,1,power_law_1.2,0.10140160322189332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,1,power_law_1.2,0.10722559690475464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,1,power_law_1.2,0.10700800418853759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,1,power_law_1.2,0.11603200435638428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,2,power_law_1.2,0.09201279878616334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,1,power_law_1.2,0.1219648003578186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,1,power_law_1.2,0.13665920495986938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,2,power_law_1.2,0.09180160164833069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,1,power_law_1.2,0.16678400039672853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,2,power_law_1.2,0.1074112057685852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,1,power_law_1.2,0.20181119441986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.2,0.10756479501724243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,2,power_law_1.2,0.11253759860992432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,2,power_law_1.2,0.11562240123748779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,2,power_law_1.2,0.1282495975494385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,1,power_law_1.2,0.25804800987243653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,2,power_law_1.2,0.12944639921188356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,2,power_law_1.2,0.147052800655365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,1,power_law_1.2,0.3109247922897339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,2,power_law_1.2,0.18114559650421141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,1,power_law_1.2,0.424505615234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.2,0.14522240161895753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,2,power_law_1.2,0.20247039794921876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,1,power_law_1.2,0.5383232116699219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,2,power_law_1.2,0.24291839599609374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,1,power_law_1.2,0.7561791896820068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,2,power_law_1.2,0.33239040374755857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.2,0.17831039428710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,1,power_law_1.2,0.9850624084472657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,2,power_law_1.2,0.428115177154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.2,0.2534847974777222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,1,power_law_1.2,1.4281087875366212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,2,power_law_1.2,0.683519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.2,0.3263040065765381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,2,power_law_1.2,0.7779839992523193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,1,power_law_1.2,1.863724708557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,4,power_law_1.01,0.022368000447750093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,4,power_law_1.01,0.021510399878025055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,4,power_law_1.01,0.021408000588417055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,4,power_law_1.01,0.02215680032968521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,4,power_law_1.01,0.023071999847888946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,4,power_law_1.01,0.023372800648212434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,4,power_law_1.01,0.029791998863220214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,4,power_law_1.01,0.025856000185012818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,4,power_law_1.01,0.05172479748725891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,4,power_law_1.01,0.050886398553848265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,4,power_law_1.01,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,4,power_law_1.01,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,4,power_law_1.01,0.05041279792785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,4,power_law_1.01,0.04032639861106872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,4,power_law_1.01,0.06318079829216003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,4,power_law_1.01,0.05258240103721619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,4,power_law_1.01,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,4,power_law_1.01,0.07286400198936463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.01,0.09216639995574952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.01,0.12430720329284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.2,0.4733888149261475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,8,power_law_1.01,0.05559679865837097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.01,0.1503424048423767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,8,power_law_1.01,0.05138559937477112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,8,power_law_1.01,0.05358719825744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,8,power_law_1.01,0.05676159858703613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,8,power_law_1.01,0.054739201068878175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,8,power_law_1.01,0.06090239882469177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,32,balanced,0.023141334454218548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,32,balanced,0.021007999777793884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,32,balanced,0.021269333859284718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,32,balanced,0.02515733242034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,32,balanced,0.021301334102948506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,32,balanced,0.022986667851607006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,32,balanced,0.024149333437283833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,32,balanced,0.02532266577084859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,32,balanced,0.0295413335164388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,32,balanced,0.029279999434947968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,32,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,32,balanced,0.043621331453323364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,32,balanced,0.04177600145339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,32,balanced,0.041893333196640015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,32,balanced,0.05109333495299021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,32,balanced,0.05172266562779745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,32,balanced,0.049728001157442726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,32,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,32,balanced,0.05914666752020518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,32,balanced,0.08481066425641377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,32,balanced,0.09646399815877278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,2,power_law_1.2,1.1310079574584961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,8,power_law_1.01,0.06050559878349304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,8,power_law_1.01,0.06053760051727295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,8,power_law_1.01,0.06279680132865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,8,power_law_1.01,0.06247680187225342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,8,power_law_1.01,0.060678398609161376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,8,power_law_1.01,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,8,power_law_1.01,0.06405760049819946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,8,power_law_1.01,0.06516479849815368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,8,power_law_1.01,0.07104640007019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.01,0.21052160263061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,8,power_law_1.01,0.06982399821281433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,8,power_law_1.01,0.07511680126190186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,8,power_law_1.01,0.08583679795265198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.01,0.0954688012599945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.01,0.1074560046195984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.01,0.13313920497894288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.01,0.15705599784851074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,32,balanced,0.12918399771054587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,32,balanced,0.15330666303634644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.01,0.1908735990524292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,32,balanced,0.2183039983113607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,32,balanced,0.2762773235638936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,32,balanced,0.39694400628407794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,32,balanced,0.5201546748479208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.01,0.26725120544433595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.01,0.2728512048721313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.01,0.3542655944824219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.01,0.47771520614624025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.01,0.6375232219696045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.01,0.3803200006484985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,2,power_law_1.01,0.01960960030555725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,2,power_law_1.01,0.01966080069541931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,2,power_law_1.01,0.020076799392700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,2,power_law_1.01,0.02120320051908493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,2,power_law_1.01,0.023481599986553192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,2,power_law_1.01,0.02393600046634674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,2,power_law_1.01,0.023571200668811798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,2,power_law_1.01,0.025068798661231996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,2,power_law_1.01,0.02691200077533722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.01,0.4961408138275146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,2,power_law_1.01,0.02540160119533539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,2,power_law_1.2,1.5064895629882813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,2,power_law_1.01,0.027935999631881713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,2,power_law_1.01,0.05053439736366272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.2,0.6216256141662597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,2,power_law_1.01,0.05059199929237366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,2,power_law_1.01,0.053311997652053834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,2,power_law_1.01,0.04348160028457641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,2,power_law_1.01,0.03889279961585999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,2,power_law_1.01,0.04225279986858368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,2,power_law_1.01,0.05578879714012146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.01,0.0608959972858429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.01,0.0827072024345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.01,0.10216959714889526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.01,0.13533439636230468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.01,0.16451200246810913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,32,balanced,0.03170666595300039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,32,balanced,0.02961066613594691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,32,balanced,0.027535999814669292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,32,balanced,0.031354665756225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,32,balanced,0.03165333221356074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,32,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,32,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,32,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,32,balanced,0.03446933378775915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,32,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,32,balanced,0.03514666606982549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,32,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,32,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,32,balanced,0.03572266548871994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,32,balanced,0.039850667119026184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,32,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,32,balanced,0.04022933294375738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,32,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,32,balanced,0.043840001026789345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,32,balanced,0.045610666275024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.01,0.23578879833221436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,32,balanced,0.04783466458320618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,32,balanced,0.059018666545550026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,32,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,32,balanced,0.07958399752775829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,32,balanced,0.09622400005658467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,32,balanced,0.11691199739774068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,32,balanced,0.14124266306559244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.01,0.2992768049240112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,2,power_law_1.2,2.1029375076293944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.01,0.730131196975708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.01,0.43703041076660154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.01,0.9594880104064941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.01,0.5726848125457764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,16,power_law_1.2,0.060812801122665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,16,power_law_1.2,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,16,power_law_1.2,0.07591040134429931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,16,power_law_1.2,0.08247680068016053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,16,power_law_1.2,0.08448640108108521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,16,power_law_1.2,0.07773439884185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,16,power_law_1.2,0.08424320220947265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,16,power_law_1.2,0.08350080251693726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,16,power_law_1.2,0.082150399684906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,16,power_law_1.2,0.08526080250740051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,16,power_law_1.2,0.08746240139007569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,16,power_law_1.2,0.08843520283699036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,16,power_law_1.2,0.09059200286865235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,16,power_law_1.2,0.09296640157699584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,16,power_law_1.2,0.10076160430908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,16,power_law_1.2,0.1020799994468689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,16,power_law_1.2,0.10738559961318969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,16,power_law_1.2,0.12453759908676147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.2,0.13725440502166747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.2,0.16684160232543946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.2,0.1823040008544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.2,0.22663679122924804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.2,0.3174272060394287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.2,0.39706881046295167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.2,0.4892159938812256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.2,0.7627007961273193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.2,0.8853568077087403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,4,power_law_1.01,0.038726401329040525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,4,power_law_1.01,0.04056319892406464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,4,power_law_1.01,0.04113920032978058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,4,power_law_1.01,0.045407998561859134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,4,power_law_1.01,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,4,power_law_1.01,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,4,power_law_1.01,0.054771202802658084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,4,power_law_1.01,0.05508480072021484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,4,power_law_1.01,0.05428479909896851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,4,power_law_1.01,0.056441599130630495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,4,power_law_1.01,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,4,power_law_1.01,0.06020479798316956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,4,power_law_1.01,0.06269440054893494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,8,power_law_1.01,0.05059840083122254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,8,power_law_1.01,0.07189760208129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,8,power_law_1.01,0.09943040013313294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,4,power_law_1.01,0.06577280163764954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,8,power_law_1.01,0.0969983994960785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,4,power_law_1.01,0.07226880192756653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,4,power_law_1.01,0.07800319790840149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,8,power_law_1.01,0.11569279432296753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,8,power_law_1.01,0.12463999986648559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,4,power_law_1.01,0.08312960267066956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,8,power_law_1.01,0.12354559898376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,8,power_law_1.01,0.1209663987159729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,8,power_law_1.01,0.11860480308532714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,8,power_law_1.01,0.12263679504394531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,8,power_law_1.01,0.1267135977745056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,4,power_law_1.01,0.0979968011379242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,8,power_law_1.01,0.1254271984100342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,4,power_law_1.01,0.1032639980316162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,8,power_law_1.01,0.13192960023880004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,4,power_law_1.01,0.13048959970474244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,4,power_law_1.01,0.14401919841766359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,8,power_law_1.01,0.1361407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,4,power_law_1.01,0.1861631989479065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,8,power_law_1.01,0.14644479751586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,4,power_law_1.01,0.22639360427856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,8,power_law_1.01,0.143449604511261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,8,power_law_1.01,0.15415680408477783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,8,power_law_1.01,0.18077440261840821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,8,power_law_1.01,0.19068800210952758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,4,power_law_1.01,0.31341440677642823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,8,power_law_1.01,0.23375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,4,power_law_1.01,0.42933120727539065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,8,power_law_1.01,0.2651007890701294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,8,power_law_1.01,0.3292351961135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,4,power_law_1.01,0.6079232215881347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,8,power_law_1.01,0.4014143943786621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,8,power_law_1.01,0.5401088237762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,1,balanced,0.03583999971548716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,1,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,1,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,8,power_law_1.01,0.677561616897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,1,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,1,balanced,0.050026665131251015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,1,balanced,0.0503359983364741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,1,balanced,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,1,balanced,0.05169600248336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,1,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,4,power_law_1.01,0.7958015918731689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,1,balanced,0.050757333636283875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,1,balanced,0.05205333232879639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,1,balanced,0.05787733197212219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,1,balanced,0.0581279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,1,balanced,0.061424002051353455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,1,balanced,0.06625066697597504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,1,balanced,0.07042133311430614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,1,balanced,0.07254933317502339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,8,power_law_1.01,1.0041407585144042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,8,power_law_1.01,1.1933888435363769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,1,balanced,0.09307733178138733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,1,balanced,0.10230400164922078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,1,balanced,0.13222400347391763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,1,balanced,0.16009599963823953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,1,balanced,0.23776533206303915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,1,balanced,0.3155253330866496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,1,balanced,0.44573867321014404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,1,balanced,0.5724746783574423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,1,balanced,0.8262346585591634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,1,balanced,1.0841866334279378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,4,power_law_1.01,0.05003520250320435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,4,power_law_1.01,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,4,power_law_1.01,0.07525759935379028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,4,power_law_1.01,0.08113279938697815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,4,power_law_1.01,0.09605119824409485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,4,power_law_1.01,0.096697598695755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,4,power_law_1.01,0.10736000537872314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,4,power_law_1.01,0.1278272032737732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,4,power_law_1.01,0.1342527985572815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,4,power_law_1.01,0.12672640085220338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,4,power_law_1.01,0.140665602684021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,4,power_law_1.01,0.14187519550323485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,4,power_law_1.01,0.13975679874420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,4,power_law_1.01,0.14558080434799195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,4,power_law_1.01,0.1571328043937683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,4,power_law_1.01,0.15776000022888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,4,power_law_1.01,0.1602687954902649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,4,power_law_1.01,0.18151040077209474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,4,power_law_1.01,0.20453119277954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,4,power_law_1.01,0.24341120719909667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,4,power_law_1.01,0.2773695945739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,4,power_law_1.01,0.3587968111038208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,4,power_law_1.01,0.4456960201263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,4,power_law_1.01,0.5921728134155273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,4,power_law_1.01,0.7075967788696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,4,power_law_1.01,1.0001343727111816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,4,power_law_1.01,1.3218367576599122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,1,balanced,0.02737066646416982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,1,balanced,0.040864000717798867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,1,balanced,0.053674668073654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,1,balanced,0.07603199779987335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,1,balanced,0.12652267018953958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,1,balanced,0.2237493395805359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,1,balanced,0.2316853404045105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,1,balanced,0.23694399992624918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,1,balanced,0.24250133832295737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,1,balanced,0.24885867039362589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,1,balanced,0.25040533145268756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,1,balanced,0.25897600253423053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,1,balanced,0.26899200677871704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,1,balanced,0.26045332352320355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,1,balanced,0.26873066027959186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,1,balanced,0.277349332968394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,1,balanced,0.29605333010355633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,1,balanced,0.33661866188049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,1,balanced,0.34470399220784503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,1,balanced,0.42347200711568195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,1,balanced,0.4692586660385132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,1,balanced,0.619493325551351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,1,balanced,0.5578560034434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,1,balanced,0.7524106502532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,1,balanced,0.9464639822642008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,1,balanced,1.354431947072347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,1,balanced,1.7726880709330242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,8,power_law_1.01,0.10457600355148315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,8,power_law_1.01,0.14776320457458497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,4,power_law_1.01,0.04103040099143982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,4,power_law_1.01,0.03297280073165894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,4,power_law_1.01,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,8,power_law_1.01,0.13670400381088257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,4,power_law_1.01,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,4,power_law_1.01,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,8,power_law_1.01,0.16236159801483155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,4,power_law_1.01,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,4,power_law_1.01,0.03875199854373932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,4,power_law_1.01,0.04192639887332916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,4,power_law_1.01,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,4,power_law_1.01,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,4,power_law_1.01,0.047628799080848695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,4,power_law_1.01,0.04826880097389221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,8,power_law_1.01,0.1805184006690979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,4,power_law_1.01,0.04984959959983826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,4,power_law_1.01,0.054655998945236206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,8,power_law_1.01,0.19217920303344727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,8,power_law_1.01,0.18888959884643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,8,power_law_1.01,0.19017599821090697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,8,power_law_1.01,0.1818176031112671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,8,power_law_1.01,0.18131840229034424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,4,power_law_1.01,0.061849600076675414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,8,power_law_1.01,0.19292160272598266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,8,power_law_1.01,0.19598079919815065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,8,power_law_1.01,0.19534720182418824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,4,power_law_1.01,0.06535680294036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,8,power_law_1.01,0.2016319990158081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,4,power_law_1.01,0.07123839855194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,8,power_law_1.01,0.21349120140075684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,4,power_law_1.01,0.08195840120315552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,8,power_law_1.01,0.21660799980163575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,4,power_law_1.01,0.08568959832191467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,8,power_law_1.01,0.22799999713897706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,4,power_law_1.01,0.09865599870681763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,8,power_law_1.01,0.25793280601501467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,8,power_law_1.01,0.2920896053314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,8,power_law_1.01,0.3619456052780151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,8,power_law_1.01,0.39346559047698976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,8,power_law_1.01,0.5285312175750733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,4,power_law_1.01,0.13315839767456056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,8,power_law_1.01,0.6127552032470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,4,power_law_1.01,0.14469759464263915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,8,power_law_1.01,0.7905792236328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,8,power_law_1.01,0.9924351692199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,8,power_law_1.01,1.361580753326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,8,power_law_1.01,1.7309375762939454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,4,power_law_1.01,0.19379839897155762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,4,power_law_1.01,0.26321280002593994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,4,power_law_1.01,0.35739519596099856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,32,balanced,0.045893331368764244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,32,balanced,0.045738667249679565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,32,balanced,0.04781333108743032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,32,balanced,0.05073066552480062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,32,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,32,balanced,0.0614026685555776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,32,balanced,0.060453335444132485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,32,balanced,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,32,balanced,0.06406400104363759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,32,balanced,0.06433600187301636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,32,balanced,0.06356800099213918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,32,balanced,0.07113066812356313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,32,balanced,0.07124799986680348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,32,balanced,0.07222933570543925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,32,balanced,0.08785067001978557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,32,balanced,0.08969066540400188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,32,balanced,0.08559999863306682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,32,balanced,0.10051199793815613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,32,balanced,0.10610133409500122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,32,balanced,0.11963733037312825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,32,balanced,0.1343946655591329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,32,balanced,0.16434666514396667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,32,balanced,0.18928533792495728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,32,balanced,0.24286399284998575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,32,balanced,0.29337066411972046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,32,balanced,0.40323734283447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,32,balanced,0.5264426469802856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,4,power_law_1.01,0.4447616100311279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,128,power_law_1.2,0.06387199759483338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,128,power_law_1.2,0.06399359703063964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,128,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,128,power_law_1.2,0.05713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,128,power_law_1.2,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,128,power_law_1.2,0.0596671998500824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,128,power_law_1.2,0.059494400024414064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,128,power_law_1.2,0.05740799903869629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,128,power_law_1.2,0.058950400352478026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,128,power_law_1.2,0.05883520245552063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,128,power_law_1.2,0.059411197900772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,128,power_law_1.2,0.06828799843788147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,128,power_law_1.2,0.06988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,128,power_law_1.2,0.07175040245056152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,128,power_law_1.2,0.08108800053596496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,128,power_law_1.2,0.08247039914131164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,128,power_law_1.2,0.09117439985275269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,128,power_law_1.2,0.10903680324554443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,128,power_law_1.2,0.11453440189361572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,128,power_law_1.2,0.1461951971054077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,128,power_law_1.2,0.1656767964363098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,128,power_law_1.2,0.21441280841827393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,128,power_law_1.2,0.24377601146697997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,128,power_law_1.2,0.3373055934906006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,4,power_law_1.01,0.5663487911224365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,128,power_law_1.2,0.3799936056137085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,128,power_law_1.2,0.5490303993225097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,128,power_law_1.2,0.6302976131439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,32,power_law_1.01,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,32,power_law_1.01,0.044787201285362246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,32,power_law_1.01,0.0440064013004303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,1,balanced,0.01932266727089882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,32,power_law_1.01,0.044095999002456664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,1,balanced,0.01941866676012675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,1,balanced,0.01897066707412402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,32,power_law_1.01,0.04524799883365631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,1,balanced,0.01904533306757609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,1,balanced,0.01937066639463107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,1,balanced,0.020917333662509918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,32,power_law_1.01,0.04838399887084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,1,balanced,0.021317332983016968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,1,balanced,0.021717332303524017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,1,balanced,0.025055999557177227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,1,balanced,0.02918400118748347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,1,balanced,0.025450666745503742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,1,balanced,0.026426665484905243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,1,balanced,0.029818666477998097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,1,balanced,0.030261332790056866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,1,balanced,0.0316746657093366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,1,balanced,0.03551999976237615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,1,balanced,0.03624533365170161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,1,balanced,0.04725866516431173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,1,balanced,0.04974933465321859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,1,balanced,0.06603200236956279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,1,balanced,0.07831466694672902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,1,balanced,0.09943999846776326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,1,balanced,0.12332266569137573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,1,balanced,0.16699200868606567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,1,balanced,0.20908800760904947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,1,balanced,0.2962826689084371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,1,balanced,0.3843253453572591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,32,power_law_1.01,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,32,power_law_1.01,0.06475520133972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,32,power_law_1.01,0.0869376003742218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,32,power_law_1.01,0.08569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,32,power_law_1.01,0.08654080033302307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,32,power_law_1.01,0.0982591986656189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,32,power_law_1.01,0.09747200012207032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,32,power_law_1.01,0.09593600034713745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,32,power_law_1.01,0.09296000003814697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,32,power_law_1.01,0.09348480105400085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,32,power_law_1.01,0.09798399806022644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,32,power_law_1.01,0.10311039686203002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,32,power_law_1.01,0.09388160109519958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,32,power_law_1.01,0.11041280031204223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,32,power_law_1.01,0.13235199451446533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,32,power_law_1.01,0.15113600492477416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,32,power_law_1.01,0.19140479564666749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,32,power_law_1.01,0.27898879051208497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,32,power_law_1.01,0.3703104019165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,32,power_law_1.01,0.4871039867401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,32,power_law_1.01,0.6753536224365234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,128,balanced,0.064560001095136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,128,balanced,0.06413333117961884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,128,balanced,0.06412800153096516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,128,balanced,0.06640533109505971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,128,balanced,0.06545599798361461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,128,balanced,0.07433600227038066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,128,balanced,0.07367466886838277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,128,balanced,0.07488533357779185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,128,balanced,0.07468800246715546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,128,balanced,0.07464000085989635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,128,balanced,0.07798933486143748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,128,balanced,0.07494399944941203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,128,balanced,0.08428266644477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,128,balanced,0.08090666433175404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,128,balanced,0.08406399687131245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,128,balanced,0.08931733171145122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,128,balanced,0.08700266480445862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,128,balanced,0.0972106655438741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,128,balanced,0.10149332880973816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,128,balanced,0.11272533734639485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,128,balanced,0.1223360002040863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,128,balanced,0.1432213286558787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,128,balanced,0.16420267025629678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,128,balanced,0.20758932828903198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,128,balanced,0.24764800071716309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,128,balanced,0.33131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,128,balanced,0.41034666697184247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,16,balanced,0.04318933188915253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,16,balanced,0.04381333291530609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,16,balanced,0.04249600072701772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,16,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,16,balanced,0.04794666667779287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,16,balanced,0.06623466809590657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,16,balanced,0.06669333577156067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,16,balanced,0.06739733119805653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,16,balanced,0.06821333368619283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,16,balanced,0.06811733543872833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,16,balanced,0.06877333422501881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,16,balanced,0.0711946686108907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,16,balanced,0.07301333546638489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,16,balanced,0.07249066730340321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,16,balanced,0.07863999903202057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,16,balanced,0.07896533111731212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,16,balanced,0.0805866668621699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,16,balanced,0.08933867017428081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,16,balanced,0.09468266367912292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,16,balanced,0.10880000392595927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,16,balanced,0.12180800239245097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,16,balanced,0.1492533286412557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,16,balanced,0.18026133378346762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,16,balanced,0.23427200317382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,16,balanced,0.27954665819803876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,16,balanced,0.3856533368428548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,16,balanced,0.4959466854731242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,4,balanced,0.039994666973749794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,4,balanced,0.03986666599909464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,4,balanced,0.04010133445262909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,4,balanced,0.04557333389918009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,4,balanced,0.054048001766204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,4,balanced,0.05604266623655955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,4,balanced,0.055786664287249245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,4,balanced,0.056048000852266945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,4,balanced,0.056799997886021934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,4,balanced,0.05610666672388712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,4,balanced,0.05789333085219065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,4,balanced,0.0618453323841095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,4,balanced,0.06201600035031637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,4,balanced,0.062277331948280334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,4,balanced,0.06569600105285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,4,balanced,0.066021333138148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,4,balanced,0.0680213322242101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,4,balanced,0.07494399944941203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,4,balanced,0.07895466685295105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,4,balanced,0.09089600046475728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,4,balanced,0.09866666793823242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,4,balanced,0.12801067034403482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,4,balanced,0.14539200067520142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,4,balanced,0.19768534104029337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,4,balanced,0.2589120070139567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,4,balanced,0.3855146567026774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,4,balanced,0.48979198932647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,2,balanced,0.06917333106199901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,2,balanced,0.09428266684214275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,2,balanced,0.14415466785430908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,2,balanced,0.24489599466323853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,2,balanced,0.4405706723531087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,2,balanced,0.8415520191192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,2,balanced,0.8418292999267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,2,balanced,0.8384373188018799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,2,balanced,0.8421493371327718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,2,balanced,0.8428746859232584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,2,balanced,0.8500746885935465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,2,balanced,0.851253350575765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,2,balanced,0.8543787002563477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,1,power_law_1.01,0.018303999304771425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,2,balanced,0.8619679609934489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,2,balanced,0.869429349899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,2,balanced,0.8738453388214111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,2,balanced,0.8869813283284506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,2,balanced,0.9147679805755615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,2,balanced,0.9583360354105631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,2,balanced,1.0026826858520508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,1,power_law_1.01,0.01796479970216751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,1,power_law_1.01,0.017843200266361235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,2,balanced,1.0378133455912273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,1,power_law_1.01,0.017849600315093993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,1,power_law_1.01,0.018905599415302277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,2,balanced,1.1169493198394775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,1,power_law_1.01,0.019097599387168884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,1,power_law_1.01,0.019942399859428406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,2,balanced,1.247226635615031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,1,power_law_1.01,0.020659199357032774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,1,power_law_1.01,0.023379200696945192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,2,balanced,1.4559094111124675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,2,balanced,1.580410639444987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,2,balanced,2.476031939188639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,1,power_law_1.01,0.028832000494003297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,2,balanced,2.563093344370524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,1,power_law_1.01,0.02456959933042526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,1,power_law_1.01,0.025312000513076784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,1,power_law_1.01,0.028921601176261903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,1,power_law_1.01,0.0293503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,1,power_law_1.01,0.030924800038337707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,1,power_law_1.01,0.034668800234794614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,1,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,1,power_law_1.01,0.045824000239372255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.01,0.04821760058403015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.01,0.06369919776916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.01,0.07684479951858521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.01,0.09737600088119507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.01,0.12030080556869507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,1,power_law_1.01,0.07208960056304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,1,power_law_1.01,0.09329919815063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,1,power_law_1.01,0.1023743987083435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,1,power_law_1.01,0.11790080070495605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,16,power_law_1.2,0.08044800162315369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,1,power_law_1.01,0.13126399517059326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,16,power_law_1.2,0.10899200439453124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,1,power_law_1.01,0.16293120384216309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,1,power_law_1.01,0.1681920051574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,1,power_law_1.01,0.17352960109710694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,1,power_law_1.01,0.17768319845199584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,1,power_law_1.01,0.1811519980430603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,1,power_law_1.01,0.18272000551223755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,1,power_law_1.01,0.18908159732818602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,16,power_law_1.2,0.10381439924240113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,1,power_law_1.01,0.1914240002632141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,1,power_law_1.01,0.19573760032653809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,16,power_law_1.2,0.11157759428024291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,1,power_law_1.01,0.1931071996688843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.01,0.16537599563598632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,16,power_law_1.2,0.11140480041503906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,16,power_law_1.2,0.1059008002281189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,16,power_law_1.2,0.11113599538803101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,1,power_law_1.01,0.19921280145645143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,1,power_law_1.01,0.20823678970336915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,1,power_law_1.01,0.22577919960021972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,1,power_law_1.01,0.24786560535430907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,1,power_law_1.01,0.2967871904373169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,1,power_law_1.01,0.3549312114715576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,16,power_law_1.2,0.10931839942932128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,1,power_law_1.01,0.4592639923095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,1,power_law_1.01,0.5822080135345459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,1,power_law_1.01,0.8102463722229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.01,0.20611200332641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,1,power_law_1.01,1.0237631797790527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,16,power_law_1.2,0.10547840595245361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,16,power_law_1.2,0.11390080451965331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,16,power_law_1.2,0.11042560338973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,16,power_law_1.2,0.11291520595550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,1,power_law_1.01,1.4682239532470702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,16,power_law_1.2,0.11575039625167846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,1,power_law_1.01,1.9373056411743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,16,power_law_1.2,0.11648000478744507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,16,power_law_1.2,0.12636799812316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,16,power_law_1.2,0.12874239683151245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,16,power_law_1.2,0.13379199504852296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,16,power_law_1.2,0.149344003200531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,16,power_law_1.2,0.1646399974822998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,16,power_law_1.2,0.18908159732818602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,16,power_law_1.2,0.22902400493621827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,16,power_law_1.2,0.2811199903488159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,16,power_law_1.2,0.3778048038482666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,16,power_law_1.2,0.5321792125701904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,16,power_law_1.2,0.6931072235107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,16,power_law_1.2,0.9301247596740723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,16,power_law_1.2,1.2453375816345216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.01,0.29361920356750487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.01,0.3808831930160522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,8,power_law_1.2,0.022438399493694305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,8,power_law_1.2,0.021984000504016877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,8,power_law_1.2,0.021510399878025055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,8,power_law_1.2,0.02152319997549057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,8,power_law_1.2,0.022316800057888032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,8,power_law_1.2,0.0247296005487442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,8,power_law_1.2,0.048767998814582825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,8,power_law_1.2,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,8,power_law_1.2,0.04245119988918304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,8,power_law_1.2,0.04046080112457275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,8,power_law_1.2,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,8,power_law_1.2,0.050963199138641356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,8,power_law_1.2,0.0496832013130188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,8,power_law_1.2,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,8,power_law_1.2,0.06244480013847351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,8,power_law_1.2,0.05271040201187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,8,power_law_1.2,0.0587007999420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,8,power_law_1.2,0.07245439887046815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.2,0.09095039963722229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.2,0.12553600072860718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.2,0.1504256010055542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.2,0.21132800579071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.2,0.2670975923538208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,2,power_law_1.01,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,2,power_law_1.01,0.04865919947624207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,2,power_law_1.01,0.055289602279663085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,2,power_law_1.01,0.06769919991493226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,2,power_law_1.01,0.07800319790840149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,2,power_law_1.01,0.08718079924583436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,2,power_law_1.01,0.10531200170516967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,2,power_law_1.01,0.10580480098724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,2,power_law_1.01,0.1128383994102478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,2,power_law_1.01,0.11162240505218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,2,power_law_1.01,0.11237119436264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,2,power_law_1.01,0.11892479658126831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,2,power_law_1.01,0.12261760234832764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,2,power_law_1.01,0.12634240388870238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,2,power_law_1.01,0.13839999437332154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,2,power_law_1.01,0.14026880264282227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,2,power_law_1.01,0.14383360147476196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,2,power_law_1.01,0.16487679481506348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,2,power_law_1.01,0.17752959728240966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,2,power_law_1.01,0.2158207893371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,2,power_law_1.01,0.23991041183471679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,2,power_law_1.01,0.30254719257354734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,2,power_law_1.01,0.35936000347137453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,2,power_law_1.01,0.47207040786743165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,2,power_law_1.01,0.6381120204925537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,2,power_law_1.01,0.8648575782775879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.2,0.38163199424743655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,2,power_law_1.01,1.132633590698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.2,0.4970047950744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,8,power_law_1.2,0.07180799841880799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,8,power_law_1.2,0.08515200018882751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,8,power_law_1.2,0.08104320168495179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,8,power_law_1.2,0.09793919920921326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,8,power_law_1.2,0.10080000162124633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,8,power_law_1.2,0.1093440055847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,8,power_law_1.2,0.11079039573669433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,8,power_law_1.2,0.11206400394439697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,8,power_law_1.2,0.10888320207595825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,8,power_law_1.2,0.11237119436264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,8,power_law_1.2,0.11196800470352172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,8,power_law_1.2,0.11415679454803467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,8,power_law_1.2,0.11575679779052735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,8,power_law_1.2,0.11733119487762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,8,power_law_1.2,0.1239359974861145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,8,power_law_1.2,0.12328319549560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,8,power_law_1.2,0.13029119968414307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,8,power_law_1.2,0.14321919679641723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.2,0.15665279626846312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.2,0.18795520067214966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.2,0.21084160804748536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.2,0.7274687767028809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.2,0.2636735916137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.2,0.3175935983657837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.2,0.4674623966217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.2,0.5732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.2,0.8616512298583985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.2,1.0425215721130372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,16,power_law_1.01,0.05075200200080872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,16,power_law_1.01,0.05651199817657471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,16,power_law_1.01,0.04702720046043396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,16,power_law_1.01,0.04938240051269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,16,power_law_1.01,0.048153600096702574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,16,power_law_1.01,0.05056639909744263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,16,power_law_1.01,0.05198720097541809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,16,power_law_1.01,0.05223039984703064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,16,power_law_1.01,0.0531328022480011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,16,power_law_1.01,0.05430399775505066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,16,power_law_1.01,0.05488640069961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,16,power_law_1.01,0.056704002618789676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,16,power_law_1.01,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,16,power_law_1.01,0.059033602476119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,16,power_law_1.01,0.06498559713363647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,16,power_law_1.01,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,16,power_law_1.01,0.0710528016090393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,16,power_law_1.01,0.0812928020954132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,16,power_law_1.01,0.08866559863090515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,16,power_law_1.01,0.11635839939117432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,16,power_law_1.01,0.15208319425582886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,16,power_law_1.01,0.19543039798736572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,16,power_law_1.01,0.2323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,16,power_law_1.01,0.3130111932754517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,16,power_law_1.01,0.43555197715759275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,16,power_law_1.01,0.6026944160461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,16,power_law_1.01,0.7117887973785401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.2,0.9570560455322266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,1,power_law_1.01,0.036934399604797365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,1,power_law_1.01,0.036723199486732486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,1,power_law_1.01,0.042630401253700254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,1,power_law_1.01,0.05061119794845581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,1,power_law_1.01,0.05985280275344849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,1,power_law_1.01,0.0775168001651764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,1,power_law_1.01,0.08028799891471863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,1,power_law_1.01,0.0828607976436615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,1,power_law_1.01,0.08627200126647949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,1,power_law_1.01,0.09018880128860474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,1,power_law_1.01,0.09548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,1,power_law_1.01,0.10138239860534667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,1,power_law_1.01,0.10693119764328003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,1,power_law_1.01,0.10620160102844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,1,power_law_1.01,0.11603840589523315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,1,power_law_1.01,0.12011519670486451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,1,power_law_1.01,0.13363200426101685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,1,power_law_1.01,0.1655295968055725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,1,power_law_1.01,0.19875199794769288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,1,power_law_1.01,0.2577343940734863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,1,power_law_1.01,0.3097088098526001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,1,power_law_1.01,0.42524161338806155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,1,power_law_1.01,0.5374720096588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,1,power_law_1.01,0.7569471836090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,1,power_law_1.01,0.9908672332763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,1,power_law_1.01,1.429644775390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,1,power_law_1.01,1.8663423538208008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,16,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,16,balanced,0.045791998505592346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,16,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,16,balanced,0.053914666175842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,16,balanced,0.05409599840641022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,16,balanced,0.06211733321348826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,16,balanced,0.06429333488146464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,16,balanced,0.06373333434263866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,16,balanced,0.0641546646753947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,16,balanced,0.06414933502674103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,16,balanced,0.06369600196679433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,16,balanced,0.0647626668214798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,16,balanced,0.06619733572006226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,16,balanced,0.06571733454863231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,16,balanced,0.07190933326880138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,16,balanced,0.07213866710662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,16,balanced,0.07403733332951863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,16,balanced,0.08042666812737782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,16,balanced,0.08518399794896443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,16,balanced,0.09683733185132344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,16,balanced,0.10531733433405559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,16,balanced,0.1276853382587433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,16,balanced,0.15035200119018555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,16,balanced,0.21022399266560873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,16,balanced,0.25726934274037677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,16,balanced,0.35226134459177655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,16,balanced,0.4506293137868245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,2,power_law_1.2,0.04978559911251068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,2,power_law_1.2,0.057580798864364624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,2,power_law_1.2,0.060550397634506224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,2,power_law_1.2,0.07830399870872498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,2,power_law_1.2,0.08210560083389282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,2,power_law_1.2,0.08471680283546448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,2,power_law_1.2,0.08673920035362244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,2,power_law_1.2,0.08773120045661927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,2,power_law_1.2,0.09806079864501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,2,power_law_1.2,0.10172799825668336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,2,power_law_1.2,0.09303680062294006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,2,power_law_1.2,0.09717760086059571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,2,power_law_1.2,0.09761919975280761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,2,power_law_1.2,0.10666240453720092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,2,power_law_1.2,0.1148095965385437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,2,power_law_1.2,0.12960000038146974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,2,power_law_1.2,0.1274623990058899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,2,power_law_1.2,0.14955519437789916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,2,power_law_1.2,0.16680959463119507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,2,power_law_1.2,0.21143679618835448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,2,power_law_1.2,0.23155200481414795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,2,power_law_1.2,0.30859520435333254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,2,power_law_1.2,0.412230396270752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,2,power_law_1.2,0.5178431987762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,16,power_law_1.2,0.09560319781303406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,16,power_law_1.2,0.13619840145111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,16,power_law_1.2,0.12314879894256592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,16,power_law_1.2,0.1319551944732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,16,power_law_1.2,0.1298624038696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,16,power_law_1.2,0.11476479768753052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,16,power_law_1.2,0.11727999448776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,16,power_law_1.2,0.12254719734191895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,16,power_law_1.2,0.1284351944923401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,16,power_law_1.2,0.1338495969772339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,16,power_law_1.2,0.13642879724502563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,1,power_law_1.2,0.07059199810028076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,1,power_law_1.2,0.09251840114593506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,1,power_law_1.2,0.10957440137863159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,2,power_law_1.2,0.6344768047332764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,1,power_law_1.2,0.15979520082473755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,1,power_law_1.2,0.16080000400543212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,1,power_law_1.2,0.16298240423202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,1,power_law_1.2,0.16298880577087402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,16,power_law_1.2,0.1315392017364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,1,power_law_1.2,0.16314239501953126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,16,power_law_1.2,0.1304255962371826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,1,power_law_1.2,0.16336640119552612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,16,power_law_1.2,0.1403839945793152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,16,power_law_1.2,0.15017600059509278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,16,power_law_1.2,0.15904639959335326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,16,power_law_1.2,0.1578495979309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,1,power_law_1.2,0.16430720090866088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,16,power_law_1.2,0.18911999464035034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,16,power_law_1.2,0.20010879039764404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,16,power_law_1.2,0.26102399826049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,16,power_law_1.2,0.25847039222717283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,16,power_law_1.2,0.3249727964401245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,16,power_law_1.2,0.39077119827270507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,1,power_law_1.2,0.1642688035964966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,16,power_law_1.2,0.5666687965393067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,1,power_law_1.2,0.1675968050956726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,16,power_law_1.2,0.6954751968383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,1,power_law_1.2,0.16931840181350707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,16,power_law_1.2,1.0413056373596192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,1,power_law_1.2,0.17417600154876708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,16,power_law_1.2,1.1814784049987792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,1,power_law_1.2,0.196070396900177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,1,power_law_1.2,0.2
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,1,power_law_1.2,0.20977280139923096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,128,balanced,0.03673599908749262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,128,balanced,0.037104000647862755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,128,balanced,0.035631999373435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,1,power_law_1.2,0.24508800506591796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,128,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,128,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,128,balanced,0.04021333406368891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,128,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,128,balanced,0.04120533416668574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,128,balanced,0.04134399940570196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,128,balanced,0.0414986660083135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,128,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,128,balanced,0.04154133299986521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,128,balanced,0.04374399781227112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,128,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,128,balanced,0.045754666129748024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,128,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,128,balanced,0.047925333182017006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,128,balanced,0.051514665285746254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,128,balanced,0.05198933184146881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,128,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,128,balanced,0.062074666221936546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,128,balanced,0.07409599920113881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,128,balanced,0.0946560005346934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,128,balanced,0.1339306632677714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,128,balanced,0.16060800353686014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,128,balanced,0.21769066651662192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,128,balanced,0.2720746596654256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.2,0.3010879993438721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,4,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,4,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,4,balanced,0.053642665346463524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,4,balanced,0.07730666796366374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,4,balanced,0.12807466586430868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,4,balanced,0.12504000465075174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,4,balanced,0.1251520017782847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.2,0.40918397903442383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,4,balanced,0.122597336769104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,4,balanced,0.12474133570988973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,4,balanced,0.12492266297340393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,4,balanced,0.12506133317947388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,4,balanced,0.12824533383051553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,4,balanced,0.12705600261688232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,4,balanced,0.13015466928482056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,4,balanced,0.1362666686375936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,4,balanced,0.13563733299573263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,4,balanced,0.1383039951324463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,4,balanced,0.14482667048772177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,4,balanced,0.15639467040697733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,4,balanced,0.16819200913111368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,4,balanced,0.18195199966430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,4,balanced,0.2302239934603373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,2,power_law_1.2,1.04333438873291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,4,balanced,0.2500373323758443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,4,balanced,0.3438986539840698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,4,balanced,0.3983093500137329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,4,balanced,0.5362133185068766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.2,0.48874878883361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,4,balanced,0.6874240239461263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.2,0.6584256172180176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.2,0.8432127952575683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,2,power_law_1.01,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,2,power_law_1.01,0.049344000220298764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,2,power_law_1.01,0.05170559883117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,2,power_law_1.01,0.06399999856948853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,2,power_law_1.01,0.06313599944114685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,2,power_law_1.01,0.06596480011940002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,2,power_law_1.01,0.06592000126838685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,2,power_law_1.01,0.06647040247917176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,2,power_law_1.01,0.0664896011352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,2,power_law_1.01,0.0668287992477417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,2,power_law_1.01,0.06736639738082886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,2,power_law_1.01,0.07496320009231568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,2,power_law_1.01,0.07992960214614868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,2,power_law_1.01,0.08796799778938294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,128,power_law_1.2,0.07129600048065185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,2,power_law_1.01,0.09489920139312744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,128,power_law_1.2,0.07139840126037597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,128,power_law_1.2,0.07107840180397033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,128,power_law_1.2,0.06683520078659058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,128,power_law_1.2,0.07056000232696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,128,power_law_1.2,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,2,power_law_1.01,0.09523199796676636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,128,power_law_1.2,0.07344639897346497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,2,power_law_1.01,0.10154880285263061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,128,power_law_1.2,0.07271040081977845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,128,power_law_1.2,0.07402240037918091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,128,power_law_1.2,0.07452160120010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,128,power_law_1.2,0.07717120051383972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,128,power_law_1.2,0.07541120052337646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,128,power_law_1.2,0.0774399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,128,power_law_1.2,0.0787775993347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,128,power_law_1.2,0.08423039913177491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,128,power_law_1.2,0.08515200018882751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,128,power_law_1.2,0.09130880236625671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,128,power_law_1.2,0.1072767972946167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,128,power_law_1.2,0.11646080017089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,128,power_law_1.2,0.15126399993896483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,128,power_law_1.2,0.18108160495758058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,128,power_law_1.2,0.25372800827026365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,2,power_law_1.01,0.12641279697418212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,128,power_law_1.2,0.3123456001281738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,128,power_law_1.2,0.4075647830963135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,128,power_law_1.2,0.5200767993927002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,128,power_law_1.2,0.7921088218688965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,128,power_law_1.2,1.0814847946166992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.2,1.1880576133728027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,2,power_law_1.01,0.14517760276794434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,2,power_law_1.01,0.18419840335845947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,2,power_law_1.01,0.23443200588226318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,2,power_law_1.2,1.4642687797546388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,2,power_law_1.01,0.2974656105041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,2,power_law_1.01,0.3702336072921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,2,power_law_1.01,0.5945792198181152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,2,power_law_1.01,0.7137663841247559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.2,1.5443008422851563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.2,2.267180824279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,2,power_law_1.01,1.1436287879943847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,2,power_law_1.01,1.4247167587280274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.2,2.996294403076172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,4,power_law_1.01,0.0785152018070221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,4,power_law_1.01,0.09959679841995239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,4,power_law_1.01,0.1116096019744873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,4,power_law_1.01,0.13575040102005004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,4,power_law_1.01,0.15470080375671386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,2,power_law_1.2,0.024134400486946105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,4,power_law_1.01,0.15516799688339233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,2,power_law_1.2,0.030297601222991945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,4,power_law_1.01,0.17114239931106567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,2,power_law_1.2,0.04160000085830688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,2,power_law_1.2,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,4,power_law_1.01,0.18210560083389282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,2,power_law_1.2,0.0592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,4,power_law_1.01,0.1825152039527893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,2,power_law_1.2,0.08232319951057435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,2,power_law_1.2,0.08566399812698364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,2,power_law_1.2,0.0876416027545929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,2,power_law_1.2,0.09307519793510437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2688,1856,6,128,1,2,power_law_1.2,0.0925055980682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2688,1856,6,128,1,2,power_law_1.2,0.09459840059280396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2688,1856,6,128,1,2,power_law_1.2,0.10063359737396241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2688,1856,6,128,1,2,power_law_1.2,0.10232319831848144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2688,1856,6,128,1,2,power_law_1.2,0.11221120357513428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2688,1856,6,128,1,2,power_law_1.2,0.12114559412002564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,4,power_law_1.01,0.18632320165634156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2688,1856,6,128,1,2,power_law_1.2,0.12993919849395752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,4,power_law_1.01,0.1776703953742981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2688,1856,6,128,1,2,power_law_1.2,0.11802239418029785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,4,power_law_1.01,0.18183039426803588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,4,power_law_1.01,0.17909120321273803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,4,power_law_1.01,0.18925440311431885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,4,power_law_1.01,0.19382400512695314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,4,power_law_1.01,0.19736959934234619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,4,power_law_1.01,0.19839999675750733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,4,power_law_1.01,0.2211008071899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,4,power_law_1.01,0.23591039180755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,4,power_law_1.01,0.2784320116043091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,4,power_law_1.01,0.3127487897872925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,4,power_law_1.01,0.38680319786071776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,4,power_law_1.01,0.46355838775634767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,4,power_law_1.01,0.6175551891326905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,4,power_law_1.01,0.8240896224975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,4,power_law_1.01,1.013804817199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,4,power_law_1.01,1.3994943618774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,32,power_law_1.01,0.09275519847869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,32,power_law_1.01,0.09442560076713562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,32,power_law_1.01,0.09052799940109253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,32,power_law_1.01,0.09505280256271362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,32,power_law_1.01,0.09393919706344604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,32,power_law_1.01,0.0964352011680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,32,power_law_1.01,0.09604480266571044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,32,power_law_1.01,0.09813119769096375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,32,power_law_1.01,0.0989247977733612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,32,power_law_1.01,0.09901440143585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,32,power_law_1.01,0.09903360009193421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,32,power_law_1.01,0.10238080024719239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,32,power_law_1.01,0.10242559909820556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,32,power_law_1.01,0.11006720066070556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,32,power_law_1.01,0.11504000425338745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,32,power_law_1.01,0.1253376007080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,32,power_law_1.01,0.12524800300598143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,32,power_law_1.01,0.1492735981941223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,32,power_law_1.01,0.14793599843978883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,1,power_law_1.01,0.022438399493694305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,32,power_law_1.01,0.18367359638214112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,32,power_law_1.01,0.20830080509185792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,32,power_law_1.01,0.2743168115615845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,32,power_law_1.01,0.3095232009887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,1,power_law_1.01,0.027718400955200194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,32,power_law_1.01,0.43939838409423826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,32,power_law_1.01,0.555020809173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,32,power_law_1.01,0.7627391815185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,32,power_law_1.01,0.9838848114013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,1,power_law_1.01,0.04826239943504333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,1,power_law_1.01,0.07097600102424621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,1,power_law_1.01,0.09774720072746276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,1,power_law_1.01,0.13772799968719482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,1,power_law_1.01,0.17653759717941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,1,power_law_1.01,0.1826367974281311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,1,power_law_1.01,0.19312000274658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,1,power_law_1.01,0.1912384033203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,1,power_law_1.01,0.20307838916778564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,1,power_law_1.01,0.2159168004989624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,1,power_law_1.01,0.22357120513916015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,1,power_law_1.01,0.2458944082260132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,1,power_law_1.01,0.23312640190124512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,1,power_law_1.01,0.24270079135894776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,1,power_law_1.01,0.27175679206848147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,1,power_law_1.01,0.26257920265197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.01,0.29184000492095946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.01,0.32170240879058837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.01,0.3551167964935303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.01,0.7580736160278321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.01,0.9067839622497559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.01,0.6676352024078369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.01,0.7773248195648194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.01,1.0360960006713866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.01,1.27706880569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,8,power_law_1.01,0.08883200287818908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,8,power_law_1.01,0.12520960569381714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,8,power_law_1.01,0.10791679620742797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,8,power_law_1.01,0.13007999658584596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,8,power_law_1.01,0.14664959907531738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,8,power_law_1.01,0.14975359439849853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,8,power_law_1.01,0.14919040203094483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,8,power_law_1.01,0.14651520252227784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,8,power_law_1.01,0.14244480133056642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,8,power_law_1.01,0.14787839651107787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,8,power_law_1.01,0.15637120008468627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,8,power_law_1.01,0.15783040523529052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,8,power_law_1.01,0.1604032039642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,8,power_law_1.01,0.1624768018722534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,8,power_law_1.01,0.17347840070724488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,8,power_law_1.01,0.1787775993347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,8,power_law_1.01,0.19266560077667236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,8,power_law_1.01,0.21916160583496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,8,power_law_1.01,0.23342719078063964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,8,power_law_1.01,0.2864448070526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,8,power_law_1.01,0.3188544034957886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,8,power_law_1.01,0.41256961822509763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,8,power_law_1.01,0.43564162254333494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,8,power_law_1.01,0.5868800163269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,8,power_law_1.01,0.7221183776855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,8,power_law_1.01,1.046656036376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,8,power_law_1.01,1.2660799980163575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,8,power_law_1.2,0.06760960221290588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,8,power_law_1.2,0.07728000283241272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,8,power_law_1.2,0.07482240200042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,8,power_law_1.2,0.0932096004486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,8,power_law_1.2,0.10053759813308716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,8,power_law_1.2,0.09560319781303406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,8,power_law_1.2,0.09808639883995056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,8,power_law_1.2,0.1020799994468689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,8,power_law_1.2,0.09844480156898498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,8,power_law_1.2,0.10504319667816162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,8,power_law_1.2,0.10086400508880615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,8,power_law_1.2,0.10059520006179809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,8,power_law_1.2,0.1050112009048462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,8,power_law_1.2,0.1063040018081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,8,power_law_1.2,0.11591039896011353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,8,power_law_1.2,0.11614079475402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,8,power_law_1.2,0.12462719678878784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,8,power_law_1.2,0.13559679985046386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,8,power_law_1.2,0.1460736036300659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,8,power_law_1.2,0.1741696000099182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,8,power_law_1.2,0.20006399154663085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,8,power_law_1.2,0.2529920101165771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,8,power_law_1.2,0.3048831939697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,8,power_law_1.2,0.4024191856384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,8,power_law_1.2,0.5323328018188477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,8,power_law_1.2,0.755737590789795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,8,power_law_1.2,0.9511743545532226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,2,power_law_1.01,0.05317760109901428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,2,power_law_1.01,0.06542720198631287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,2,power_law_1.01,0.08264960050582885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,2,power_law_1.01,0.11558400392532349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,2,power_law_1.01,0.1305343985557556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,2,power_law_1.01,0.18503040075302124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,2,power_law_1.01,0.18883199691772462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,2,power_law_1.01,0.19155839681625367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,2,power_law_1.01,0.1916159987449646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,2,power_law_1.01,0.20215039253234862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,2,power_law_1.01,0.2065471887588501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,2,power_law_1.01,0.21107840538024902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,2,power_law_1.01,0.21409280300140382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,2,power_law_1.01,0.21955199241638185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,2,power_law_1.01,0.24009599685668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,2,power_law_1.01,0.24651520252227782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,2,power_law_1.01,0.24747519493103026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,2,power_law_1.01,0.27351040840148927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,2,power_law_1.01,0.29976959228515626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,2,power_law_1.01,0.35973761081695554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,2,power_law_1.01,0.404095983505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,2,power_law_1.01,0.5115007877349853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,2,power_law_1.01,0.6000703811645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,2,power_law_1.01,0.7851647853851318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,2,power_law_1.01,1.04335355758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,2,power_law_1.01,1.4224191665649415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,2,power_law_1.01,1.8750783920288085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,1,power_law_1.01,0.12691199779510498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,1,power_law_1.01,0.21552000045776368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,1,power_law_1.01,0.25689599514007566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,1,power_law_1.01,0.37472000122070315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,1,power_law_1.01,0.38030080795288085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,1,power_law_1.01,0.3981055974960327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,1,power_law_1.01,0.4077951908111572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,1,power_law_1.01,0.40322561264038087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,1,power_law_1.01,0.42060160636901855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,1,power_law_1.01,0.4215871810913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,1,power_law_1.01,0.4380928039550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,1,power_law_1.01,0.45348482131958007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,1,power_law_1.01,0.4578688144683838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,1,power_law_1.01,0.4578432083129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,1,power_law_1.01,0.4999551773071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,1,power_law_1.01,0.5142144203186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,1,power_law_1.01,0.5445248126983643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,1,power_law_1.01,0.6382912158966064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,1,power_law_1.01,0.7237887859344483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,1,power_law_1.01,0.8956031799316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,1,power_law_1.01,1.1081855773925782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,1,power_law_1.01,1.474124813079834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,1,power_law_1.01,1.8539968490600587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,1,power_law_1.01,2.6121984481811524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,1,power_law_1.01,3.4916606903076173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,8,power_law_1.2,0.055353599786758426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,8,power_law_1.2,0.0523967981338501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,8,power_law_1.2,0.058374398946762086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,8,power_law_1.2,0.062041598558425906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,8,power_law_1.2,0.06455039978027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,8,power_law_1.2,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,8,power_law_1.2,0.07670400142669678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,8,power_law_1.2,0.07555840015411378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,8,power_law_1.2,0.07898240089416504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,8,power_law_1.2,0.07953919768333435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,8,power_law_1.2,0.07923200130462646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,8,power_law_1.2,0.07965440154075623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,8,power_law_1.2,0.08034560084342957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,8,power_law_1.2,0.0844223976135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,8,power_law_1.2,0.0934656023979187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,8,power_law_1.2,0.09806079864501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,8,power_law_1.2,0.09301120042800903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,8,power_law_1.2,0.10332800149917602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,8,power_law_1.2,0.12340480089187622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,8,power_law_1.2,0.14704639911651612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,8,power_law_1.2,0.1537279963493347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,8,power_law_1.2,0.20869119167327882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,8,power_law_1.2,0.22715520858764648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,8,power_law_1.2,0.29443840980529784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,8,power_law_1.2,0.4235519886016846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,1,power_law_1.01,5.0278270721435545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,1,power_law_1.01,6.627801513671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,8,power_law_1.2,0.5525760173797607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,8,power_law_1.2,0.7501503944396972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,1,power_law_1.2,0.07393280267715455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,1,power_law_1.2,0.09560319781303406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,1,power_law_1.2,0.14349440336227418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,1,power_law_1.2,0.2043328046798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,1,power_law_1.2,0.19682559967041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,1,power_law_1.2,0.2063999891281128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,1,power_law_1.2,0.2095871925354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,1,power_law_1.2,0.21010560989379884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,1,power_law_1.2,0.216812801361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,1,power_law_1.2,0.21862399578094482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,1,power_law_1.2,0.21892480850219725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,1,power_law_1.2,0.2371135950088501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,1,power_law_1.2,0.23953919410705565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,1,power_law_1.2,0.23875839710235597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,1,power_law_1.2,0.25433599948883057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,1,power_law_1.2,0.26423680782318115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,1,power_law_1.2,0.282150411605835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,1,power_law_1.2,0.340012788772583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,1,power_law_1.01,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,1,power_law_1.01,0.05307520031929016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,1,power_law_1.01,0.06454399824142457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,1,power_law_1.01,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,1,power_law_1.01,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,1,power_law_1.01,0.13464319705963135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,1,power_law_1.2,0.3587007999420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,1,power_law_1.01,0.14232319593429565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,1,power_law_1.01,0.1490496039390564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,1,power_law_1.01,0.15292799472808838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,1,power_law_1.01,0.1598528027534485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,1,power_law_1.01,0.1671872019767761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,1,power_law_1.01,0.1772096037864685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,1,power_law_1.01,0.184716796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,1,power_law_1.01,0.18691840171813964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,1,power_law_1.01,0.2004607915878296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,1,power_law_1.01,0.20872960090637208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,1,power_law_1.01,0.2250368118286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,1,power_law_1.01,0.27651200294494627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,1,power_law_1.01,0.3231359958648682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,1,power_law_1.01,0.41327362060546874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,1,power_law_1.01,0.49195518493652346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,32,power_law_1.01,0.13296639919281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,1,power_law_1.01,0.6535935878753663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,32,power_law_1.01,0.15835520029067993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,1,power_law_1.01,0.8576767921447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,1,power_law_1.2,0.44763522148132323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,1,power_law_1.01,1.163488006591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,1,power_law_1.01,1.5912832260131835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,32,power_law_1.01,0.1577023983001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,1,power_law_1.01,2.1860864639282225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,32,power_law_1.01,0.15623040199279786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,32,power_law_1.01,0.16073600053787232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,32,power_law_1.01,0.1483456015586853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,1,power_law_1.01,2.8168256759643553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,32,power_law_1.01,0.14832639694213867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,32,power_law_1.01,0.13969919681549073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,32,power_law_1.01,0.15387519598007202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,1,power_law_1.2,0.5264768123626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,32,power_law_1.01,0.1620736002922058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,32,power_law_1.01,0.15572479963302613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,32,power_law_1.01,0.15709439516067505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,32,power_law_1.01,0.1692479968070984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,32,power_law_1.01,0.15839359760284424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,32,power_law_1.01,0.16790399551391602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,32,power_law_1.01,0.1746240019798279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,32,power_law_1.01,0.17646080255508423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,32,power_law_1.01,0.18858879804611206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,32,power_law_1.01,0.2028480052947998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,32,power_law_1.01,0.23939840793609618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,32,power_law_1.01,0.27331840991973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,32,power_law_1.01,0.3347968101501465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,32,power_law_1.01,0.3972543954849243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,32,power_law_1.01,0.5556352138519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,1,power_law_1.2,0.6941311836242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,32,power_law_1.01,0.6839231967926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,32,power_law_1.01,0.9956031799316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,32,power_law_1.01,1.1889535903930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,1,power_law_1.2,0.8810560226440429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,1,power_law_1.2,1.2240511894226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,2,power_law_1.01,0.03130879998207092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,2,power_law_1.01,0.03157120048999786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,2,power_law_1.01,0.03201279938220978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,2,power_law_1.01,0.03500159978866577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,2,power_law_1.01,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,2,power_law_1.01,0.036160001158714296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,2,power_law_1.01,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,2,power_law_1.01,0.0364544004201889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,2,power_law_1.01,0.03642880022525787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,2,power_law_1.01,0.03664000034332275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,2,power_law_1.01,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,2,power_law_1.01,0.04116480052471161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,2,power_law_1.01,0.04519680142402649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,2,power_law_1.01,0.0471807986497879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,2,power_law_1.01,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,2,power_law_1.01,0.05514240264892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,2,power_law_1.01,0.06175360083580017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,2,power_law_1.01,0.07059199810028076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,2,power_law_1.01,0.08886399865150452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,2,power_law_1.01,0.1053887963294983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,1,power_law_1.2,1.5424127578735352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,2,power_law_1.01,0.1394752025604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,2,power_law_1.01,0.18109439611434935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,2,power_law_1.01,0.22559359073638915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,64,power_law_1.01,0.049728000164031984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,64,power_law_1.01,0.04774399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,64,power_law_1.01,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,64,power_law_1.01,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,64,power_law_1.01,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,64,power_law_1.01,0.054176002740859985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,64,power_law_1.01,0.05453439950942993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,64,power_law_1.01,0.055315202474594115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,64,power_law_1.01,0.05609599947929382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,64,power_law_1.01,0.05748479962348938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,64,power_law_1.01,0.05751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,2,power_law_1.01,0.34119040966033937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,64,power_law_1.01,0.060703998804092406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,1,power_law_1.2,2.2016576766967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,64,power_law_1.01,0.061964797973632815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,64,power_law_1.01,0.06636160016059875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,64,power_law_1.01,0.0749504029750824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,64,power_law_1.01,0.0795199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,64,power_law_1.01,0.07789440155029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,64,power_law_1.01,0.09961599707603455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,64,power_law_1.01,0.10767999887466431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,64,power_law_1.01,0.12465920448303222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,64,power_law_1.01,0.14476799964904785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,64,power_law_1.01,0.18362239599227906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,64,power_law_1.01,0.2221951961517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,64,power_law_1.01,0.30363519191741944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,64,power_law_1.01,0.37512960433959963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,64,power_law_1.01,0.5961919784545898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,64,power_law_1.01,0.7820608139038085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,2,power_law_1.01,0.4455296039581299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,2,power_law_1.01,0.6323840141296386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,1,power_law_1.2,2.88143367767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,2,power_law_1.01,0.8829888343811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,2,power_law_1.2,0.09383040070533752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,2,power_law_1.2,0.14442239999771117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,2,power_law_1.2,0.18218239545822143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,2,power_law_1.2,0.2598016023635864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,2,power_law_1.2,0.37029759883880614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,2,power_law_1.2,0.47322878837585447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,2,power_law_1.2,0.656774377822876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,2,power_law_1.2,0.6668992042541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,2,power_law_1.2,0.6962495803833008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,2,power_law_1.2,0.6818175792694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,2,power_law_1.2,0.7163648128509521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,2,power_law_1.2,0.7281919956207276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,2,power_law_1.2,0.7608320236206054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,2,power_law_1.2,0.7325952053070068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,2,power_law_1.2,0.7920703887939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,2,power_law_1.2,0.8162752151489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,2,power_law_1.2,0.8775808334350585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,2,power_law_1.2,0.9586560249328613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,2,power_law_1.2,0.9646080017089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,2,power_law_1.2,1.059059238433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,2,power_law_1.2,1.1002431869506837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,2,power_law_1.2,1.2786879539489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,2,power_law_1.2,1.471673583984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,2,power_law_1.2,1.8121856689453124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,2,power_law_1.2,2.169536018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,2,power_law_1.2,2.786457633972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,2,power_law_1.2,3.3255935668945313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,1,power_law_1.01,0.05976319909095764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,1,power_law_1.01,0.0630400002002716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,1,power_law_1.01,0.08908799886703492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,1,power_law_1.01,0.11598080396652222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,1,power_law_1.01,0.1665343999862671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,1,power_law_1.01,0.2119488000869751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,1,power_law_1.01,0.2390592098236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,1,power_law_1.01,0.2929408073425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,1,power_law_1.01,0.31015040874481203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,1,power_law_1.01,0.31607038974761964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,1,power_law_1.01,0.32357120513916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,1,power_law_1.01,0.34041600227355956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,1,power_law_1.01,0.35601279735565183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,1,power_law_1.01,0.378982400894165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,1,power_law_1.01,0.4023551940917969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,1,power_law_1.01,0.42515201568603517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,1,power_law_1.01,0.4583424091339111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,2,power_law_1.2,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,1,power_law_1.01,0.5505407810211181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,2,power_law_1.2,0.051948797702789304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,1,power_law_1.01,0.5610688209533692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,1,power_law_1.01,0.674067211151123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,1,power_law_1.01,0.7528192043304444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,1,power_law_1.01,0.9657728195190429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,2,power_law_1.2,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,2,power_law_1.2,0.06570879817008972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,1,power_law_1.01,1.1793343544006347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,1,power_law_1.01,1.591443157196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,2,power_law_1.2,0.0757695972919464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,2,power_law_1.2,0.08577280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,2,power_law_1.2,0.10736639499664306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,1,power_law_1.01,2.0358144760131838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,2,power_law_1.2,0.10988800525665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,2,power_law_1.2,0.11270400285720825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,2,power_law_1.2,0.11403520107269287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,1,power_law_1.01,2.771487998962402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,2,power_law_1.2,0.12475520372390747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,2,power_law_1.2,0.12873599529266358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,2,power_law_1.2,0.133024001121521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,2,power_law_1.2,0.143014395236969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,2,power_law_1.2,0.15708160400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,1,power_law_1.01,3.5611392974853517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,2,power_law_1.2,0.16416640281677247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,2,power_law_1.2,0.16684800386428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,2,power_law_1.2,0.19043840169906617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,2,power_law_1.2,0.22709119319915771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,2,power_law_1.2,0.2820800065994263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,2,power_law_1.2,0.3307391881942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,2,power_law_1.2,0.4616384029388428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,2,power_law_1.2,0.5897727966308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,2,power_law_1.2,0.768940782546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,2,power_law_1.2,0.9865856170654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,2,power_law_1.2,1.458240032196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,2,power_law_1.2,1.9884544372558595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,64,balanced,0.04983466863632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,64,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,64,balanced,0.052906667192777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,64,balanced,0.05619733532269796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,64,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,64,balanced,0.05795733133951823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,64,balanced,0.0582826683918635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,64,balanced,0.060047999024391174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,64,balanced,0.05996799965699514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,64,balanced,0.059936001896858215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,64,balanced,0.06058666606744131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,64,balanced,0.06402133405208588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,64,balanced,0.06440000236034393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,64,balanced,0.06599999964237213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,64,balanced,0.0724480003118515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,64,balanced,0.08312533299128215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,64,balanced,0.07897066573301952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,64,balanced,0.10257599751154582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,64,balanced,0.10502933462460835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,64,balanced,0.13268799583117166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,64,balanced,0.14900267124176025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,64,balanced,0.19820266962051392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,64,balanced,0.1967946688334147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,64,balanced,0.26365866263707477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,64,balanced,0.293887992699941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,64,balanced,0.39441601435343426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,64,balanced,0.4819999933242798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,256,power_law_1.2,0.05644800066947937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,256,power_law_1.2,0.05983359813690185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,256,power_law_1.2,0.056524801254272464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,256,power_law_1.2,0.06076800227165222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,256,power_law_1.2,0.062457597255706786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,256,power_law_1.2,0.06395519971847534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,256,power_law_1.2,0.06335359811782837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,256,power_law_1.2,0.06449919939041138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,256,power_law_1.2,0.06448000073432922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,256,power_law_1.2,0.06465280055999756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,256,power_law_1.2,0.06348159909248352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,256,power_law_1.2,0.0652288019657135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,256,power_law_1.2,0.06528000235557556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,256,power_law_1.2,0.06352000236511231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,256,power_law_1.2,0.06768640279769897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,256,power_law_1.2,0.06622080206871032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,256,power_law_1.2,0.07916160225868225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,256,power_law_1.2,0.08090239763259888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.2,0.09733120203018189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.2,0.1198591947555542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.2,0.14181120395660402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.2,0.18705919981002808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.2,0.22835841178894042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.2,0.32264320850372313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.2,0.42926721572875975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.2,0.681932783126831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.2,0.8068415641784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,4,power_law_1.2,0.022470399737358093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,4,power_law_1.2,0.021798400580883025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,4,power_law_1.2,0.021196800470352172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,4,power_law_1.2,0.021062399446964263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,4,power_law_1.2,0.021887999773025513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,4,power_law_1.2,0.025305598974227905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,4,power_law_1.2,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,4,power_law_1.01,0.04803839921951294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,4,power_law_1.2,0.028921601176261903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,4,power_law_1.01,0.04956159889698029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,4,power_law_1.01,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,4,power_law_1.01,0.061350399255752565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,4,power_law_1.01,0.06120319962501526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,4,power_law_1.2,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,4,power_law_1.01,0.07486079931259156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,4,power_law_1.2,0.03187839984893799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,4,power_law_1.01,0.07463039755821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,4,power_law_1.2,0.03478400111198425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,4,power_law_1.2,0.04044159948825836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,4,power_law_1.2,0.0438975989818573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,4,power_law_1.2,0.046105599403381346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,4,power_law_1.01,0.07535359859466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,4,power_law_1.2,0.06320000290870667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,4,power_law_1.01,0.07437440156936645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,4,power_law_1.01,0.08075519800186157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,4,power_law_1.2,0.055225598812103274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,4,power_law_1.01,0.08246399760246277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,4,power_law_1.01,0.08832640051841736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,4,power_law_1.01,0.0955136001110077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,4,power_law_1.01,0.09112319946289063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,4,power_law_1.01,0.10049920082092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,4,power_law_1.2,0.06220160126686096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,4,power_law_1.01,0.10488320589065551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,4,power_law_1.01,0.11253119707107544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,4,power_law_1.01,0.12845439910888673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,4,power_law_1.01,0.1487936019897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,4,power_law_1.01,0.18376319408416747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,4,power_law_1.01,0.23148798942565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,4,power_law_1.2,0.0830847978591919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,4,power_law_1.01,0.3052095890045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,4,power_law_1.01,0.38587520122528074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,4,power_law_1.01,0.523360013961792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.2,0.0983295977115631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,4,power_law_1.01,0.6851583957672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,4,power_law_1.01,0.9898431777954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,4,power_law_1.01,1.2436287879943848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.2,0.1370303988456726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.2,0.1716096043586731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.2,0.23963520526885987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.2,0.30005760192871095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.2,0.43415040969848634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.2,0.5639808177947998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.2,0.8300864219665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,2,balanced,0.031152000029881794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,2,balanced,0.03141333411137263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,2,balanced,0.0340693344672521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,2,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,2,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,2,balanced,0.03676266719897588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,2,balanced,0.03603733330965042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,2,balanced,0.039349332451820374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,2,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,2,balanced,0.03955733279387156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,2,balanced,0.041573333243529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,2,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,2,balanced,0.04383466641108195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,2,balanced,0.04728533327579498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,2,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,2,balanced,0.06164266665776571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,2,balanced,0.06262933214505513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,2,balanced,0.07414933542410533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,2,balanced,0.08002666632334392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,2,balanced,0.10097600022951762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,2,balanced,0.11541866262753804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,2,balanced,0.15043733517328897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,2,balanced,0.18389866749445596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,2,balanced,0.24889065821965536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,2,balanced,0.3125706712404887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,2,balanced,0.44254398345947266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,2,balanced,0.5702293316523234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,1,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,1,balanced,0.051925331354141235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,1,balanced,0.05598933498064677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,1,balanced,0.06647466619809468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,1,balanced,0.08893866340319316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,1,balanced,0.11993066469828288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,1,balanced,0.12246933579444885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,1,balanced,0.12343466281890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,1,balanced,0.12463999787966411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,1,balanced,0.12803733348846436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,1,balanced,0.129530668258667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,1,balanced,0.13473066687583923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,1,balanced,0.13851199547449747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,1,balanced,0.13989333311716715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,1,balanced,0.15037332971890768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,1,balanced,0.15542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,1,balanced,0.1653279960155487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,1,balanced,0.19342400630315146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,1,balanced,0.21502399444580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,1,balanced,0.26632533470789593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,1,balanced,0.32654933134714764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,1,balanced,0.4287733236948649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,1,balanced,0.5256853501001993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,1,balanced,0.7819519837697347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,1,balanced,0.986687978108724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,1,balanced,1.4389386177062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,1,balanced,1.891642729441325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.2,1.091327953338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,16,balanced,0.062047998110453285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,16,balanced,0.06278400123119354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,16,balanced,0.07242133220036824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,16,balanced,0.09507733583450317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,16,balanced,0.1253653367360433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,16,balanced,0.20300267140070596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,16,balanced,0.19343467553456625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,16,balanced,0.19638933738072714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,16,balanced,0.19450666507085165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,16,balanced,0.18959999084472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,16,balanced,0.19340266784032187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,16,balanced,0.1859253247578939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,16,balanced,0.1895680030186971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,16,balanced,0.19162132342656454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,16,balanced,0.1905919909477234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,16,balanced,0.207914670308431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,16,balanced,0.18685867389043173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,16,balanced,0.20586667458216348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,16,balanced,0.20127999782562256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,16,balanced,0.22230400641759238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,16,balanced,0.22740266720453897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,16,balanced,0.24361600478490195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,16,balanced,0.2737119992574056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,16,balanced,0.41934935251871747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,16,balanced,0.4509600003560384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,16,balanced,0.6230773528416952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,16,balanced,0.802074670791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,32,2,power_law_1.2,0.02122880071401596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,32,2,power_law_1.2,0.022444799542427063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,32,2,power_law_1.2,0.024710400402545928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,32,2,power_law_1.2,0.029292801022529603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,32,2,power_law_1.2,0.037574398517608645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,32,2,power_law_1.2,0.04035840034484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,32,2,power_law_1.2,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,32,2,power_law_1.2,0.04803839921951294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,32,2,power_law_1.2,0.05219200253486633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,32,2,power_law_1.2,0.05585920214653015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,32,2,power_law_1.2,0.06297600269317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,32,2,power_law_1.2,0.06640639901161194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,32,2,power_law_1.2,0.06959360241889953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,32,2,power_law_1.2,0.07671679854393006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,32,2,power_law_1.2,0.07180799841880799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,32,2,power_law_1.2,0.0769536018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,32,2,power_law_1.2,0.0844543993473053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,32,2,power_law_1.2,0.10584319829940796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,32,2,power_law_1.2,0.11822079420089722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,32,2,power_law_1.2,0.18294399976730347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,32,2,power_law_1.2,0.19792640209197998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,32,2,power_law_1.2,0.22846078872680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,32,2,power_law_1.2,0.25932159423828127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,32,2,power_law_1.2,0.31738240718841554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,32,2,power_law_1.2,0.39687039852142336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,32,2,power_law_1.2,0.5247424125671387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,32,2,power_law_1.2,0.6531072139739991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,2,balanced,0.048469334840774536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,2,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,2,balanced,0.054197331269582115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,2,balanced,0.07062399884064992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,2,balanced,0.09476266304651897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,2,balanced,0.13762666781743368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,2,balanced,0.17706666390101114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,2,balanced,0.17717333634694418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,2,balanced,0.17941866318384805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,2,balanced,0.1818880041440328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,2,balanced,0.18216532468795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,2,balanced,0.18893333276112875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,2,balanced,0.1918933391571045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,2,balanced,0.19605332612991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,2,balanced,0.20775467157363892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,2,balanced,0.21172799666722616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,2,balanced,0.2230666677157084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,2,balanced,0.25414933760960895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,2,balanced,0.27491732438405353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,2,balanced,0.3309119939804077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,2,balanced,0.3715413411458333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,2,balanced,0.4771093527475993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,2,balanced,0.6012853384017944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,2,balanced,0.8010613123575846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,2,balanced,1.066480000813802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,2,balanced,1.5044906934102376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,2,balanced,1.9580799738566081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,4,power_law_1.01,0.03828479945659637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,4,power_law_1.01,0.041766399145126344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,4,power_law_1.01,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,4,power_law_1.01,0.05902720093727112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,4,power_law_1.01,0.06027519702911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,4,power_law_1.01,0.07528960108757018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,4,power_law_1.01,0.07577599883079529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,4,power_law_1.01,0.07811200022697448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,4,power_law_1.01,0.07921280264854431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,4,power_law_1.01,0.07909119725227357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,4,power_law_1.01,0.08266879916191101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,4,power_law_1.01,0.08300799727439881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,4,power_law_1.01,0.08553599715232849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,4,power_law_1.01,0.0902463972568512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,4,power_law_1.01,0.09692800045013428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,4,power_law_1.01,0.10271999835968018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,4,power_law_1.01,0.10057599544525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,4,power_law_1.01,0.10763520002365112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,4,power_law_1.01,0.12492799758911133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,4,power_law_1.01,0.1478335976600647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,4,power_law_1.01,0.16571520566940307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,4,power_law_1.01,0.21889278888702393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,4,power_law_1.01,0.2476288080215454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,4,power_law_1.01,0.32728960514068606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,4,power_law_1.01,0.4509888172149658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,4,power_law_1.01,0.5910592079162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,4,power_law_1.01,0.8012928009033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,4,balanced,0.08762666583061218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,4,balanced,0.1269813378651937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,4,balanced,0.19840532541275024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,4,balanced,0.35496000448862713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,4,balanced,0.61736532052358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,4,balanced,0.8447039922078451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,16,power_law_1.01,0.12350720167160034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,4,balanced,0.8484106858571371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,4,balanced,0.851482629776001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,16,power_law_1.01,0.15184639692306517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,4,balanced,0.8542239665985107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,16,power_law_1.01,0.15152000188827514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,16,power_law_1.01,0.15282560586929322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,16,power_law_1.01,0.15653120279312133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,16,power_law_1.01,0.14906879663467407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,4,balanced,0.8556533654530843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,4,balanced,0.8659733136494955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,4,balanced,0.8683626651763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,4,balanced,0.8725759983062744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,4,balanced,0.8760586579640707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,4,balanced,0.8846720059712728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,4,balanced,0.8944479624430338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,4,balanced,0.9035360018412272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,4,balanced,1.0735893249511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,4,balanced,0.9635252952575684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,4,balanced,1.2509013017018635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,16,power_law_1.01,0.14306559562683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,16,power_law_1.01,0.14653439521789552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,16,power_law_1.01,0.14328960180282593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,16,power_law_1.01,0.1491520047187805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,16,power_law_1.01,0.149619197845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,16,power_law_1.01,0.1587648034095764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,4,balanced,1.074954668680827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,16,power_law_1.01,0.15226880311965943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,4,balanced,1.565989335378011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,16,power_law_1.01,0.15807360410690308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,16,power_law_1.01,0.16348799467086791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,4,balanced,1.3055306275685628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,16,power_law_1.01,0.17201919555664064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,16,power_law_1.01,0.17541120052337647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,4,balanced,2.1319680213928223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,16,power_law_1.01,0.20160000324249266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,16,power_law_1.01,0.21853439807891845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,4,balanced,1.9449653625488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,16,power_law_1.01,0.28110079765319823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,16,power_law_1.01,0.3036223888397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,4,balanced,3.3698507944742837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,4,balanced,3.0089546839396157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,16,power_law_1.01,0.3849087953567505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,16,power_law_1.01,0.4016767978668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,16,power_law_1.01,0.5056384086608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,16,power_law_1.01,0.6253503799438477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,16,power_law_1.01,0.8239871978759765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,16,power_law_1.01,0.9358976364135743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,16,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,16,balanced,0.0313226655125618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,16,balanced,0.02926933268706004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,16,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,16,balanced,0.034634667138258614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,16,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,16,balanced,0.03852800031503042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,16,balanced,0.03834133346875509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,16,balanced,0.03833599885304769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,16,balanced,0.03904533386230469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,16,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,16,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,16,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,16,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,16,balanced,0.043568000197410583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,16,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,16,balanced,0.04242133100827535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,16,balanced,0.052330667773882546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,16,balanced,0.04788800080617269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,16,balanced,0.057775999108950295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,16,balanced,0.055018668373425804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,16,balanced,0.07014399766921997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,16,balanced,0.07130133112271626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,16,balanced,0.09513066212336223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,16,balanced,0.09821333487828572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,16,balanced,0.14109333356221518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,16,balanced,0.13844799995422363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,128,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,128,balanced,0.03925866633653641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,128,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,128,balanced,0.03957333415746689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,128,balanced,0.04053333401679993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,128,balanced,0.043509334325790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,128,balanced,0.04190400242805481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,128,balanced,0.041850666205088295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,128,balanced,0.044031997521718345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,128,balanced,0.0424586683511734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,128,balanced,0.042821332812309265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,128,balanced,0.042597333590189614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,128,balanced,0.04489066700140635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,128,balanced,0.04292800029118856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,128,balanced,0.04590400060017904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,128,balanced,0.045738667249679565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,128,balanced,0.045834665497144066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,128,balanced,0.04786133269468943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,128,balanced,0.049327999353408813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,128,balanced,0.04994133114814758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,128,balanced,0.051594664653142296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,128,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,128,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,128,power_law_1.01,0.0561024010181427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,128,power_law_1.01,0.05356159806251526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,128,power_law_1.01,0.05505920052528381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,128,power_law_1.01,0.05717759728431702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,128,balanced,0.06771733363469441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,128,power_law_1.01,0.05711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,128,power_law_1.01,0.058739197254180905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,128,power_law_1.01,0.05878400206565857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,128,power_law_1.01,0.058713597059249875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,128,power_law_1.01,0.0585919976234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,128,power_law_1.01,0.059334397315979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,128,balanced,0.07391466697057088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,128,power_law_1.01,0.05934720039367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,128,balanced,0.09175466497739156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,128,balanced,0.10738133390744527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,128,power_law_1.01,0.06268799901008607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,128,power_law_1.01,0.0630079984664917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,128,power_law_1.01,0.06342399716377259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,128,power_law_1.01,0.07079039812088013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,128,power_law_1.01,0.07252479791641235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,128,power_law_1.01,0.07897599935531616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,128,power_law_1.01,0.09102079868316651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,128,power_law_1.01,0.10061440467834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,128,power_law_1.01,0.12648320198059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,128,power_law_1.01,0.14513280391693115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,128,power_law_1.01,0.19848959445953368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,128,power_law_1.01,0.19935359954833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,128,power_law_1.01,0.25340800285339354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,128,power_law_1.01,0.3192895889282227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,128,power_law_1.01,0.4269248008728027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,128,power_law_1.01,0.5074624061584473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,4,power_law_1.2,0.04431999921798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,4,power_law_1.2,0.04533120095729828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,4,power_law_1.2,0.04413439929485321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,4,power_law_1.2,0.04895359873771667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,4,power_law_1.2,0.0496832013130188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,4,power_law_1.2,0.051558399200439455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,4,power_law_1.2,0.05503360033035278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,128,power_law_1.01,0.06378880143165588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,4,power_law_1.2,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,128,power_law_1.01,0.06350079774856568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,4,power_law_1.2,0.057145601511001586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,4,power_law_1.2,0.05612800121307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,128,power_law_1.01,0.061612802743911746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,4,power_law_1.2,0.05805439949035644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,128,power_law_1.01,0.06445440053939819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,4,power_law_1.2,0.06033920049667359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,128,power_law_1.01,0.06327040195465088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,128,power_law_1.01,0.06659839749336242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,128,power_law_1.01,0.0665727972984314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,128,power_law_1.01,0.06694399714469909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,4,power_law_1.2,0.061868798732757566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,128,power_law_1.01,0.06791679859161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,4,power_law_1.2,0.06401919722557067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,128,power_law_1.01,0.06754559874534607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,128,power_law_1.01,0.06938880085945129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,128,power_law_1.01,0.06934400200843811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,128,power_law_1.01,0.06935679912567139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,128,power_law_1.01,0.07319679856300354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,128,power_law_1.01,0.07687680125236511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,4,power_law_1.2,0.07020800113677979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,128,power_law_1.01,0.07830399870872498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,4,power_law_1.2,0.07467520236968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,128,power_law_1.01,0.08551679849624634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,4,power_law_1.2,0.07875199913978577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,128,power_law_1.01,0.09723520278930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,4,power_law_1.2,0.09223039746284485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,128,power_law_1.01,0.11285760402679443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,4,power_law_1.2,0.1131327986717224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,128,power_law_1.01,0.13906559944152833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,4,power_law_1.2,0.139027202129364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,128,power_law_1.01,0.16124800443649293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,4,power_law_1.2,0.14227839708328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,128,power_law_1.01,0.212172794342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,4,power_law_1.2,0.1770815968513489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,128,power_law_1.01,0.26693120002746584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,4,power_law_1.2,0.24140799045562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,128,power_law_1.01,0.3651135921478271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,4,power_law_1.2,0.3310080051422119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,128,power_law_1.01,0.46179838180541993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,4,power_law_1.2,0.4019199848175049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,128,power_law_1.01,0.6532864093780517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,4,power_law_1.2,0.5719552040100098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,128,power_law_1.01,0.8522879600524902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,4,power_law_1.2,0.7476799964904786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,8,power_law_1.01,0.046239998936653134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,8,power_law_1.01,0.050521600246429446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,8,power_law_1.01,0.048172798752784726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,8,power_law_1.01,0.054150402545928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,8,power_law_1.01,0.0531711995601654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,8,power_law_1.01,0.05550079941749573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,8,power_law_1.01,0.05573760271072388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,8,power_law_1.01,0.05740799903869629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,8,power_law_1.01,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,8,power_law_1.01,0.060729598999023436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,8,power_law_1.01,0.06176000237464905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,8,power_law_1.01,0.06817920207977295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,8,power_law_1.01,0.07139840126037597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,8,power_law_1.01,0.0704255998134613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,8,power_law_1.01,0.07568640112876893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,8,power_law_1.01,0.08264960050582885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,8,power_law_1.01,0.08596479892730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,8,power_law_1.01,0.09886080026626587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,8,power_law_1.01,0.11717760562896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,8,power_law_1.01,0.14740480184555055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,8,power_law_1.01,0.16852480173110962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,8,power_law_1.01,0.2316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,8,power_law_1.01,0.30517120361328126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,8,power_law_1.01,0.3892352104187012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,8,power_law_1.01,0.5107264041900634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,8,power_law_1.01,0.6815423965454102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,8,power_law_1.01,0.9426495552062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,32,power_law_1.01,0.05400320291519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,32,power_law_1.01,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,32,power_law_1.01,0.04583680033683777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,32,power_law_1.01,0.046988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,32,power_law_1.01,0.04828799962997436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,32,power_law_1.01,0.04874880015850067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,32,power_law_1.01,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,32,power_law_1.01,0.050963199138641356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,32,power_law_1.01,0.051769602298736575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,32,power_law_1.01,0.05233280062675476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,32,power_law_1.01,0.053913599252700804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,32,power_law_1.01,0.056524801254272464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,32,power_law_1.01,0.05852159857749939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,32,power_law_1.01,0.06266239881515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,32,power_law_1.01,0.06862080097198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,32,power_law_1.01,0.07146880030632019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,32,power_law_1.01,0.07596799731254578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,32,power_law_1.01,0.09273599982261657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,32,power_law_1.01,0.0998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,32,power_law_1.01,0.12453759908676147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,32,power_law_1.01,0.14442239999771117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,32,power_law_1.01,0.17747199535369873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,32,power_law_1.01,0.22673919200897216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,32,power_law_1.01,0.302508807182312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,32,power_law_1.01,0.38562560081481934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,32,power_law_1.01,0.5546815872192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,32,power_law_1.01,0.7091839790344239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,8,balanced,0.03875733415285746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,8,balanced,0.03826133410135905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,8,balanced,0.04358933369318644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,8,balanced,0.04872000217437744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,8,balanced,0.06448533137639363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,8,balanced,0.06414400041103363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,8,balanced,0.062277331948280334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,8,balanced,0.06224533418814341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,8,balanced,0.06005866825580597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,8,balanced,0.062352001667022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,8,balanced,0.06206400195757548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,8,balanced,0.06029333174228668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,8,balanced,0.062394668658574425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,8,balanced,0.06205866734186808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,8,balanced,0.06611200173695882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,8,balanced,0.0665280024210612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,8,balanced,0.06884266436100006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,8,balanced,0.07469866673151652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,8,balanced,0.07682133217652638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,8,balanced,0.08772800366083781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,8,balanced,0.09817066788673401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,8,balanced,0.13062399625778198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,8,balanced,0.14833600322405496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,8,balanced,0.1934773325920105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,8,balanced,0.24046399195988974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,8,balanced,0.33672531445821124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,8,balanced,0.433135986328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,8,power_law_1.01,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,2,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,2,balanced,0.07414933542410533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,2,balanced,0.10262933373451233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,2,balanced,0.15235732992490134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,2,balanced,0.25884799162546795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,2,balanced,0.3123679955800374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,2,balanced,0.31438400348027545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,2,balanced,0.3165439963340759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,2,balanced,0.32003732522328693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,2,balanced,0.3180053234100342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,2,balanced,0.32387200991312665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,2,balanced,0.3261760075887044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,2,balanced,0.32846933603286743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,2,balanced,0.3301173249880473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,2,balanced,0.3423413435618083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,2,balanced,0.3455413182576497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,2,balanced,0.3559199968973796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,2,balanced,0.3789866765340169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,2,balanced,0.39120535055796307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,2,balanced,0.49085867404937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,2,balanced,0.47596800327301025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,2,balanced,0.6726933320363363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,2,balanced,0.6458293199539185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,2,balanced,0.9035680294036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,8,power_law_1.01,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,2,balanced,1.0380853017171223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,8,power_law_1.01,0.03281280100345611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,8,power_law_1.01,0.035571199655532834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,2,balanced,1.5761119524637859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,8,power_law_1.01,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,8,power_law_1.01,0.03594880104064942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,2,balanced,1.9326133728027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,8,power_law_1.01,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,8,power_law_1.01,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,8,power_law_1.01,0.03666560053825378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,8,power_law_1.01,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,8,power_law_1.01,0.03787519931793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,8,power_law_1.2,0.08440319895744323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,8,power_law_1.2,0.1000704050064087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,8,power_law_1.01,0.04095360040664673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,8,power_law_1.2,0.11032960414886475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,8,power_law_1.01,0.04238080084323883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,8,power_law_1.2,0.13074560165405275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,8,power_law_1.2,0.1201856017112732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,8,power_law_1.2,0.14831360578536987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,8,power_law_1.2,0.1468287944793701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,8,power_law_1.2,0.1450495958328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,8,power_law_1.01,0.04364160001277924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,8,power_law_1.2,0.1530176043510437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,8,power_law_1.2,0.14689919948577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,8,power_law_1.2,0.15489920377731323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,8,power_law_1.2,0.16064640283584594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,8,power_law_1.2,0.16366080045700074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,8,power_law_1.2,0.16872960329055786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,8,power_law_1.2,0.17944320440292358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,8,power_law_1.01,0.052960002422332765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,8,power_law_1.2,0.18364800214767457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,8,power_law_1.01,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,8,power_law_1.2,0.2020416021347046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,8,power_law_1.2,0.2340480089187622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,8,power_law_1.2,0.2405951976776123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,8,power_law_1.2,0.29820799827575684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,8,power_law_1.2,0.3438656091690063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,8,power_law_1.01,0.06327040195465088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,8,power_law_1.2,0.44303359985351565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,8,power_law_1.2,0.47708802223205565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,8,power_law_1.2,0.6249599933624268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,8,power_law_1.01,0.06575999855995178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,8,power_law_1.2,0.8458432197570801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,8,power_law_1.2,1.1509759902954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,8,power_law_1.2,1.3020992279052734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,8,power_law_1.01,0.09306880235671997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,8,power_law_1.01,0.11813119649887086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,8,power_law_1.01,0.14975359439849853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,8,power_law_1.01,0.1971392035484314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,8,power_law_1.01,0.23265280723571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,8,power_law_1.01,0.3869312047958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,8,power_law_1.01,0.5080639839172363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,8,power_law_1.01,0.7109824180603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,32,power_law_1.01,0.0797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,32,power_law_1.01,0.06819199919700622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,32,power_law_1.01,0.0664255976676941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,32,power_law_1.01,0.07147520184516906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,32,power_law_1.01,0.07108479738235474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,32,power_law_1.01,0.06741759777069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,32,power_law_1.01,0.06856319904327393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,32,power_law_1.01,0.07192959785461425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,32,power_law_1.01,0.07309439778327942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,32,power_law_1.01,0.07227519750595093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,32,power_law_1.01,0.07464960217475891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,32,power_law_1.01,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,8,power_law_1.01,1.143449592590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,32,power_law_1.01,0.0739135980606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,32,power_law_1.01,0.07811840176582337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,32,power_law_1.01,0.08486400246620178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,32,power_law_1.01,0.0894976019859314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,32,power_law_1.01,0.08981760144233704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,32,power_law_1.01,0.10097279548645019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,32,power_law_1.01,0.10547200441360474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,32,power_law_1.01,0.12710399627685548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,32,power_law_1.01,0.14026880264282227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,32,power_law_1.01,0.17349120378494262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,32,power_law_1.01,0.2091007947921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,32,power_law_1.01,0.27048320770263673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,32,power_law_1.01,0.3402688026428223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,32,power_law_1.01,0.4578944206237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,32,power_law_1.01,0.5849279880523681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,64,power_law_1.01,0.022707200050354003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,64,power_law_1.01,0.023270399868488313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,64,power_law_1.01,0.02197760045528412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,64,power_law_1.01,0.023532800376415253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,64,power_law_1.01,0.022649599611759184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,64,power_law_1.01,0.025203201174736022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,64,power_law_1.01,0.025139200687408447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,64,power_law_1.01,0.02515200078487396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,64,power_law_1.01,0.027980801463127137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,64,power_law_1.01,0.02929919958114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,64,power_law_1.01,0.03000960052013397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,64,power_law_1.01,0.030380800366401672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,64,power_law_1.01,0.03038719892501831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,64,power_law_1.01,0.029977598786354066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,64,power_law_1.01,0.032902398705482484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,64,power_law_1.01,0.033555200695991515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,64,power_law_1.01,0.034835198521614076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,64,power_law_1.01,0.036345601081848145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,64,power_law_1.01,0.03768959939479828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,64,power_law_1.01,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,64,power_law_1.01,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,64,power_law_1.01,0.05286399722099304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,64,power_law_1.01,0.05623679757118225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,64,power_law_1.01,0.07544320225715637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,64,power_law_1.01,0.08909440040588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,64,power_law_1.01,0.11943680047988892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,64,power_law_1.01,0.14260480403900147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,4,power_law_1.01,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,4,power_law_1.01,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,4,power_law_1.01,0.04259200096130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,4,power_law_1.01,0.056601601839065555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,4,power_law_1.01,0.05749120116233826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,4,power_law_1.01,0.06046720147132874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,4,power_law_1.01,0.06477439999580384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,4,power_law_1.01,0.06364160180091857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,4,power_law_1.01,0.0648256003856659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,4,power_law_1.01,0.0646336019039154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,4,power_law_1.01,0.06889600157737732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,4,power_law_1.01,0.06878719925880432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,4,power_law_1.01,0.07030400037765502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,4,power_law_1.01,0.07905279994010925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,4,power_law_1.01,0.08292480111122132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,4,power_law_1.01,0.08579199910163879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,4,power_law_1.01,0.10119680166244507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,4,power_law_1.01,0.12125439643859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,4,power_law_1.01,0.13075200319290162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,4,power_law_1.01,0.1724992036819458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,4,power_law_1.01,0.18303999900817872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,4,power_law_1.01,0.22860798835754395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,4,power_law_1.01,0.28124799728393557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,4,power_law_1.01,0.4235519886016846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,4,power_law_1.01,0.5980991840362548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,4,power_law_1.01,0.7028096199035645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,4,power_law_1.01,0.9935040473937988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,32,power_law_1.2,0.03540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,32,power_law_1.2,0.0359360009431839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,32,power_law_1.2,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,32,power_law_1.2,0.035743999481201175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,32,power_law_1.2,0.03619199991226196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,32,power_law_1.2,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,32,power_law_1.2,0.03785600066184998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,32,power_law_1.2,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,32,power_law_1.2,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,32,power_law_1.2,0.038335999846458434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,32,power_law_1.2,0.03885439932346344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,32,power_law_1.2,0.04196479916572571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,32,power_law_1.2,0.042124798893928526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,32,power_law_1.2,0.04323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,32,power_law_1.2,0.04760960042476654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,32,power_law_1.2,0.04894720017910004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,32,power_law_1.2,0.049292799830436704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,32,power_law_1.2,0.05720959901809693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.2,0.05877760052680969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.2,0.07032319903373718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.2,0.07915520071983337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.2,0.10240000486373901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.2,0.13093119859695435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.2,0.1968832015991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,64,balanced,0.023941333095232647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,64,balanced,0.02006400004029274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,64,balanced,0.019952000429232914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,64,balanced,0.021397332350413006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,64,balanced,0.023082666099071503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,64,balanced,0.026191999514897663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,64,balanced,0.02518400053183238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.2,0.22899200916290283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,64,balanced,0.039813332259655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,64,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,64,balanced,0.03850133220354716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,64,balanced,0.0384853333234787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,64,balanced,0.03681600093841553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,64,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,64,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,64,balanced,0.045082668463389076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,64,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,64,balanced,0.039290666580200195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,64,balanced,0.044405331214269005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,64,balanced,0.05410666763782501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,64,balanced,0.06304533282915752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,64,balanced,0.07796800136566162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,64,balanced,0.10191999872525533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,64,balanced,0.12688000003496805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,64,balanced,0.1722559928894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,64,balanced,0.2197386622428894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,64,balanced,0.3134933312733968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,64,balanced,0.40909334023793537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.2,0.38048639297485354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,16,power_law_1.2,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,16,power_law_1.2,0.07099519968032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,16,power_law_1.2,0.08326399922370911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,16,power_law_1.2,0.08669440150260925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,16,power_law_1.2,0.08921599984169007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,16,power_law_1.2,0.08618879914283753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,16,power_law_1.2,0.0781503975391388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,16,power_law_1.2,0.0874176025390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,16,power_law_1.2,0.08135039806365967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,16,power_law_1.2,0.08698880076408386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,16,power_law_1.2,0.08739839792251587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,16,power_law_1.2,0.08655999898910523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,16,power_law_1.2,0.08973439931869506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,16,power_law_1.2,0.09603840112686157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,16,power_law_1.2,0.1004480004310608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,16,power_law_1.2,0.10000640153884888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,16,power_law_1.2,0.1078976035118103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,16,power_law_1.2,0.11969280242919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,16,power_law_1.2,0.13358720541000366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,16,power_law_1.2,0.15633920431137086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,16,power_law_1.2,0.17817599773406984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,16,power_law_1.2,0.22652161121368408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,16,power_law_1.2,0.28488960266113283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,16,power_law_1.2,0.3699903964996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,16,power_law_1.2,0.48744959831237794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,16,power_law_1.2,0.656057596206665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,16,power_law_1.2,0.8594623565673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.2,0.44657278060913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,8,balanced,0.05018133421738943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,8,balanced,0.05217599868774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,8,balanced,0.05434666574001312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,8,balanced,0.0602453351020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,8,balanced,0.06425599753856659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,8,balanced,0.07675200204054515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,8,balanced,0.07663466533025105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,8,balanced,0.07740266621112823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,8,balanced,0.07852800190448761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,8,balanced,0.0794293334086736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,8,balanced,0.0794293334086736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,8,balanced,0.08176533381144206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,8,balanced,0.08259200056393941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,8,balanced,0.08284799754619598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,8,balanced,0.09071999788284302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,8,balanced,0.09119466940561931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,8,balanced,0.09515733520189922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,8,balanced,0.10576533277829488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,8,balanced,0.11406399806340535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,8,balanced,0.13266133268674216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,8,balanced,0.1524853308995565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,8,balanced,0.1897760033607483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,8,balanced,0.23201600710550943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,8,balanced,0.3207520047823588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,8,balanced,0.3905706803003947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,8,balanced,0.55676798025767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,8,balanced,0.7178880373636881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,4,balanced,0.029487999776999157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,4,balanced,0.02739733209212621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,4,balanced,0.03323733309904734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,4,balanced,0.050288001696268715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,4,balanced,0.07683733105659485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,4,balanced,0.07721066474914551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,4,balanced,0.07852266728878021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,4,balanced,0.07658133407433827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,4,balanced,0.07858666777610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,4,balanced,0.07867733140786488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,4,balanced,0.07754133145014445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,4,balanced,0.07932800054550171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,4,balanced,0.07919999957084656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,4,balanced,0.08032000064849854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,4,balanced,0.08682133754094441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,4,balanced,0.08620267113049825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,4,balanced,0.08851733803749084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,4,balanced,0.09275199969609578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,4,balanced,0.10682132840156555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,4,balanced,0.12009066343307495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,4,balanced,0.11546666423479716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,4,balanced,0.14405333002408346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,4,balanced,0.1542026698589325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,4,balanced,0.21450134118398032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,4,balanced,0.2472053368886312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,4,balanced,0.3384053309758504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,4,balanced,0.4230399926503499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,8,balanced,0.023717333873112995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,8,balanced,0.025221332907676697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,8,balanced,0.0271573339899381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,8,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,8,balanced,0.0516533354918162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,8,balanced,0.07634133100509644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,8,balanced,0.07736533383528392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,8,balanced,0.07924266656239827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,8,balanced,0.09086400270462036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,8,balanced,0.08894399801890056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,8,balanced,0.09161067008972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,8,balanced,0.12345600128173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,8,balanced,0.11924800276756287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,8,balanced,0.12000532944997151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,8,balanced,0.16840000947316489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,8,balanced,0.16978667179743448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,8,balanced,0.1790026624997457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,8,balanced,0.18175466855367026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,8,balanced,0.1888373295466105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,8,balanced,0.20895999670028687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,8,balanced,0.23094399770100912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,8,balanced,0.2727893392244975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,8,balanced,0.3148266673088074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,8,balanced,0.38948265711466473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,8,balanced,0.46594667434692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,8,balanced,0.7278133233388265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,8,balanced,0.8871680100758871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,2,power_law_1.01,0.028915199637413024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,2,power_law_1.01,0.030067199468612672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,2,power_law_1.01,0.028825598955154418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,2,power_law_1.01,0.032332798838615416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,2,power_law_1.01,0.03250559866428375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,2,power_law_1.01,0.033452799916267394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,2,power_law_1.01,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,2,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,2,power_law_1.01,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,2,power_law_1.01,0.03841919898986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,2,power_law_1.01,0.037651199102401736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,2,power_law_1.01,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,2,power_law_1.01,0.04254080057144165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,2,power_law_1.01,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,2,power_law_1.01,0.04927999973297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,2,power_law_1.01,0.05240960121154785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,2,power_law_1.01,0.05859839916229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,2,power_law_1.01,0.06555520296096802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,2,power_law_1.01,0.07685760259628296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,2,power_law_1.01,0.09967359900474548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,2,power_law_1.01,0.11632000207901001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,2,power_law_1.01,0.14746880531311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,2,power_law_1.01,0.17537920475006102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,2,power_law_1.01,0.2636224031448364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,256,balanced,0.05106666684150696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,256,balanced,0.05323733389377594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,256,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,256,balanced,0.05583466589450836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,256,balanced,0.05417599777380625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,256,balanced,0.055162668228149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,256,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,256,balanced,0.05830933153629303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,256,balanced,0.05598400036493937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,256,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,256,balanced,0.0577706644932429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,256,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,256,balanced,0.05994666616121928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,256,balanced,0.06087466577688853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,256,balanced,0.06287466486295064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,256,balanced,0.06410133341948192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,256,balanced,0.0647626668214798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,256,balanced,0.07041599849859874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,256,balanced,0.07249600191911061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,256,balanced,0.08089066545168559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,256,balanced,0.0869653324286143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,256,balanced,0.10136533776919048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,256,balanced,0.1174720029036204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,256,balanced,0.150709331035614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,256,balanced,0.18088533480962118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,256,balanced,0.24871466557184854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,256,balanced,0.30876266956329346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,2,power_law_1.01,0.30919680595397947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,8,balanced,0.02769600103298823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,8,balanced,0.02698666602373123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,8,balanced,0.025546667476495106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,8,balanced,0.025434667865435284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,8,balanced,0.027232001225153606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,8,balanced,0.029418667157491047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,8,balanced,0.05830933153629303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,8,balanced,0.05894933144251505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,8,balanced,0.060677334666252136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,8,balanced,0.05453866720199585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,8,balanced,0.04577066500981649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,8,balanced,0.08452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,8,balanced,0.07985599835713704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,8,balanced,0.062080000837643944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,8,balanced,0.09201066692670186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,8,balanced,0.0746506651242574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,8,balanced,0.09681600332260132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,8,balanced,0.11978666981061299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,8,balanced,0.14427733421325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,8,balanced,0.20410666863123575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,8,balanced,0.25961599747339886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,8,balanced,0.3622879981994629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,8,balanced,0.47117865085601807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,8,balanced,0.67795197168986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,8,balanced,0.8895040353139242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,8,balanced,1.3131732940673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,8,balanced,1.7308425903320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,2,power_law_1.01,0.47109122276306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,2,power_law_1.01,0.06213120222091675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,2,power_law_1.01,0.07006719708442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,2,power_law_1.01,0.08416640162467956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,2,power_law_1.01,0.10636160373687745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,2,power_law_1.01,0.13335039615631103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,2,power_law_1.01,0.14643839597702027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,2,power_law_1.01,0.18443520069122316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,2,power_law_1.01,0.1866752028465271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,2,power_law_1.01,0.1906559944152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,2,power_law_1.01,0.1917248010635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,2,power_law_1.01,0.6027967929840088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,2,power_law_1.01,0.20127999782562256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,2,power_law_1.01,0.20193920135498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,2,power_law_1.01,0.2016063928604126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,2,power_law_1.01,0.2160192012786865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,2,power_law_1.01,0.2200000047683716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,2,power_law_1.01,0.21953918933868408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,2,power_law_1.01,0.23479039669036866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,2,power_law_1.01,0.2557120084762573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,2,power_law_1.01,0.2904896020889282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,2,power_law_1.01,0.34708480834960936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,2,power_law_1.01,0.39415678977966306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,2,power_law_1.01,0.4947840213775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,2,power_law_1.01,0.5944896221160889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,2,power_law_1.01,0.838924789428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,2,power_law_1.01,1.0296832084655763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,2,power_law_1.01,1.4557503700256347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,2,power_law_1.01,1.9110143661499024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,8,balanced,0.046181331078211464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,8,balanced,0.046240001916885376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,8,balanced,0.05225066840648651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,8,balanced,0.07036266724268596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,8,balanced,0.09829333424568176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,8,balanced,0.14856533209482828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,8,balanced,0.1495519975821177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,8,balanced,0.15229866902033487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,8,balanced,0.15134933590888977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,8,balanced,0.15331733226776123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,8,balanced,0.15316266814867655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,8,balanced,0.15709867080052695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,8,balanced,0.1561973293622335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,8,balanced,0.15822399655977884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,8,balanced,0.16334399580955505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,8,balanced,0.16470932960510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,8,balanced,0.16895467042922974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,8,balanced,0.1809920072555542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,8,balanced,0.18577067057291666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,8,balanced,0.20371200640996298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,8,balanced,0.22539732853571573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,8,balanced,0.26134934027989704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,8,balanced,0.2956213355064392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,8,balanced,0.36689066886901855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,8,balanced,0.43400001525878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,8,balanced,0.6083199977874756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,8,balanced,0.7405493259429932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,1,power_law_1.01,0.12043520212173461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,1,power_law_1.01,0.17531520128250122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,1,power_law_1.01,0.24516479969024657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,1,power_law_1.01,0.3830591917037964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,1,power_law_1.01,0.3702336072921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,1,power_law_1.01,0.3821183919906616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,1,power_law_1.01,0.37112960815429685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,1,power_law_1.01,0.3918015956878662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,1,power_law_1.01,0.39331200122833254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,1,power_law_1.01,0.4017471790313721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,1,power_law_1.01,0.3939647912979126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,1,power_law_1.01,0.4228032112121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,1,power_law_1.01,0.4273215770721436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,1,power_law_1.01,0.44564480781555177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,1,power_law_1.01,0.4648575782775879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,1,power_law_1.01,0.4848127841949463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,1,power_law_1.01,0.5324480056762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,1,power_law_1.01,0.5819712162017823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,1,power_law_1.01,0.7037055969238282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,1,power_law_1.01,0.8347904205322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,1,power_law_1.01,0.9813823699951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,1,power_law_1.01,1.3137920379638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,64,power_law_1.2,0.0651199996471405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,1,power_law_1.01,1.693017578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,64,power_law_1.2,0.06234239935874939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,64,power_law_1.2,0.06109439730644226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,64,power_law_1.2,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,64,power_law_1.2,0.060185599327087405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,64,power_law_1.2,0.06285439729690552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,64,power_law_1.2,0.0627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,64,power_law_1.2,0.06550400257110596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,64,power_law_1.2,0.06622080206871032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,64,power_law_1.2,0.06499199867248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,64,power_law_1.2,0.06626560091972351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,64,power_law_1.2,0.07010560035705567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,64,power_law_1.2,0.0677183985710144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,64,power_law_1.2,0.06780800223350525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,64,power_law_1.2,0.07315840125083924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,128,power_law_1.01,0.018348799645900728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,64,power_law_1.2,0.0730239987373352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,128,power_law_1.01,0.018195199966430663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,64,power_law_1.2,0.07857919931411743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,128,power_law_1.01,0.018675200641155243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,64,power_law_1.2,0.08677759766578674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,128,power_law_1.01,0.021107199788093566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.2,0.09581440091133117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.2,0.12327040433883667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.2,0.13519359827041627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,128,power_law_1.01,0.03870719969272614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,128,power_law_1.01,0.029420799016952513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,128,power_law_1.01,0.02961280047893524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,128,power_law_1.01,0.03020159900188446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,128,power_law_1.01,0.03017599880695343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,128,power_law_1.01,0.030924800038337707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,128,power_law_1.01,0.030156800150871278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,128,power_law_1.01,0.031302401423454286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,128,power_law_1.01,0.03132160007953644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,128,power_law_1.01,0.03187839984893799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,128,power_law_1.01,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,128,power_law_1.01,0.03463680148124695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,128,power_law_1.01,0.03506560027599335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,128,power_law_1.01,0.0379584014415741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.01,0.04007680118083954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.01,0.04550400078296661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.01,0.05023360252380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.01,0.06007680296897888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,1,power_law_1.01,2.409190368652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.2,0.1795904040336609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.2,0.21716480255126952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.01,0.07240319848060608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.2,0.29210240840911866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.01,0.0989247977733612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.2,0.3641792058944702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.01,0.12499840259552002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.01,0.17374080419540405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.2,0.5420928001403809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.01,0.22140159606933593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.2,0.7162496089935303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,1,power_law_1.01,3.0814016342163084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,1,power_law_1.01,4.329670333862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,32,power_law_1.2,0.1306496024131775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,32,power_law_1.2,0.14840960502624512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,32,power_law_1.2,0.14442880153656007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,32,power_law_1.2,0.14871679544448851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,32,power_law_1.2,0.14778239727020265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,32,power_law_1.2,0.147161602973938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,32,power_law_1.2,0.13829760551452636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,32,power_law_1.2,0.14375679492950438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,32,power_law_1.2,0.14552320241928102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,32,power_law_1.2,0.14404480457305907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,32,power_law_1.2,0.14859520196914672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,32,power_law_1.2,0.15193599462509155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,32,power_law_1.2,0.1506816029548645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,32,power_law_1.2,0.14246400594711303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,32,power_law_1.2,0.16058239936828614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,32,power_law_1.2,0.16663680076599122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,32,power_law_1.2,0.17406079769134522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,32,power_law_1.2,0.1890112042427063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,32,power_law_1.2,0.19530240297317505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,32,power_law_1.2,0.22714879512786865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,8,power_law_1.01,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,32,power_law_1.2,0.26680960655212405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,8,power_law_1.01,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,8,power_law_1.01,0.035334399342536925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,32,power_law_1.2,0.3490943908691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,8,power_law_1.01,0.03767040073871612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,8,power_law_1.01,0.03758719861507416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,8,power_law_1.01,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,8,power_law_1.01,0.04629760086536407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,8,power_law_1.01,0.04912000000476837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,8,power_law_1.01,0.0530239999294281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,8,power_law_1.01,0.055718398094177245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,8,power_law_1.01,0.05475199818611145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,8,power_law_1.01,0.0576960027217865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,8,power_law_1.01,0.06053760051727295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,8,power_law_1.01,0.06442239880561829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,8,power_law_1.01,0.07736960053443909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,8,power_law_1.01,0.07787520289421082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,8,power_law_1.01,0.08558080196380616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,32,power_law_1.2,0.405017614364624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,8,power_law_1.01,0.10983680486679077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,32,power_law_1.2,0.6035711765289307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,32,power_law_1.2,0.7458879947662354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,8,power_law_1.01,0.1293887972831726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,32,power_law_1.2,1.0822784423828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,2,balanced,0.12352533141771953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,2,balanced,0.1285760005315145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,2,balanced,0.13772799571355185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,8,power_law_1.01,0.16671359539031982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,2,balanced,0.15476266543070474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,2,balanced,0.17589867115020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,2,balanced,0.22406399250030518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,2,balanced,0.2731626629829407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,2,balanced,0.2747946580251058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,2,balanced,0.2763520081837972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,2,balanced,0.2757173379262288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,2,balanced,0.28143999973932904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,2,balanced,0.2870826721191406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,2,balanced,0.2902560035387675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,2,balanced,0.2966880003611247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,32,power_law_1.2,1.526688003540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,8,power_law_1.01,0.18511999845504762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,8,power_law_1.01,0.26716160774230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,8,power_law_1.01,0.32685439586639403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,8,power_law_1.01,0.4672383785247803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,1,power_law_1.01,5.668262481689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,8,power_law_1.01,0.5811903953552247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,2,balanced,0.04529599845409393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,2,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,2,balanced,0.05675200124581655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,2,balanced,0.08356799681981404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,2,balanced,0.11342933773994446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,2,balanced,0.13385066390037537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,2,balanced,0.13538666566212973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,2,balanced,0.13730133573214212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,2,balanced,0.14006400108337402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,2,balanced,0.14114133516947427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,2,balanced,0.13954133788744608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,2,balanced,0.1458399991194407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,2,balanced,0.15119999647140503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,2,balanced,0.15347199638684592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,2,balanced,0.16216533382733664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,2,balanced,0.16756800810496011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,2,balanced,0.17567465702692667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,2,balanced,0.20122667153676352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,2,balanced,0.21901333332061768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,2,balanced,0.2841920057932536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,2,balanced,0.3049546678860982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,2,balanced,0.4366613229115804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,2,balanced,0.5055413246154785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,2,balanced,0.7729173501332601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,2,balanced,0.9108959833780924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,2,balanced,1.3555413881937664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,2,balanced,1.701877276102702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,8,power_law_1.01,0.8544192314147949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,8,power_law_1.01,1.0852288246154784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,8,power_law_1.2,0.047628799080848695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,8,power_law_1.2,0.050380802154541014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,8,power_law_1.2,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,8,power_law_1.2,0.062054401636123656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,8,power_law_1.2,0.06977919936180114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,8,power_law_1.2,0.08068479895591736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,8,power_law_1.2,0.07912319898605347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,8,power_law_1.2,0.07827839851379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,8,power_law_1.2,0.07391999959945679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,8,power_law_1.2,0.07802240252494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,8,power_law_1.2,0.07796480059623719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,8,power_law_1.2,0.08263040184974671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,8,power_law_1.2,0.08697599768638611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,8,power_law_1.2,0.09377920031547546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,8,power_law_1.2,0.10529279708862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,8,power_law_1.2,0.10624639987945557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,8,power_law_1.2,0.11665920019149781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,8,power_law_1.2,0.13507200479507447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,8,power_law_1.2,0.14526079893112182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,8,power_law_1.2,0.18592000007629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,8,power_law_1.2,0.22293760776519775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,8,power_law_1.2,0.2862272024154663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,8,power_law_1.2,0.3823999881744385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,8,power_law_1.2,0.52674560546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,8,power_law_1.2,0.6830912113189698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,8,power_law_1.2,0.923635196685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,8,power_law_1.2,1.1583552360534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,64,power_law_1.2,0.0673088014125824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,64,power_law_1.2,0.054150402545928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,64,power_law_1.2,0.04533120095729828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,64,power_law_1.2,0.04840959906578064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,64,power_law_1.2,0.04883840084075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,64,power_law_1.2,0.04962559938430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,64,power_law_1.2,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,64,power_law_1.2,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,64,power_law_1.2,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,64,power_law_1.2,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,64,power_law_1.2,0.05211520195007324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,64,power_law_1.2,0.05348479747772217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,64,power_law_1.2,0.05546240210533142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,64,power_law_1.2,0.05861120223999024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,64,power_law_1.2,0.06513280272483826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,64,power_law_1.2,0.06794880032539367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,64,power_law_1.2,0.07052159905433655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,64,power_law_1.2,0.08092799782752991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,64,power_law_1.2,0.09107199907302857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,64,power_law_1.2,0.10784640312194824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,64,power_law_1.2,0.12227200269699097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,64,power_law_1.2,0.16222720146179198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,64,power_law_1.2,0.18227200508117675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,64,power_law_1.2,0.2598207950592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,64,power_law_1.2,0.30398719310760497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,64,power_law_1.2,0.4670976161956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,64,power_law_1.2,0.6129536151885986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,8,power_law_1.01,0.05575680136680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,8,power_law_1.01,0.06931840181350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,8,power_law_1.01,0.09160959720611572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,8,power_law_1.01,0.08751360177993775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,8,power_law_1.01,0.09848319888114929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,8,power_law_1.01,0.10906879901885987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,4,power_law_1.01,0.051020801067352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,8,power_law_1.01,0.11039999723434449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,4,power_law_1.01,0.05400959849357605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,8,power_law_1.01,0.12042239904403687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,4,power_law_1.01,0.0590719997882843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,4,power_law_1.01,0.06380800008773804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,4,power_law_1.01,0.06458240151405334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,4,power_law_1.01,0.06548479795455933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,4,power_law_1.01,0.06599680185317994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,4,power_law_1.01,0.0662015974521637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,8,power_law_1.01,0.10892159938812256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,4,power_law_1.01,0.06670079827308655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,8,power_law_1.01,0.10919680595397949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,4,power_law_1.01,0.06740480065345764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,8,power_law_1.01,0.11194239854812622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,4,power_law_1.01,0.06738560199737549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,8,power_law_1.01,0.11653759479522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,8,power_law_1.01,0.11354240179061889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,8,power_law_1.01,0.11570559740066529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,8,power_law_1.01,0.12972160577774047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,8,power_law_1.01,0.13000320196151732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,4,power_law_1.01,0.0772607982158661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,8,power_law_1.01,0.1398527979850769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,8,power_law_1.01,0.15323519706726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,8,power_law_1.01,0.16851840019226075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,8,power_law_1.01,0.19487359523773193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,8,power_law_1.01,0.22542080879211426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,8,power_law_1.01,0.28015360832214353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,4,power_law_1.01,0.08280959725379944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,8,power_law_1.01,0.34545919895172117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,4,power_law_1.01,0.08841599822044373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,8,power_law_1.01,0.45041279792785643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,8,power_law_1.01,0.6155263900756835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,8,power_law_1.01,0.830777645111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,8,power_law_1.01,1.083027172088623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,4,power_law_1.01,0.10439039468765259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,4,power_law_1.01,0.1041152000427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,4,power_law_1.01,0.12602239847183228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,4,power_law_1.01,0.15749119520187377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,4,power_law_1.01,0.19309439659118652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,4,power_law_1.01,0.239136004447937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,4,power_law_1.01,0.32235519886016845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,4,power_law_1.01,0.36097280979156493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,4,power_law_1.01,0.554310417175293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,4,power_law_1.01,0.7664319992065429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,4,power_law_1.01,1.1079999923706054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,4,power_law_1.01,1.5742079734802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,16,power_law_1.01,0.05975679755210876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,16,power_law_1.01,0.057785600423812866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,16,power_law_1.01,0.05671039819717407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,16,power_law_1.01,0.05937280058860779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,16,power_law_1.01,0.05813760161399841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,16,power_law_1.01,0.06382079720497132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,16,power_law_1.01,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,16,power_law_1.01,0.06537600159645081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,16,power_law_1.01,0.06405760049819946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,16,power_law_1.01,0.0648576021194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,16,power_law_1.01,0.06558079719543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,16,power_law_1.01,0.06899200081825256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,16,power_law_1.01,0.06880000233650208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,16,power_law_1.01,0.07057920098304749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,16,power_law_1.01,0.07663999795913697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,16,power_law_1.01,0.07921280264854431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,16,power_law_1.01,0.08850560188293458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,16,power_law_1.01,0.0991104006767273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,16,power_law_1.01,0.10487680435180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,4,power_law_1.01,2.2208768844604494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,16,power_law_1.01,0.12392319440841675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,16,power_law_1.01,0.14841599464416505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,16,power_law_1.01,0.19498239755630492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,16,power_law_1.01,0.23573760986328124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,16,power_law_1.01,0.3248831987380981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,16,power_law_1.01,0.46062722206115725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,16,power_law_1.01,0.6880640029907227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,16,power_law_1.01,1.127507209777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,2,balanced,0.05596266686916351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,2,balanced,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,2,balanced,0.0904906690120697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,2,balanced,0.13481066624323526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,2,balanced,0.20518932739893594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,2,balanced,0.2676960031191508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,2,balanced,0.2737013300259908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,2,balanced,0.2718133330345154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,2,balanced,0.27322665850321454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,2,balanced,0.27553067604700726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,2,balanced,0.2799733281135559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,2,balanced,0.28488000233968097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,2,balanced,0.2861973245938619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,2,balanced,0.2892799973487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,2,balanced,0.29949333270390827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,2,balanced,0.30426132678985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,2,balanced,0.31888532638549805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,2,balanced,0.3740533192952474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,2,balanced,0.3712853193283081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,2,balanced,0.5248159964879354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,2,balanced,0.5245120127995809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,2,balanced,0.8020053704579672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,2,balanced,0.8108747005462646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,2,balanced,1.1995200316111247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,2,balanced,1.3736054102579753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,2,balanced,2.0527733167012534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,2,balanced,2.514911969502767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,8,power_law_1.2,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,8,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,8,power_law_1.2,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,8,power_law_1.2,0.03973760008811951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,8,power_law_1.2,0.03956480026245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,8,power_law_1.2,0.03989759981632233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,8,power_law_1.2,0.039103999733924866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,8,power_law_1.2,0.038975998759269714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,8,power_law_1.2,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,8,power_law_1.2,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,8,power_law_1.2,0.03775359988212586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,8,power_law_1.2,0.04140160083770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,8,power_law_1.2,0.04306559860706329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,8,power_law_1.2,0.04134399890899658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,8,power_law_1.2,0.048256000876426695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,8,power_law_1.2,0.05264000296592712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,8,power_law_1.2,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,8,power_law_1.2,0.058822399377822875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.2,0.06614400148391723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.2,0.07939839959144593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.2,0.09575039744377137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.2,0.12463359832763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.2,0.1538432002067566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.2,0.21638400554656984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.2,0.2753535985946655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,64,balanced,0.08349866668383281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,64,balanced,0.0827893316745758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,64,balanced,0.08455999692281087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,64,balanced,0.088319996992747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,64,balanced,0.09713066617647807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,64,balanced,0.13059199849764505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,64,balanced,0.16890132427215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,64,balanced,0.16682666540145874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,64,balanced,0.16432000199953714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,64,balanced,0.16778665781021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,64,balanced,0.17118932803471884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,64,balanced,0.16831467549006143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,64,balanced,0.17052799463272095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,64,balanced,0.18709866205851236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,64,balanced,0.18523732821146646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,64,balanced,0.175653338432312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,64,balanced,0.17736534277598062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,64,balanced,0.1835520068804423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,64,balanced,0.19364267587661743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,64,balanced,0.21312532822291055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,64,balanced,0.21651732921600342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,64,balanced,0.2428213357925415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.2,0.397273588180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,64,balanced,0.2700693408648173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,64,balanced,0.32764265934626263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,64,balanced,0.4280639886856079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,64,balanced,0.5186986525853475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,64,balanced,0.6766506830851237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,8,power_law_1.01,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,8,power_law_1.01,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,8,power_law_1.01,0.03315199911594391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,8,power_law_1.01,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,8,power_law_1.01,0.037036800384521486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,8,power_law_1.01,0.03882879912853241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,8,power_law_1.01,0.038899201154708865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,8,power_law_1.01,0.040863999724388124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,8,power_law_1.01,0.03914879858493805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,8,power_law_1.01,0.040140798687934874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,8,power_law_1.01,0.03978880047798157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,8,power_law_1.01,0.040505599975585935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,8,power_law_1.01,0.042694398760795595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,8,power_law_1.01,0.04343039989471435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,8,power_law_1.01,0.049747198820114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,8,power_law_1.01,0.049779200553894044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,8,power_law_1.01,0.05585920214653015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,8,power_law_1.01,0.060012799501419065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,8,power_law_1.01,0.0669376015663147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,8,power_law_1.01,0.08393599987030029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,8,power_law_1.01,0.09768959879875183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,8,power_law_1.01,0.12209279537200927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,8,power_law_1.01,0.1509376049041748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,8,power_law_1.01,0.2105151891708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,8,power_law_1.01,0.28469760417938234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,8,power_law_1.01,0.41656317710876467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,8,power_law_1.01,0.5747839927673339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.2,0.5158271789550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,2,balanced,0.031498665610949196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,2,balanced,0.029178666571776073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,2,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,2,balanced,0.05332266787687937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,2,balanced,0.07930133243401845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,2,balanced,0.08196799953778584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,2,balanced,0.08251733581225078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,2,balanced,0.08245866497357686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,2,balanced,0.08253333469231923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,2,balanced,0.08436800042788188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,2,balanced,0.08346666892369588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,2,balanced,0.08616532882054646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,2,balanced,0.08582400282224019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,2,balanced,0.08783466617266338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,2,balanced,0.0953546663125356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,2,balanced,0.09710933764775594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,2,balanced,0.10017599662144978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,2,balanced,0.11036800344785054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,2,balanced,0.11479467153549194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,2,balanced,0.13537599643071493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,2,balanced,0.14166399836540222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,2,balanced,0.18228799104690552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,2,balanced,0.21809067328770956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,2,balanced,0.2983413338661194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,2,balanced,0.36213334401448566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,2,balanced,0.5043946504592896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,2,balanced,0.6600746711095175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,1,power_law_1.01,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,1,power_law_1.01,0.05377280116081238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,1,power_law_1.01,0.07345920205116271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,1,power_law_1.01,0.10175360441207885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,1,power_law_1.01,0.12568960189819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,1,power_law_1.01,0.18273279666900635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,1,power_law_1.01,0.1927616000175476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,1,power_law_1.01,0.19463679790496827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,1,power_law_1.01,0.19857920408248902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,1,power_law_1.01,0.20385921001434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,1,power_law_1.01,0.20853760242462158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,1,power_law_1.01,0.21357440948486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,1,power_law_1.01,0.21813759803771973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,1,power_law_1.01,0.2283008098602295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,1,power_law_1.01,0.2380352020263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,1,power_law_1.01,0.24552319049835206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,1,power_law_1.01,0.2582655906677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,1,power_law_1.01,0.29006080627441405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,1,power_law_1.01,0.32274560928344725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,1,power_law_1.01,0.3899327993392944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,1,power_law_1.01,0.4373631954193115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,1,power_law_1.01,0.5558080196380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,1,power_law_1.01,0.6663551807403565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,1,power_law_1.01,0.8870207786560058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,1,power_law_1.01,1.1048831939697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,1,power_law_1.01,1.545900821685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,1,power_law_1.01,1.9852415084838868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,4,balanced,0.03179199993610382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,4,balanced,0.02938133229811986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,4,balanced,0.03160000095764796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,4,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,4,balanced,0.059290667374928795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,4,balanced,0.05946133534113566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,4,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,4,balanced,0.06053866446018219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,4,balanced,0.06049066781997681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,4,balanced,0.06033066908518473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,4,balanced,0.06253333389759064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,4,balanced,0.06262933214505513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,4,balanced,0.06542933483918507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,4,balanced,0.06709866722424825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,4,balanced,0.07255466779073079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,4,balanced,0.07525333265463512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,4,balanced,0.07637866834799449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,4,balanced,0.08308266599973042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,4,balanced,0.09518933296203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,4,balanced,0.10909866293271382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,4,balanced,0.11746133367220561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,4,balanced,0.1476533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,4,balanced,0.17055465777715048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,4,balanced,0.24243199825286865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,4,balanced,0.295306662718455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,4,balanced,0.42556265989939374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,4,balanced,0.5505439837773641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,4,power_law_1.2,0.13267840147018434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,4,power_law_1.2,0.18205440044403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,4,power_law_1.2,0.19011839628219604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,4,power_law_1.2,0.24712960720062255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,4,power_law_1.2,0.3211456060409546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,4,power_law_1.2,0.39302399158477785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,4,power_law_1.2,0.42471680641174314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,4,power_law_1.2,0.44728960990905764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,4,power_law_1.2,0.44099841117858884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,4,power_law_1.2,0.40711679458618166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,4,power_law_1.2,0.4184000015258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,4,power_law_1.2,0.4423679828643799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,4,power_law_1.2,0.45217280387878417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,4,power_law_1.2,0.4803455829620361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,4,power_law_1.2,0.47185277938842773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,4,power_law_1.2,0.45759358406066897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,4,power_law_1.2,0.48248958587646484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,4,power_law_1.2,0.5104447841644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,4,power_law_1.2,0.5622848033905029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,4,power_law_1.2,0.6265791893005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,4,power_law_1.2,0.6738880157470704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,4,power_law_1.2,0.809721565246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,4,power_law_1.2,0.9462976455688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,4,power_law_1.2,1.255571174621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,4,power_law_1.2,1.4899968147277831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,4,power_law_1.2,2.1106496810913087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,4,power_law_1.2,2.590336036682129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,2,power_law_1.2,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,2,power_law_1.2,0.03384959995746613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,2,power_law_1.2,0.03464959859848023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,2,power_law_1.2,0.03790079951286316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,2,power_law_1.2,0.037939199805259706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,2,power_law_1.2,0.03858560025691986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,2,power_law_1.2,0.038675200939178464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,2,power_law_1.2,0.03922559916973114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,2,power_law_1.2,0.040275201201438904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,2,power_law_1.2,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,2,power_law_1.2,0.04095999896526337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,2,power_law_1.2,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,2,power_law_1.2,0.045798400044441225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,2,power_law_1.2,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,2,power_law_1.2,0.056601601839065555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,2,power_law_1.2,0.05834239721298218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,2,power_law_1.2,0.061247998476028444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,2,power_law_1.2,0.0713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.2,0.07958400249481201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.2,0.10586240291595458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.2,0.11683839559555054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.2,0.16238080263137816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.2,0.2061824083328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,2,power_law_1.01,0.03956480026245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,2,power_law_1.01,0.043059200048446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,2,power_law_1.01,0.053964799642562865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,2,power_law_1.01,0.057766401767730714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,2,power_law_1.01,0.05886080265045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,2,power_law_1.01,0.06046720147132874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,2,power_law_1.01,0.06103039979934692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,2,power_law_1.01,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,2,power_law_1.01,0.06387839913368225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,2,power_law_1.01,0.06576640009880066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,2,power_law_1.01,0.06781439781188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,2,power_law_1.01,0.07009919881820678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,2,power_law_1.01,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.2,0.2712831974029541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,2,power_law_1.01,0.07619839906692505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,2,power_law_1.01,0.08509439826011658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,2,power_law_1.01,0.09134719967842102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,2,power_law_1.01,0.08969600200653076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,2,power_law_1.01,0.10730880498886108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,2,power_law_1.01,0.11388159990310669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,1,power_law_1.01,0.036857599020004274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,1,power_law_1.01,0.04296320080757141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,1,power_law_1.01,0.058739197254180905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,1,power_law_1.01,0.07015680074691773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,1,power_law_1.01,0.08624640107154846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,1,power_law_1.01,0.11907839775085449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,1,power_law_1.01,0.12254079580307006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,1,power_law_1.01,0.13007999658584596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,1,power_law_1.01,0.13114240169525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,1,power_law_1.01,0.13239680528640746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,1,power_law_1.01,0.14094079732894899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,1,power_law_1.01,0.1486207962036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,1,power_law_1.01,0.15478399991989136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,1,power_law_1.01,0.15734399557113649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,1,power_law_1.01,0.1664896011352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,1,power_law_1.01,0.1711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,1,power_law_1.01,0.18426239490509033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,1,power_law_1.01,0.21278080940246583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,1,power_law_1.01,0.23819520473480224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,2,power_law_1.01,0.15110399723052978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,2,power_law_1.01,0.17738879919052125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,1,power_law_1.01,0.2947776079177856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,1,power_law_1.01,0.3410304069519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,1,power_law_1.01,0.45210881233215333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,1,power_law_1.01,0.5653056144714356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,1,power_law_1.01,0.7599743843078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,1,power_law_1.01,0.9495360374450683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,2,power_law_1.01,0.24233601093292237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,1,power_law_1.01,1.343513584136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,1,power_law_1.01,1.78088321685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.2,0.2967103958129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,2,power_law_1.01,0.26817920207977297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,2,power_law_1.01,0.41326079368591306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,2,power_law_1.01,0.46341118812561033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,4,balanced,0.040048000713189445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,4,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,4,balanced,0.045834665497144066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,4,balanced,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,4,balanced,0.045781334241231285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,4,balanced,0.04567466676235199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,4,balanced,0.04571733375390371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,4,balanced,0.04984533290068308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,4,balanced,0.04729066789150238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,4,balanced,0.04576000074545542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,4,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,4,balanced,0.045797333121299744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,4,balanced,0.050154666105906166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,4,balanced,0.04605866471926371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,4,balanced,0.048112000028292336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,4,balanced,0.050383999943733215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,4,balanced,0.0521919975678126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,4,balanced,0.06201066573460897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,4,balanced,0.06035199761390686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,4,balanced,0.07425066828727722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,4,balanced,0.08285333216190338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,4,balanced,0.10733333230018616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,4,balanced,0.12572266658147177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,4,balanced,0.16847999890645346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,4,balanced,0.20670400063196817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,4,balanced,0.28868265946706134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,4,balanced,0.3693813482920329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.2,0.4127488136291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,2,power_law_1.01,0.6462975978851319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,2,power_law_1.01,0.9287487983703613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,1,power_law_1.2,0.09345279932022095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,1,power_law_1.2,0.10056960582733154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,1,power_law_1.2,0.11468160152435303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,1,power_law_1.2,0.15066879987716675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,1,power_law_1.2,0.19385600090026855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,1,power_law_1.2,0.2375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,1,power_law_1.2,0.267411208152771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,1,power_law_1.2,0.3356544017791748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,1,power_law_1.2,0.34461441040039065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,1,power_law_1.2,0.35019519329071047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,1,power_law_1.2,0.365337610244751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,64,power_law_1.2,0.043654400110244754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,1,power_law_1.2,0.3764096021652222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,64,power_law_1.2,0.043558400869369504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,1,power_law_1.2,0.3866559982299805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,64,power_law_1.2,0.04289279878139496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,1,power_law_1.2,0.40510082244873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,64,power_law_1.2,0.04613119959831238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,1,power_law_1.2,0.41800317764282224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,64,power_law_1.2,0.048492801189422605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,1,power_law_1.2,0.43118720054626464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,64,power_law_1.2,0.04869759976863861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,1,power_law_1.2,0.4480576038360596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,64,power_law_1.2,0.04924159944057464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,64,power_law_1.2,0.050335997343063356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,1,power_law_1.2,0.5160192012786865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.2,0.5594751834869385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.2,0.6593023777008057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.2,0.7509632110595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.2,0.9489215850830078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.2,1.143404769897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,64,power_law_1.2,0.05570560097694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,64,power_law_1.2,0.05621119737625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.2,1.5278400421142577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,64,power_law_1.2,0.05731840133666992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,1,power_law_1.2,2.0550912857055663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,1,power_law_1.2,2.8220287322998048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,1,power_law_1.2,3.4283073425292967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,64,power_law_1.2,0.054739201068878175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,64,power_law_1.2,0.05555840134620667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,64,power_law_1.2,0.060838401317596436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,64,power_law_1.2,0.06476799845695495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,64,power_law_1.2,0.06380159854888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,64,power_law_1.2,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,64,power_law_1.2,0.08208000063896179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,64,power_law_1.2,0.08505600094795226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,64,power_law_1.2,0.10983040332794189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,64,power_law_1.2,0.12842240333557128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,64,power_law_1.2,0.1739135980606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,64,power_law_1.2,0.20604159832000732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,64,power_law_1.2,0.30570240020751954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,64,power_law_1.2,0.38772480487823485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,64,power_law_1.2,0.6059328079223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,64,power_law_1.2,0.8480064392089843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.2,0.5739776134490967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,1,power_law_1.01,0.15648640394210817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,1,power_law_1.01,0.25411200523376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,1,power_law_1.01,0.4514304161071777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,1,power_law_1.01,0.8385279655456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,1,power_law_1.01,1.4695424079895019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,1,power_law_1.01,2.097305679321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,1,power_law_1.01,2.440019226074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,1,power_law_1.01,3.173971176147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,1,power_law_1.01,3.2871551513671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,1,power_law_1.01,3.3643520355224608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,1,power_law_1.01,3.545702362060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,1,power_law_1.01,3.7535552978515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,1,power_law_1.01,3.7632640838623046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,1,power_law_1.01,3.8339775085449217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,32,power_law_1.2,0.19213440418243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,1,power_law_1.01,4.000729751586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,32,power_law_1.2,0.25490560531616213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,32,power_law_1.2,0.25180160999298096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,1,power_law_1.01,4.126553726196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,32,power_law_1.2,0.250380802154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,32,power_law_1.2,0.2521791934967041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,32,power_law_1.2,0.21870720386505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,1,power_law_1.01,4.374361419677735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,32,power_law_1.2,0.22772479057312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,32,power_law_1.2,0.2440000057220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,1,power_law_1.01,4.592512130737305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,1,power_law_1.01,4.859142303466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,32,power_law_1.2,0.23939840793609618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,32,power_law_1.2,0.23788158893585204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,32,power_law_1.2,0.2310015916824341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,32,power_law_1.2,0.2510335922241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,32,power_law_1.2,0.23697919845581056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,32,power_law_1.2,0.23503360748291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,1,power_law_1.01,5.434444808959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,32,power_law_1.2,0.2438271999359131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,32,power_law_1.2,0.245849609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,32,power_law_1.2,0.2514303922653198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,1,power_law_1.01,5.205759811401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,32,power_law_1.2,0.2755712032318115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,32,power_law_1.2,0.30271360874176023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,32,power_law_1.2,0.3277951955795288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,32,power_law_1.2,0.36110720634460447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,1,power_law_1.01,5.777695846557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,32,power_law_1.2,0.48137598037719725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,32,power_law_1.2,0.5621632099151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,32,power_law_1.2,0.7348608016967774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,32,power_law_1.2,0.883244800567627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,1,power_law_1.01,6.935174560546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,32,power_law_1.2,1.2703488349914551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,32,power_law_1.2,1.605824089050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,1,power_law_1.01,8.68545913696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,1,power_law_1.01,8.492665863037109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,1,power_law_1.01,10.746854400634765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,32,balanced,0.043621331453323364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,32,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,32,balanced,0.04725866516431173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,32,balanced,0.04284266630808512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,32,balanced,0.043434664607048035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,1,power_law_1.01,14.189613342285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,32,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,32,balanced,0.04805333415667216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,32,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,32,balanced,0.06196266909440359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,32,balanced,0.06330666442712148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,32,balanced,0.08396800359090169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,32,balanced,0.11031466722488403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,32,balanced,0.11136000355084737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,32,balanced,0.11356266339619954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,32,balanced,0.14038933316866556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,32,balanced,0.1418453355630239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,32,balanced,0.17057067155838013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,32,balanced,0.2513226668039958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,32,balanced,0.3150239984194438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,32,balanced,0.4536373217900594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,32,balanced,0.21450134118398032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,32,balanced,0.29577600955963135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,32,balanced,0.38053866227467853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,32,balanced,0.5478933254877726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,32,balanced,0.7197333176930746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,32,balanced,1.0624053478240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,32,balanced,1.4030879338582356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,4,power_law_1.2,0.020851199328899384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,4,power_law_1.2,0.02011519968509674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,4,power_law_1.2,0.019808000326156615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,4,power_law_1.2,0.020236800611019134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,4,power_law_1.2,0.020601600408554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,4,power_law_1.2,0.0208639994263649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,4,power_law_1.2,0.02479359954595566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,4,power_law_1.2,0.023980799317359924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,4,power_law_1.2,0.048614400625228885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,4,power_law_1.2,0.04810880124568939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,4,power_law_1.2,0.04970879852771759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,4,power_law_1.2,0.04211840033531189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,4,power_law_1.2,0.03968639969825745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,4,power_law_1.2,0.03688960075378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,4,power_law_1.2,0.043136000633239746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,4,power_law_1.2,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,4,power_law_1.2,0.04267520010471344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,4,power_law_1.2,0.05541759729385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.2,0.060211199522018435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.2,0.08242560029029847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.2,0.10209920406341552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.2,0.13660800457000732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.2,0.1647104024887085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.2,0.2360383987426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,64,power_law_1.2,0.029529601335525513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,64,power_law_1.2,0.029471999406814574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,64,power_law_1.2,0.028812798857688903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,64,power_law_1.2,0.030963200330734252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,64,power_law_1.2,0.030931198596954347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,64,power_law_1.2,0.03155840039253235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,16,power_law_1.01,0.02540160119533539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,16,power_law_1.01,0.023257599771022798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,64,power_law_1.2,0.03201920092105866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,64,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,16,power_law_1.01,0.02393600046634674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,64,power_law_1.2,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,64,power_law_1.2,0.033670398592948916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,64,power_law_1.2,0.034860798716545106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,64,power_law_1.2,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,64,power_law_1.2,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,64,power_law_1.2,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,64,power_law_1.2,0.042694398760795595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,16,power_law_1.01,0.023635199666023253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,64,power_law_1.2,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,16,power_law_1.01,0.024006399512290954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,64,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,64,power_law_1.2,0.0530239999294281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,64,power_law_1.2,0.055846399068832396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,64,power_law_1.2,0.06906239986419678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,64,power_law_1.2,0.08042880296707153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,16,power_law_1.01,0.08216320276260376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,16,power_law_1.01,0.09807999730110169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,16,power_law_1.01,0.02635520100593567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,16,power_law_1.01,0.0255295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,16,power_law_1.01,0.1099776029586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,16,power_law_1.01,0.027871999144554137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,16,power_law_1.01,0.0310591995716095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,16,power_law_1.01,0.03080959916114807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,16,power_law_1.01,0.03261440098285675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,16,power_law_1.01,0.032576000690460204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,64,power_law_1.2,0.11936639547348023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,16,power_law_1.01,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,16,power_law_1.01,0.032671999931335446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,16,power_law_1.01,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,16,power_law_1.01,0.03544319868087768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,16,power_law_1.01,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,16,power_law_1.01,0.11041280031204223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,16,power_law_1.01,0.11308799982070923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,16,power_law_1.01,0.11259520053863525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,64,power_law_1.2,0.14632960557937622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.2,0.299782395362854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,16,power_law_1.01,0.10232959985733033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,16,power_law_1.01,0.11045119762420655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,16,power_law_1.01,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,16,power_law_1.01,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,16,power_law_1.01,0.10881279706954956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,16,power_law_1.01,0.05141119956970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,16,power_law_1.01,0.10654079914093018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,16,power_law_1.01,0.110316801071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,16,power_law_1.01,0.11056640148162841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,16,power_law_1.01,0.11512960195541382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,16,power_law_1.01,0.11473920345306396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,16,power_law_1.01,0.05514879822731018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,16,power_law_1.01,0.12873599529266358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,16,power_law_1.01,0.12621439695358277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,16,power_law_1.01,0.1357759952545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,16,power_law_1.01,0.1431615948677063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,16,power_law_1.01,0.16274559497833252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,16,power_law_1.01,0.06565759778022766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,16,power_law_1.01,0.1905344009399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,16,power_law_1.01,0.07222399711608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,16,power_law_1.01,0.09626240134239197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,16,power_law_1.01,0.2051392078399658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,16,power_law_1.01,0.1338047981262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,16,power_law_1.01,0.26571519374847413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,16,power_law_1.01,0.1898368000984192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,16,power_law_1.01,0.2065664052963257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,16,power_law_1.01,0.33196799755096434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,16,power_law_1.01,0.4457791805267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,16,power_law_1.01,0.5586880207061767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,16,power_law_1.01,0.848089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,16,power_law_1.01,1.1082880020141601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,64,power_law_1.2,0.20076799392700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,64,power_law_1.2,0.2515007972717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,1,power_law_1.01,0.05618559718132019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,1,power_law_1.01,0.06116480231285095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,1,power_law_1.01,0.0781440019607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,1,power_law_1.01,0.10573439598083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,1,power_law_1.01,0.13452160358428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,1,power_law_1.01,0.163046395778656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,1,power_law_1.01,0.20334079265594482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,1,power_law_1.01,0.21328001022338866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,1,power_law_1.01,0.22469758987426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,1,power_law_1.01,0.2240000009536743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,1,power_law_1.01,0.23838720321655274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,1,power_law_1.01,0.2495743989944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,1,power_law_1.01,0.2587007999420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,1,power_law_1.01,0.2746815919876099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,1,power_law_1.01,0.2937664031982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,1,power_law_1.01,0.3108992099761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,1,power_law_1.01,0.3136255979537964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,1,power_law_1.01,0.37036159038543703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,1,power_law_1.01,0.40601601600646975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,1,power_law_1.01,0.5101759910583497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,1,power_law_1.01,0.5955327987670899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,1,power_law_1.01,0.7676991939544677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,1,power_law_1.01,0.9333888053894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,1,power_law_1.01,1.2640447616577148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,1,power_law_1.01,1.5930879592895508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,1,power_law_1.01,2.2303295135498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.2,0.4352255821228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,1,power_law_1.01,2.899852752685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,64,power_law_1.2,0.34888319969177245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.2,0.5710207939147949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,64,power_law_1.2,0.4907072067260742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,2,balanced,0.04855999847253164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,2,balanced,0.06505600114663442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,16,power_law_1.2,0.020665599405765532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,2,balanced,0.08859733740488689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,16,power_law_1.2,0.02111999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,2,balanced,0.13595199584960938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,2,balanced,0.22192533810933432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,2,balanced,0.3908799886703491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,2,balanced,0.39270933469136554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,2,balanced,0.39601067701975506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,2,balanced,0.3987733523050944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,16,power_law_1.2,0.021004800498485566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,16,power_law_1.2,0.021804800629615782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,16,power_law_1.2,0.023187200725078582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,16,power_law_1.2,0.022860799729824067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,16,power_law_1.2,0.023814399540424348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,16,power_law_1.2,0.023532800376415253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,16,power_law_1.2,0.027584001421928406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,16,power_law_1.2,0.02890239953994751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,16,power_law_1.2,0.029260799288749695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,16,power_law_1.2,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,16,power_law_1.2,0.052223998308181765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,16,power_law_1.2,0.05256959795951843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,16,power_law_1.2,0.048153600096702574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,16,power_law_1.2,0.048127999901771544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,2,balanced,0.4015626509984334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,16,power_law_1.2,0.05015040040016174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,2,balanced,0.4042293230692546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,2,balanced,0.41042133172353107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,2,balanced,0.4117813507715861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,2,balanced,0.416650652885437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,2,balanced,0.42846401532491046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,2,balanced,0.4331626494725545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,2,balanced,0.4421866734822591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,2,balanced,0.4701546827952067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,2,balanced,0.49502400557200116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,2,balanced,0.5340266625086466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,16,power_law_1.2,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,2,balanced,0.5994933446248373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,16,power_law_1.2,0.05901439785957337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,2,balanced,0.6906932989756266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,16,power_law_1.2,0.06783999800682068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,2,balanced,0.7854932943979899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,16,power_law_1.2,0.07803519964218139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,2,balanced,1.0559413433074951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,16,power_law_1.2,0.09711359739303589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,16,power_law_1.2,0.11356159448623657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,16,power_law_1.2,0.15004160404205322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,16,power_law_1.2,0.20253438949584962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,16,power_law_1.2,0.26828799247741697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,16,power_law_1.2,0.362444806098938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,2,balanced,1.204090674718221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,2,balanced,1.7298720677693684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,2,balanced,2.150266647338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,2,power_law_1.2,0.12518399953842163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,2,power_law_1.2,0.16536959409713745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,2,power_law_1.2,0.23900160789489747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,2,power_law_1.2,0.3275968074798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,2,power_law_1.2,0.41262078285217285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,2,power_law_1.2,0.6633024215698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,2,power_law_1.2,0.6603072166442872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,2,power_law_1.2,0.703334379196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,1,power_law_1.01,0.05440639853477478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,2,power_law_1.2,0.7192959785461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,2,power_law_1.2,0.7209152221679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,2,power_law_1.2,0.7782847881317139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,2,power_law_1.2,0.8093503952026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,2,power_law_1.2,0.8066559791564941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,2,power_law_1.2,0.8595071792602539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,2,power_law_1.2,0.8911040306091309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,2,power_law_1.2,0.8974143981933593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,1,power_law_1.01,0.07397119998931885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,1,power_law_1.01,0.1021888017654419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,1,power_law_1.01,0.14314240217208862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,1,power_law_1.01,0.1784127950668335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,1,power_law_1.01,0.2671679973602295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,2,power_law_1.2,0.9679679870605469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,1,power_law_1.01,0.2899712085723877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,1,power_law_1.01,0.2967103958129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,2,power_law_1.2,1.0549375534057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,1,power_law_1.01,0.3026432037353516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,1,power_law_1.01,0.3105664014816284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,1,power_law_1.01,0.3150784015655518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,2,power_law_1.2,1.0993087768554688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,1,power_law_1.01,0.32914559841156005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,1,power_law_1.01,0.34275839328765867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,2,power_law_1.2,1.2590271949768066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,1,power_law_1.01,0.34523520469665525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,1,power_law_1.01,0.36092801094055177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,2,power_law_1.2,1.410694408416748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,1,power_law_1.01,0.3772671937942505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,1,power_law_1.01,0.40695037841796877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,2,power_law_1.2,1.7016704559326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,1,power_law_1.01,0.4677055835723877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,1,power_law_1.01,0.49482879638671873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,2,power_law_1.2,1.7832832336425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,1,power_law_1.01,0.6095168113708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,1,power_law_1.01,0.7430399894714356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,2,power_law_1.2,2.1998720169067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,1,power_law_1.01,0.951148796081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,2,power_law_1.2,2.9239295959472655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,1,power_law_1.01,1.087500762939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,1,power_law_1.01,1.4312704086303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,1,power_law_1.01,1.85849609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,2,power_law_1.2,3.8165313720703127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,1,power_law_1.01,2.563443183898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,1,power_law_1.01,3.209382247924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,2,power_law_1.2,4.288486480712891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,1,power_law_1.01,0.05772799849510193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,1,power_law_1.01,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,1,power_law_1.01,0.06959999799728393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,1,power_law_1.01,0.09795839786529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,1,power_law_1.01,0.1387392044067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,1,power_law_1.01,0.17039999961853028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,1,power_law_1.01,0.1892351984977722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,1,power_law_1.01,0.22287359237670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,1,power_law_1.01,0.23771519660949708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,1,power_law_1.01,0.24343678951263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,1,power_law_1.01,0.2539263963699341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,1,power_law_1.01,0.26428799629211425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,1,power_law_1.01,0.27998080253601076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,1,power_law_1.01,0.2937279939651489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,1,power_law_1.01,0.3158400058746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,1,power_law_1.01,0.3337023973464966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,1,power_law_1.01,0.3334399938583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,1,power_law_1.01,0.39317760467529295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,1,power_law_1.01,0.4343679904937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,1,power_law_1.01,0.5357183933258056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,1,power_law_1.01,0.6262335777282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,1,power_law_1.01,0.8231424331665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,1,power_law_1.01,1.0270912170410156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,1,power_law_1.01,1.3897151947021484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,1,power_law_1.01,1.756211280822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,1,power_law_1.01,2.486950492858887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,1,power_law_1.01,3.2058174133300783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,1,power_law_1.2,0.05333120226860046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,1,power_law_1.2,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,1,power_law_1.2,0.0585919976234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,1,power_law_1.2,0.06277120113372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,1,power_law_1.2,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,1,power_law_1.2,0.09658240079879761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,1,power_law_1.2,0.10073599815368653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,1,power_law_1.2,0.10212479829788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,1,power_law_1.2,0.10193920135498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,1,power_law_1.2,0.10581760406494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,1,power_law_1.2,0.10947840213775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,1,power_law_1.2,0.11195520162582398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,1,power_law_1.2,0.11363199949264527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,1,power_law_1.2,0.11886080503463745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,1,power_law_1.2,0.12871040105819703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,1,power_law_1.2,0.13375359773635864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,1,power_law_1.2,0.14723199605941772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,1,power_law_1.2,0.17789440155029296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,1,power_law_1.2,0.20052480697631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,1,power_law_1.2,0.2535552024841309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,1,power_law_1.2,0.3082815885543823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,1,power_law_1.2,0.4229311943054199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,1,power_law_1.2,0.5383232116699219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,1,power_law_1.2,0.7706624031066894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,4,balanced,0.021514666577180225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,4,balanced,0.021935999393463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,4,balanced,0.02370133250951767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,4,balanced,0.027717334528764088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,4,balanced,0.04211199780305227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,4,balanced,0.061850666999816895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,4,balanced,0.06507200002670288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,4,balanced,0.06566399832566579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,4,balanced,0.06795200208822887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,4,balanced,0.06834666430950165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,4,balanced,0.06961599985758464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,4,balanced,0.08241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,4,balanced,0.0804319977760315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,4,balanced,0.08166400094827016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,4,balanced,0.11411199967066447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,4,balanced,0.11101333300272624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,4,balanced,0.11321600278218587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,4,balanced,0.14122666915257773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,4,balanced,0.15762666861216226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,4,balanced,0.17398399114608765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,4,balanced,0.18785067399342856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,4,balanced,0.2018453280131022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,4,balanced,0.21887467304865518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,4,balanced,0.23466134071350098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,4,balanced,0.26153600215911865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,4,balanced,0.4337386687596639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,4,balanced,0.47592000166575116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,1,power_law_1.2,0.9967807769775391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,1,power_law_1.2,1.4404543876647948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,32,power_law_1.01,0.1601215958595276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,32,power_law_1.01,0.15557119846343995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,32,power_law_1.01,0.152729594707489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,32,power_law_1.01,0.15872000455856322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,32,power_law_1.01,0.15255039930343628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,32,power_law_1.01,0.14885120391845702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,32,power_law_1.01,0.1583616018295288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,32,power_law_1.01,0.16121599674224854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,32,power_law_1.01,0.16184959411621094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,64,power_law_1.2,0.03829759955406189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,64,power_law_1.2,0.03754239976406097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,64,power_law_1.2,0.03824000060558319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,64,power_law_1.2,0.03943679928779602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,64,power_law_1.2,0.03920640051364899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,64,power_law_1.2,0.0406143993139267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,64,power_law_1.2,0.039603200554847715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,64,power_law_1.2,0.04041599929332733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,64,power_law_1.2,0.04026240110397339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,64,power_law_1.2,0.04028800129890442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,64,power_law_1.2,0.041894400119781496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,64,power_law_1.2,0.04163840115070343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,128,power_law_1.2,0.051046401262283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,64,power_law_1.2,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,64,power_law_1.2,0.041094401478767396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,128,power_law_1.2,0.05187199711799621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,64,power_law_1.2,0.046054399013519286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,64,power_law_1.2,0.048416000604629514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,64,power_law_1.2,0.047295999526977536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,32,power_law_1.01,0.1623103976249695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,128,power_law_1.2,0.051609599590301515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,64,power_law_1.2,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,32,power_law_1.01,0.16780799627304077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,128,power_law_1.2,0.052767997980117796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,64,power_law_1.2,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,128,power_law_1.2,0.0536575973033905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,64,power_law_1.2,0.06680319905281067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,128,power_law_1.2,0.054118400812149046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,64,power_law_1.2,0.07782400250434876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,128,power_law_1.2,0.054636800289154054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,64,power_law_1.2,0.08969600200653076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,128,power_law_1.2,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,64,power_law_1.2,0.11781760454177856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,128,power_law_1.2,0.05692160129547119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,64,power_law_1.2,0.15595519542694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,128,power_law_1.2,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,128,power_law_1.2,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,32,power_law_1.01,0.17044479846954347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,128,power_law_1.2,0.056435197591781616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,32,power_law_1.01,0.16658560037612916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,128,power_law_1.2,0.058284801244735715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,32,power_law_1.01,0.17226239442825317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,128,power_law_1.2,0.05817599892616272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,32,power_law_1.01,0.17438080310821533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,128,power_law_1.2,0.06298239827156067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,32,power_law_1.01,0.17469439506530762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,32,power_law_1.01,0.1852031946182251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,64,power_law_1.2,0.20616960525512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,32,power_law_1.01,0.21337599754333497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,32,power_law_1.01,0.23079679012298585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,128,power_law_1.2,0.06255360245704651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,128,power_law_1.2,0.06679040193557739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,128,power_law_1.2,0.0733568012714386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,128,power_law_1.2,0.08024960160255432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,128,power_law_1.2,0.09368320107460022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,128,power_law_1.2,0.10687999725341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,32,power_law_1.01,0.2870975971221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,128,power_law_1.2,0.13793280124664306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,32,power_law_1.01,0.3322751998901367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,128,power_law_1.2,0.1720896005630493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,128,power_law_1.2,0.22294399738311768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,32,power_law_1.01,0.427180814743042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,64,power_law_1.2,0.3015872001647949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,128,power_law_1.2,0.2744704008102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,32,power_law_1.01,0.538585615158081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,128,power_law_1.2,0.4568192005157471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,32,power_law_1.01,0.7579008102416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,128,power_law_1.2,0.5300032138824463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,32,power_law_1.01,0.9926207542419434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,32,power_law_1.01,1.4290559768676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,32,power_law_1.01,1.812588882446289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,64,power_law_1.2,0.4090688228607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,1,power_law_1.2,1.8787904739379884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,1,power_law_1.2,0.018464000523090364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,1,power_law_1.2,0.024345600605010988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,1,power_law_1.2,0.028224000334739686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,1,power_law_1.2,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,1,power_law_1.2,0.041075199842453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,1,power_law_1.2,0.05622400045394897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,1,power_law_1.2,0.05979520082473755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,1,power_law_1.2,0.06307839751243591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,1,power_law_1.2,0.06502400040626526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,1,power_law_1.2,0.06438400149345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,1,power_law_1.2,0.06938880085945129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,1,power_law_1.2,0.0681984007358551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,1,power_law_1.2,0.06972799897193908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,1,power_law_1.2,0.07335039973258972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,1,power_law_1.2,0.07111679911613464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,1,power_law_1.2,0.07374719977378845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,1,power_law_1.2,0.07622399926185608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,1,power_law_1.2,0.08447999954223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,1,power_law_1.2,0.09437440037727356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,1,power_law_1.2,0.18491519689559938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,1,power_law_1.2,0.2147968053817749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,1,power_law_1.2,0.17416319847106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,1,power_law_1.2,0.1972607970237732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,16,balanced,0.05026666820049286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,16,balanced,0.05036266644795736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,16,balanced,0.05202133456865946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,16,balanced,0.05823466678460439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,16,balanced,0.07083733379840851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,16,balanced,0.07843199868996938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,16,balanced,0.07793599863847096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,16,balanced,0.07798400024573009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,16,balanced,0.08040533463160197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,16,balanced,0.07921066880226135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,16,balanced,0.08030400176843007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,16,balanced,0.08144533137480418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,16,balanced,0.08861866593360901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,16,balanced,0.09055466453234355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,16,balanced,0.09805333614349365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,1,power_law_1.2,0.24907519817352294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,16,balanced,0.09442666172981262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,16,balanced,0.10122666756312053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,16,balanced,0.10473066568374634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,16,balanced,0.11508267124493916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,16,balanced,0.12560000022252402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,16,balanced,0.1395199994246165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,16,balanced,0.17016534010569254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,16,balanced,0.19476266702016196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,16,balanced,0.26690133412679035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,16,balanced,0.32846933603286743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,16,balanced,0.4621973435084025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,16,balanced,0.5959839820861816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,16,power_law_1.2,0.05182719826698303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,16,power_law_1.2,0.03671039938926697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,16,power_law_1.2,0.03463039994239807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,1,power_law_1.2,0.30468480587005614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,16,power_law_1.2,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,16,power_law_1.2,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,16,power_law_1.2,0.03857919871807099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,16,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,16,power_law_1.2,0.03824000060558319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,16,power_law_1.2,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,16,power_law_1.2,0.03937279880046844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,16,power_law_1.2,0.04188160002231598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,16,power_law_1.2,0.04246399998664856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,16,power_law_1.2,0.044249600172042845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,16,power_law_1.2,0.0471231997013092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,16,power_law_1.2,0.05297279953956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,16,power_law_1.2,0.055270397663116456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,16,power_law_1.2,0.05735039710998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,16,power_law_1.2,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,16,power_law_1.2,0.06798080205917359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,16,power_law_1.2,0.08597760200500489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,16,power_law_1.2,0.09071360230445862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,16,power_law_1.2,0.11871999502182007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,16,power_law_1.2,0.1272063970565796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,16,power_law_1.2,0.16877440214157105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,16,power_law_1.2,0.2343616008758545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,16,power_law_1.2,0.30884480476379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,1,power_law_1.2,0.4155519962310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,4,power_law_1.01,0.02176000028848648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,4,power_law_1.01,0.025779199600219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,4,power_law_1.01,0.03678719997406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,4,power_law_1.01,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,4,power_law_1.01,0.05122560262680054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,4,power_law_1.01,0.056620800495147706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,4,power_law_1.01,0.06267520189285278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,4,power_law_1.01,0.06803839802742004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,4,power_law_1.01,0.07139840126037597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,4,power_law_1.01,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,4,power_law_1.01,0.07247359752655029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,4,power_law_1.01,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,4,power_law_1.01,0.0732479989528656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,4,power_law_1.01,0.07657600045204163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,4,power_law_1.01,0.08125439882278443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,4,power_law_1.01,0.08570880293846131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,4,power_law_1.01,0.08842880129814149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,4,power_law_1.01,0.09752320051193238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,4,power_law_1.01,0.10957440137863159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,4,power_law_1.01,0.13114240169525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,4,power_law_1.01,0.17161600589752196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,4,power_law_1.01,0.19541759490966798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,4,power_law_1.01,0.21950719356536866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,4,power_law_1.01,0.2759167909622192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,4,power_law_1.01,0.3433151960372925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,4,power_law_1.01,0.455295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,4,power_law_1.01,0.5929408073425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,16,power_law_1.2,0.41436161994934084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,1,power_law_1.2,0.5231359958648681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,1,power_law_1.2,0.05611519813537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,1,power_law_1.2,0.06401280164718628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,1,power_law_1.2,0.09707520008087159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,1,power_law_1.2,0.1277184009552002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,1,power_law_1.2,0.13198720216751098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,1,power_law_1.2,0.1344831943511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,1,power_law_1.2,0.13605120182037353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,1,power_law_1.2,0.13730560541152953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,1,power_law_1.2,0.14079999923706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,1,power_law_1.2,0.1430848002433777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,128,power_law_1.2,0.03620480000972748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,1,power_law_1.2,0.1421183943748474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,1,power_law_1.2,0.14840960502624512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,1,power_law_1.2,0.1535871982574463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,1,power_law_1.2,0.16046080589294434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,1,power_law_1.2,0.1745471954345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,128,power_law_1.2,0.034790399670600894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,128,power_law_1.2,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,1,power_law_1.2,0.1855936050415039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,1,power_law_1.2,0.1859328031539917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,128,power_law_1.2,0.036550399661064145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,128,power_law_1.2,0.03726719915866852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,128,power_law_1.2,0.0385343998670578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,128,power_law_1.2,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,2,balanced,0.030640001098314922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,2,balanced,0.03180799881617228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,2,balanced,0.03249066571394602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,2,balanced,0.035317334036032356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,1,power_law_1.2,0.2109312057495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,128,power_law_1.2,0.038815999031066896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,128,power_law_1.2,0.038438400626182555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,128,power_law_1.2,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,128,power_law_1.2,0.0383103996515274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,2,balanced,0.037317333122094475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,2,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,2,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,2,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,2,balanced,0.03632533301909765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,1,power_law_1.2,0.23710720539093016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,2,balanced,0.03740799923737844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,2,balanced,0.03631466627120972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,2,balanced,0.039408000806967415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,128,power_law_1.2,0.038873600959777835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,2,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,128,power_law_1.2,0.039103999733924866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,128,power_law_1.2,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,128,power_law_1.2,0.04379520118236542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,128,power_law_1.2,0.0456063985824585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,128,power_law_1.2,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,2,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,128,power_law_1.2,0.04837760031223297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,128,power_law_1.2,0.05443840026855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,128,power_law_1.2,0.06108160018920898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,128,power_law_1.2,0.0679040014743805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,128,power_law_1.2,0.082259202003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,1,power_law_1.2,0.29945600032806396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,128,power_law_1.2,0.10603519678115844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,2,balanced,0.04746133089065552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,2,balanced,0.047925333182017006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,2,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,128,power_law_1.2,0.14663679599761964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,128,power_law_1.2,0.18789119720458985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,1,power_law_1.2,0.3378432035446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,2,balanced,0.06317333380381267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,128,power_law_1.2,0.2752896070480347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,16,power_law_1.2,0.06725119948387145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,16,power_law_1.2,0.09367679953575134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,16,power_law_1.2,0.08491520285606384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,16,power_law_1.2,0.09321600198745728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,2,balanced,0.06851199766000111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,16,power_law_1.2,0.09581440091133117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,2,balanced,0.07763200004895528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,2,balanced,0.09497066338857015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,16,power_law_1.2,0.09073280096054077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,2,balanced,0.1281599998474121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,2,balanced,0.15005333224932352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,2,balanced,0.20836265881856283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,1,power_law_1.2,0.44040961265563966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,2,balanced,0.2803200085957845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,2,balanced,0.4139253298441569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,2,balanced,0.5328373511632284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,16,power_law_1.2,0.09284480214118958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,16,power_law_1.2,0.0964735984802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,16,power_law_1.2,0.08850560188293458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,16,power_law_1.2,0.09408640265464782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,16,power_law_1.2,0.09971839785575867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,1,power_law_1.2,0.5413311958312989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,16,power_law_1.2,0.09921919703483581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,16,power_law_1.2,0.1023743987083435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,16,power_law_1.2,0.10464639663696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,16,power_law_1.2,0.10951039791107178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,16,power_law_1.2,0.1155519962310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,16,power_law_1.2,0.11861759424209595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,16,power_law_1.2,0.13701119422912597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,16,power_law_1.2,0.13971199989318847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,16,power_law_1.2,0.16607359647750855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,16,power_law_1.2,0.18933759927749633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,16,power_law_1.2,0.2315135955810547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,16,power_law_1.2,0.29809279441833497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,16,power_law_1.2,0.3982399940490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,16,power_law_1.2,0.4779967784881592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,16,power_law_1.2,0.6691904067993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,16,power_law_1.2,0.9500224113464355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,1,power_law_1.2,0.7547455787658691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,128,power_law_1.2,0.35023999214172363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,1,power_law_1.2,0.9477055549621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,1,power_law_1.2,1.3565312385559083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,2,power_law_1.2,0.03456639945507049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,2,power_law_1.2,0.03790079951286316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,2,power_law_1.2,0.03818239867687225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,2,power_law_1.2,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,2,power_law_1.2,0.045824000239372255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,2,power_law_1.2,0.045414400100708005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,2,power_law_1.2,0.04737919867038727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,2,power_law_1.2,0.047200000286102294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,2,power_law_1.2,0.04642559885978699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,2,power_law_1.2,0.04726400077342987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,2,power_law_1.2,0.04776960015296936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,2,power_law_1.2,0.053913599252700804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,2,power_law_1.2,0.055878400802612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,2,power_law_1.2,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,1,power_law_1.2,1.762335968017578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,2,power_law_1.2,0.0703935980796814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,64,power_law_1.01,0.06093440055847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,2,power_law_1.2,0.07368320226669312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,64,power_law_1.01,0.041254401206970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,64,power_law_1.01,0.04106239974498749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,64,power_law_1.01,0.04379520118236542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,64,power_law_1.01,0.04506239891052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,64,power_law_1.01,0.04534400105476379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,64,power_law_1.01,0.04615679979324341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,2,power_law_1.2,0.08362879753112792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,64,power_law_1.01,0.04690560102462769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,64,power_law_1.01,0.04666880071163178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,64,power_law_1.01,0.047225600481033324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,64,power_law_1.01,0.04840959906578064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,64,power_law_1.01,0.049344000220298764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,64,power_law_1.01,0.050809597969055174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,64,power_law_1.01,0.05260159969329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,64,power_law_1.01,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,64,power_law_1.01,0.05756160020828247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,64,power_law_1.01,0.06197119951248169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,64,power_law_1.01,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,64,power_law_1.01,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,64,power_law_1.01,0.09193599820137024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,64,power_law_1.01,0.09936000108718872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,64,power_law_1.01,0.13174400329589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,64,power_law_1.01,0.15740159749984742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,64,power_law_1.01,0.21215999126434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,64,power_law_1.01,0.27370879650115965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,64,power_law_1.01,0.39531519412994387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,2,power_law_1.2,0.09098240137100219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,64,power_law_1.01,0.49237761497497556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,2,power_law_1.2,0.1057919979095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,128,power_law_1.01,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,128,power_law_1.01,0.034969601035118106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,128,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,128,power_law_1.01,0.037049600481987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,128,power_law_1.01,0.03710080087184906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,128,power_law_1.01,0.038412800431251524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,128,power_law_1.01,0.03857919871807099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,2,power_law_1.2,0.1288256049156189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,128,power_law_1.01,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,128,power_law_1.01,0.03976320028305054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,128,power_law_1.01,0.040300801396369934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,128,power_law_1.01,0.04078719913959503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,128,power_law_1.01,0.041606399416923526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,128,power_law_1.01,0.04184960126876831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,128,power_law_1.01,0.04172160029411316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,128,power_law_1.01,0.044563201069831845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,128,power_law_1.01,0.046060800552368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,128,power_law_1.01,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,128,power_law_1.01,0.051532799005508424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,128,power_law_1.01,0.05455999970436096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,128,power_law_1.01,0.06382079720497132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,128,power_law_1.01,0.07052800059318542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,128,power_law_1.01,0.10456960201263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,128,power_law_1.01,0.12995200157165526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,128,power_law_1.01,0.17696640491485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,128,power_law_1.01,0.2161344051361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,2,power_law_1.2,0.17166080474853515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,128,power_law_1.01,0.3090303897857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,128,power_law_1.01,0.4167424201965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,2,power_law_1.2,0.2595455884933472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,2,power_law_1.2,0.35676159858703616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,32,power_law_1.01,0.07543039917945862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,2,power_law_1.2,0.46611838340759276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,32,power_law_1.01,0.06666880249977111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,32,power_law_1.01,0.06569600105285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,32,power_law_1.01,0.07317759990692138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,32,power_law_1.01,0.07109119892120361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,32,power_law_1.01,0.07339519858360291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,32,power_law_1.01,0.0729088008403778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,32,power_law_1.01,0.07185919880867005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,32,power_law_1.01,0.0744383990764618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,32,power_law_1.01,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,32,power_law_1.01,0.07646080255508422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,32,power_law_1.01,0.07978879809379577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,32,power_law_1.01,0.08033279776573181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,32,power_law_1.01,0.08194559812545776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,32,power_law_1.01,0.09304959774017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,32,power_law_1.01,0.09544320106506347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,32,power_law_1.01,0.10058239698410035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,32,power_law_1.01,0.11570559740066529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,32,power_law_1.01,0.13084800243377687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,32,power_law_1.01,0.16486400365829468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,32,power_law_1.01,0.19511040449142455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,32,power_law_1.01,0.25693440437316895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,32,power_law_1.01,0.268723201751709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,32,power_law_1.01,0.34452478885650634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,32,power_law_1.01,0.4415296077728271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,32,power_law_1.01,0.6073472023010253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,32,power_law_1.01,0.6962687969207764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,2,power_law_1.2,0.5803775787353516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,4,balanced,0.12590400377909342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,4,balanced,0.1167733371257782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,4,balanced,0.21307732661565146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,4,balanced,0.21303999423980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,4,balanced,0.21223467588424683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,4,balanced,0.21416000525156656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,4,balanced,0.21481066942214966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,1,balanced,0.05428266525268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,1,balanced,0.06843733290831248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,1,balanced,0.09127466877301534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,1,balanced,0.10462400317192078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,1,balanced,0.09473599990208943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,1,balanced,0.09872532884279887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,1,balanced,0.10215466221173604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,1,balanced,0.09367466966311137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,1,balanced,0.09704533219337463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,1,balanced,0.09705600142478943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,1,balanced,0.09292266766230266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,1,balanced,0.09506666660308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,1,balanced,0.0922986666361491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,1,balanced,0.0940053363641103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,1,balanced,0.10069333513577779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,1,balanced,0.10052800178527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,1,balanced,0.10455466310183208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,1,balanced,0.14108799894650778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,1,balanced,0.14410666624704996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,1,balanced,0.1823093295097351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,4,balanced,0.21514666080474854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,1,balanced,0.21994133790334067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,4,balanced,0.21678400039672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,1,balanced,0.29870400826136273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,1,balanced,0.361135999361674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,1,balanced,0.5211893320083618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,1,balanced,0.667253335316976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,1,balanced,0.9781386852264404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,1,balanced,1.2982079982757568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,4,balanced,0.21780800819396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,4,balanced,0.24105600516001383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,4,balanced,0.22316267093022665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,4,balanced,0.22405866781870523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,4,balanced,0.2267626722653707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,4,balanced,0.261845330397288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,4,balanced,0.2648959954579671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,4,balanced,0.2523840069770813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,4,balanced,0.33296000957489014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,4,balanced,0.2988213300704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,4,balanced,0.4264479875564575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,4,balanced,0.42574934164683026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,4,balanced,0.5896106561024984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,4,balanced,0.7406240304311117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,4,balanced,1.0401866436004639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,4,balanced,1.4625280698140461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,4,balanced,2.094202677408854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,4,balanced,2.753877321879069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,2,power_law_1.2,0.8745087623596192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,2,power_law_1.2,1.2038335800170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,16,power_law_1.2,0.0705407977104187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,16,power_law_1.2,0.07831040024757385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,16,power_law_1.2,0.07004799842834472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,16,power_law_1.2,0.0750976026058197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,16,power_law_1.2,0.0732096016407013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,16,power_law_1.2,0.0719488024711609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,16,power_law_1.2,0.07496320009231568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,16,power_law_1.2,0.07405440211296081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,16,power_law_1.2,0.0728767991065979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,16,power_law_1.2,0.0772159993648529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,16,power_law_1.2,0.07642880082130432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,16,power_law_1.2,0.07933440208435058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,16,power_law_1.2,0.08056960105895997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,16,power_law_1.2,0.08191360235214233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,16,power_law_1.2,0.0874176025390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,16,power_law_1.2,0.08837119936943054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,16,power_law_1.2,0.09256960153579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,16,power_law_1.2,0.10666240453720092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.2,0.11913599967956542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.2,0.13781119585037233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.2,0.15567359924316407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.2,0.20942718982696534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.2,0.24657919406890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.2,0.32398080825805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.2,0.456115198135376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.2,0.6157248020172119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.2,0.848691177368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,4,power_law_1.01,0.09442560076713562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,4,power_law_1.01,0.1436352014541626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,4,power_law_1.01,0.18264960050582885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,4,power_law_1.01,0.2148224115371704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,4,power_law_1.01,0.2716223955154419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,4,power_law_1.01,0.2717695951461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,4,power_law_1.01,0.37338240146636964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,4,power_law_1.01,0.3731775999069214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,4,power_law_1.01,0.39655680656433107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,4,power_law_1.01,0.406009578704834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,4,power_law_1.01,0.3926719903945923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,4,power_law_1.01,0.39667201042175293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,4,power_law_1.01,0.39504640102386473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,4,power_law_1.01,0.4123136043548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,4,power_law_1.01,0.43672962188720704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,4,power_law_1.01,0.4531519889831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,4,power_law_1.01,0.4683135986328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,4,power_law_1.01,0.5152192115783691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,4,power_law_1.01,0.5322239875793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,4,power_law_1.01,0.6040383815765381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,4,power_law_1.01,0.6243072032928467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,4,power_law_1.01,0.7418047904968261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,4,power_law_1.01,0.8690815925598144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,4,power_law_1.01,1.188096046447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,4,power_law_1.01,1.3111231803894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,4,power_law_1.01,1.7411392211914063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,4,power_law_1.01,2.257702445983887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,4,power_law_1.2,0.07237759828567505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,4,power_law_1.2,0.081632000207901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,4,power_law_1.2,0.08195840120315552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,4,power_law_1.2,0.10613119602203369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,4,power_law_1.2,0.1049407958984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,4,power_law_1.2,0.13114880323410033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,4,power_law_1.2,0.14384000301361083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,4,power_law_1.2,0.14762879610061647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,4,power_law_1.2,0.14112000465393065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,4,power_law_1.2,0.13576960563659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,4,power_law_1.2,0.13646719455718995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,4,power_law_1.2,0.15279359817504884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,4,power_law_1.2,0.15192960500717162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,4,power_law_1.2,0.14911359548568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,4,power_law_1.2,0.1599679946899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,4,power_law_1.2,0.16723840236663817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,4,power_law_1.2,0.16888320446014404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,4,power_law_1.2,0.18977279663085939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,4,power_law_1.2,0.20848639011383058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,4,power_law_1.2,0.24185600280761718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,4,power_law_1.2,0.2724096059799194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,4,power_law_1.2,0.35422720909118655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,4,power_law_1.2,0.4006591796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,4,power_law_1.2,0.586521577835083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,4,power_law_1.2,0.745472002029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,4,power_law_1.2,1.0345024108886718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,4,power_law_1.2,1.3200127601623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,1,balanced,0.05282666782538096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,1,balanced,0.07796266674995422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,1,balanced,0.12026133139928182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,1,balanced,0.20598934094111124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,1,balanced,0.37489600976308185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,1,balanced,0.37534932295481366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,1,balanced,0.3768213192621867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,1,balanced,0.3785119851430257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,1,balanced,0.38464534282684326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,1,balanced,0.3855786720911662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,1,balanced,0.38704534371693927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,1,balanced,0.39230934778849286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,1,balanced,0.39420799414316815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,1,balanced,0.40060798327128094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,1,balanced,0.41012267271677655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,1,balanced,0.41570134957631427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,1,balanced,0.42796266078948975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,1,balanced,0.45613332589467365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,1,balanced,0.4919999837875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,1,balanced,0.5264799992243449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,1,balanced,0.5809653202692667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,1,balanced,0.7330293655395508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,1,balanced,0.7993066310882568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,1,balanced,1.118938684463501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,1,balanced,1.344901402791341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,1,balanced,1.914197285970052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,1,balanced,2.527946631113688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,8,power_law_1.2,0.04960640072822571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,8,power_law_1.2,0.05203199982643127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,8,power_law_1.2,0.052134400606155394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,8,power_law_1.2,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,8,power_law_1.2,0.060710400342941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,8,power_law_1.2,0.05859839916229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,8,power_law_1.2,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,8,power_law_1.2,0.061791998147964475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,8,power_law_1.2,0.06399359703063964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,8,power_law_1.2,0.06487680077552796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,8,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,8,power_law_1.2,0.06520959734916687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,8,power_law_1.2,0.06725760102272034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,8,power_law_1.2,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,8,power_law_1.2,0.07701119780540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,8,power_law_1.2,0.07739520072937012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,8,power_law_1.2,0.08390399813652039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,8,power_law_1.2,0.09717119932174682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,8,power_law_1.2,0.10505599975585937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,8,power_law_1.2,0.12969599962234496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,8,power_law_1.2,0.1473471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,8,power_law_1.2,0.19594240188598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,8,power_law_1.2,0.22955520153045655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,8,power_law_1.2,0.32279040813446047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,8,power_law_1.2,0.3891200065612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,8,power_law_1.2,0.5575232028961181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,8,power_law_1.2,0.7117440223693847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,32,power_law_1.2,0.04357120096683502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,32,power_law_1.2,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,32,power_law_1.2,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,32,power_law_1.2,0.03475840091705322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,32,power_law_1.2,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,32,power_law_1.2,0.03648639917373657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,32,power_law_1.2,0.036601600050926206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,32,power_law_1.2,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,32,power_law_1.2,0.03964160084724426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,32,power_law_1.2,0.0414463996887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,32,power_law_1.2,0.044947201013565065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,32,power_law_1.2,0.04808320105075836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,32,power_law_1.2,0.04924159944057464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,32,power_law_1.2,0.05160319805145264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,32,power_law_1.2,0.0579584002494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,32,power_law_1.2,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,32,power_law_1.2,0.06316159963607788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,32,power_law_1.2,0.07869439721107482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,32,power_law_1.2,0.08139520287513732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,32,power_law_1.2,0.1009600043296814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,32,power_law_1.2,0.11901439428329467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,32,power_law_1.2,0.14999040365219116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,32,power_law_1.2,0.18765439987182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,32,power_law_1.2,0.2816960096359253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,32,power_law_1.2,0.364902400970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,2,power_law_1.01,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,2,power_law_1.01,0.05859839916229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,2,power_law_1.01,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,2,power_law_1.01,0.07514879703521729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,2,power_law_1.01,0.07665280103683472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,2,power_law_1.01,0.07918720245361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,2,power_law_1.01,0.08115199804306031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,2,power_law_1.01,0.07999359965324401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,2,power_law_1.01,0.0838271975517273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,2,power_law_1.01,0.08551040291786194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,2,power_law_1.01,0.08844159841537476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,2,power_law_1.01,0.08989440202713013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,2,power_law_1.01,0.08840320110321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,2,power_law_1.01,0.09897599816322326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,2,power_law_1.01,0.10963200330734253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,2,power_law_1.01,0.11330560445785523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,2,power_law_1.01,0.12284799814224243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,2,power_law_1.01,0.13877760171890258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,2,power_law_1.01,0.16208640336990357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,2,power_law_1.01,0.20635519027709961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,2,power_law_1.01,0.22948479652404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,32,power_law_1.2,0.5561215877532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,2,power_law_1.01,0.34050559997558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,2,power_law_1.01,0.3987519979476929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,4,power_law_1.01,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,4,power_law_1.01,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,4,power_law_1.01,0.04012799859046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,4,power_law_1.01,0.04302720129489899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,4,power_law_1.01,0.042335999011993405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,4,power_law_1.01,0.0435263991355896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,4,power_law_1.01,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,4,power_law_1.01,0.044588801264762876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,4,power_law_1.01,0.04979200065135956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,4,power_law_1.01,0.05141760110855102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,4,power_law_1.01,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,4,power_law_1.01,0.05594239830970764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,4,power_law_1.01,0.058713597059249875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,4,power_law_1.01,0.0630079984664917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,4,power_law_1.01,0.07036160230636597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,4,power_law_1.01,0.07439360022544861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,4,power_law_1.01,0.07779840230941773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,4,power_law_1.01,0.09110400080680847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.01,0.10225919485092164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.01,0.13393280506134034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,32,power_law_1.2,0.6325183868408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.01,0.1497215986251831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,2,power_law_1.01,0.6352575778961181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,2,power_law_1.01,0.7209152221679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.01,0.2377216100692749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.01,0.2612031936645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.01,0.36206080913543703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,2,power_law_1.01,1.0352255821228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.01,0.5296832084655761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,128,power_law_1.01,0.09098240137100219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,128,power_law_1.01,0.08888959884643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,128,power_law_1.01,0.08973439931869506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,128,power_law_1.01,0.09326720237731934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,128,power_law_1.01,0.09408640265464782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,128,power_law_1.01,0.10168319940567017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,128,power_law_1.01,0.10455679893493652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,128,power_law_1.01,0.0987775981426239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,128,power_law_1.01,0.09774079918861389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,128,power_law_1.01,0.10019199848175049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,128,power_law_1.01,0.10117759704589843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,128,power_law_1.01,0.09717119932174682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,128,power_law_1.01,0.1008255958557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,128,power_law_1.01,0.09721599817276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,128,power_law_1.01,0.10105600357055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,128,power_law_1.01,0.1044927954673767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,128,power_law_1.01,0.10172799825668336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,128,power_law_1.01,0.11890560388565063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,128,power_law_1.01,0.12362879514694214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,128,power_law_1.01,0.1530943989753723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,128,power_law_1.01,0.1789888024330139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,128,power_law_1.01,0.23658881187438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,128,power_law_1.01,0.27406721115112304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,128,power_law_1.01,0.35212159156799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,128,power_law_1.01,0.46403841972351073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,128,power_law_1.01,0.6417407989501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,128,power_law_1.01,0.8537471771240235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.01,0.6217599868774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,2,power_law_1.01,1.4246463775634766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.01,0.9199295997619629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,2,power_law_1.2,0.10071040391921997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,2,power_law_1.2,0.1484928011894226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,32,balanced,0.0583840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,32,balanced,0.044026667873064675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,32,balanced,0.045738667249679565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,32,balanced,0.04994666576385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,32,balanced,0.07654933134714763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,32,balanced,0.08222400148709615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,32,balanced,0.08384000261624654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,32,balanced,0.0827946662902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,32,balanced,0.08317333459854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,32,balanced,0.08456533153851827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,32,balanced,0.08453866839408875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,32,balanced,0.08477333188056946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,32,balanced,0.08624533812204997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,32,balanced,0.08900266885757446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,32,balanced,0.09065066774686177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,32,balanced,0.09284800291061401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,32,balanced,0.09345066547393799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,32,balanced,0.09886399904886882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,32,balanced,0.10169600447018941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,32,balanced,0.11788266897201538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,32,balanced,0.11692266662915547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,32,balanced,0.14893866578737894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,32,balanced,0.1523413360118866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,32,balanced,0.21554666757583618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,32,balanced,0.22978132963180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,32,balanced,0.3151786724726359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,32,balanced,0.3785119851430257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,2,power_law_1.2,0.214630389213562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,2,power_law_1.2,0.3312319993972778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,2,power_law_1.2,0.452345609664917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,2,power_law_1.2,0.6080639839172364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,2,power_law_1.2,0.6959104061126709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,2,power_law_1.2,0.9184191703796387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,2,power_law_1.2,0.8990015983581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,2,power_law_1.2,0.9299776077270507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,2,power_law_1.2,0.899449634552002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,2,power_law_1.2,0.9722559928894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,2,power_law_1.2,0.9684160232543946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,2,power_law_1.2,1.09169921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,2,power_law_1.2,1.0575167655944824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,2,power_law_1.2,1.1172927856445312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,2,power_law_1.2,1.1364224433898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,2,power_law_1.2,1.2723072052001954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,2,power_law_1.2,1.3200960159301758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,2,power_law_1.2,1.5216256141662599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,2,power_law_1.2,1.5837568283081054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,2,power_law_1.2,1.9354560852050782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,2,power_law_1.2,1.8841728210449218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,2,power_law_1.2,2.299923133850098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,2,power_law_1.2,2.7215999603271483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,2,power_law_1.2,3.5972862243652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,2,power_law_1.2,4.379193496704102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,64,power_law_1.01,0.1217087984085083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,64,power_law_1.01,0.07210239768028259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,64,power_law_1.01,0.07055360078811646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,64,power_law_1.01,0.07457280158996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,64,power_law_1.01,0.07631999850273133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,64,power_law_1.01,0.07498239874839782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,64,power_law_1.01,0.07411199808120728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,64,power_law_1.01,0.07902079820632935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,64,power_law_1.01,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,64,power_law_1.01,0.0787392020225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,64,power_law_1.01,0.07953919768333435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,64,power_law_1.01,0.07836160063743591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,64,power_law_1.01,0.08056960105895997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,64,power_law_1.01,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,64,power_law_1.01,0.09015039801597595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,64,power_law_1.01,0.0904640018939972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,64,power_law_1.01,0.09539200067520141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,64,power_law_1.01,0.10499199628829955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,64,power_law_1.01,0.11537920236587525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,64,power_law_1.01,0.13587839603424073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,64,power_law_1.01,0.15709439516067505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,64,power_law_1.01,0.20205440521240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,64,power_law_1.01,0.23508479595184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,64,power_law_1.01,0.3195519924163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,64,power_law_1.01,0.3850496053695679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,64,power_law_1.01,0.5697535991668701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,64,power_law_1.01,0.7041855812072754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,32,balanced,0.0557226687669754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,32,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,32,balanced,0.05601066847642263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,32,balanced,0.06417599817117055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,32,balanced,0.06839466591676076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,32,balanced,0.10068800052007039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,32,balanced,0.09653866291046143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,32,balanced,0.09744000434875488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,32,balanced,0.09442133704821269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,32,balanced,0.09399466713269551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,32,balanced,0.09202667077382405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,32,balanced,0.0929813285668691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,32,balanced,0.09788800279299419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,32,balanced,0.09309867024421692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,32,balanced,0.09917866190274556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,32,balanced,0.10089600086212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,32,balanced,0.10651733477910359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,32,balanced,0.11530133088429768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,32,balanced,0.11595732967058818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,32,balanced,0.1325866679350535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,32,balanced,0.1409119963645935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,32,balanced,0.16849066813786825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,32,balanced,0.1937546730041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,32,balanced,0.26519999901453656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,32,balanced,0.3152479926745097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,32,balanced,0.43584001064300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,32,balanced,0.5570506652196249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,8,balanced,0.017071999609470367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,8,balanced,0.017093333105246227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,8,balanced,0.017194667210181553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,8,balanced,0.017194667210181553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,8,balanced,0.019018666197856266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,8,balanced,0.0210506667693456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,8,balanced,0.021231998999913532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,8,balanced,0.02130666623512904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,8,balanced,0.022997332115968067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,8,balanced,0.023370665808518726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,8,balanced,0.023024000227451324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,8,balanced,0.023168000082174938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,8,balanced,0.025114665428797405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,8,balanced,0.027461332579453785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,8,balanced,0.029472000896930695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,8,balanced,0.029296000798543293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,8,balanced,0.03989866624275843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,8,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,8,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,8,balanced,0.04874666531880697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,8,balanced,0.056517332792282104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,8,balanced,0.06622933348019917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,8,balanced,0.08032000064849854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,8,balanced,0.10732266306877136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,8,balanced,0.13198399543762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,8,balanced,0.1813653310139974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,8,balanced,0.23100266853968301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,8,power_law_1.01,0.07412480115890503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,8,power_law_1.01,0.10853760242462158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,8,power_law_1.01,0.13461120128631593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,8,power_law_1.01,0.1439296007156372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,8,power_law_1.01,0.1515455961227417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,8,power_law_1.01,0.15233919620513917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,8,power_law_1.01,0.15245440006256103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,8,power_law_1.01,0.16227200031280517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,8,power_law_1.01,0.15008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,8,power_law_1.01,0.16599680185317994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,8,power_law_1.01,0.16496000289916993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,8,power_law_1.01,0.16298240423202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,8,power_law_1.01,0.16580480337142944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,8,power_law_1.01,0.16737279891967774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,8,power_law_1.01,0.17658239603042603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,8,power_law_1.01,0.17849600315093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,8,power_law_1.01,0.19040000438690186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,8,power_law_1.01,0.21342079639434813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,8,power_law_1.01,0.20899200439453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,8,power_law_1.01,0.24268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,8,power_law_1.01,0.26382720470428467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,8,power_law_1.01,0.32780799865722654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,8,power_law_1.01,0.34452478885650634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,8,power_law_1.01,0.42430720329284666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,8,power_law_1.01,0.5195775985717773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,8,power_law_1.01,0.7169472217559815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,8,power_law_1.01,0.8616703987121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,2,power_law_1.01,0.021721599996089934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,2,power_law_1.01,0.022752000391483305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,2,power_law_1.01,0.02869119942188263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,2,power_law_1.01,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,2,power_law_1.01,0.05503360033035278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,2,power_law_1.01,0.061791998147964475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,2,power_law_1.01,0.07326719760894776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,2,power_law_1.01,0.0774399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,2,power_law_1.01,0.08033279776573181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,2,power_law_1.01,0.08122239708900451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,2,power_law_1.01,0.0906175971031189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,2,power_law_1.01,0.08802559971809387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,2,power_law_1.01,0.09099519848823548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,2,power_law_1.01,0.09664000272750854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,2,power_law_1.01,0.09541760087013244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,2,power_law_1.01,0.09865599870681763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,2,power_law_1.01,0.10544639825820923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,2,power_law_1.01,0.13207679986953735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.01,0.13407360315322875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.01,0.23714559078216552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.01,0.2621119976043701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.01,0.25968639850616454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.01,0.2881088018417358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.01,0.34746880531311036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.01,0.4254464149475098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.01,0.5659455776214599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.01,0.7392255783081054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,4,power_law_1.2,0.035206401348114015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,4,power_law_1.2,0.049414399266242984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,4,power_law_1.2,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,4,power_law_1.2,0.05429760217666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,4,power_law_1.2,0.05467519760131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,4,power_law_1.2,0.05854079723358154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,4,power_law_1.2,0.06056960225105286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,4,power_law_1.2,0.05964159965515137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,4,power_law_1.2,0.06282879710197449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,4,power_law_1.2,0.06284160017967225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,4,power_law_1.2,0.06378880143165588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,4,power_law_1.2,0.06783360242843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,4,power_law_1.2,0.06844159960746765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,4,power_law_1.2,0.07390720248222352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,4,power_law_1.2,0.08515840172767639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,4,power_law_1.2,0.08929280042648316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,4,power_law_1.2,0.09566079974174499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,4,power_law_1.2,0.11864320039749146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,4,power_law_1.2,0.1356735944747925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,4,power_law_1.2,0.15830399990081787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,4,power_law_1.2,0.19900799989700318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,4,power_law_1.2,0.26640000343322756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,4,power_law_1.2,0.327891206741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,4,power_law_1.2,0.5190207958221436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,4,power_law_1.2,0.6097023963928223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,4,power_law_1.2,0.08671360015869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,4,power_law_1.2,0.10043519735336304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,4,power_law_1.2,0.11134719848632812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,4,power_law_1.2,0.13203200101852416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,4,power_law_1.2,0.16147840023040771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,4,power_law_1.2,0.18447999954223632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,4,power_law_1.2,0.18911999464035034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,4,power_law_1.2,0.18523520231246948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,4,power_law_1.2,0.19749120473861695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,4,power_law_1.2,0.20199038982391357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,4,power_law_1.2,0.21110401153564454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,4,power_law_1.2,0.22035200595855714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,4,power_law_1.2,0.2268928050994873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,4,power_law_1.2,0.2307584047317505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,4,power_law_1.2,0.24885120391845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,4,power_law_1.2,0.25961599349975584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,4,power_law_1.2,0.2720128059387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,4,power_law_1.2,0.32497920989990237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,4,power_law_1.2,0.379532790184021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,4,power_law_1.2,0.47875199317932127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,4,power_law_1.2,0.5204800128936767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,4,power_law_1.2,0.6713280200958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,4,power_law_1.2,0.8443584442138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,4,power_law_1.2,1.1419455528259277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,4,power_law_1.2,1.456339168548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,4,power_law_1.2,2.0331327438354494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,4,power_law_1.2,0.8195136070251465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,4,power_law_1.2,2.4920448303222655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,4,balanced,0.04061333338419596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,4,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,4,balanced,0.031178665657838184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,4,balanced,0.03370666752258936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,4,balanced,0.033258666594823204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,4,balanced,0.035258665680885315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,4,balanced,0.03502399971087774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,4,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,4,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,4,balanced,0.03576533248027166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,4,balanced,0.03532266616821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,4,balanced,0.03738133360942205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,4,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,4,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,4,balanced,0.044154668847719826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,4,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,4,balanced,0.04358933369318644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,4,balanced,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,4,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,4,balanced,0.06197333335876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,4,balanced,0.06836266815662384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,4,balanced,0.09065066774686177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,4,balanced,0.10282666484514873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,4,balanced,0.1361066699028015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,4,balanced,0.15372799833615622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,4,balanced,0.2205866575241089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,4,balanced,0.2590720057487488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,4,power_law_1.2,1.0296256065368652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,16,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,16,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,16,balanced,0.05445333321889242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,16,balanced,0.05620799958705902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,16,balanced,0.057909334699312844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,16,balanced,0.07188799977302551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,16,balanced,0.08889066179593404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,16,balanced,0.09115200241406758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,16,balanced,0.09257599711418152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,16,balanced,0.09390399853388469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,16,balanced,0.09127466877301534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,16,balanced,0.09479999542236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,16,balanced,0.09593600034713745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,16,balanced,0.09687999884287517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,16,balanced,0.10307733217875163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,16,balanced,0.10313066840171814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,16,balanced,0.1111946702003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,16,balanced,0.11940266688664754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,16,balanced,0.12729600071907043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,16,balanced,0.1470080018043518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,16,balanced,0.16338666280110678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,16,balanced,0.2005173365275065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,16,balanced,0.23494933048884073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,16,balanced,0.3035999933878581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,16,balanced,0.3924906651178996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,16,balanced,0.5262773434321085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,16,balanced,0.6806026299794515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,4,balanced,0.03373866776625315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,4,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,4,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,4,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,4,balanced,0.03745600084463755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,4,balanced,0.03566399961709976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,4,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,4,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,4,balanced,0.036501333117485046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,4,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,4,balanced,0.03640533238649368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,4,balanced,0.04102399945259094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,4,balanced,0.040133332212766014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,4,balanced,0.040048000713189445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,4,balanced,0.04949333270390829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,4,balanced,0.04971200227737427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,4,balanced,0.05421333511670431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,4,balanced,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,4,balanced,0.06587199866771698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,4,balanced,0.07956799864768982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,4,balanced,0.08820266524950664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,4,balanced,0.12356799840927124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,4,balanced,0.14685866236686707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,4,balanced,0.1991999944051107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,4,balanced,0.25490132967631024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,4,balanced,0.3864373366038005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,4,balanced,0.5112213293711344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,64,balanced,0.04890666902065277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,64,balanced,0.050154666105906166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,64,balanced,0.05389333268006643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,64,balanced,0.07271466652552287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,64,balanced,0.08381866415341695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,64,balanced,0.0827466646830241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,64,balanced,0.08239466448624928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,64,balanced,0.08307733138402303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,64,balanced,0.08311999837557475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,64,balanced,0.08462400237719218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,64,balanced,0.08551999926567078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,64,balanced,0.08764800429344177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,64,balanced,0.08898666501045227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,64,balanced,0.08919999996821086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,64,balanced,0.09537600477536519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,64,balanced,0.0979360044002533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,64,balanced,0.10087999701499939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,64,balanced,0.11542399724324544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,64,balanced,0.12199999888737996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,64,balanced,0.15011200308799744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,64,balanced,0.1646773318449656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,64,balanced,0.21314666668574014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,64,balanced,0.21025067567825317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,64,balanced,0.27773332595825195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,64,balanced,0.31522132953008014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,64,balanced,0.4220800002415975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,64,balanced,0.5165546735127767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,64,power_law_1.01,0.040243199467658995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,64,power_law_1.01,0.0395583987236023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,64,power_law_1.01,0.03904640078544617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,64,power_law_1.01,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,64,power_law_1.01,0.042822399735450746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,64,power_law_1.01,0.045158401131629944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,64,power_law_1.01,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,64,power_law_1.01,0.04595839977264404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,64,power_law_1.01,0.04725759923458099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,64,power_law_1.01,0.0470335990190506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,64,power_law_1.01,0.04780159890651703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,64,power_law_1.01,0.048563200235366824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,64,power_law_1.01,0.048793599009513855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,64,power_law_1.01,0.05189759731292724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,4,balanced,0.0481279989083608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,4,balanced,0.043978666265805565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,4,balanced,0.06333333253860474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,4,balanced,0.06693333387374878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,4,balanced,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,4,balanced,0.06824533144632976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,4,balanced,0.06816533207893372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,4,balanced,0.07076266904671986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,4,balanced,0.07110400001207988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,4,balanced,0.07027733325958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,4,balanced,0.06967466572920482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,64,power_law_1.01,0.057030397653579715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,4,balanced,0.06968533496061961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,4,balanced,0.07187200089295705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,4,balanced,0.07249066730340321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,64,power_law_1.01,0.05719040036201477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,4,balanced,0.08458133538564046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,4,balanced,0.08781333764394124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,64,power_law_1.01,0.06169599890708923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,64,power_law_1.01,0.06750720143318176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,64,power_law_1.01,0.0741823971271515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,64,power_law_1.01,0.08882560133934021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,64,power_law_1.01,0.10135040283203126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,64,power_law_1.01,0.1344383955001831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,4,balanced,0.08789867162704468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,4,balanced,0.10470400253931682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,4,balanced,0.10539199908574422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,4,balanced,0.12707199652989706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,4,balanced,0.1462453305721283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,4,balanced,0.18491200606028238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,4,balanced,0.22414400180180868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,4,balanced,0.2994239926338196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,4,balanced,0.3887626727422078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,4,balanced,0.5296746492385864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,4,balanced,0.6958026885986328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,64,power_law_1.01,0.16227840185165404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,64,power_law_1.01,0.22033920288085937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,64,power_law_1.01,0.2797823905944824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,64,power_law_1.01,0.4039487838745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,64,power_law_1.01,0.51211519241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,4,balanced,0.03436800092458725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,4,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,4,balanced,0.05337599913279215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,4,balanced,0.07946133116881053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,4,balanced,0.12357866764068604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,4,balanced,0.1260426640510559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,4,balanced,0.12363200386365254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,4,balanced,0.12449066837628682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,4,balanced,0.1264533301194509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,4,balanced,0.12557333707809448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,4,balanced,0.1267093320687612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,4,balanced,0.12879467010498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,4,balanced,0.1322879989941915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,4,balanced,0.13404800494511923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,4,balanced,0.13822933038075766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,4,balanced,0.14085333546002707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,4,balanced,0.14528000354766846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,4,balanced,0.15639467040697733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,4,balanced,0.1662986675898234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,4,balanced,0.19430400927861533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,4,balanced,0.20886399348576865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,4,balanced,0.2649386723836263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,4,balanced,0.30401599407196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,4,balanced,0.4236160119374593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,4,balanced,0.49982933203379315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,4,balanced,0.7043893337249756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,4,balanced,0.8997120062510172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,4,power_law_1.2,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,4,power_law_1.2,0.037599998712539676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,4,power_law_1.2,0.03923200070858002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,4,power_law_1.2,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,4,power_law_1.2,0.04291839897632599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,4,power_law_1.2,0.043628799915313723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,4,power_law_1.2,0.04368639886379242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,4,power_law_1.2,0.04479359984397888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,4,power_law_1.2,0.04529280066490173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,4,power_law_1.2,0.04679679870605469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,4,power_law_1.2,0.046086400747299194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,4,power_law_1.2,0.05799040198326111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,4,power_law_1.2,0.05930240154266357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,4,power_law_1.2,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,4,power_law_1.2,0.07941120266914367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,4,power_law_1.2,0.08186879754066467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,4,power_law_1.2,0.09704959988594056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,4,power_law_1.2,0.07803519964218139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,4,power_law_1.2,0.08787199854850769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,4,power_law_1.2,0.10335359573364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,4,power_law_1.2,0.11438080072402954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,4,power_law_1.2,0.11635839939117432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,4,power_law_1.2,0.12711039781570435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,4,power_law_1.2,0.12104320526123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,4,power_law_1.2,0.1406399965286255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,4,power_law_1.2,0.1430848002433777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,4,power_law_1.2,0.15872000455856322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,4,power_law_1.2,0.1549183964729309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,4,power_law_1.2,0.161190402507782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,4,power_law_1.2,0.15124479532241822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,4,power_law_1.2,0.162009596824646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,4,power_law_1.2,0.1606528043746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,4,power_law_1.2,0.17776000499725342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,4,power_law_1.2,0.17414400577545167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,4,power_law_1.2,0.1772096037864685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,4,power_law_1.2,0.18556159734725952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,4,power_law_1.2,0.1905727982521057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,4,power_law_1.2,0.205401611328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,4,power_law_1.2,0.23167359828948975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,4,power_law_1.2,0.28767359256744385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,4,power_law_1.2,0.3316351890563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,4,power_law_1.2,0.21340160369873046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,4,power_law_1.2,0.43488640785217286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,4,power_law_1.2,0.5742144107818603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,4,power_law_1.2,0.7383679866790771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,4,power_law_1.2,0.9833600044250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,4,power_law_1.2,0.28342399597167967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,4,power_law_1.2,1.4499263763427734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,4,power_law_1.2,1.8490175247192382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,4,power_law_1.2,0.3870527982711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,4,power_law_1.2,0.6277760028839111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,1,balanced,0.06006933252016703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,1,balanced,0.06285866598288219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,1,balanced,0.06822933256626129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,1,balanced,0.08594133456548055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,1,balanced,0.12985600034395853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,1,balanced,0.18764267365137735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,1,balanced,0.2510506709416707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,1,balanced,0.25355732440948486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,1,balanced,0.257205327351888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,1,balanced,0.25692800680796307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,1,balanced,0.26209600766499835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,1,balanced,0.2688213388125102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,1,balanced,0.27355732520421344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,1,balanced,0.28067199389139813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,1,balanced,0.28904000918070477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,1,balanced,0.2982133428255717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,1,balanced,0.3158346613248189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,1,balanced,0.369706670443217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,1,balanced,0.4066559871037801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,4,power_law_1.2,0.8556544303894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,1,balanced,0.4894186655680339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,1,balanced,0.5535360177357992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,1,balanced,0.7091466585795084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,1,balanced,0.9147573312123617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,1,balanced,1.2319786548614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,1,balanced,1.707632064819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,1,balanced,2.353391965230306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,1,balanced,3.147424062093099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,4,power_law_1.2,1.3312959671020508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,1,power_law_1.2,0.09104639887809754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,1,power_law_1.2,0.10855679512023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,1,power_law_1.2,0.12738560438156127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,1,power_law_1.2,0.15463680028915405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,1,power_law_1.2,0.17882239818572998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,1,power_law_1.2,0.26031999588012694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,1,power_law_1.2,0.27242240905761717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,1,power_law_1.2,0.2725248098373413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,4,power_law_1.2,1.5870335578918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,1,power_law_1.01,0.05134080052375793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,1,power_law_1.01,0.06555520296096802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,1,power_law_1.01,0.07910400032997131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,1,power_law_1.01,0.11637760400772094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,1,power_law_1.01,0.14704639911651612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,1,power_law_1.01,0.18967039585113527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,1,power_law_1.01,0.2460927963256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,1,power_law_1.01,0.2515455961227417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,1,power_law_1.01,0.2650559902191162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,1,power_law_1.01,0.2732608079910278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,1,power_law_1.01,0.2797312021255493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,1,power_law_1.01,0.287174391746521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,1,power_law_1.01,0.29857280254364016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,1,power_law_1.01,0.3070847988128662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,1,power_law_1.01,0.31653759479522703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,1,power_law_1.01,0.32987520694732664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,1,power_law_1.01,0.34533119201660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,1,power_law_1.01,0.3917952060699463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,1,power_law_1.01,0.41495041847229003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,1,power_law_1.01,0.4881472110748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,1,power_law_1.01,0.5476160049438477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,1,power_law_1.01,0.6890624046325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,1,power_law_1.01,0.8334464073181153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,1,power_law_1.01,1.1156543731689452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,1,power_law_1.01,1.4151935577392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,1,power_law_1.01,1.9687807083129882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,1,power_law_1.01,2.512620735168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,128,power_law_1.01,0.06968960165977478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,128,power_law_1.01,0.06895999908447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,128,power_law_1.01,0.06829439997673034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,128,power_law_1.01,0.073471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,128,power_law_1.01,0.06924160122871399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,128,power_law_1.01,0.0754688024520874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,128,power_law_1.01,0.07354239821434021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,128,power_law_1.01,0.08377599716186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,128,power_law_1.01,0.08380799889564514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,128,power_law_1.01,0.08520320057868958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,128,power_law_1.01,0.07949439883232116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,128,power_law_1.01,0.0886784017086029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,128,power_law_1.01,0.09028480052947999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,128,power_law_1.01,0.08889600038528442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,128,power_law_1.01,0.1017151951789856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,128,power_law_1.01,0.10047999620437623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,128,power_law_1.01,0.10558079481124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,128,power_law_1.01,0.11632640361785888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,128,power_law_1.01,0.12896640300750734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,128,power_law_1.01,0.15501439571380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,128,power_law_1.01,0.17191679477691652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,128,power_law_1.01,0.22491519451141356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,128,power_law_1.01,0.2799743890762329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,128,power_law_1.01,0.375654411315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,128,power_law_1.01,0.49735679626464846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,128,power_law_1.01,0.7069056034088135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,128,power_law_1.01,0.9485952377319335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,1,power_law_1.2,0.09976959824562073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,1,power_law_1.2,0.1494655966758728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,1,power_law_1.2,0.2425920009613037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,1,power_law_1.2,0.4120063781738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,1,power_law_1.2,0.5754047870635987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,1,power_law_1.2,0.8073408126831054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,1,power_law_1.2,1.1624320030212403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,1,power_law_1.2,1.191481590270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,1,power_law_1.2,1.2686847686767577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,1,power_law_1.2,1.3173952102661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,1,power_law_1.2,1.3594112396240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,1,power_law_1.2,1.3813311576843261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,1,power_law_1.2,1.4257792472839355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,1,power_law_1.2,1.455123233795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,1,power_law_1.2,1.5194047927856444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,1,power_law_1.2,1.536844825744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,1,power_law_1.2,1.6684864044189454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,1,power_law_1.2,1.8043519973754882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,1,power_law_1.2,1.8521600723266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,1,power_law_1.2,2.086419105529785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,1,power_law_1.2,2.1062463760375976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,1,power_law_1.2,2.467865562438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,1,power_law_1.2,2.7142335891723635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,1,power_law_1.2,3.271372985839844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,1,power_law_1.2,3.7490238189697265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,1,power_law_1.2,4.9212799072265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,1,power_law_1.2,6.094291305541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,2,power_law_1.2,0.0709824025630951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,2,power_law_1.2,0.07524480223655701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,2,power_law_1.2,0.08113279938697815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,2,power_law_1.2,0.0921280026435852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,2,power_law_1.2,0.10706559419631959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,2,power_law_1.2,0.11688319444656373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,2,power_law_1.2,0.14042880535125732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,2,power_law_1.2,0.1457856059074402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,2,power_law_1.2,0.14543360471725464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,2,power_law_1.2,0.15011199712753295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,2,power_law_1.2,0.1506943941116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,2,power_law_1.2,0.15529600381851197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,2,power_law_1.2,0.1565824031829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,2,power_law_1.2,0.16073600053787232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,2,power_law_1.2,0.1681920051574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,2,power_law_1.2,0.171833598613739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,2,power_law_1.2,0.18083200454711915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,2,power_law_1.2,0.20595839023590087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,2,power_law_1.2,0.220198392868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,2,power_law_1.2,0.268121600151062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,2,power_law_1.2,0.3049920082092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,2,power_law_1.2,0.3874239921569824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,2,power_law_1.2,0.4836863994598389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,2,power_law_1.2,0.6408512115478515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,2,power_law_1.2,0.8498815536499024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,2,power_law_1.2,1.2233471870422363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,2,power_law_1.2,1.5052160263061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,16,power_law_1.01,0.052691197395324706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,16,power_law_1.01,0.062067198753356936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,16,power_law_1.01,0.06120319962501526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,16,power_law_1.01,0.060134398937225345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,16,power_law_1.01,0.05829120278358459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,16,power_law_1.01,0.0572160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,16,power_law_1.01,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,16,power_law_1.01,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,16,power_law_1.01,0.061817598342895505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,16,power_law_1.01,0.0611519992351532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,16,power_law_1.01,0.06099200248718262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,16,power_law_1.01,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,16,power_law_1.01,0.0654911994934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,16,power_law_1.01,0.06577919721603394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,16,power_law_1.01,0.07121279835700989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,16,power_law_1.01,0.07431679964065552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,16,power_law_1.01,0.07969279885292054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,16,power_law_1.01,0.08963199853897094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,16,power_law_1.01,0.10134400129318237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,16,power_law_1.01,0.12288000583648681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,16,power_law_1.01,0.13778560161590575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,16,power_law_1.01,0.18063360452651978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,16,power_law_1.01,0.22136321067810058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,16,power_law_1.01,0.2881216049194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,16,power_law_1.01,0.3597887992858887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,16,power_law_1.01,0.497983980178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,16,power_law_1.01,0.6492671966552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,16,power_law_1.01,0.060083198547363284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,16,power_law_1.01,0.0838591992855072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,16,power_law_1.01,0.06876159906387329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,16,power_law_1.01,0.06885120272636414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,16,power_law_1.01,0.06873599886894226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,16,power_law_1.01,0.07211520075798035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,16,power_law_1.01,0.07125759720802308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,16,power_law_1.01,0.07054719924926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,16,power_law_1.01,0.07215999960899352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,16,power_law_1.01,0.07218559980392455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,16,power_law_1.01,0.07308160066604615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,16,power_law_1.01,0.0767359972000122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,16,power_law_1.01,0.07533439993858337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,16,power_law_1.01,0.07976959943771363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,16,power_law_1.01,0.08684160113334656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,16,power_law_1.01,0.08931840062141419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,16,power_law_1.01,0.09548159837722778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,16,power_law_1.01,0.10913920402526855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,16,power_law_1.01,0.12200959920883178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,16,power_law_1.01,0.15072640180587768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,16,power_law_1.01,0.17015039920806885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,16,power_law_1.01,0.22504959106445313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,16,power_law_1.01,0.2603071928024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,16,power_law_1.01,0.33841280937194823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,16,power_law_1.01,0.4391168117523193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,16,power_law_1.01,0.6139391899108887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,16,power_law_1.01,0.8077823638916015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,2,power_law_1.2,0.05363199710845947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,2,power_law_1.2,0.06620799899101257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,2,power_law_1.2,0.08277119994163513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,2,power_law_1.2,0.10503040552139283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,2,power_law_1.2,0.12199679613113404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,2,power_law_1.2,0.15854719877243043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,2,power_law_1.2,0.17492480278015138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,2,power_law_1.2,0.17253119945526124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,2,power_law_1.2,0.18399360179901122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,2,power_law_1.2,0.1864575982093811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,2,power_law_1.2,0.1887935996055603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,2,power_law_1.2,0.19341440200805665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,2,power_law_1.2,0.19757440090179443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,2,power_law_1.2,0.20848639011383058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,2,power_law_1.2,0.2153856039047241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,2,power_law_1.2,0.22423040866851807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,2,power_law_1.2,0.2466048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,2,power_law_1.2,0.29137918949127195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,2,power_law_1.2,0.3101311922073364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,2,power_law_1.2,0.38268160820007324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,2,power_law_1.2,0.4456575870513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,2,power_law_1.2,0.5513792037963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,2,power_law_1.2,0.6836671829223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,2,power_law_1.2,0.9204352378845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,2,power_law_1.2,1.164575958251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,2,power_law_1.2,1.6019647598266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,2,power_law_1.2,2.0528640747070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,2,power_law_1.01,0.07376639842987061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,2,power_law_1.01,0.09467520117759705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,2,power_law_1.01,0.11279360055923462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,2,power_law_1.01,0.15190399885177613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,2,power_law_1.01,0.1723456025123596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,2,power_law_1.01,0.2726016044616699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,2,power_law_1.01,0.2808511972427368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,2,power_law_1.01,0.28161919116973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,2,power_law_1.01,0.28739840984344484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,2,power_law_1.01,0.2998719930648804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,2,power_law_1.01,0.3025536060333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,2,power_law_1.01,0.3076416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,2,power_law_1.01,0.32042880058288575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,2,power_law_1.01,0.33102080821990965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,2,power_law_1.01,0.341215991973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,2,power_law_1.01,0.35231359004974366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,2,power_law_1.01,0.37450881004333497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,2,power_law_1.01,0.4312575817108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,2,power_law_1.01,0.436736011505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,2,power_law_1.01,0.5408959865570069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,2,power_law_1.01,0.5627327919006347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,2,power_law_1.01,0.7019392013549804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,2,power_law_1.01,0.8262271881103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,2,power_law_1.01,1.095308780670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,2,power_law_1.01,1.6058752059936523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,2,power_law_1.01,2.0864704132080076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,2,power_law_1.01,2.6798847198486326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,32,power_law_1.2,0.1198464035987854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,32,power_law_1.2,0.11421439647674561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,32,power_law_1.2,0.1143231987953186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,32,power_law_1.2,0.11640959978103638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,32,power_law_1.2,0.1128383994102478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,32,power_law_1.2,0.11252479553222657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,32,power_law_1.2,0.10712319612503052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,32,power_law_1.2,0.1090175986289978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,32,power_law_1.2,0.11289600133895875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,32,power_law_1.2,0.1142848014831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,32,power_law_1.2,0.11369600296020507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,32,power_law_1.2,0.11561599969863892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,32,power_law_1.2,0.11079039573669433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,32,power_law_1.2,0.11722240447998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,32,power_law_1.2,0.12140799760818481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,32,power_law_1.2,0.12331520318984986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,32,power_law_1.2,0.12917759418487548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,32,power_law_1.2,0.15009280443191528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,32,power_law_1.2,0.16072959899902345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,32,power_law_1.2,0.18499840497970582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,32,power_law_1.2,0.21183359622955322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,32,power_law_1.2,0.27869439125061035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,32,power_law_1.2,0.34686079025268557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,32,power_law_1.2,0.4125823974609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,32,power_law_1.2,0.5257855892181397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,32,power_law_1.2,0.7537600040435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,32,power_law_1.2,1.098310375213623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,4,balanced,0.03162666658560435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,4,balanced,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,4,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,4,balanced,0.057189335425694786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,4,balanced,0.08044266700744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,4,balanced,0.08231466511885326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,4,balanced,0.08262399832407634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,4,balanced,0.08064533273379008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,4,balanced,0.08203200002511342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,4,balanced,0.08272533118724823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,4,balanced,0.08270933230717976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,4,balanced,0.084906667470932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,4,balanced,0.08462400237719218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,4,balanced,0.08789867162704468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,4,balanced,0.09499733646710713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,4,balanced,0.09481599926948547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,4,balanced,0.09815999865531921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,4,balanced,0.10425600409507751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,4,balanced,0.11416533589363098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,4,balanced,0.12638399998346964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,4,balanced,0.14244799812634787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,4,balanced,0.1753173271814982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,4,balanced,0.20465066035588583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,4,balanced,0.2718186577161153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,4,balanced,0.33318400382995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,4,balanced,0.4564799865086873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,4,balanced,0.5774986743927002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,64,balanced,0.01758933315674464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,64,balanced,0.016938666502634685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,64,balanced,0.01720533271630605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,64,balanced,0.018042666216691334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,64,balanced,0.0210506667693456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,64,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,64,balanced,0.027466667195161183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,64,balanced,0.02736533433198929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,64,balanced,0.027461332579453785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,64,balanced,0.02736533433198929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,64,balanced,0.027456000447273254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,64,balanced,0.027471999327341717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,64,balanced,0.029146666328112285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,64,balanced,0.029152000943819683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,64,balanced,0.02957333376010259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,64,balanced,0.029391999046007793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,64,balanced,0.03152533372243246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,64,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,64,balanced,0.035301332672437034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,64,balanced,0.04314666489760081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,64,balanced,0.04520533482233683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,64,balanced,0.05193600058555603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,64,balanced,0.062277331948280334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,64,balanced,0.0803413341442744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,64,balanced,0.10077866911888123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,64,balanced,0.1368053356806437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,64,balanced,0.17386666933695474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,8,balanced,0.043978666265805565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,8,balanced,0.043866669138272606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,8,balanced,0.046906664967536926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,8,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,8,balanced,0.07388799885908763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,8,balanced,0.10345066587130229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,8,balanced,0.10505599776903789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,8,balanced,0.10635200142860413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,8,balanced,0.10571733117103577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,8,balanced,0.10872000455856323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,8,balanced,0.10885866483052571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,8,balanced,0.10916800300280254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,8,balanced,0.11148800452550252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,8,balanced,0.1139306624730428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,8,balanced,0.11667199929555257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,8,balanced,0.12045333782831828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,8,balanced,0.12172800302505493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,8,balanced,0.1339413324991862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,8,balanced,0.13828800121943155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,8,balanced,0.15654399991035461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,8,balanced,0.17679466803868613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,8,balanced,0.20942399899164835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,8,balanced,0.24627200762430826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,8,balanced,0.315013329188029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,8,balanced,0.3889546791712443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,8,balanced,0.5377493302027384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,8,balanced,0.6725599765777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,4,balanced,0.021530665457248688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,4,balanced,0.021029333273569744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,4,balanced,0.023215999205907185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,4,balanced,0.02756800005833308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,4,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,4,balanced,0.05105599761009216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,4,balanced,0.05108266572157542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,4,balanced,0.05574400226275126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,4,balanced,0.0565280020236969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,4,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,4,balanced,0.06121066709359487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,4,balanced,0.06357333560784657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,4,balanced,0.06600533425807953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,4,balanced,0.06685333450635274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,4,balanced,0.07054399947325389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,4,balanced,0.07246933380762736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,4,balanced,0.08412800232569377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,4,balanced,0.09646399815877278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,4,balanced,0.10957333445549011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,4,balanced,0.12520000338554382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,4,balanced,0.1906773249308268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,4,balanced,0.1936639944712321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,4,balanced,0.23058666785558066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,4,balanced,0.27380800247192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,4,balanced,0.32181866963704425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,4,balanced,0.4965386788050334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,4,balanced,0.5956053336461385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,8,power_law_1.2,0.02617599964141846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,8,power_law_1.2,0.024736000597476958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,8,power_law_1.2,0.024153600633144378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,8,power_law_1.2,0.024223999679088594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,8,power_law_1.2,0.024966399371623992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,8,power_law_1.2,0.02917119860649109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,8,power_law_1.2,0.05761280059814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,8,power_law_1.2,0.05845119953155518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,8,power_law_1.2,0.05790079832077026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,8,power_law_1.2,0.05288959741592407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,8,power_law_1.2,0.044268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,8,power_law_1.2,0.08187519907951354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,8,power_law_1.2,0.07784960269927979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,8,power_law_1.2,0.05975679755210876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,8,power_law_1.2,0.0920960009098053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,8,power_law_1.2,0.072953599691391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,8,power_law_1.2,0.09377279877662659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,8,power_law_1.2,0.11701760292053223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.2,0.14115840196609497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.2,0.20044798851013185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.2,0.25541119575500487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.2,0.3572864055633545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.2,0.4672959804534912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,16,balanced,0.019589333484570186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,16,balanced,0.020917333662509918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,16,balanced,0.02092266579469045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,16,balanced,0.021274665991465252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,16,balanced,0.02606400102376938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,16,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,16,balanced,0.04658666749795278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,16,balanced,0.045312002301216125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,16,balanced,0.045461331804593406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,16,balanced,0.033200000723203026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,16,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,16,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,16,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,16,balanced,0.041936000188191734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,16,balanced,0.043061330914497375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,16,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,16,balanced,0.043509334325790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,16,balanced,0.04368533194065094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,16,balanced,0.0432586669921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,16,balanced,0.043866669138272606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,16,balanced,0.04363733530044556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,16,balanced,0.045514668027559914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,16,balanced,0.0443146675825119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,16,balanced,0.046623999873797096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,16,balanced,0.04870399832725525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,16,balanced,0.04915200173854828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,16,balanced,0.05226133267084757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,16,balanced,0.052245333790779114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,4,power_law_1.2,0.0489984005689621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,4,power_law_1.2,0.06216959953308106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,4,power_law_1.2,0.06071680188179016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,4,power_law_1.2,0.07582079768180847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,4,power_law_1.2,0.08432000279426574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,4,power_law_1.2,0.08733440041542054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,4,power_law_1.2,0.11110399961471558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,4,power_law_1.2,0.10299520492553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,4,power_law_1.2,0.11006720066070556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,4,power_law_1.2,0.10767999887466431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,4,power_law_1.2,0.11274880170822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,4,power_law_1.2,0.11939200162887573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,4,power_law_1.2,0.12016639709472657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,4,power_law_1.2,0.12068480253219604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,4,power_law_1.2,0.13026560544967652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,4,power_law_1.2,0.1345471978187561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,16,balanced,0.05842133363087972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,16,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,16,balanced,0.07400533556938171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,16,balanced,0.08078399797280629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,16,balanced,0.10533866286277771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,16,balanced,0.1283573309580485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,16,balanced,0.16808533668518066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,16,balanced,0.2103839914004008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,4,power_law_1.2,0.1442752003669739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,4,power_law_1.2,0.16652799844741822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,4,power_law_1.2,0.17541120052337647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,4,power_law_1.2,0.19663360118865966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,4,power_law_1.2,0.22517120838165283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,4,power_law_1.2,0.29964160919189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.2,0.6753727912902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,4,power_law_1.2,0.3524928092956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,4,power_law_1.2,0.5018176078796387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,4,power_law_1.2,0.6036352157592774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,4,power_law_1.2,0.900268840789795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,4,power_law_1.2,1.0546175956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.2,0.8865280151367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,64,balanced,0.05821866790453593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,64,balanced,0.04952533543109894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,64,balanced,0.04985600213209788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,64,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,64,balanced,0.05235733091831207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,64,balanced,0.06474133332570393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,64,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,64,balanced,0.06620799998442332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,64,balanced,0.06599999964237213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,64,balanced,0.06634666522343953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,64,balanced,0.06659733255704244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,64,balanced,0.07006399830182393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,64,balanced,0.07294933497905731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,64,balanced,0.07246399919191997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,64,balanced,0.07835199932257335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,64,balanced,0.07999466856320699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,64,balanced,0.08312533299128215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,64,balanced,0.09033600489298503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,64,balanced,0.09635200103123982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,64,balanced,0.10725333293279012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,64,balanced,0.12688533465067545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,64,balanced,0.14918933312098184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,64,balanced,0.1684160033861796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,64,balanced,0.21401600042978922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,64,balanced,0.2548266649246216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,64,balanced,0.343178669611613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,64,balanced,0.4342079957326253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.2,1.3185919761657714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.2,1.7275711059570313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,1,power_law_1.2,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,1,power_law_1.2,0.03578880131244659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,1,power_law_1.2,0.039084801077842714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,1,power_law_1.2,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,1,power_law_1.2,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,1,power_law_1.2,0.0788096010684967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,1,power_law_1.2,0.08270080089569092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,1,power_law_1.2,0.0861631989479065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,1,power_law_1.2,0.08930559754371643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,1,power_law_1.2,0.09047039747238159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,1,power_law_1.2,0.095551997423172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,1,power_law_1.2,0.09860479831695557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,1,power_law_1.2,0.1028480052947998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,1,power_law_1.2,0.11029119491577148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,1,power_law_1.2,0.12046719789505005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,1,power_law_1.2,0.12725119590759276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,1,power_law_1.2,0.13420159816741944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,1,power_law_1.2,0.1587391972541809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,1,power_law_1.2,0.17772159576416016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,1,power_law_1.2,0.22306559085845948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,1,power_law_1.2,0.2613248109817505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,1,power_law_1.2,0.3573888063430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,1,power_law_1.2,0.44714879989624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,1,power_law_1.2,0.6349440097808838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,64,power_law_1.2,0.07676159739494323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,64,power_law_1.2,0.07361279726028443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,64,power_law_1.2,0.07285119891166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,64,power_law_1.2,0.07594879865646362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,64,balanced,0.029509333272775013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,64,balanced,0.029296000798543293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,64,balanced,0.029440000653266907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,64,balanced,0.02994133283694585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,64,balanced,0.033717334270477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,64,balanced,0.03347733368476232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,64,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,64,balanced,0.033802665770053864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,64,balanced,0.03403733422358831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,64,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,64,balanced,0.03534399966398875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,64,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,64,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,64,balanced,0.03764266769091288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,64,power_law_1.2,0.0721343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,64,power_law_1.2,0.07786880135536194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,64,power_law_1.2,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,64,power_law_1.2,0.07811840176582337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,64,power_law_1.2,0.08048639893531799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,64,power_law_1.2,0.07918720245361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,64,balanced,0.03976000100374222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,64,balanced,0.041749333341916404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,64,power_law_1.2,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,64,power_law_1.2,0.08495360016822814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,64,power_law_1.2,0.08616960048675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,64,power_law_1.2,0.08652799725532531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,64,power_law_1.2,0.09658240079879761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,64,power_law_1.2,0.09916160106658936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,64,power_law_1.2,0.1049280047416687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,64,balanced,0.041893333196640015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,64,balanced,0.043706665436426796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,64,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,64,power_law_1.2,0.12770559787750244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,64,balanced,0.05040533343950907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,64,balanced,0.05286933481693268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,64,balanced,0.07229866584142049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,64,balanced,0.09116799632708232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,64,balanced,0.11963199575742085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,64,balanced,0.14615999658902487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,64,balanced,0.1965173284212748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,64,balanced,0.24921600023905435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,1,power_law_1.2,0.8129983901977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,64,power_law_1.2,0.1332800030708313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,64,power_law_1.2,0.1710911989212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,64,power_law_1.2,0.1791424036026001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,64,power_law_1.2,0.24243199825286865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,64,power_law_1.2,0.24332799911499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,64,power_law_1.2,0.3396735906600952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,64,power_law_1.2,0.3946943998336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,64,power_law_1.2,0.5888383865356446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,64,power_law_1.2,0.7336832046508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,1,power_law_1.2,1.1721343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,2,power_law_1.01,0.1478976011276245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,2,power_law_1.01,0.2425920009613037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,2,power_law_1.01,0.3499392032623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,2,power_law_1.01,0.5555327892303467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,2,power_law_1.01,0.7606207847595214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,2,power_law_1.01,0.9086848258972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,2,power_law_1.01,1.271827220916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,2,power_law_1.01,1.2416064262390136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,2,power_law_1.01,1.325817584991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,2,power_law_1.01,1.3701248168945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,2,power_law_1.01,1.3869248390197755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,2,power_law_1.01,1.3037440299987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,2,power_law_1.01,1.3778816223144532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,2,power_law_1.01,1.5774527549743653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,2,power_law_1.01,1.5184384346008302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,2,power_law_1.01,1.5355775833129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,1,power_law_1.2,1.531884765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,2,power_law_1.01,1.607276725769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,2,power_law_1.01,1.7760128021240233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,2,power_law_1.01,1.732352066040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,2,power_law_1.01,1.9876096725463868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,2,power_law_1.01,1.8577280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,2,power_law_1.01,2.145011138916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,2,power_law_1.01,2.5580928802490233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,2,power_law_1.01,2.9721471786499025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,2,power_law_1.01,3.4475582122802733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,2,power_law_1.01,4.589913558959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,2,power_law_1.01,5.530879974365234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,8,power_law_1.01,0.06260480284690857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,8,power_law_1.01,0.077920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,8,power_law_1.01,0.07841280102729797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,8,power_law_1.01,0.08631680011749268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,8,power_law_1.01,0.09078400135040283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,8,power_law_1.01,0.09307519793510437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,8,power_law_1.01,0.09483519792556763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,8,power_law_1.01,0.0960640013217926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,8,power_law_1.01,0.09568639993667602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,8,power_law_1.01,0.09455999732017517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,8,power_law_1.01,0.09881600141525268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,8,power_law_1.01,0.0982208013534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,8,power_law_1.01,0.10519039630889893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,8,power_law_1.01,0.10505599975585937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,8,power_law_1.01,0.11115520000457764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,8,power_law_1.01,0.11336959600448608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,8,power_law_1.01,0.11745280027389526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,8,power_law_1.01,0.13276159763336182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,8,power_law_1.01,0.141702401638031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,8,power_law_1.01,0.16980479955673217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,8,power_law_1.01,0.19608960151672364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,8,power_law_1.01,0.2718656063079834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,8,power_law_1.01,0.3414463996887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,8,power_law_1.01,0.4537087917327881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,8,power_law_1.01,0.605625581741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,8,power_law_1.01,0.8420224189758301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,8,power_law_1.01,1.1252096176147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,1,power_law_1.2,0.05052800178527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,1,power_law_1.2,0.05416960120201111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,1,power_law_1.2,0.06786559820175171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,1,power_law_1.2,0.0859008014202118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,1,power_law_1.2,0.10073599815368653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,1,power_law_1.2,0.14742399454116822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,1,power_law_1.2,0.1592512011528015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,1,power_law_1.2,0.16317440271377565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,1,power_law_1.2,0.16803200244903566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,1,power_law_1.2,0.17593599557876588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,1,power_law_1.2,0.17507840394973756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,1,power_law_1.2,0.17991039752960206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,1,power_law_1.2,0.18691200017929077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,1,power_law_1.2,0.1971519947052002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,1,power_law_1.2,0.20298240184783936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,1,power_law_1.2,0.20977919101715087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,1,power_law_1.2,0.22381439208984374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,1,power_law_1.2,0.2578687906265259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.2,0.28789119720458983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.2,0.3524735927581787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.2,0.3875711917877197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.2,0.5012991905212403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.2,0.6151679992675781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.2,0.8251456260681153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.2,1.0347904205322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.2,1.4560000419616699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.2,1.886911964416504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,64,power_law_1.01,0.08695039749145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,64,power_law_1.01,0.0855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,64,power_law_1.01,0.08340479731559754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,64,power_law_1.01,0.08871039748191833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,64,power_law_1.01,0.0831488013267517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,64,power_law_1.01,0.08546559810638428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,64,power_law_1.01,0.08943359851837158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,64,power_law_1.01,0.09271039962768554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,64,power_law_1.01,0.08791040182113648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,64,power_law_1.01,0.08610560297966004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,64,power_law_1.01,0.09322239756584168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,64,power_law_1.01,0.09409279823303222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,64,power_law_1.01,0.09634559750556945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,64,power_law_1.01,0.09573119878768921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,64,power_law_1.01,0.10130560398101807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,64,power_law_1.01,0.10324480533599853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,32,balanced,0.043007999658584595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,64,power_law_1.01,0.10792959928512573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,32,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,32,balanced,0.0418453315893809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,64,power_law_1.01,0.12296960353851319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,32,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,32,balanced,0.04568533102671305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,64,power_law_1.01,0.14058239459991456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,64,power_law_1.01,0.16563839912414552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,64,power_law_1.01,0.19052799940109252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,64,power_law_1.01,0.24228479862213134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,64,power_law_1.01,0.31125760078430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,64,power_law_1.01,0.412659215927124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,32,balanced,0.0499946673711141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,64,power_law_1.01,0.523526382446289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,32,balanced,0.049770668148994446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,32,balanced,0.0516480008761088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,32,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,64,power_law_1.01,0.7413951873779296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,32,balanced,0.05221333106358846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,32,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,32,balanced,0.054048001766204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,32,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,64,power_law_1.01,0.9781824111938476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,32,balanced,0.05615466833114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,32,balanced,0.06072533130645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,32,balanced,0.06166933476924896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,32,balanced,0.062394668658574425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,32,balanced,0.07978133360544841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,32,balanced,0.07844799757003784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,32,balanced,0.10595732927322388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,32,balanced,0.11941867073376973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,32,balanced,0.15678933262825012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,32,balanced,0.1796320080757141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,32,balanced,0.24412800868352255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,32,balanced,0.28628265857696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,32,balanced,0.40563734372456867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,32,balanced,0.510533332824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,128,power_law_1.01,0.037651199102401736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,128,power_law_1.01,0.036908799409866334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,128,power_law_1.01,0.037011200189590455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,128,power_law_1.01,0.038764798641204835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,128,power_law_1.01,0.03948799967765808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,128,power_law_1.01,0.04023039937019348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,128,power_law_1.01,0.040454399585723874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,128,power_law_1.01,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,128,power_law_1.01,0.039705601334571836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,128,power_law_1.01,0.0409855991601944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,128,power_law_1.01,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,128,power_law_1.01,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,128,power_law_1.01,0.04192639887332916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,128,power_law_1.01,0.04181120097637177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,32,power_law_1.01,0.06216959953308106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,128,power_law_1.01,0.04454399943351746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,32,power_law_1.01,0.05950719714164734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,128,power_law_1.01,0.04415999948978424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,32,power_law_1.01,0.05926399827003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,128,power_law_1.01,0.045491200685501096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,32,power_law_1.01,0.059961599111557004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,32,power_law_1.01,0.06051200032234192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,32,power_law_1.01,0.062067198753356936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,32,power_law_1.01,0.06122879981994629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,32,power_law_1.01,0.061561602354049685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,128,power_law_1.01,0.045798400044441225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,128,power_law_1.01,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,128,power_law_1.01,0.054073601961135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,128,power_law_1.01,0.055327999591827395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,128,power_law_1.01,0.06895359754562377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,128,power_law_1.01,0.0804032027721405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,128,power_law_1.01,0.10440319776535034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,128,power_law_1.01,0.12870399951934813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,32,power_law_1.01,0.06241919994354248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,128,power_law_1.01,0.1902783989906311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,32,power_law_1.01,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,128,power_law_1.01,0.21724801063537597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,32,power_law_1.01,0.06282240152359009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,32,power_law_1.01,0.06437119841575623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,32,power_law_1.01,0.06531199812889099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,32,power_law_1.01,0.06578559875488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,32,power_law_1.01,0.07061759829521179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,32,power_law_1.01,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,32,power_law_1.01,0.07423359751701356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,32,power_law_1.01,0.08126720190048217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,32,power_law_1.01,0.0906495988368988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,32,power_law_1.01,0.11156480312347412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,32,power_law_1.01,0.12679040431976318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,32,power_law_1.01,0.1589184045791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,32,power_law_1.01,0.19169280529022217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,32,power_law_1.01,0.26632959842681886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,32,power_law_1.01,0.32217600345611574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,32,power_law_1.01,0.4451648235321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,32,power_law_1.01,0.5704703807830811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,8,balanced,0.03329599897066752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,8,balanced,0.03329066683848699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,8,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,8,balanced,0.035114665826161705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,8,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,8,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,8,balanced,0.03562133262554804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,8,balanced,0.03728533287843069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,8,balanced,0.03572800010442734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,8,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,8,balanced,0.03588266670703888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,8,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,8,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,8,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,8,balanced,0.04586133360862732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,8,balanced,0.04852266609668732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,8,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,8,balanced,0.05243733525276184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,8,balanced,0.058335999647776283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,8,balanced,0.0664160003264745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,8,balanced,0.07351466516653697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,8,balanced,0.09076799949010213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,8,balanced,0.11322133739789327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,8,balanced,0.1607253352801005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,8,balanced,0.19724265734354654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,8,balanced,0.2691733241081238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,8,balanced,0.35226134459177655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,4,power_law_1.2,0.07626240253448487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,4,power_law_1.2,0.082150399684906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,4,power_law_1.2,0.08339200019836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,4,power_law_1.2,0.09182720184326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,4,power_law_1.2,0.10750720500946045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,4,power_law_1.2,0.11458560228347778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,4,power_law_1.2,0.13144320249557495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,4,power_law_1.2,0.14013439416885376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,4,power_law_1.2,0.13868800401687623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,4,power_law_1.2,0.1345728039741516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,4,power_law_1.2,0.1414720058441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,4,power_law_1.2,0.13868160247802735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,4,power_law_1.2,0.14046080112457277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,4,power_law_1.2,0.14840960502624512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,4,power_law_1.2,0.15447039604187013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,4,power_law_1.2,0.15742720365524293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,4,power_law_1.2,0.16528639793395997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,4,power_law_1.2,0.18272639513015748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,4,power_law_1.2,0.19836800098419188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,4,power_law_1.2,0.23467519283294677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,4,power_law_1.2,0.2690239906311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,4,power_law_1.2,0.35493760108947753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,4,power_law_1.2,0.41716480255126953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,4,power_law_1.2,0.5775807857513428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,4,power_law_1.2,0.7613759994506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,4,power_law_1.2,1.1285887718200684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,4,power_law_1.2,1.375667190551758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,32,power_law_1.01,0.029868799448013305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,32,power_law_1.01,0.027699199318885804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,32,power_law_1.01,0.02773759961128235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,32,power_law_1.01,0.030067199468612672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,32,power_law_1.01,0.03086079955101013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,32,power_law_1.01,0.03221760094165802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,32,power_law_1.01,0.032313600182533264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,32,power_law_1.01,0.03333759903907776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,32,power_law_1.01,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,32,power_law_1.01,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,32,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,8,power_law_1.01,0.10531200170516967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,8,power_law_1.01,0.13879040479660035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,8,power_law_1.01,0.1258687973022461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,8,power_law_1.01,0.15015039443969727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,32,power_law_1.01,0.03594239950180054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,8,power_law_1.01,0.13991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,8,power_law_1.01,0.1669376015663147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,8,power_law_1.01,0.16227840185165404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,8,power_law_1.01,0.16785279512405396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,8,power_law_1.01,0.17704319953918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,8,power_law_1.01,0.1697535991668701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,32,power_law_1.01,0.03640320003032684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,8,power_law_1.01,0.17777279615402222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,8,power_law_1.01,0.18417279720306395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,8,power_law_1.01,0.18110079765319825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,8,power_law_1.01,0.18857599496841432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,8,power_law_1.01,0.1996224045753479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,32,power_law_1.01,0.03841919898986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,8,power_law_1.01,0.2088831901550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,8,power_law_1.01,0.21446399688720702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,8,power_law_1.01,0.237939190864563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,8,power_law_1.01,0.2518336057662964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,8,power_law_1.01,0.29569919109344484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,8,power_law_1.01,0.31390719413757323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,8,power_law_1.01,0.38178560733795164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,8,power_law_1.01,0.4296383857727051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,8,power_law_1.01,0.5986688137054443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,8,power_law_1.01,0.8025152206420898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,8,power_law_1.01,0.9603391647338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,8,power_law_1.01,1.4885760307312013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,32,power_law_1.01,0.04284160137176514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,32,power_law_1.01,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,32,power_law_1.01,0.04527359902858734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,32,power_law_1.01,0.05051519870758057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,32,power_law_1.01,0.05390080213546753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,32,power_law_1.01,0.06792960166931153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,32,power_law_1.01,0.07484800219535828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,32,power_law_1.01,0.10338560342788697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,32,power_law_1.01,0.12323839664459228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,32,power_law_1.01,0.17292799949645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,32,power_law_1.01,0.2318272113800049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,32,power_law_1.01,0.3397952079772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,32,power_law_1.01,0.4732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,8,power_law_1.01,0.045459198951721194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,8,power_law_1.01,0.06872320175170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,8,power_law_1.01,0.08060160279273987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,8,power_law_1.01,0.08944000005722046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,8,power_law_1.01,0.10889600515365601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,8,power_law_1.01,0.11204479932785034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,8,power_law_1.01,0.10812159776687622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,8,power_law_1.01,0.10739200115203858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,8,power_law_1.01,0.11834880113601684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,8,power_law_1.01,0.12135679721832275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,8,power_law_1.01,0.11700479984283448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,8,power_law_1.01,0.11599359512329102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,8,power_law_1.01,0.11567360162734985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,8,power_law_1.01,0.12798080444335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,8,power_law_1.01,0.13619199991226197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,8,power_law_1.01,0.13980799913406372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,8,power_law_1.01,0.1420799970626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,8,power_law_1.01,0.1614016056060791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,8,power_law_1.01,0.31550719738006594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,8,power_law_1.01,0.38974719047546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,8,power_law_1.01,0.2455615997314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,8,power_law_1.01,0.29015040397644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,8,power_law_1.01,0.33159680366516114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,8,power_law_1.01,0.44871039390563966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,8,power_law_1.01,0.529804801940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,8,power_law_1.01,0.7117055892944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,8,power_law_1.01,0.9198399543762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,1,balanced,0.04990933338801066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,1,balanced,0.05317866802215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,1,balanced,0.06480533381303151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,1,balanced,0.09397332866986592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,1,balanced,0.13821867108345032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,1,balanced,0.15996266404787698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,1,balanced,0.16226133704185486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,1,balanced,0.16423466801643372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,1,balanced,0.1653546690940857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,1,balanced,0.1685333251953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,1,balanced,0.17147733767827353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,1,balanced,0.17543466885884604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,1,balanced,0.17804267009099325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,1,balanced,0.1859253247578939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,1,balanced,0.19187732537587485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,1,balanced,0.20061333974202475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,1,balanced,0.21037866671880087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,1,balanced,0.2510506709416707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,1,balanced,0.27498666445414227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,1,balanced,0.39678935209910077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,1,balanced,0.4137173493703206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,1,balanced,0.623744010925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,1,balanced,0.6928213437398275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,1,balanced,1.0365173021952312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,1,balanced,1.267685333887736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,1,balanced,1.9341227213541667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,1,balanced,2.3768107096354165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,2,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,2,balanced,0.051701332132021584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,2,balanced,0.07221866647402446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,2,balanced,0.10590400298436482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,2,balanced,0.1783626675605774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,2,balanced,0.18797866503397623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,2,balanced,0.19014400243759155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,2,balanced,0.19125866889953613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,2,balanced,0.19065600633621216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,2,balanced,0.1930560072263082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,2,balanced,0.19818133115768433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,2,balanced,0.2008906602859497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,2,balanced,0.203658660252889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,2,balanced,0.20627200603485107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,2,balanced,0.21308799584706625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,2,balanced,0.21935999393463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,2,balanced,0.22581332921981812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,2,balanced,0.24963200092315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,2,balanced,0.2614826758702596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,2,balanced,0.3227519989013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,2,balanced,0.3441493511199951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,2,balanced,0.4873973528544108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,2,balanced,0.5173013210296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,2,balanced,0.7917280197143555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,2,balanced,0.8671466509501139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,2,balanced,1.297434647878011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,2,balanced,1.582357406616211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,8,power_law_1.01,0.05777279734611511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,8,power_law_1.01,0.0646016001701355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,4,power_law_1.01,0.057683199644088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,8,power_law_1.01,0.07155200242996215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,8,power_law_1.01,0.07084800004959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,8,power_law_1.01,0.07856000065803528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,8,power_law_1.01,0.08634240031242371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,4,balanced,0.06514133512973785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,8,power_law_1.01,0.08618879914283753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,4,balanced,0.08564800024032593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,4,balanced,0.12984533111254373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,8,power_law_1.01,0.08657280206680298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,4,balanced,0.2213653326034546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,4,balanced,0.38521599769592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,8,power_law_1.01,0.08600320219993592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,4,balanced,0.722821315129598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,8,power_law_1.01,0.08407679796218873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,8,power_law_1.01,0.08749439716339111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,8,power_law_1.01,0.09200000166893005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,4,power_law_1.01,0.062035202980041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,8,power_law_1.01,0.09360640048980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,4,power_law_1.01,0.06813439726829529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,4,power_law_1.01,0.07311999797821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,4,power_law_1.01,0.08542079925537109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,4,balanced,0.7283679644266764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,4,power_law_1.01,0.09540479779243469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,4,balanced,0.7276053428649902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,4,power_law_1.01,0.1003775954246521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,4,power_law_1.01,0.1015552043914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,4,power_law_1.01,0.10087039470672607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,4,power_law_1.01,0.10239360332489014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,4,power_law_1.01,0.10392960309982299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,4,power_law_1.01,0.10735360383987427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,4,power_law_1.01,0.10704640150070191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,4,balanced,0.7292586962381998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,4,power_law_1.01,0.11033600568771362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,4,balanced,0.7273386319478353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,4,power_law_1.01,0.11450239419937133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,4,power_law_1.01,0.11493760347366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,8,power_law_1.01,0.0936959981918335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,4,power_law_1.01,0.12492159605026246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,8,power_law_1.01,0.10753920078277587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,4,power_law_1.01,0.14262399673461915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,8,power_law_1.01,0.10925439596176148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,4,power_law_1.01,0.15723520517349243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,4,power_law_1.01,0.1828927993774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,4,balanced,0.732634703318278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,4,balanced,0.7398826281229655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,4,power_law_1.01,0.22330238819122314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,4,balanced,0.7487893104553223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,4,balanced,0.027749332288901012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,4,balanced,0.7436373233795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,4,balanced,0.0330826664964358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,4,balanced,0.04554666578769684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,4,balanced,0.0583840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,4,balanced,0.0867146650950114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,4,balanced,0.10544533530871074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,8,power_law_1.01,0.11936000585556031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,8,power_law_1.01,0.1337407946586609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,8,power_law_1.01,0.15511679649353027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,8,power_law_1.01,0.1866495966911316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,4,power_law_1.01,0.27971839904785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,8,power_law_1.01,0.19416320323944092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,4,balanced,0.7498400211334229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,4,power_law_1.01,0.37007999420166016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,4,balanced,0.7559519608815511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,4,balanced,0.10518399874369304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,4,balanced,0.10914666453997295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,4,balanced,0.7572106520334879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,4,balanced,0.1123466690381368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,4,balanced,0.12326932946840923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,4,balanced,0.7733919620513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,4,balanced,0.7991253534952799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,4,balanced,0.8244053522745768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,4,balanced,0.8606133460998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,8,power_law_1.01,0.2508671998977661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,4,balanced,0.9092693328857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,8,power_law_1.01,0.30203518867492674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,4,power_law_1.01,0.5174208164215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,4,balanced,0.12550933162371317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,8,power_law_1.01,0.4058239936828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,4,power_law_1.01,0.6676415920257568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,8,power_law_1.01,0.5594240188598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,4,power_law_1.01,0.9980928421020507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,8,power_law_1.01,0.7881216049194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,4,balanced,1.008789300918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,4,power_law_1.01,1.2783743858337402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,4,balanced,1.1165013313293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,4,balanced,1.2076373100280762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,4,balanced,1.7379412651062012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,4,balanced,1.8939414024353027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,4,balanced,0.12761599818865457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,4,balanced,0.17926400899887085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,4,balanced,0.17412267128626505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,4,balanced,0.16949333747227988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,4,balanced,0.23173866669336954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,4,balanced,0.2579519947369893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,4,balanced,0.23441600799560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,4,balanced,0.24744532505671182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,4,balanced,0.26825066407521564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,4,balanced,0.2916266719500224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,8,power_law_1.01,1.0118144035339356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,4,balanced,0.3319680094718933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,4,balanced,0.3602026700973511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,4,balanced,0.566975990931193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,4,balanced,0.6461546818415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,4,balanced,0.9510293006896973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,4,balanced,1.2375786304473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,32,balanced,0.018186666071414948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,32,balanced,0.01720000058412552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,32,balanced,0.01741333305835724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,32,balanced,0.01724799970785777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,32,balanced,0.019226666539907455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,32,balanced,0.019333332777023315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,32,balanced,0.01934933289885521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,32,balanced,0.021322667598724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,32,balanced,0.021509334444999695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,32,balanced,0.023317334552605946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,32,balanced,0.029290666182835896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,32,balanced,0.02852799991766612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,32,balanced,0.029461334149042766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,32,balanced,0.03051200012365977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,32,balanced,0.029440000653266907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,32,balanced,0.02943466603755951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,32,balanced,0.03224000086386999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,32,balanced,0.036144000788529716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,32,balanced,0.0377813329299291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,32,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,32,balanced,0.04779199759165446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,32,balanced,0.06235733131567637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,32,balanced,0.07169066866238911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,32,balanced,0.09504533807436626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,32,balanced,0.11824533343315125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,32,balanced,0.1644480029741923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,32,balanced,0.20576000213623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,64,power_law_1.2,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,64,power_law_1.2,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,64,power_law_1.2,0.044486400485038755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,64,power_law_1.2,0.04723199903964996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,64,power_law_1.2,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,64,power_law_1.2,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,64,power_law_1.2,0.05156480073928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,64,power_law_1.2,0.05272960066795349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,64,power_law_1.2,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,64,power_law_1.2,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,64,power_law_1.2,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,64,power_law_1.2,0.05852800011634827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,64,power_law_1.2,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,64,power_law_1.2,0.061945599317550656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,64,power_law_1.2,0.07048320174217224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,64,power_law_1.2,0.07285760045051574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,64,power_law_1.2,0.07687680125236511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,64,power_law_1.2,0.0943615972995758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,64,power_law_1.2,0.10295039415359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,64,power_law_1.2,0.12478079795837402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,64,power_law_1.2,0.14465919733047486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,64,power_law_1.2,0.19413119554519653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,64,power_law_1.2,0.24769279956817628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,64,power_law_1.2,0.3387712001800537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,64,power_law_1.2,0.44389119148254397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,64,power_law_1.2,0.6917312145233154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,64,power_law_1.2,1.0464256286621094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,8,power_law_1.01,0.07587199807167053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,8,power_law_1.01,0.08911359906196595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,8,power_law_1.01,0.0876800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,8,power_law_1.01,0.09162880182266235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,8,power_law_1.01,0.09873279929161072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,8,power_law_1.01,0.10623359680175781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,8,power_law_1.01,0.10780800580978393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,8,power_law_1.01,0.10409599542617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,8,power_law_1.01,0.1108672022819519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,8,power_law_1.01,0.10867840051651001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,8,power_law_1.01,0.11444480419158935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,8,power_law_1.01,0.11237759590148926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,8,power_law_1.01,0.11784319877624512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,8,power_law_1.01,0.12215679883956909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,8,power_law_1.01,0.13152639865875243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,8,power_law_1.01,0.13396480083465576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,8,power_law_1.01,0.13560960292816163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,8,power_law_1.01,0.15158400535583497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,8,power_law_1.01,0.1514623999595642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,8,power_law_1.01,0.19728000164031984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,8,power_law_1.01,0.21984639167785644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,8,power_law_1.01,0.2798847913742065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,8,power_law_1.01,0.35034239292144775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,8,power_law_1.01,0.46286721229553224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,8,power_law_1.01,0.4609216213226318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,8,power_law_1.01,0.7951168060302735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,8,power_law_1.01,0.9260543823242188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,32,power_law_1.2,0.08714240193367004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,32,power_law_1.2,0.0875711977481842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,32,power_law_1.2,0.0823743999004364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,32,power_law_1.2,0.08583679795265198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,32,power_law_1.2,0.06968320012092591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,32,power_law_1.2,0.0800000011920929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,32,power_law_1.2,0.08566399812698364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,32,power_law_1.2,0.08144639730453491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,32,power_law_1.2,0.08568959832191467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,32,power_law_1.2,0.082259202003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,32,power_law_1.2,0.08563200235366822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,32,power_law_1.2,0.08776320219039917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,32,power_law_1.2,0.08733440041542054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,32,power_law_1.2,0.09565439820289612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,32,power_law_1.2,0.10087039470672607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,32,power_law_1.2,0.10433919429779052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,32,power_law_1.2,0.11228799819946289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,32,power_law_1.2,0.1275712013244629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,32,power_law_1.2,0.12916480302810668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,32,power_law_1.2,0.16263680458068847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,32,power_law_1.2,0.156876802444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,32,power_law_1.2,0.20531198978424073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,32,power_law_1.2,0.2553024053573608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,8,power_law_1.01,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,32,power_law_1.2,0.3330303907394409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,8,power_law_1.01,0.049830400943756105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,32,power_law_1.2,0.4329216003417969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,8,power_law_1.01,0.0572160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,32,power_law_1.2,0.6236224174499512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,8,power_law_1.01,0.06354560256004334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,8,power_law_1.01,0.0691968023777008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,32,power_law_1.2,0.827302360534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,8,power_law_1.01,0.07862399816513062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,8,power_law_1.2,0.037049600481987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,8,power_law_1.01,0.08085119724273682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,8,power_law_1.2,0.0377920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,8,power_law_1.2,0.045612800121307376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,8,power_law_1.01,0.08348159790039063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,8,power_law_1.2,0.05043839812278748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,8,power_law_1.01,0.0784063994884491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,8,power_law_1.2,0.04332799911499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,8,power_law_1.01,0.08161919713020324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,8,power_law_1.2,0.05559039711952209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,8,power_law_1.2,0.056601601839065555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,8,power_law_1.01,0.07870079874992371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,8,power_law_1.2,0.05863680243492127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,8,power_law_1.01,0.0852735996246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,8,power_law_1.2,0.058195197582244874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,8,power_law_1.2,0.05802239775657654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,8,power_law_1.2,0.061952000856399535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,8,power_law_1.2,0.06341760158538819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,8,power_law_1.2,0.06664320230484008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,8,power_law_1.2,0.06943359971046448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,8,power_law_1.2,0.07561600208282471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,8,power_law_1.2,0.07951359748840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,8,power_law_1.2,0.0807039976119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,8,power_law_1.01,0.08791679739952088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,8,power_law_1.01,0.09390079975128174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,8,power_law_1.01,0.10174720287322998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,8,power_law_1.01,0.10675840377807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,8,power_law_1.01,0.11179519891738891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,8,power_law_1.01,0.127839994430542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,8,power_law_1.2,0.09182720184326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,8,power_law_1.01,0.14336639642715454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,8,power_law_1.2,0.1090880036354065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,8,power_law_1.01,0.179692804813385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,8,power_law_1.2,0.13242239952087403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,8,power_law_1.01,0.20943999290466309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,8,power_law_1.2,0.14766080379486085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,8,power_law_1.01,0.2735743999481201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,8,power_law_1.2,0.19224319458007813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,8,power_law_1.2,0.2486720085144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,8,power_law_1.2,0.35727999210357664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,8,power_law_1.01,0.3587519884109497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,8,power_law_1.01,0.47399678230285647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,8,power_law_1.01,0.592844820022583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,8,power_law_1.01,0.889459228515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,8,power_law_1.01,1.0933247566223145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,8,power_law_1.2,0.5057727813720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,8,power_law_1.2,0.6296448230743408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,8,power_law_1.2,0.757203197479248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,2,power_law_1.01,0.06606720089912414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,2,power_law_1.01,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,2,power_law_1.01,0.07914239764213563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,2,power_law_1.01,0.09291520118713378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,2,power_law_1.01,0.10629119873046874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,2,power_law_1.01,0.12477439641952515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,2,power_law_1.01,0.1472000002861023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,2,power_law_1.01,0.153164803981781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,2,power_law_1.01,0.15918079614639283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,2,power_law_1.01,0.1588096022605896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,2,power_law_1.01,0.1641088008880615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,2,power_law_1.01,0.1673535943031311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,2,power_law_1.01,0.16947200298309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,2,power_law_1.01,0.17382400035858153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,2,power_law_1.01,0.1814463973045349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,2,power_law_1.01,0.18855040073394774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,2,power_law_1.01,0.19661439657211305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,2,power_law_1.01,0.22159359455108643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.01,0.25257599353790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.01,0.30721919536590575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.01,0.3566143989562988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.01,0.46450557708740237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.01,0.5726336002349853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.01,0.7853312015533447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.01,1.0174079895019532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.01,1.4276800155639648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.01,1.9438335418701171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,8,power_law_1.2,0.10215679407119752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,8,power_law_1.2,0.10736000537872314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,8,power_law_1.2,0.14458240270614625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,8,power_law_1.2,0.15260159969329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,8,power_law_1.2,0.1708672046661377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,8,power_law_1.2,0.15924479961395263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,8,power_law_1.2,0.16416640281677247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,8,power_law_1.2,0.16458239555358886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,8,power_law_1.2,0.1685312032699585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,8,power_law_1.2,0.1683776021003723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,8,power_law_1.2,0.1783552050590515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,8,power_law_1.2,0.17929600477218627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,8,power_law_1.2,0.18216320276260375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,8,power_law_1.2,0.19763200283050536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,8,power_law_1.2,0.198636794090271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,8,power_law_1.2,0.21143040657043458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,8,power_law_1.2,0.05735039710998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,8,power_law_1.2,0.23043839931488036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,8,power_law_1.2,0.07343360185623168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,8,power_law_1.2,0.07518079876899719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,8,power_law_1.2,0.08871039748191833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,8,power_law_1.2,0.09347839951515198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,8,power_law_1.2,0.09880319833755494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,8,power_law_1.2,0.2573247909545898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,8,power_law_1.2,0.10648319721221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,8,power_law_1.2,0.10651520490646363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,8,power_law_1.2,0.1072767972946167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,8,power_law_1.2,0.10314240455627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,8,power_law_1.2,0.10777599811553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,8,power_law_1.2,0.26576640605926516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,8,power_law_1.2,0.10559999942779541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,8,power_law_1.2,0.3131648063659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,8,power_law_1.2,0.10782719850540161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,8,power_law_1.2,0.3401727914810181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,8,power_law_1.2,0.11197439432144166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,8,power_law_1.2,0.3959359884262085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,8,power_law_1.2,0.12001919746398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,8,power_law_1.2,0.11790080070495605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,8,power_law_1.2,0.5126783847808838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,8,power_law_1.2,0.1275264024734497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,8,power_law_1.2,0.13974399566650392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,8,power_law_1.2,0.677836799621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,8,power_law_1.2,0.1467136025428772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,8,power_law_1.2,0.1690111994743347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,8,power_law_1.2,0.19799679517745972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,8,power_law_1.2,0.2524096012115479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,8,power_law_1.2,0.2582207918167114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,8,power_law_1.2,0.3303936004638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,8,power_law_1.2,0.4428095817565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,8,power_law_1.2,0.6066112041473388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,8,power_law_1.2,0.7188928127288818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,8,power_law_1.2,0.8290047645568848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,8,power_law_1.2,1.1241279602050782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,4,power_law_1.01,0.05196160078048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,4,power_law_1.01,0.06632959842681885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,4,power_law_1.01,0.06467199921607972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,4,power_law_1.01,0.08084480166435241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,4,power_law_1.01,0.08769919872283935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,4,power_law_1.01,0.10439039468765259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,4,power_law_1.01,0.11397759914398194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,4,power_law_1.01,0.11288959980010986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,4,power_law_1.01,0.11235200166702271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,4,power_law_1.01,0.11579519510269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,4,power_law_1.01,0.1173632025718689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,4,power_law_1.01,0.11457279920578003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,4,power_law_1.01,0.12190079689025879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,4,power_law_1.01,0.12826240062713623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,4,power_law_1.01,0.13489919900894165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,4,power_law_1.01,0.13791999816894532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,4,power_law_1.01,0.15000959634780883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,4,power_law_1.01,0.16758400201797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,4,power_law_1.01,0.18483200073242187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,4,power_law_1.01,0.2231231927871704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,4,power_law_1.01,0.25410559177398684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,4,power_law_1.01,0.3226111888885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,4,power_law_1.01,0.37562239170074463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,4,power_law_1.01,0.5059264183044434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,4,power_law_1.01,0.637830400466919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,4,power_law_1.01,0.8683967590332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,4,power_law_1.01,1.1852992057800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,8,power_law_1.2,1.5855039596557616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,2,power_law_1.01,0.05275520086288452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,2,power_law_1.01,0.06517120003700257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,2,power_law_1.01,0.0907584011554718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,2,power_law_1.01,0.10455679893493652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,2,power_law_1.01,0.12725119590759276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,2,power_law_1.01,0.18427519798278807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,2,power_law_1.01,0.19178880453109742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,2,power_law_1.01,0.18890880346298217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,2,power_law_1.01,0.19514240026474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,2,power_law_1.01,0.19494400024414063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,2,power_law_1.01,0.1990399956703186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,2,power_law_1.01,0.20436480045318603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,2,power_law_1.01,0.1998271942138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,2,power_law_1.01,0.21551361083984374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,2,power_law_1.01,0.23135359287261964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,2,power_law_1.01,0.23425281047821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,2,power_law_1.01,0.2386687994003296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,2,power_law_1.01,0.26552319526672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,2,power_law_1.01,0.285427188873291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,2,power_law_1.01,0.3452543973922729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,2,power_law_1.01,0.3670207977294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,2,power_law_1.01,0.4447360038757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,2,power_law_1.01,0.5305600166320801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,2,power_law_1.01,0.6717951774597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,2,power_law_1.01,0.8760767936706543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,2,power_law_1.01,1.1721023559570312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,2,power_law_1.01,1.4350208282470702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,32,4,power_law_1.2,0.04471679925918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,32,4,power_law_1.2,0.04874880015850067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,32,4,power_law_1.2,0.04782719910144806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,32,4,power_law_1.2,0.05484799742698669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,32,4,power_law_1.2,0.05772160291671753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,32,4,power_law_1.2,0.06334720253944397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,32,4,power_law_1.2,0.07347840070724487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,32,4,power_law_1.2,0.07429119944572449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,32,4,power_law_1.2,0.07637760043144226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,32,4,power_law_1.2,0.07723519802093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,32,4,power_law_1.2,0.07869439721107482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,32,4,power_law_1.2,0.08293120265007019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,32,4,power_law_1.2,0.0865343987941742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,32,4,power_law_1.2,0.09335039854049683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,32,4,power_law_1.2,0.10086400508880615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,32,4,power_law_1.2,0.10723199844360351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,32,4,power_law_1.2,0.10900479555130005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,32,4,power_law_1.2,0.13232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,32,4,power_law_1.2,0.14633599519729615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,32,4,power_law_1.2,0.1810431957244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,32,4,power_law_1.2,0.21452159881591798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,32,4,power_law_1.2,0.2775039911270142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,32,4,power_law_1.2,0.3711616039276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,32,4,power_law_1.2,0.504966402053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,32,4,power_law_1.2,0.6676032066345214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,32,4,power_law_1.2,0.9816767692565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,32,4,power_law_1.2,1.2844287872314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,32,power_law_1.2,0.06331520080566407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,32,power_law_1.2,0.053881597518920896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,32,power_law_1.2,0.05121279954910278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,32,power_law_1.2,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,32,power_law_1.2,0.051712000370025636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,32,power_law_1.2,0.05389440059661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,32,power_law_1.2,0.05559679865837097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,32,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,32,power_law_1.2,0.05766400098800659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,32,power_law_1.2,0.054118400812149046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,32,power_law_1.2,0.056966400146484374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,32,power_law_1.2,0.059628802537918094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,32,power_law_1.2,0.05989120006561279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,32,power_law_1.2,0.0607807993888855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,32,power_law_1.2,0.06670079827308655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,32,power_law_1.2,0.07068799734115601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,32,power_law_1.2,0.07448959946632386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,32,power_law_1.2,0.08181759715080261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,32,power_law_1.2,0.09109119772911071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,32,power_law_1.2,0.10844800472259522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,32,power_law_1.2,0.13244160413742065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,32,power_law_1.2,0.18362879753112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,32,power_law_1.2,0.20347518920898439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,4,power_law_1.01,0.07655680179595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,4,power_law_1.01,0.08511360287666321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,4,power_law_1.01,0.12254079580307006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,4,power_law_1.01,0.12870399951934813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,4,power_law_1.01,0.12920960187911987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,4,power_law_1.01,0.18384000062942504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,4,power_law_1.01,0.18543360233306885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,32,power_law_1.2,0.2993088006973267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,4,power_law_1.01,0.1996415972709656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,4,power_law_1.01,0.19413119554519653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,4,power_law_1.01,0.1952191948890686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,4,power_law_1.01,0.20570878982543944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,4,power_law_1.01,0.20236799716949463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,4,power_law_1.01,0.21068160533905028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,4,power_law_1.01,0.21877760887145997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,4,power_law_1.01,0.2384255886077881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,4,power_law_1.01,0.24555521011352538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,4,power_law_1.01,0.2352895975112915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,4,power_law_1.01,0.26389760971069337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,4,power_law_1.01,0.295033597946167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,4,power_law_1.01,0.347161602973938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,4,power_law_1.01,0.33824639320373534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,4,power_law_1.01,0.41063680648803713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,4,power_law_1.01,0.47178239822387696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,4,power_law_1.01,0.6468031883239747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,4,power_law_1.01,0.7883776187896728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,4,power_law_1.01,1.0271936416625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,4,power_law_1.01,1.2565695762634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,32,power_law_1.2,0.34908161163330076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,32,power_law_1.2,0.49936637878417967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,32,power_law_1.2,0.7116159915924072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,1,power_law_1.01,0.20424959659576417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,1,power_law_1.01,0.33895680904388426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,1,power_law_1.01,0.5969984054565429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,1,power_law_1.01,1.0098496437072755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,1,power_law_1.01,1.3982975959777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,1,power_law_1.01,2.408198356628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,1,power_law_1.01,2.5477888107299806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,1,power_law_1.01,2.6633792877197267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,1,power_law_1.01,2.7122495651245115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,1,power_law_1.01,2.732313537597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,1,power_law_1.01,2.847417640686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,1,power_law_1.01,2.8705215454101562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,1,power_law_1.01,2.9677568435668946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,1,power_law_1.01,3.0563968658447265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,1,power_law_1.01,3.150060844421387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,1,power_law_1.01,3.2134078979492187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,1,power_law_1.01,3.3534015655517577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,1,power_law_1.01,3.6527809143066405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,1,power_law_1.01,3.878092956542969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,1,power_law_1.01,4.270771026611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,1,power_law_1.01,4.481753540039063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,1,power_law_1.01,5.327104187011718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,1,power_law_1.01,5.737484741210937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,1,power_law_1.01,7.012947082519531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,1,power_law_1.01,9.764109039306641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,1,power_law_1.01,12.0080322265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,1,power_law_1.01,13.050341796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,1,power_law_1.01,0.04986880123615265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,1,power_law_1.01,0.056531202793121335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,1,power_law_1.01,0.06652799844741822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,1,power_law_1.01,0.09178879857063293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,1,power_law_1.01,0.09098880290985108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,1,power_law_1.01,0.09169920086860657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,1,power_law_1.01,0.09330559968948364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,1,power_law_1.01,0.09418879747390747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,1,power_law_1.01,0.09430400133132935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,1,power_law_1.01,0.09464960098266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,1,power_law_1.01,0.09609599709510804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,1,power_law_1.01,0.09872639775276185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,1,power_law_1.01,0.10911999940872193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,1,power_law_1.01,0.11062400341033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,1,power_law_1.01,0.12694400548934937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,1,power_law_1.01,0.13114240169525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,1,power_law_1.01,0.134662401676178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,1,power_law_1.01,0.16005120277404786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.01,0.1900480031967163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.01,0.25639679431915285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.01,0.3180351972579956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.01,0.42534399032592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.01,0.5482367992401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,2,power_law_1.2,0.07057920098304749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,2,power_law_1.2,0.09676799774169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,2,power_law_1.2,0.11345920562744141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,2,power_law_1.2,0.16532479524612426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,2,power_law_1.2,0.19431040287017823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,2,power_law_1.2,0.25911679267883303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,2,power_law_1.2,0.3541055917739868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,2,power_law_1.2,0.3722304105758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,2,power_law_1.2,0.37218561172485354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,2,power_law_1.2,0.39166719913482667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,2,power_law_1.2,0.3836159944534302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,2,power_law_1.2,0.3979840040206909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,2,power_law_1.2,0.41350398063659666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,2,power_law_1.2,0.43317761421203616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.01,0.7748479843139648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,2,power_law_1.2,0.43953919410705566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,2,power_law_1.2,0.4584640026092529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,2,power_law_1.2,0.4927999973297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,2,power_law_1.2,0.5481855869293213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,2,power_law_1.2,0.5760191917419434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,2,power_law_1.2,0.6845568180084228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,2,power_law_1.2,0.7126463890075684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,2,power_law_1.2,0.8561216354370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,2,power_law_1.2,1.0207167625427247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,2,power_law_1.2,1.3449088096618653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,2,power_law_1.2,1.5776896476745605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,2,power_law_1.2,2.1624832153320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,2,power_law_1.2,2.86246395111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.01,1.0137920379638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.01,1.451136016845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.01,1.9334592819213867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,1,balanced,0.0210506667693456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,1,balanced,0.022986667851607006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,1,balanced,0.026730666557947796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,1,balanced,0.03995199998219808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,1,balanced,0.062181333700815834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,1,balanced,0.10028266906738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,1,balanced,0.10248532891273499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,1,balanced,0.1051573355992635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,1,balanced,0.10764799515406291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,1,balanced,0.11126933495203654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,1,balanced,0.11061333616574605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,1,balanced,0.11373333136240642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,1,balanced,0.11939199765523274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,1,balanced,0.12157332897186279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,1,balanced,0.1283253331979116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,1,balanced,0.1339040001233419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,1,balanced,0.13526399930318198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,1,balanced,0.15223466356595358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,1,balanced,0.16983999808629355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,1,balanced,0.2270080049832662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,1,balanced,0.22327999273935953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,1,balanced,0.2976106603940328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,1,balanced,0.2714293400446574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,1,balanced,0.36559466520945233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,1,balanced,0.4599306583404541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,1,balanced,0.6566773255666097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,1,balanced,0.8620800177256266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,32,2,power_law_1.2,0.04983679950237274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,32,2,power_law_1.2,0.05317760109901428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,32,2,power_law_1.2,0.054579198360443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,32,2,power_law_1.2,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,32,2,power_law_1.2,0.07144960165023803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,32,2,power_law_1.2,0.08476799726486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,32,2,power_law_1.2,0.10222079753875732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,32,2,power_law_1.2,0.10092799663543701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,32,2,power_law_1.2,0.10316799879074097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,32,2,power_law_1.2,0.10497920513153076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,32,2,power_law_1.2,0.10867199897766114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,32,2,power_law_1.2,0.11576319932937622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,32,2,power_law_1.2,0.12129919528961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,32,2,power_law_1.2,0.13338240385055541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,32,2,power_law_1.2,0.1414463996887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,32,2,power_law_1.2,0.15202560424804687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,32,2,power_law_1.2,0.14863359928131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,32,2,power_law_1.2,0.17926399707794188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,32,2,power_law_1.2,0.20262401103973388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,32,2,power_law_1.2,0.2540287971496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,32,2,power_law_1.2,0.3038912057876587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,32,2,power_law_1.2,0.4039936065673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,32,2,power_law_1.2,0.521068811416626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,32,2,power_law_1.2,0.7099264144897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,32,2,power_law_1.2,0.9132927894592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,32,2,power_law_1.2,1.3349568367004394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,32,2,power_law_1.2,1.661414337158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,4,power_law_1.2,0.05731840133666992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,4,power_law_1.2,0.06310399770736694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,4,power_law_1.2,0.06572800278663635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,4,power_law_1.2,0.07505279779434204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,4,power_law_1.2,0.0803264021873474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,4,power_law_1.2,0.08943359851837158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,4,power_law_1.2,0.1079103946685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,4,power_law_1.2,0.10266239643096924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,4,power_law_1.2,0.1008255958557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,4,power_law_1.2,0.10090880393981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,4,power_law_1.2,0.1051584005355835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,4,power_law_1.2,0.11400320529937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,4,power_law_1.2,0.11117440462112427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,4,power_law_1.2,0.11960320472717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,4,power_law_1.2,0.1256127953529358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,4,power_law_1.2,0.12942080497741698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,4,power_law_1.2,0.13642879724502563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,4,power_law_1.2,0.15749119520187377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.2,0.17443840503692626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.2,0.20476160049438477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.2,0.23337600231170655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.2,0.3047231912612915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.2,0.4126336097717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.2,0.5684288024902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.2,0.6865087985992432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.2,0.9937727928161622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.2,1.375980758666992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,16,power_law_1.01,0.02227199971675873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,16,power_law_1.01,0.02818560004234314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,16,power_law_1.01,0.027718400955200194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,16,power_law_1.01,0.02839680016040802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,16,power_law_1.01,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,16,power_law_1.01,0.030559998750686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,16,power_law_1.01,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,16,power_law_1.01,0.03377279937267304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,16,power_law_1.01,0.048127999901771544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,16,power_law_1.01,0.04776319861412048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,16,power_law_1.01,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,16,power_law_1.01,0.06602879762649536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,16,power_law_1.01,0.0671999990940094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,16,power_law_1.01,0.06862080097198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,16,power_law_1.01,0.07592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,16,power_law_1.01,0.07761920094490052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,16,power_law_1.01,0.07553279995918274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,16,power_law_1.01,0.0812287986278534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.01,0.08542079925537109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.01,0.09925119876861573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.01,0.10055040121078491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.01,0.12168320417404174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.01,0.14905600547790526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.01,0.2022655963897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.01,0.25384960174560545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.01,0.354367995262146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.01,0.4542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,4,balanced,0.04572799801826477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,4,balanced,0.05078400174776713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,4,balanced,0.04827733337879181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,4,balanced,0.054154664278030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,4,balanced,0.06038400034109751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,4,balanced,0.0888320008913676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,4,balanced,0.09027199943860371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,4,balanced,0.08898666501045227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,4,balanced,0.09089600046475728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,4,balanced,0.09027199943860371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,4,balanced,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,4,balanced,0.09516800443331401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,4,balanced,0.09579199552536011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,4,balanced,0.0983840028444926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,4,balanced,0.10598400235176086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,4,balanced,0.1076639990011851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,4,balanced,0.11892799536387126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,4,balanced,0.1360640029112498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,4,balanced,0.1449013352394104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,4,balanced,0.17279465993245444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,4,balanced,0.1978293259938558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,4,balanced,0.26772799094518024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,4,balanced,0.31650133927663165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,4,balanced,0.4526826540629069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,4,balanced,0.5697386662165324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,4,balanced,0.822709321975708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,4,balanced,1.0676906903584797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,32,power_law_1.2,0.06702719926834107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,32,power_law_1.2,0.04620159864425659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,32,power_law_1.2,0.04679040014743805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,32,power_law_1.2,0.050521600246429446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,32,power_law_1.2,0.049267199635505673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,32,power_law_1.2,0.051545602083206174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,32,power_law_1.2,0.053388798236846925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,32,power_law_1.2,0.05516800284385681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,32,power_law_1.2,0.05425919890403748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,32,power_law_1.2,0.05509759783744812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,32,power_law_1.2,0.057683199644088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,32,power_law_1.2,0.057868802547454835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,32,power_law_1.2,0.06042879819869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,32,power_law_1.2,0.06277120113372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,32,power_law_1.2,0.06826879978179931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,32,power_law_1.2,0.07183359861373902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,32,power_law_1.2,0.07764480113983155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,32,power_law_1.2,0.0923200011253357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,32,power_law_1.2,0.09681280255317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,32,power_law_1.2,0.12051199674606324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,32,power_law_1.2,0.13292800188064574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,32,power_law_1.2,0.17384320497512817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,32,power_law_1.2,0.20936961174011232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,32,power_law_1.2,0.30743041038513186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,32,power_law_1.2,0.4017471790313721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,32,power_law_1.2,0.5365952014923095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,32,power_law_1.2,0.5939839839935303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,1,power_law_1.01,0.08766720294952393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,1,power_law_1.01,0.09373440146446228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,1,power_law_1.01,0.10375679731369018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,1,power_law_1.01,0.13282560110092162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,1,power_law_1.01,0.15511679649353027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,1,power_law_1.01,0.17904640436172486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,1,power_law_1.01,0.21626880168914794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,1,power_law_1.01,0.22121601104736327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,1,power_law_1.01,0.2332223892211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,1,power_law_1.01,0.23271679878234863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,1,power_law_1.01,0.23726720809936525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,1,power_law_1.01,0.24669439792633058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,1,power_law_1.01,0.24876160621643068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,1,power_law_1.01,0.2568703889846802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,1,power_law_1.01,0.2660799980163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,1,power_law_1.01,0.2716928005218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,1,power_law_1.01,0.2908351898193359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,1,power_law_1.01,0.32721920013427735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,1,power_law_1.01,0.36474881172180174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,1,power_law_1.01,0.43697919845581057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,1,power_law_1.01,0.5087615966796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,1,power_law_1.01,0.6566463947296143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,1,power_law_1.01,0.8042880058288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,1,power_law_1.01,1.093932819366455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,1,power_law_1.01,1.386297607421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,1,power_law_1.01,1.9737472534179688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,1,power_law_1.01,2.5511232376098634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,32,balanced,0.05428266525268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,32,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,32,balanced,0.04301333427429199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,32,balanced,0.04609066744645437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,32,balanced,0.04685866832733154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,32,balanced,0.04970133304595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,32,balanced,0.05000533163547516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,32,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,32,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,32,balanced,0.05189866820971171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,32,balanced,0.051813334226608276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,32,balanced,0.05170666674772898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,32,balanced,0.05226133267084757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,32,balanced,0.05630399783452352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,32,balanced,0.0561706672112147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,32,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,32,balanced,0.0625600020090739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,32,balanced,0.06663466493288676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,32,balanced,0.0666240006685257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,32,balanced,0.0796319991350174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,32,balanced,0.09129599730173747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,32,balanced,0.11820800105730693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,32,balanced,0.12814399600028992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,32,balanced,0.16944533586502075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,32,balanced,0.18210132916768393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,32,balanced,0.2582613428433736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,32,balanced,0.3046506643295288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,1,power_law_1.01,0.07514879703521729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,1,power_law_1.01,0.09934719800949096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,1,power_law_1.01,0.13763840198516847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,1,power_law_1.01,0.22164480686187743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,1,power_law_1.01,0.3193599939346313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,1,power_law_1.01,0.40687360763549807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,1,power_law_1.01,0.5505727767944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,1,power_law_1.01,0.5737919807434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,1,power_law_1.01,0.5921599864959717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,1,power_law_1.01,0.6036992073059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,1,power_law_1.01,0.622873592376709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,1,power_law_1.01,0.6569600105285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,1,power_law_1.01,0.6708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,1,power_law_1.01,0.6824192047119141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,1,power_law_1.01,0.7050752162933349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,1,power_law_1.01,0.7271999835968017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,1,power_law_1.01,0.7670144081115723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,1,power_law_1.01,0.8334015846252442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,1,power_law_1.01,0.8796863555908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,1,power_law_1.01,1.0106047630310058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,1,power_law_1.01,1.065004825592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,1,power_law_1.01,1.2929471969604491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,1,power_law_1.01,1.5089792251586913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,1,power_law_1.01,1.9640575408935548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,1,power_law_1.01,2.381977653503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,1,power_law_1.01,3.2382015228271483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,1,power_law_1.01,4.1180671691894535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,4,balanced,0.033813332517941795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,4,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,4,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,4,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,4,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,4,balanced,0.037674665451049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,4,balanced,0.03931200007597605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,4,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,4,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,4,balanced,0.039359999199708305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,4,balanced,0.03953066716591517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,4,balanced,0.04196266829967499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,4,balanced,0.04147200038035711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,4,balanced,0.04190400242805481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,4,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,4,balanced,0.04957866668701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,4,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,4,balanced,0.0666720022757848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,4,balanced,0.06948266426722209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,4,balanced,0.088319996992747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,4,balanced,0.09464533130327861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,4,balanced,0.123690664768219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,4,balanced,0.16098666191101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,4,balanced,0.20944533745447794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,4,balanced,0.2707680066426595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,4,balanced,0.40334399541219074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,4,balanced,0.5543520053227743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,128,power_law_1.01,0.03278079926967621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,128,power_law_1.01,0.03139199912548065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,128,power_law_1.01,0.030924800038337707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,128,power_law_1.01,0.033529600501060484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,128,power_law_1.01,0.03311359882354736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,128,power_law_1.01,0.034969601035118106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,128,power_law_1.01,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,128,power_law_1.01,0.03468160033226013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,128,power_law_1.01,0.03516800105571747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,128,power_law_1.01,0.035148799419403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,128,power_law_1.01,0.03608959913253784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,128,power_law_1.01,0.03637759983539581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,128,power_law_1.01,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,128,power_law_1.01,0.03738240003585815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,128,power_law_1.01,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,128,power_law_1.01,0.03944959938526153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,128,power_law_1.01,0.04081920087337494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,128,power_law_1.01,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,128,power_law_1.01,0.04577920138835907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,16,power_law_1.01,0.040191999077796935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,16,power_law_1.01,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,16,power_law_1.01,0.032543998956680295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,16,power_law_1.01,0.03445119857788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,16,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,16,power_law_1.01,0.03488639891147614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,16,power_law_1.01,0.035795199871063235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,16,power_law_1.01,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,16,power_law_1.01,0.036006399989128114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,16,power_law_1.01,0.03597440123558045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,16,power_law_1.01,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,16,power_law_1.01,0.037862399220466615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,16,power_law_1.01,0.04094719886779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,16,power_law_1.01,0.04227840006351471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,128,power_law_1.01,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,16,power_law_1.01,0.048403200507164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,128,power_law_1.01,0.05549439787864685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,16,power_law_1.01,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,128,power_law_1.01,0.06566399931907654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,16,power_law_1.01,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,128,power_law_1.01,0.07477120161056519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,16,power_law_1.01,0.05610880255699158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,128,power_law_1.01,0.10088959932327271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,16,power_law_1.01,0.060140800476074216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,128,power_law_1.01,0.11471359729766846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,16,power_law_1.01,0.0709119975566864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,128,power_law_1.01,0.1568384051322937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,16,power_law_1.01,0.0712768018245697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,16,power_law_1.01,0.08379520177841186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,16,power_law_1.01,0.10028159618377686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,16,power_law_1.01,0.11644799709320068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,16,power_law_1.01,0.15650559663772584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,16,power_law_1.01,0.26384639739990234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,16,power_law_1.01,0.2532991886138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,128,power_law_1.01,0.20258560180664062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,1,power_law_1.01,0.016627199947834015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,1,power_law_1.01,0.01823360025882721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,1,power_law_1.01,0.02220800071954727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,1,power_law_1.01,0.025900799036026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,1,power_law_1.01,0.029440000653266907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,1,power_law_1.01,0.03841919898986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,1,power_law_1.01,0.040217599272727965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,1,power_law_1.01,0.042847999930381776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,1,power_law_1.01,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,1,power_law_1.01,0.04423680007457733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,1,power_law_1.01,0.047513601183891294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,1,power_law_1.01,0.04796159863471985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,1,power_law_1.01,0.04912000000476837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,1,power_law_1.01,0.05258880257606506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,1,power_law_1.01,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,1,power_law_1.01,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,1,power_law_1.01,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,1,power_law_1.01,0.06079999804496765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,1,power_law_1.01,0.06645119786262513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,1,power_law_1.01,0.1197119951248169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,1,power_law_1.01,0.1386623978614807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,1,power_law_1.01,0.12906240224838256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,1,power_law_1.01,0.1450111985206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,1,power_law_1.01,0.18464640378952027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,1,power_law_1.01,0.22640640735626222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,1,power_law_1.01,0.30427520275115966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,1,power_law_1.01,0.3878720045089722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,2,power_law_1.2,0.031999999284744264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,2,power_law_1.2,0.032671999931335446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,2,power_law_1.2,0.03780480027198792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,2,power_law_1.2,0.044563201069831845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,2,power_law_1.2,0.04997760057449341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,2,power_law_1.2,0.07018880248069763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,2,power_law_1.2,0.0738752007484436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,2,power_law_1.2,0.07648640275001525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,2,power_law_1.2,0.07837439775466919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,2,power_law_1.2,0.07997440099716187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,2,power_law_1.2,0.08323839902877808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,2,power_law_1.2,0.08370559811592101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,2,power_law_1.2,0.0849727988243103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,2,power_law_1.2,0.08730239868164062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,2,power_law_1.2,0.0959231972694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,2,power_law_1.2,0.09932159781455993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,2,power_law_1.2,0.10178560018539429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,2,power_law_1.2,0.11486079692840576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,2,power_law_1.2,0.12407040596008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,2,power_law_1.2,0.1408128023147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,2,power_law_1.2,0.15882879495620728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,2,power_law_1.2,0.2008960008621216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,2,power_law_1.2,0.2622783899307251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,2,power_law_1.2,0.3642879962921143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,2,power_law_1.2,0.42122879028320315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,2,power_law_1.2,0.5980544090270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,8,power_law_1.2,0.0752128005027771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,8,power_law_1.2,0.10778239965438843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,8,power_law_1.2,0.10385279655456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,8,power_law_1.2,0.13045120239257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,8,power_law_1.2,0.13543039560317993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,8,power_law_1.2,0.1438591957092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,8,power_law_1.2,0.1427072048187256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,2,power_law_1.2,0.7754752159118652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,8,power_law_1.2,0.16003199815750122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,8,power_law_1.2,0.1608191967010498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,8,power_law_1.2,0.15588480234146118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,8,power_law_1.2,0.1578112006187439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,8,power_law_1.2,0.16647039651870726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,8,power_law_1.2,0.16832000017166138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,8,power_law_1.2,0.16803840398788453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,8,power_law_1.2,0.17182719707489014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,8,power_law_1.2,0.17230720520019532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,8,power_law_1.2,0.19871360063552856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,8,power_law_1.2,0.21762559413909913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,8,power_law_1.2,0.2163520097732544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,8,power_law_1.2,0.26725120544433595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,8,power_law_1.2,0.27616000175476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,8,power_law_1.2,0.3384704113006592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,8,power_law_1.2,0.36386559009552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,8,power_law_1.2,0.4748991966247559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,8,power_law_1.2,0.5476928234100342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,8,power_law_1.2,0.7146240234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,8,power_law_1.2,1.0076864242553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,1,power_law_1.01,0.018163199722766876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,1,power_law_1.01,0.019398400187492372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,1,power_law_1.01,0.022950400412082673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,1,power_law_1.01,0.02810240089893341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,1,power_law_1.01,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,1,power_law_1.01,0.044659200310707095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,1,power_law_1.01,0.0470335990190506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,1,power_law_1.01,0.04922240078449249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,1,power_law_1.01,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,1,power_law_1.01,0.05395200252532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,1,power_law_1.01,0.05655679702758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,1,power_law_1.01,0.05679360032081604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,1,power_law_1.01,0.06033279895782471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,1,power_law_1.01,0.06626560091972351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,1,power_law_1.01,0.06236799955368042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,1,power_law_1.01,0.06507520079612732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,1,power_law_1.01,0.07192320227622986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,1,power_law_1.01,0.08312320113182067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.01,0.0944703996181488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.01,0.15909759998321532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.01,0.18081920146942138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.01,0.1719488024711609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.01,0.1928447961807251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.01,0.24396800994873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.01,0.30487680435180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.01,0.4141439914703369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.01,0.5274752140045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,8,power_law_1.01,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,8,power_law_1.01,0.03903999924659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,8,power_law_1.01,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,8,power_law_1.01,0.0377920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,8,power_law_1.01,0.03771519958972931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,8,power_law_1.01,0.037222400307655334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,8,power_law_1.01,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,8,power_law_1.01,0.03848319947719574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,8,power_law_1.01,0.03880960047245026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,8,power_law_1.01,0.040249601006507874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,8,power_law_1.01,0.041356799006462094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,8,power_law_1.01,0.04172160029411316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,8,power_law_1.01,0.044627198576927186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,8,power_law_1.01,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,8,power_law_1.01,0.05283839702606201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,8,power_law_1.01,0.05550079941749573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,8,power_law_1.01,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,8,power_law_1.01,0.06262400150299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,8,power_law_1.01,0.06465920209884643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,8,power_law_1.01,0.07747200131416321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,8,power_law_1.01,0.08392959833145142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,8,power_law_1.01,0.10380159616470337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,8,power_law_1.01,0.12329599857330323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,8,power_law_1.01,0.16596479415893556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,8,power_law_1.01,0.19460480213165282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,8,power_law_1.01,0.27474560737609866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,8,power_law_1.01,0.3214207887649536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,8,power_law_1.2,0.031590399146080014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,8,power_law_1.2,0.03343999981880188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,8,power_law_1.2,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,8,power_law_1.2,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,8,power_law_1.2,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,32,power_law_1.2,0.01640319973230362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,8,power_law_1.2,0.037599998712539676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,8,power_law_1.2,0.037459200620651244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,8,power_law_1.2,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,8,power_law_1.2,0.0392767995595932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,8,power_law_1.2,0.040300801396369934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,8,power_law_1.2,0.04126079976558685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,8,power_law_1.2,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,8,power_law_1.2,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,8,power_law_1.2,0.04715520143508911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,32,power_law_1.2,0.01578879952430725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,8,power_law_1.2,0.05364480018615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,8,power_law_1.2,0.05618559718132019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,8,power_law_1.2,0.05816320180892944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,8,power_law_1.2,0.06473600268363952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,8,power_law_1.2,0.06773120164871216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,8,power_law_1.2,0.08440319895744323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,8,power_law_1.2,0.09057919979095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,8,power_law_1.2,0.10789120197296143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,32,power_law_1.2,0.016064000129699708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,8,power_law_1.2,0.13400959968566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,32,power_law_1.2,0.017100800573825837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,8,power_law_1.2,0.16817920207977294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,32,power_law_1.2,0.017612800002098083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,32,power_law_1.2,0.01998720020055771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,8,power_law_1.2,0.22353920936584473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,32,power_law_1.2,0.031488001346588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,8,power_law_1.2,0.29295361042022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,32,power_law_1.2,0.031276801228523256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,32,power_law_1.2,0.02826879918575287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,32,power_law_1.2,0.02826879918575287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,32,power_law_1.2,0.027980801463127137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,2,balanced,0.03048533449570338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,2,balanced,0.0360959991812706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,32,power_law_1.2,0.028543999791145323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,32,power_law_1.2,0.028409600257873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,32,power_law_1.2,0.02800000011920929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,32,power_law_1.2,0.028883200883865357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,32,power_law_1.2,0.02922239899635315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,32,power_law_1.2,0.030124801397323608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,32,power_law_1.2,0.032601600885391234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,32,power_law_1.2,0.03661440014839172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,2,balanced,0.05006400247414907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,2,balanced,0.07512533167997996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,2,balanced,0.11900267004966736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,2,balanced,0.12177600463231404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,2,balanced,0.1272213359673818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,2,balanced,0.13339199622472128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,2,balanced,0.1299679974714915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,2,balanced,0.1318880021572113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,2,balanced,0.13595199584960938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,2,balanced,0.16234133640925089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,2,balanced,0.15923733512560526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,2,balanced,0.1547040045261383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,2,balanced,0.23944532871246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,2,balanced,0.23291200399398804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,32,power_law_1.2,0.04162560105323791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,32,power_law_1.2,0.04560000002384186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,32,power_law_1.2,0.05467519760131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,2,balanced,0.2162826657295227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,32,power_law_1.2,0.06779519915580749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,2,balanced,0.31943466265996295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,2,balanced,0.3546453317006429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,2,balanced,0.3425439993540446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,2,balanced,0.35133334000905353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,2,balanced,0.3335839907328288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,2,balanced,0.34073599179585773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,2,balanced,0.6106026570002238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,2,balanced,0.59879998366038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,2,balanced,0.8639039993286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,2,balanced,1.1328907012939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,8,power_law_1.2,0.40459518432617186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,32,power_law_1.2,0.0870848000049591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,32,power_law_1.2,0.10602240562438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,32,power_law_1.2,0.14528640508651733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,4,power_law_1.01,0.032025599479675294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,4,power_law_1.01,0.03191039860248566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,4,power_law_1.01,0.03250559866428375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,4,power_law_1.01,0.03470720052719116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,4,power_law_1.01,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,4,power_law_1.01,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,4,power_law_1.01,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,4,power_law_1.01,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,4,power_law_1.01,0.03749760091304779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,4,power_law_1.01,0.03803519904613495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,4,power_law_1.01,0.039577600359916684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,4,power_law_1.01,0.04108799993991852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,4,power_law_1.01,0.04524799883365631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,4,power_law_1.01,0.04650880098342895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,4,power_law_1.01,0.05569919943809509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,4,power_law_1.01,0.059059202671051025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,4,power_law_1.01,0.07310079932212829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,4,power_law_1.01,0.08481919765472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,4,power_law_1.01,0.11143039464950562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,4,power_law_1.01,0.12894079685211182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,32,power_law_1.2,0.1927616000175476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,4,power_law_1.01,0.17182719707489014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,4,power_law_1.01,0.23860480785369872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,4,power_law_1.01,0.3413183927536011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,4,power_law_1.01,0.39592320919036866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,4,power_law_1.01,0.5494783878326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,256,balanced,0.05601066847642263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,256,balanced,0.057999998331069946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,256,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,256,balanced,0.06180266539255778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,256,balanced,0.05820266902446747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,256,balanced,0.06426666676998138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,256,balanced,0.0653546651204427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,256,balanced,0.06234666705131531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,256,balanced,0.06229333579540253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,256,balanced,0.06543466448783875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,256,balanced,0.062496001521746315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,256,balanced,0.06687999765078227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,256,balanced,0.06749866902828217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,256,balanced,0.06612800061702728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,256,balanced,0.07228800157705943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,256,balanced,0.07249600191911061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,256,balanced,0.07532800237337749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,256,balanced,0.0825973351796468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,256,balanced,0.08630399902661641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,256,balanced,0.09703999757766724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,256,balanced,0.10541866223017375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,256,balanced,0.13292266925175986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,256,balanced,0.15897066394488016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,256,balanced,0.20998932917912802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,256,balanced,0.2572159965833028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,256,balanced,0.352234681447347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,256,balanced,0.44513599077860516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,4,power_law_1.01,0.9861632347106933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,16,balanced,0.05226666728655497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,16,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,16,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,16,balanced,0.04783466458320618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,16,balanced,0.047770669062932335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,16,balanced,0.05592533449331919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,16,balanced,0.057536001006762184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,16,balanced,0.05815466741720835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,16,balanced,0.057904000083605446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,16,balanced,0.060234665870666504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,16,balanced,0.06005333364009857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,16,balanced,0.06170133252938589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,16,balanced,0.06201600035031637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,16,balanced,0.06532800197601318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,16,balanced,0.06842666864395142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,16,balanced,0.07135466734568278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,16,balanced,0.07595199843247731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,16,balanced,0.0819893330335617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,16,balanced,0.08889599641164143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,16,balanced,0.10316266616185506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,4,power_law_1.01,1.2077695846557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,16,balanced,0.11205866932868958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,16,balanced,0.13833600282669067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,16,balanced,0.16529066363970438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,16,balanced,0.21192532777786255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,16,balanced,0.2667306661605835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,16,balanced,0.370959997177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,16,balanced,0.45345067977905273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,2,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,2,balanced,0.07658133407433827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,2,balanced,0.10149866342544556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,2,balanced,0.15205867091814676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,2,balanced,0.2505706747372945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,2,balanced,0.4542880058288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,2,balanced,0.6523093382517496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,2,balanced,0.6502666473388672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,2,balanced,0.6564319928487142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,2,balanced,0.6596800088882446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,2,balanced,0.6578613519668579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,2,balanced,0.6648693482081095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,2,balanced,0.6709919770558676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,2,balanced,0.6769013404846191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,2,balanced,0.6889653205871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,2,balanced,0.6942613124847412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,2,balanced,0.7097226778666178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,2,balanced,0.7405333518981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,2,balanced,0.7630826632181803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,2,balanced,0.8259359995524088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,2,balanced,0.8713546593983968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,2,balanced,0.9944693247477213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,2,balanced,1.1001226902008057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,2,balanced,1.317850669225057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,2,balanced,1.626207987467448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,2,balanced,2.1119200388590493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,2,balanced,2.74291197458903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,32,1,power_law_1.2,0.052179199457168576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,32,1,power_law_1.2,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,32,1,power_law_1.2,0.060601598024368285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,32,1,power_law_1.2,0.07779200077056884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,32,1,power_law_1.2,0.09508479833602905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,32,1,power_law_1.2,0.11037440299987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,32,1,power_law_1.2,0.1366528034210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,32,1,power_law_1.2,0.14309120178222656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,32,1,power_law_1.2,0.14892159700393676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,32,1,power_law_1.2,0.1515776038169861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,32,1,power_law_1.2,0.16367360353469848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,32,1,power_law_1.2,0.17507840394973756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,8,power_law_1.2,0.04407680034637451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,32,1,power_law_1.2,0.18442239761352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,8,power_law_1.2,0.04787200093269348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,32,1,power_law_1.2,0.1968384027481079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,8,power_law_1.2,0.04704639911651611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,32,1,power_law_1.2,0.21304960250854493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,8,power_law_1.2,0.051571202278137204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,32,1,power_law_1.2,0.23032960891723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,8,power_law_1.2,0.05568640232086182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,32,1,power_law_1.2,0.22812159061431886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,8,power_law_1.2,0.05780479907989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,32,1,power_law_1.2,0.2774784088134766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,8,power_law_1.2,0.05982720255851746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,32,1,power_law_1.2,0.31731839179992677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,8,power_law_1.2,0.06099200248718262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,32,1,power_law_1.2,0.40067200660705565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,8,power_law_1.2,0.061945599317550656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,32,1,power_law_1.2,0.47431039810180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,8,power_law_1.2,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,32,1,power_law_1.2,0.6336959838867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,8,power_law_1.2,0.066457599401474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,8,power_law_1.2,0.06826879978179931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,32,1,power_law_1.2,0.8039936065673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,8,power_law_1.2,0.07127040028572082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,8,power_law_1.2,0.07638400197029113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,32,1,power_law_1.2,1.1041600227355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,8,power_law_1.2,0.08600320219993592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,8,power_law_1.2,0.08819199800491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,32,1,power_law_1.2,1.4029376029968261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,8,power_law_1.2,0.09374719858169556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,8,power_law_1.2,0.11048959493637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,8,power_law_1.2,0.12222080230712891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,32,1,power_law_1.2,2.0020864486694334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,8,power_law_1.2,0.15122560262680054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,8,power_law_1.2,0.18179839849472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,8,power_law_1.2,0.22933120727539064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,32,1,power_law_1.2,2.6263999938964844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,8,power_law_1.2,0.29526400566101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,8,power_law_1.2,0.4374527931213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,8,power_law_1.2,0.5303232192993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,8,power_law_1.2,0.8268032073974609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,8,power_law_1.2,1.0917951583862304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,128,power_law_1.2,0.05340800285339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,128,power_law_1.2,0.06588159799575806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,128,power_law_1.2,0.04622719883918762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,128,power_law_1.2,0.04580479860305786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,128,power_law_1.2,0.04831359982490539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,128,power_law_1.2,0.04849919974803925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,128,power_law_1.2,0.04927360117435455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,128,power_law_1.2,0.050323200225830075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,128,power_law_1.2,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,128,power_law_1.2,0.05111039876937866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,128,power_law_1.2,0.05795199871063232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,128,power_law_1.2,0.05914880037307739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,128,power_law_1.2,0.06530560255050659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,128,power_law_1.2,0.07014399766921997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,128,power_law_1.2,0.07771520018577575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,128,power_law_1.2,0.08250880241394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,128,power_law_1.2,0.0840448021888733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,128,power_law_1.2,0.10370559692382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,128,power_law_1.2,0.10559999942779541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,128,power_law_1.2,0.13397760391235353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,128,power_law_1.2,0.16280319690704345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,128,power_law_1.2,0.21073920726776124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,128,power_law_1.2,0.24992001056671143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,128,power_law_1.2,0.3499200105667114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,128,power_law_1.2,0.4502272129058838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,128,power_law_1.2,0.6442048072814941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,128,power_law_1.2,0.8986495971679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,8,balanced,0.033514666060606636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,8,balanced,0.03576533248027166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,8,balanced,0.037418665985266365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,8,balanced,0.04399466514587402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,8,balanced,0.051269332567850746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,8,balanced,0.051072001457214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,8,balanced,0.049728001157442726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,8,balanced,0.05100800096988678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,8,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,8,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,8,balanced,0.052000001072883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,8,balanced,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,8,balanced,0.05307200054327647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,8,balanced,0.053114667534828186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,8,balanced,0.05421333511670431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,8,balanced,0.05816000203291575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,8,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,8,balanced,0.062405332922935486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,8,balanced,0.06402666866779327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,8,balanced,0.07083199918270111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,8,balanced,0.07643733421961467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,8,balanced,0.0969599982102712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,8,balanced,0.1093280017375946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,8,balanced,0.14678933223088583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,8,balanced,0.18080532550811768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,8,balanced,0.2510666648546855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,8,balanced,0.3173653284708659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,16,power_law_1.01,0.018649600446224213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,16,power_law_1.01,0.019871999323368073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,16,power_law_1.01,0.019808000326156615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,64,power_law_1.2,0.10528000593185424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,16,power_law_1.01,0.01993599981069565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,64,power_law_1.2,0.08923519849777221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,16,power_law_1.01,0.020934399962425233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,64,power_law_1.2,0.08960639834403991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,16,power_law_1.01,0.022547200322151184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,64,power_law_1.2,0.09103999733924865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,16,power_law_1.01,0.024672000110149382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,64,power_law_1.2,0.09230719804763794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,16,power_law_1.01,0.024377599358558655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,64,power_law_1.2,0.0881663978099823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,64,power_law_1.2,0.09751679897308349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,64,power_law_1.2,0.09374719858169556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,64,power_law_1.2,0.09715200066566468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,16,power_law_1.01,0.02922239899635315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,64,power_law_1.2,0.089273601770401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,16,power_law_1.01,0.028883200883865357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,64,power_law_1.2,0.09746559858322143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,16,power_law_1.01,0.028518399596214293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,16,power_law_1.01,0.060089600086212155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,64,power_law_1.2,0.09870719909667969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,16,power_law_1.01,0.04177919924259186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,16,power_law_1.01,0.059622400999069215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,64,power_law_1.2,0.10504319667816162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,16,power_law_1.01,0.041631999611854556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,16,power_law_1.01,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,64,power_law_1.2,0.09787520170211791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,16,power_law_1.01,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,64,power_law_1.2,0.1118656039237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,64,power_law_1.2,0.11603840589523315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,16,power_law_1.01,0.04348799884319306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,64,power_law_1.2,0.12345600128173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,16,power_law_1.01,0.04535039961338043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,64,power_law_1.2,0.1568384051322937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,16,power_law_1.01,0.04570879936218262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,16,power_law_1.01,0.04747520089149475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,16,power_law_1.01,0.04805119931697845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,16,power_law_1.01,0.057651197910308837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,16,power_law_1.01,0.050419199466705325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,16,power_law_1.01,0.05675519704818725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,16,power_law_1.01,0.05935360193252563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,16,power_law_1.01,0.061561602354049685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,64,power_law_1.2,0.15333759784698486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,16,power_law_1.01,0.05875200033187866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,64,power_law_1.2,0.1939903974533081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,16,power_law_1.01,0.06312320232391358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,16,power_law_1.01,0.06111360192298889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,16,power_law_1.01,0.06272000074386597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,16,power_law_1.01,0.07392640113830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,16,power_law_1.01,0.07214720249176025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,16,power_law_1.01,0.07226240038871765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,16,power_law_1.01,0.07807999849319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,16,power_law_1.01,0.08081279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,16,power_law_1.01,0.12541439533233642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,16,power_law_1.01,0.10561920404434204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,64,power_law_1.2,0.2121216058731079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,16,power_law_1.01,0.1348031997680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,64,power_law_1.2,0.2667840003967285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,16,power_law_1.01,0.13694080114364623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,16,power_law_1.01,0.15480320453643798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,16,power_law_1.01,0.18401919603347777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.01,0.22663679122924804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.01,0.31550719738006594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.01,0.1569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,16,power_law_1.01,0.12799999713897706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.01,0.21627519130706788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,16,power_law_1.01,0.17200640439987183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.01,0.25395200252532957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,16,power_law_1.01,0.2222208023071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.01,0.36701440811157227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.01,0.4784832000732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.01,0.7193727970123291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.01,0.8902336120605469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,64,power_law_1.2,0.31308159828186033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,64,power_law_1.2,0.42878079414367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,64,power_law_1.2,0.5562560081481933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,64,power_law_1.2,0.8280192375183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,64,power_law_1.2,1.0808575630187989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,32,power_law_1.01,0.1060096025466919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,32,power_law_1.01,0.10277119874954224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,32,power_law_1.01,0.10238080024719239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,32,power_law_1.01,0.09290879964828491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,32,power_law_1.01,0.0932096004486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,32,power_law_1.01,0.10267519950866699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,32,power_law_1.01,0.10346879959106445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,32,power_law_1.01,0.10899200439453124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,32,power_law_1.01,0.10323840379714966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,32,power_law_1.01,0.10677119493484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,32,power_law_1.01,0.10498559474945068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,32,power_law_1.01,0.10718719959259033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,32,power_law_1.01,0.10645760297775268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,32,power_law_1.01,0.10343040227890014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,32,power_law_1.01,0.11136640310287475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,32,power_law_1.01,0.1110975980758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,32,power_law_1.01,0.11514879465103149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,32,power_law_1.01,0.13362560272216797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.01,0.14704639911651612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.01,0.17957760095596315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.01,0.1941375970840454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.01,0.27224318981170653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.01,0.3278656005859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.01,0.4118207931518555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.01,0.5354623794555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.01,0.7290431976318359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.01,0.8835200309753418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,1,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,1,power_law_1.2,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,1,power_law_1.2,0.0648256003856659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,1,power_law_1.2,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,1,power_law_1.2,0.08440319895744323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,1,power_law_1.2,0.10828160047531128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,1,power_law_1.2,0.11393920183181763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,1,power_law_1.2,0.1158911943435669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,1,power_law_1.2,0.11630719900131226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,1,power_law_1.2,0.11643520593643189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,1,power_law_1.2,0.12208640575408936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,1,power_law_1.2,0.12372479438781739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,1,power_law_1.2,0.12667520046234132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,1,power_law_1.2,0.13367040157318116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,1,power_law_1.2,0.13998719453811645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,1,power_law_1.2,0.1469696044921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,1,power_law_1.2,0.15936000347137452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,1,power_law_1.2,0.18599679470062255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,1,power_law_1.2,0.21422080993652343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,1,power_law_1.2,0.26743040084838865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,1,power_law_1.2,0.3180095911026001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,1,power_law_1.2,0.431276798248291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,1,power_law_1.2,0.5418047904968262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,1,power_law_1.2,0.7625728130340577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,1,power_law_1.2,0.9823936462402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,1,power_law_1.2,1.4204416275024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,1,power_law_1.2,1.8499200820922852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,1,power_law_1.2,0.08697599768638611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,1,power_law_1.2,0.10720000267028809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,1,power_law_1.2,0.13523839712142943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,1,power_law_1.2,0.1707327961921692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,1,power_law_1.2,0.20500481128692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,1,power_law_1.2,0.2813184022903442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,1,power_law_1.2,0.30629758834838866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,1,power_law_1.2,0.30290560722351073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,1,power_law_1.2,0.3123840093612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,1,power_law_1.2,0.31617279052734376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,1,power_law_1.2,0.3282047986984253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,1,power_law_1.2,0.32907519340515134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,1,power_law_1.2,0.3404223918914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,1,power_law_1.2,0.33900160789489747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,1,power_law_1.2,0.35611519813537595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,1,power_law_1.2,0.36438400745391847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,1,power_law_1.2,0.3884864091873169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,1,power_law_1.2,0.44382081031799314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.2,0.4817471981048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.2,0.5785280227661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.2,0.666649580001831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.2,0.8662079811096192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.2,1.051795196533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.2,1.436684799194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.2,1.8255359649658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.2,2.585011291503906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.2,3.3313087463378905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,1,power_law_1.2,0.07579519748687744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,1,power_law_1.2,0.11753599643707276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,1,power_law_1.2,0.17985919713974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,1,power_law_1.2,0.26483840942382814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,1,power_law_1.2,0.347763204574585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,1,power_law_1.2,0.5938176155090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,1,power_law_1.2,0.6366655826568604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,1,power_law_1.2,0.6536896228790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,1,power_law_1.2,0.6674560070037842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,1,power_law_1.2,0.6866943836212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,1,power_law_1.2,0.698579216003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,1,power_law_1.2,0.6979584217071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,1,power_law_1.2,0.713427209854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,1,power_law_1.2,0.7269055843353271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,1,power_law_1.2,0.7805312156677247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,1,power_law_1.2,0.782374382019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,1,power_law_1.2,0.8038080215454102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,1,power_law_1.2,0.8787136077880859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,1,power_law_1.2,0.9490943908691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,1,power_law_1.2,1.0681407928466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,1,power_law_1.2,1.1294400215148925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,1,power_law_1.2,1.3183615684509278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,1,power_law_1.2,1.524550437927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,1,power_law_1.2,1.9126848220825194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,1,power_law_1.2,2.2838783264160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,1,power_law_1.2,3.0429248809814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,4,power_law_1.01,0.022617599368095397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,4,power_law_1.01,0.0281792014837265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,4,power_law_1.01,0.041152000427246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,4,power_law_1.01,0.051712000370025636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,4,power_law_1.01,0.0591871976852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,4,power_law_1.01,0.06401280164718628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,4,power_law_1.01,0.07263360023498536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,4,power_law_1.01,0.0759872019290924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,4,power_law_1.01,0.07585279941558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,4,power_law_1.01,0.07932800054550171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,4,power_law_1.01,0.07961599826812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,4,power_law_1.01,0.08193280100822449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,4,power_law_1.01,0.08479999899864196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,4,power_law_1.01,0.08908799886703492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,4,power_law_1.01,0.08991360068321227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,4,power_law_1.01,0.09801599979400635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,4,power_law_1.01,0.09763200283050537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,4,power_law_1.01,0.10828800201416015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.01,0.1196671962738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.01,0.14527360200881959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.01,0.18945280313491822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.01,0.21891839504241944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.01,0.2550719976425171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.01,0.3196608066558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,1,power_law_1.2,4.0876415252685545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.01,0.37937920093536376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.01,0.525548791885376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.01,0.7025216102600098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,64,balanced,0.021327999730904896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,64,balanced,0.02124800036350886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,64,balanced,0.021269333859284718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,64,balanced,0.021002667645613354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,64,balanced,0.02311466634273529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,64,balanced,0.026133333643277485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,64,balanced,0.04837333162625631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,64,balanced,0.049786667029062905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,64,balanced,0.043968002001444496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,64,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,64,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,64,balanced,0.043925335009892784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,64,balanced,0.041696002086003624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,64,balanced,0.04123199979464213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,64,balanced,0.05179200073083242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,64,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,64,balanced,0.04599999884764353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,64,balanced,0.050554667909940086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,64,balanced,0.059936001896858215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,64,balanced,0.08272533118724823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,64,balanced,0.09326933821042378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,64,balanced,0.1293226679166158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,64,balanced,0.16238933801651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,64,balanced,0.230186661084493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,8,power_law_1.01,0.01854719966650009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,8,power_law_1.01,0.020531199872493744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,8,power_law_1.01,0.02192640006542206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,8,power_law_1.01,0.022950400412082673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,8,power_law_1.01,0.02417919933795929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,8,power_law_1.01,0.023545600473880768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,8,power_law_1.01,0.02789120078086853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,64,balanced,0.2888960043589274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,8,power_law_1.01,0.026688000559806822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,64,balanced,0.4215466578801473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,8,power_law_1.01,0.03056640028953552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,64,balanced,0.5534559885660807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,8,power_law_1.01,0.03059839904308319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,8,power_law_1.01,0.030348798632621764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,8,power_law_1.01,0.039705601334571836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,8,power_law_1.01,0.03736959993839264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,8,power_law_1.01,0.0394239991903305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,8,power_law_1.01,0.058323198556900026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,8,power_law_1.01,0.05857920050621033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,8,power_law_1.01,0.058719998598098753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,8,power_law_1.01,0.061766397953033444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,8,power_law_1.01,0.06330879926681518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,8,power_law_1.01,0.0679040014743805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,8,power_law_1.01,0.06972159743309021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,8,power_law_1.01,0.08313599824905396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,8,power_law_1.01,0.09688320159912109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,8,power_law_1.01,0.12078720331192017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,8,power_law_1.01,0.1466431975364685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,8,power_law_1.01,0.20746879577636718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,8,power_law_1.01,0.25680639743804934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,8,balanced,0.043968002001444496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,8,balanced,0.04791999856630961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,8,balanced,0.0625600020090739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,8,balanced,0.08873066306114197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,8,balanced,0.1328159968058268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,8,balanced,0.2172000010808309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,8,balanced,0.21880000829696655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,8,balanced,0.22057066361109415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,8,balanced,0.2216106653213501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,8,balanced,0.22290132443110147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,8,balanced,0.22175999482472739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,8,balanced,0.22099200884501138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,8,balanced,0.2243946592013041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,8,balanced,0.22736533482869467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,8,balanced,0.23060800631841025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,8,balanced,0.2328480084737142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,8,balanced,0.23627199729283652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,8,balanced,0.24853867292404175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,8,balanced,0.2572159965833028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,8,balanced,0.2754133343696594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,8,balanced,0.2950986623764038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,8,balanced,0.33326399326324463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,8,balanced,0.3545759916305542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,8,balanced,0.44096533457438153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,8,balanced,0.49561067422231037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,8,balanced,0.6905706723531088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,8,balanced,0.778719981511434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,8,power_law_1.2,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,8,power_law_1.2,0.07134720087051391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,8,power_law_1.2,0.08328959941864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,8,power_law_1.2,0.0906175971031189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,8,power_law_1.2,0.10528639554977418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,8,power_law_1.2,0.12760319709777831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,8,power_law_1.2,0.124236798286438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,8,power_law_1.2,0.11682560443878173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,8,power_law_1.2,0.11634559631347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,8,power_law_1.2,0.12062079906463623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,8,power_law_1.2,0.12052479982376099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,8,power_law_1.2,0.12564480304718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,8,power_law_1.2,0.1277119994163513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,8,power_law_1.2,0.1324928045272827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,8,power_law_1.2,0.14513920545578002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,8,power_law_1.2,0.14852479696273804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,8,power_law_1.2,0.16090879440307618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,8,power_law_1.2,0.1853376030921936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,8,power_law_1.2,0.20079360008239747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,8,power_law_1.2,0.24520959854125976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,8,power_law_1.2,0.26883840560913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,8,power_law_1.2,0.35211520195007323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,8,power_law_1.2,0.41445121765136717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,8,power_law_1.2,0.5550591945648193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,8,power_law_1.2,0.7086976051330567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,8,power_law_1.2,1.0476415634155274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,8,power_law_1.2,1.3523648262023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,128,power_law_1.2,0.05453439950942993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,128,power_law_1.2,0.05750399827957153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,128,power_law_1.2,0.04844799935817719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,128,power_law_1.2,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,128,power_law_1.2,0.050387197732925416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,128,power_law_1.2,0.05002239942550659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,128,power_law_1.2,0.050918400287628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,128,power_law_1.2,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,128,power_law_1.2,0.05189759731292724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,128,power_law_1.2,0.052179199457168576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,128,power_law_1.2,0.05257599949836731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,128,power_law_1.2,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,128,power_law_1.2,0.05472000241279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,128,power_law_1.2,0.055283200740814206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,128,power_law_1.2,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,128,power_law_1.2,0.06607360243797303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,128,power_law_1.2,0.07404159903526306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,128,power_law_1.2,0.08362879753112792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,128,power_law_1.2,0.09185919761657715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,128,power_law_1.2,0.11107840538024902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,128,power_law_1.2,0.13716479539871215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,128,power_law_1.2,0.1713536024093628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,128,power_law_1.2,0.20168960094451904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,128,power_law_1.2,0.27717759609222414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,128,power_law_1.2,0.3471807956695557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,128,power_law_1.2,0.48497920036315917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,128,power_law_1.2,0.7037504196166993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,1,balanced,0.09907199939092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,1,balanced,0.10707199573516846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,1,balanced,0.12825066844622293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,1,balanced,0.17353065808614096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,1,balanced,0.24684266249338785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,1,balanced,0.39316264788309735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,1,balanced,0.38605864842732746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,1,balanced,0.37889599800109863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,1,balanced,0.38309868176778156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,1,balanced,0.3839679956436157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,1,balanced,0.3877813418706258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,1,balanced,0.3911306858062744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,16,power_law_1.01,0.09029759764671326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,16,power_law_1.01,0.12817280292510985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,16,power_law_1.01,0.12466559410095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,16,power_law_1.01,0.1292672038078308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,16,power_law_1.01,0.1313088059425354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,16,power_law_1.01,0.11907199621200562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,16,power_law_1.01,0.12956160306930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,16,power_law_1.01,0.1279360055923462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,128,balanced,0.039605334401130676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,128,balanced,0.03789866715669632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,128,balanced,0.03978666663169861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,16,power_law_1.01,0.13160959482192994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,128,balanced,0.04060266663630804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,128,balanced,0.042362665136655174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,128,balanced,0.043605332573254905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,16,power_law_1.01,0.12378879785537719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,16,power_law_1.01,0.12718720436096193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,16,power_law_1.01,0.13624320030212403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,1,balanced,0.39324267705281574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,1,balanced,0.39984532197316486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,16,power_law_1.01,0.136627197265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,16,power_law_1.01,0.1439743995666504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,16,power_law_1.01,0.17167359590530396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,128,balanced,0.04182933270931244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,16,power_law_1.01,0.16033920049667358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,16,power_law_1.01,0.14491519927978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,16,power_law_1.01,0.1667456030845642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,16,power_law_1.01,0.1778496026992798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,1,balanced,0.40622933705647785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,1,balanced,0.4118400017420451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,1,balanced,0.4252320130666097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,1,balanced,0.4684000015258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,1,balanced,0.5018080075581869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,1,balanced,0.5736960172653198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,1,balanced,0.6311253309249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,16,power_law_1.01,0.20256640911102294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,1,balanced,0.7882133324941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,16,power_law_1.01,0.21719040870666503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,1,balanced,0.9296373526255289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,16,power_law_1.01,0.2603391885757446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,1,balanced,1.4142452875773113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,16,power_law_1.01,0.3020096063613892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,16,power_law_1.01,0.38847999572753905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,1,balanced,1.7037760416666667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,16,power_law_1.01,0.4937856197357178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,1,balanced,2.4488746325174966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,16,power_law_1.01,0.6384640216827393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,1,balanced,3.198261260986328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,16,power_law_1.01,0.8364928245544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,128,balanced,0.04295999805132548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,128,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,128,balanced,0.043552001317342125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,128,balanced,0.04349866509437561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,128,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,128,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,128,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,128,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,128,balanced,0.04827199876308441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,128,balanced,0.050901333491007485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,128,balanced,0.052005335688591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,128,balanced,0.052426666021347046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,128,balanced,0.05990933378537496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,128,balanced,0.062224000692367554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,128,balanced,0.07889600098133087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,128,balanced,0.10190932949384053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,128,balanced,0.13732266426086426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,128,balanced,0.16335999965667725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,128,balanced,0.22053333123524985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,128,balanced,0.2820533315340678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,8,power_law_1.2,0.04368639886379242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,8,power_law_1.2,0.0472896009683609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,8,power_law_1.2,0.04312320053577423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,8,power_law_1.2,0.04619520008563995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,8,power_law_1.2,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,8,power_law_1.2,0.049414399266242984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,8,power_law_1.2,0.05057920217514038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,8,power_law_1.2,0.04912000000476837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,8,power_law_1.2,0.051123201847076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,8,power_law_1.2,0.05063040256500244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,8,power_law_1.2,0.05144320130348205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,8,power_law_1.2,0.05240319967269898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,8,power_law_1.2,0.05496320128440857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,8,power_law_1.2,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,8,power_law_1.2,0.06288639903068542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,8,power_law_1.2,0.06396160125732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,8,power_law_1.2,0.07004160284996033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,8,power_law_1.2,0.08184959888458251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,8,power_law_1.2,0.08677759766578674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,8,power_law_1.2,0.10477440357208252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,8,power_law_1.2,0.11943039894104004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,8,power_law_1.2,0.15370240211486816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,8,power_law_1.2,0.1872704029083252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,8,power_law_1.2,0.2548543930053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,8,power_law_1.2,0.3159424066543579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,8,power_law_1.2,0.4709887981414795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,8,power_law_1.2,0.5536064147949219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,2,power_law_1.2,0.07624959945678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,2,power_law_1.2,0.10024960041046142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,2,power_law_1.2,0.13008639812469483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,2,power_law_1.2,0.1863103985786438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,2,power_law_1.2,0.2524480104446411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,2,power_law_1.2,0.2960319995880127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,2,power_law_1.2,0.347379207611084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,2,power_law_1.2,0.4458047866821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,2,power_law_1.2,0.4873216152191162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,32,8,power_law_1.2,0.045337599515914914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,2,power_law_1.2,0.4724287986755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,2,power_law_1.2,0.5011775970458985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,2,power_law_1.2,0.5420351982116699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,2,power_law_1.2,0.5362751960754395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,2,power_law_1.2,0.5339968204498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,32,8,power_law_1.2,0.04615679979324341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,2,power_law_1.2,0.5421696186065674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,32,8,power_law_1.2,0.04684160053730011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,32,8,power_law_1.2,0.04962559938430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,32,8,power_law_1.2,0.054995197057724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,32,8,power_law_1.2,0.05913599729537964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,32,8,power_law_1.2,0.057817602157592775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,32,8,power_law_1.2,0.058278399705886844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,32,8,power_law_1.2,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,2,power_law_1.2,0.5789440155029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,32,8,power_law_1.2,0.05940480232238769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,2,power_law_1.2,0.6153408050537109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,32,8,power_law_1.2,0.06107519865036011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,32,8,power_law_1.2,0.06292480230331421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,2,power_law_1.2,0.6567615985870361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,32,8,power_law_1.2,0.06371200084686279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,2,power_law_1.2,0.709222412109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,32,8,power_law_1.2,0.06677119731903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,32,8,power_law_1.2,0.07368320226669312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,2,power_law_1.2,0.8487423896789551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,32,8,power_law_1.2,0.07830399870872498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,2,power_law_1.2,0.880947208404541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,32,8,power_law_1.2,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,32,8,power_law_1.2,0.10181759595870972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,2,power_law_1.2,1.0806464195251464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,32,8,power_law_1.2,0.10750080347061157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,32,8,power_law_1.2,0.1296704053878784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,2,power_law_1.2,1.2080191612243651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,32,8,power_law_1.2,0.15011839866638182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,2,power_law_1.2,1.5580096244812012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,2,power_law_1.2,1.74835205078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,32,8,power_law_1.2,0.19379199743270875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,2,power_law_1.2,2.2805376052856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,32,8,power_law_1.2,0.26547839641571047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,32,8,power_law_1.2,0.37810559272766114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,32,8,power_law_1.2,0.4602560043334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,2,power_law_1.2,2.9624576568603516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,32,8,power_law_1.2,0.6348735809326171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,32,8,power_law_1.2,0.7657087802886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,128,balanced,0.051130667328834534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,128,balanced,0.05421866476535797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,128,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,128,balanced,0.04701866706212362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,128,balanced,0.04558399816354116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,128,balanced,0.04983466863632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,128,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,128,balanced,0.04996799925963084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,128,balanced,0.05012799799442291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,128,balanced,0.05173333485921224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,128,balanced,0.05171200136343638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,128,balanced,0.052671998739242554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,128,balanced,0.0552106648683548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,128,balanced,0.054192001620928444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,128,balanced,0.06025066475073496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,128,balanced,0.06015466650327047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,128,balanced,0.06408533453941345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,128,balanced,0.07252266506354015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,128,balanced,0.07778133451938629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,128,balanced,0.10117333134015401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,128,balanced,0.12166399757067363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,128,balanced,0.15414399902025858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,128,balanced,0.1830986738204956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,128,balanced,0.24657599131266275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,128,balanced,0.3083146611849467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,128,balanced,0.43157867590586346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,128,balanced,0.5445760091145834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,16,power_law_1.01,0.10247039794921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,16,power_law_1.01,0.15359359979629517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,64,power_law_1.01,0.07166720032691956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,16,power_law_1.01,0.14312959909439088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,64,power_law_1.01,0.062118399143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,16,power_law_1.01,0.14617600440979003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,64,power_law_1.01,0.043558400869369504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,16,power_law_1.01,0.145632004737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,64,power_law_1.01,0.04599039852619171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,16,power_law_1.01,0.13676799535751344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,64,power_law_1.01,0.04688639938831329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,16,power_law_1.01,0.1320255994796753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,64,power_law_1.01,0.04730879962444305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,16,power_law_1.01,0.13866879940032958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,16,power_law_1.01,0.13627519607543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,16,power_law_1.01,0.1444416046142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,16,power_law_1.01,0.1373247981071472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,16,power_law_1.01,0.14703359603881835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,16,power_law_1.01,0.15327359437942506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,16,power_law_1.01,0.1501952052116394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,64,power_law_1.01,0.048403200507164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,16,power_law_1.01,0.15896960496902465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,64,power_law_1.01,0.04888319969177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,16,power_law_1.01,0.16840319633483886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,64,power_law_1.01,0.04903680086135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,64,power_law_1.01,0.04901759922504425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,64,power_law_1.01,0.05010560154914856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,64,power_law_1.01,0.05136640071868896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,64,power_law_1.01,0.05539199709892273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,64,power_law_1.01,0.055238401889801024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,64,power_law_1.01,0.06260480284690857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,16,power_law_1.01,0.17176320552825927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,64,power_law_1.01,0.06499840021133423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,64,power_law_1.01,0.06928640007972717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,64,power_law_1.01,0.07952640056610108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,64,power_law_1.01,0.0831167995929718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,16,power_law_1.01,0.20682239532470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,64,power_law_1.01,0.10468480587005616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,16,power_law_1.01,0.21115520000457763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,64,power_law_1.01,0.1133952021598816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,64,power_law_1.01,0.14538880586624145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,64,power_law_1.01,0.1604416012763977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,64,power_law_1.01,0.2044287919998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,64,power_law_1.01,0.26012799739837644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,16,power_law_1.01,0.25055360794067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,64,power_law_1.01,0.3693056106567383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,16,power_law_1.01,0.2712575912475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,64,power_law_1.01,0.4764544010162354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,16,power_law_1.01,0.3380095958709717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,16,power_law_1.01,0.4199999809265137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,16,power_law_1.01,0.5127935886383057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,16,power_law_1.01,0.6434112071990967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,16,power_law_1.01,0.9036928176879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,16,power_law_1.01,1.1397439956665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,128,balanced,0.04494933287302653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,128,balanced,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,128,balanced,0.04384533564249674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,128,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,128,balanced,0.04706133405367533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,128,balanced,0.05009066561857859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,128,balanced,0.052149335543314614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,128,balanced,0.05204799771308899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,128,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,128,balanced,0.05421333511670431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,128,balanced,0.056314667065938316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,128,balanced,0.060559997955958046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,128,balanced,0.0605973352988561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,128,balanced,0.06038933495680491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,128,balanced,0.06880000233650208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,128,balanced,0.0716480016708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,128,balanced,0.07258133093516032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,128,balanced,0.08268266419569652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,128,balanced,0.08990400036176045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,128,balanced,0.10364799698193868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,128,balanced,0.11530133088429768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,128,balanced,0.14844266573588052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,128,balanced,0.17459734280904135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,128,balanced,0.2267679969469706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,128,balanced,0.2755840023358663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,128,balanced,0.37681599458058673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,128,balanced,0.4698079824447632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,4,balanced,0.07242133220036824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,4,balanced,0.07420800129572551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,4,balanced,0.08092799782752991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,4,balanced,0.0951039989789327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,4,balanced,0.11531200011571248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,64,power_law_1.01,0.10175360441207885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,64,power_law_1.01,0.09889919757843017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,64,power_law_1.01,0.10100480318069457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,4,balanced,0.15382933616638184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,64,power_law_1.01,0.1019968032836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,4,balanced,0.15421332915623984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,4,balanced,0.15416533748308817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,64,power_law_1.01,0.10028799772262573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,64,power_law_1.01,0.10285439491271972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,64,power_law_1.01,0.10354559421539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,64,power_law_1.01,0.10376319885253907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,64,power_law_1.01,0.10165760517120362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,64,power_law_1.01,0.102566397190094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,4,balanced,0.1569813291231791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,4,balanced,0.15628266334533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,64,power_law_1.01,0.10501760244369507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,4,balanced,0.1579093337059021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,4,balanced,0.16062933206558228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,64,power_law_1.01,0.10651520490646363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,4,balanced,0.1606826682885488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,4,balanced,0.16637866695721945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,4,balanced,0.17293866475423178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,4,balanced,0.17354132731755575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,4,balanced,0.17834667364756265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,4,balanced,0.1956053376197815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,4,balanced,0.20929600795110068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,4,balanced,0.24775999784469604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,4,balanced,0.28138667345046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,64,power_law_1.01,0.1106943964958191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,64,power_law_1.01,0.10830719470977783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,64,power_law_1.01,0.11941759586334229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,64,power_law_1.01,0.11930880546569825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,64,power_law_1.01,0.13313920497894288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,64,power_law_1.01,0.1551103949546814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,64,power_law_1.01,0.15884799957275392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,4,balanced,0.35150400797526044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,64,power_law_1.01,0.21334400177001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,4,balanced,0.4209119876225789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,64,power_law_1.01,0.22282240390777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,64,power_law_1.01,0.2814719915390015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,4,balanced,0.6057173411051432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,64,power_law_1.01,0.2912447929382324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,4,balanced,0.7457866668701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,64,power_law_1.01,0.3714368104934692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,4,balanced,1.0702613194783528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,4,balanced,1.4099253018697102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,64,power_law_1.01,0.49674239158630373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,64,power_law_1.01,0.6623040199279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,64,power_law_1.01,0.8146944046020508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,32,balanced,0.03403199960788091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,32,balanced,0.032986665765444435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,32,balanced,0.032272001107533775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,32,balanced,0.033941333492596946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,32,balanced,0.03774933268626531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,32,balanced,0.03777066618204117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,32,balanced,0.03578133384386698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,32,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,32,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,32,balanced,0.03765333443880081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,32,balanced,0.03740799923737844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,32,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,32,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,32,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,32,balanced,0.0415040006240209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,32,balanced,0.042768001556396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,32,balanced,0.0435146689414978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,32,balanced,0.043951998154322304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,32,balanced,0.04373333354791006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,32,balanced,0.04795733094215393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,32,balanced,0.04978133241335551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,32,balanced,0.054154664278030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,32,balanced,0.05795733133951823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,32,balanced,0.07333866755167644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,32,balanced,0.08557333548863728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,32,balanced,0.10937066872914632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,32,balanced,0.1376533309618632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,8,balanced,0.06642666459083557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,8,balanced,0.05444799860318502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,8,balanced,0.05382933219273885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,8,balanced,0.05619733532269796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,8,balanced,0.055813332398732506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,8,balanced,0.05745066702365875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,8,balanced,0.056202664971351624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,8,balanced,0.055717334151268005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,8,balanced,0.0580213318268458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,8,balanced,0.05738666653633118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,8,balanced,0.05788266658782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,8,balanced,0.05985599756240845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,8,balanced,0.0577706644932429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,8,balanced,0.06563200056552887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,8,balanced,0.06836799780527751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,8,balanced,0.06596800188223521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,8,balanced,0.06842666864395142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,8,balanced,0.08169599870840709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,8,balanced,0.08004266520341237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,8,balanced,0.09676266709963481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,8,balanced,0.1069546639919281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,8,balanced,0.13132799665133157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,8,balanced,0.1578879952430725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,8,balanced,0.20761599143346152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,8,balanced,0.24482667446136475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,8,balanced,0.34246933460235596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,8,balanced,0.43225598335266113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,1,power_law_1.2,0.03699840009212494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,1,power_law_1.2,0.0374208003282547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,1,power_law_1.2,0.03922559916973114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,1,power_law_1.2,0.04674560129642487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,1,power_law_1.2,0.04938240051269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,1,power_law_1.2,0.05004159808158874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,1,power_law_1.2,0.04992640018463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,1,power_law_1.2,0.049446401000022885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,1,power_law_1.2,0.05125759840011597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,1,power_law_1.2,0.051475197076797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,1,power_law_1.2,0.05194240212440491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,1,power_law_1.2,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,1,power_law_1.2,0.05816320180892944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,1,power_law_1.2,0.061414402723312375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,1,power_law_1.2,0.0753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,1,power_law_1.2,0.07735040187835693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,1,power_law_1.2,0.08288000226020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,1,power_law_1.2,0.09297279715538025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,4,balanced,0.08120533327261607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.2,0.11509759426116943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,4,balanced,0.08030400176843007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,4,balanced,0.12774933377901712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,4,balanced,0.1281760036945343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,4,balanced,0.12947199741999307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,4,balanced,0.1300373375415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,4,balanced,0.1304373343785604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,4,balanced,0.13115200400352478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,4,balanced,0.13794133067131042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,4,balanced,0.13769066333770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,4,balanced,0.13958932956059775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,4,balanced,0.14203733205795288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,4,balanced,0.14486933747927347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,4,balanced,0.1495626668135325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,4,balanced,0.15871466199556986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,4,balanced,0.1628159979979197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,4,balanced,0.15985600153605142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,4,balanced,0.20750399430592856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,4,balanced,0.1987946629524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,4,balanced,0.2860746582349141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.2,0.14180480241775512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,4,balanced,0.2946026722590129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,4,balanced,0.4002506732940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,4,balanced,0.4991840124130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,4,balanced,0.7076746622721354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,4,balanced,0.9447466532389323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,4,balanced,1.3543200492858887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,4,balanced,1.797034740447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.2,0.1665984034538269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.2,0.22724480628967286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.2,0.30639359951019285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.2,0.4440447807312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.2,0.5949503898620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.2,0.862553596496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,32,2,balanced,0.04641066491603851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,32,2,balanced,0.04740799963474274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,32,2,balanced,0.05177066723505656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,32,2,balanced,0.060746664802233376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,32,2,balanced,0.07478933533032735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,32,2,balanced,0.11025599638621013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,32,2,balanced,0.11001599828402202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,32,2,balanced,0.11136000355084737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,32,2,balanced,0.1125866671403249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,32,2,balanced,0.11373866597811381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,32,2,balanced,0.1151146690050761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,32,2,balanced,0.11800533533096313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,32,2,balanced,0.12264532844225566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,32,2,balanced,0.12434666355450948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,32,2,balanced,0.13388267159461975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,32,2,balanced,0.13667733470598856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,32,2,balanced,0.14664000272750854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,32,2,balanced,0.16852267583211264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,32,2,balanced,0.18590933084487915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,32,2,balanced,0.22543466091156006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,32,2,balanced,0.2615893284479777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,32,2,balanced,0.38099201520284015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,32,2,balanced,0.44471466541290283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,32,2,balanced,0.6595893303553263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,32,2,balanced,0.8254826863606771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,32,2,balanced,1.2049439748128254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,32,2,balanced,1.5802826881408691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,4,power_law_1.01,0.017977599799633027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,4,power_law_1.01,0.01973759979009628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,4,power_law_1.01,0.022188800573349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,4,power_law_1.01,0.025920000672340394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,4,power_law_1.01,0.027968001365661622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,4,power_law_1.01,0.03084160089492798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,4,power_law_1.01,0.03600000143051148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,4,power_law_1.01,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,4,power_law_1.01,0.03727999925613403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,4,power_law_1.01,0.04126079976558685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,4,power_law_1.01,0.03971199989318848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,4,power_law_1.01,0.043244799971580504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,4,power_law_1.01,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,256,balanced,0.054416000843048096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,4,power_law_1.01,0.04405120015144348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,256,balanced,0.053802669048309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,256,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,256,balanced,0.05985599756240845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,4,power_law_1.01,0.05615360140800476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,4,power_law_1.01,0.05306239724159241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,4,power_law_1.01,0.05333120226860046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,4,power_law_1.01,0.07781760096549988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,4,power_law_1.01,0.0889855980873108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,4,power_law_1.01,0.09545599818229675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,4,power_law_1.01,0.09752960205078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,4,power_law_1.01,0.10750080347061157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,256,balanced,0.06604266663392384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,4,power_law_1.01,0.12210559844970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,4,power_law_1.01,0.15618560314178467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,4,power_law_1.01,0.18573440313339235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,4,power_law_1.01,0.2473151922225952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,4,power_law_1.01,0.30896639823913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,256,balanced,0.07060799996058147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,256,balanced,0.06681600213050842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,256,balanced,0.07053333520889282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,256,balanced,0.06824533144632976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,256,balanced,0.07050133248170216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,256,balanced,0.07226133346557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,256,balanced,0.07830933233102162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,256,balanced,0.07829866806666057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,256,balanced,0.07861333092053731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,256,balanced,0.0853653351465861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,256,balanced,0.0867680013179779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,256,balanced,0.09518933296203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,256,balanced,0.10730666915575664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,256,balanced,0.12165333827336629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,256,balanced,0.14630400141080221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,256,balanced,0.17284266153971353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,256,balanced,0.23246399561564127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,256,balanced,0.239738663037618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,256,balanced,0.30914666255315143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,256,balanced,0.37457064787546795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,256,balanced,0.5218133529027303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,256,balanced,0.6497653325398763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.2,1.1602304458618165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,4,power_law_1.01,0.09086080193519593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,4,power_law_1.01,0.10133119821548461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,4,power_law_1.01,0.11605119705200195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,4,power_law_1.01,0.1252992033958435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,4,power_law_1.01,0.12510080337524415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,4,power_law_1.01,0.12127360105514526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,4,power_law_1.01,0.12136319875717164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,4,power_law_1.01,0.11880320310592651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,4,power_law_1.01,0.11937919855117798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,4,power_law_1.01,0.11626240015029907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,4,power_law_1.01,0.11022080183029175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,4,power_law_1.01,0.1302016019821167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,4,power_law_1.01,0.13808000087738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,4,power_law_1.01,0.12832000255584716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,4,power_law_1.01,0.14708479642868041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,4,power_law_1.01,0.16762239933013917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,4,power_law_1.01,0.18852479457855226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,4,power_law_1.01,0.20412800312042237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.01,0.23936638832092286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.01,0.354041600227356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.01,0.3575680017471313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.01,0.5533184051513672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.01,0.6463551998138428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.01,1.1536319732666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.01,1.6831424713134766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.01,2.0826623916625975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,32,power_law_1.2,0.09064319729804993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,32,power_law_1.2,0.10200320482254029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,32,power_law_1.2,0.09893760085105896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,32,power_law_1.2,0.10154880285263061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,32,power_law_1.2,0.10247039794921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,32,power_law_1.2,0.10051840543746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,32,power_law_1.2,0.09765120148658753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,16,balanced,0.05118933320045471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,32,power_law_1.2,0.10177279710769653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,16,balanced,0.043840001026789345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,16,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,16,balanced,0.049685334165891014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,16,balanced,0.06783999999364217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,32,power_law_1.2,0.10351999998092651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,16,balanced,0.08943466345469157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,16,balanced,0.09308800101280212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,16,balanced,0.09303999940554301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,32,power_law_1.2,0.10245120525360107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,16,balanced,0.0923520028591156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,16,balanced,0.09520533680915833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,16,balanced,0.09452799956003825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,32,power_law_1.2,0.09887999892234803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,32,power_law_1.2,0.10337920188903808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,32,power_law_1.2,0.10493439435958862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,16,balanced,0.0956053336461385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,32,power_law_1.2,0.10790400505065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,16,balanced,0.09638933340708415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,32,power_law_1.2,0.11834880113601684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,16,balanced,0.09787733356157939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,16,balanced,0.10207999746004741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,32,power_law_1.2,0.11889280080795288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,16,balanced,0.10506666700045268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,16,balanced,0.10657599568367004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,32,power_law_1.2,0.12789119482040406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,16,balanced,0.11496000488599141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,16,balanced,0.11556800206502278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,32,power_law_1.2,0.14018559455871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,32,power_law_1.2,0.1528831958770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,32,power_law_1.2,0.1876736044883728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,32,power_law_1.2,0.20828800201416015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,32,power_law_1.2,0.27701120376586913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.01,2.350476837158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,32,power_law_1.2,0.32833280563354494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,32,power_law_1.2,0.4629055976867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,32,power_law_1.2,0.6840447902679443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,16,balanced,0.12797333796819052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,16,balanced,0.1436799963315328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,16,balanced,0.168122669061025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,16,balanced,0.19312000274658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,32,power_law_1.2,0.9450176239013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,32,power_law_1.2,1.3017855644226075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,16,balanced,0.2397973338762919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,16,balanced,0.29419734080632526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,16,balanced,0.39738134543100995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,16,balanced,0.49267733097076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,8,power_law_1.01,0.01820800006389618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,8,power_law_1.01,0.020710399746894835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,8,power_law_1.01,0.022310400009155275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,8,power_law_1.01,0.02160000056028366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,8,power_law_1.01,0.021638399362564086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,8,power_law_1.01,0.022655999660491942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,8,power_law_1.01,0.023820799589157105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,8,power_law_1.01,0.02388480007648468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,8,power_law_1.01,0.025011199712753295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,8,power_law_1.01,0.024928000569343568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,8,power_law_1.01,0.027091199159622194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,8,power_law_1.01,0.027532801032066345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,8,power_law_1.01,0.028799998760223388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,8,power_law_1.01,0.03290880024433136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,8,power_law_1.01,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,8,power_law_1.01,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,8,power_law_1.01,0.04636160135269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,8,power_law_1.01,0.053529602289199826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,8,power_law_1.01,0.058739197254180905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,8,power_law_1.01,0.0775168001651764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,8,power_law_1.01,0.05999360084533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,8,power_law_1.01,0.07757440209388733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,8,power_law_1.01,0.08669440150260925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,8,power_law_1.01,0.11735039949417114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,8,power_law_1.01,0.1412160038948059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,8,power_law_1.01,0.19676159620285033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,8,power_law_1.01,0.2494271993637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,1,power_law_1.01,0.051571202278137204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,1,power_law_1.01,0.06676480174064636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,1,power_law_1.01,0.08547199964523315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,1,power_law_1.01,0.11688319444656373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,1,power_law_1.01,0.11936000585556031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,1,power_law_1.01,0.12384639978408814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,1,power_law_1.01,0.1344704031944275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,1,power_law_1.01,0.12261760234832764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,1,power_law_1.01,0.12436480522155761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,1,power_law_1.01,0.12643840312957763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,1,power_law_1.01,0.1251263976097107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,1,power_law_1.01,0.13029119968414307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,1,power_law_1.01,0.13144960403442382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,1,power_law_1.01,0.14025599956512452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,1,power_law_1.01,0.15537919998168945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,1,power_law_1.01,0.15664000511169435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,1,power_law_1.01,0.15748480558395386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,1,power_law_1.01,0.18732160329818726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,1,power_law_1.01,0.2184704065322876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,1,power_law_1.01,0.25591039657592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,1,balanced,0.023472001155217487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,1,balanced,0.023018665611743927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,1,balanced,0.021344001094500225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,1,balanced,0.021189334491888683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,1,balanced,0.02309866746266683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,1,balanced,0.02532266577084859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,1,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,1,balanced,0.028437333802382152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,1,balanced,0.03697066754102707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,1,balanced,0.033333333830038704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,1,balanced,0.03589866558710734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,1,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,1,balanced,0.04247466723124186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,1,balanced,0.04332800209522247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,1,balanced,0.06262933214505513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,1,balanced,0.05555200080076853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,1,balanced,0.061535999178886414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,1,balanced,0.07437866429487865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,1,balanced,0.09114133318265279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,1,balanced,0.12706133723258972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,1,balanced,0.15435199936230978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,1,balanced,0.21274133523305258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,1,balanced,0.2707680066426595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,1,balanced,0.38581868012746173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,1,balanced,0.502458651860555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,1,balanced,0.7303840319315592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,1,balanced,0.9615519841512045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,1,power_law_1.01,0.3047935962677002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,1,power_law_1.01,0.4001023769378662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,1,power_law_1.01,0.48504319190979006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,1,power_law_1.01,0.6586368083953857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,1,power_law_1.01,0.8280832290649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,1,power_law_1.01,1.18023681640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,1,power_law_1.01,1.521440029144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,1,balanced,0.019493332753578823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,1,balanced,0.020351999749739964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,1,balanced,0.023141334454218548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,1,balanced,0.027141332626342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,1,balanced,0.043605332573254905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,1,balanced,0.06599999964237213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,1,balanced,0.06875200072924297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,1,balanced,0.07009066641330719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,1,balanced,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,1,balanced,0.07209066549936931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,1,balanced,0.07629866898059845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,1,balanced,0.078575998544693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,1,balanced,0.08424533406893413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,1,balanced,0.09128000338872273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,1,balanced,0.08656000097592671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,1,balanced,0.09087999661763509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,1,balanced,0.09941866993904114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,1,balanced,0.09915199875831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,1,balanced,0.11005866527557373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,1,balanced,0.11558933059374492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,1,balanced,0.12878933548927307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,1,balanced,0.22330133120218912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,1,balanced,0.257146676381429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,1,balanced,0.2414720058441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,1,balanced,0.25659199555714923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,1,balanced,0.44205331802368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,1,balanced,0.46668799718221027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,4,balanced,0.031925333042939506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,4,balanced,0.03169066707293192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,4,balanced,0.03765333443880081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,4,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,4,balanced,0.08213866750399272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,4,balanced,0.08595732847849528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,4,balanced,0.0848586658636729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,4,balanced,0.08502399921417236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,4,balanced,0.08479467034339905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,4,balanced,0.08530666430791219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,4,balanced,0.08680533369382222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,4,balanced,0.08874666690826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,4,balanced,0.09174399574597676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,4,balanced,0.0942133367061615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,4,balanced,0.09897599617640178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,4,balanced,0.1020960013071696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,4,balanced,0.10713600118954976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,4,balanced,0.11549333731333415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,256,balanced,0.11035733421643575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,256,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,256,balanced,0.05607999861240387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,256,balanced,0.058133333921432495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,256,balanced,0.055871998270352684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,256,balanced,0.058149332801500954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,256,balanced,0.05811200042565664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,256,balanced,0.057818666100502014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,256,balanced,0.0580320010582606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,256,balanced,0.05855466425418854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,256,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,256,balanced,0.062277331948280334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,256,balanced,0.06257600088914235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,256,balanced,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,256,balanced,0.07005333403746287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,256,balanced,0.07052266597747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,256,balanced,0.07450133562088013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,4,balanced,0.1258080005645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,256,balanced,0.09493866562843323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,4,balanced,0.15320000052452087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,256,balanced,0.09271466732025146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,4,balanced,0.16248533129692078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,256,balanced,0.11935466527938843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,4,balanced,0.21227733294169107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,256,balanced,0.1272533337275187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,4,balanced,0.24741866191228232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,256,balanced,0.1686826745669047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,4,balanced,0.35099200407663983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,256,balanced,0.1493119994799296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,4,balanced,0.432586669921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,256,balanced,0.1911626656850179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,4,balanced,0.6090293327967325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,256,balanced,0.21235734224319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,256,balanced,0.2783733407656352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,256,balanced,0.32281599442164105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,4,balanced,0.7873439788818359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,1,balanced,0.09480533003807068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,1,balanced,0.14520000418027243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,1,balanced,0.2444960077603658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,1,balanced,0.4476960102717082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,1,balanced,0.8368533452351888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,1,balanced,1.6303307215372722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,1,balanced,1.6330933570861816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,1,balanced,1.6349813143412273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,1,balanced,1.6437759399414062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,1,balanced,1.6455200513203938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,1,balanced,1.6586772600809734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,1,balanced,1.6664479573567708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,1,balanced,1.6790399551391602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,1,balanced,1.6854027112325032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,1,balanced,1.7026666005452473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,1,balanced,1.7110506693522136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,1,balanced,1.7363039652506511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,1,balanced,1.7856052716573079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,1,balanced,1.8530346552530925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,1,balanced,1.9434773127237956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,1,balanced,2.0576160748799643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,1,balanced,2.216207981109619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,1,balanced,2.390885353088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,1,balanced,3.1397387186686196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,1,balanced,3.236565272013346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,1,balanced,5.173978805541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,1,balanced,5.326698621114095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,2,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,2,balanced,0.07520533104737599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,2,balanced,0.11930666367212932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,2,balanced,0.2029973268508911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,2,balanced,0.3731679916381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,2,balanced,0.3768800099690755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,2,balanced,0.37755731741587323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,2,balanced,0.3755093415578206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,2,balanced,0.37856535116831463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,2,balanced,0.38153600692749023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,2,balanced,0.38148268063863117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,2,balanced,0.38514665762583417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,2,balanced,0.38815999031066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,2,balanced,0.39348268508911133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,2,balanced,0.40057067076365155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,2,balanced,0.4038986762364705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,8,power_law_1.2,0.032742398977279666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,2,balanced,0.41071466604868573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,2,balanced,0.42430933316548664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,8,power_law_1.2,0.031167998909950256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,2,balanced,0.45431466897328693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,8,power_law_1.2,0.030355200171470642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,8,power_law_1.2,0.03306879997253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,8,power_law_1.2,0.0320576012134552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,8,power_law_1.2,0.03409920036792755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,8,power_law_1.2,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,8,power_law_1.2,0.035104000568389894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,8,power_law_1.2,0.03546879887580871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,2,balanced,0.48556800683339435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,2,balanced,0.5075893402099609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,2,balanced,0.6142666737238566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,2,balanced,0.6377760171890259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,2,balanced,0.9177599747975668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,2,balanced,0.9694080352783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,2,balanced,1.3577440579732258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,8,power_law_1.2,0.03570559918880463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,8,power_law_1.2,0.038278400897979736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,2,balanced,1.7281492551167805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,8,power_law_1.2,0.04035840034484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,8,power_law_1.2,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,8,power_law_1.2,0.0409280002117157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,8,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,8,power_law_1.2,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,8,power_law_1.2,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,8,power_law_1.2,0.05680000185966492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,8,power_law_1.2,0.06401280164718628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,8,power_law_1.2,0.07642880082130432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,8,power_law_1.2,0.08286719918251037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,8,power_law_1.2,0.11638400554656983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,8,power_law_1.2,0.13308800458908082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,8,power_law_1.2,0.20205440521240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,8,power_law_1.2,0.24533119201660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,8,power_law_1.2,0.37845759391784667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,1,power_law_1.2,0.04748800098896026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,1,power_law_1.2,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,1,power_law_1.2,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,1,power_law_1.2,0.07152640223503112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,1,power_law_1.2,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,1,power_law_1.2,0.07124480009078979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,1,power_law_1.2,0.06936320066452026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,1,power_law_1.2,0.06886399984359741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,1,power_law_1.2,0.06972799897193908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,1,power_law_1.2,0.06933119893074036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,1,power_law_1.2,0.06940159797668458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,1,power_law_1.2,0.07129600048065185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,1,power_law_1.2,0.07457919716835022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,1,power_law_1.2,0.07701119780540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,1,power_law_1.2,0.08780800104141236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,1,power_law_1.2,0.09015039801597595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,1,power_law_1.2,0.09303039908409119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,1,power_law_1.2,0.10798720121383668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.2,0.12256640195846558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.2,0.14986879825592042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,8,power_law_1.2,0.036236798763275145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,8,power_law_1.2,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,8,power_law_1.2,0.03587839901447296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,8,power_law_1.2,0.0381056010723114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,8,power_law_1.2,0.03834240138530731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,8,power_law_1.2,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,8,power_law_1.2,0.03893760144710541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,8,power_law_1.2,0.03880960047245026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,8,power_law_1.2,0.48672637939453123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,8,power_law_1.2,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,8,power_law_1.2,0.03925760090351105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,8,power_law_1.2,0.03943679928779602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,8,power_law_1.2,0.040601599216461184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,8,power_law_1.2,0.04111360013484955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,8,power_law_1.2,0.042387199401855466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,8,power_law_1.2,0.04699519872665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,8,power_law_1.2,0.047116801142692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,8,power_law_1.2,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.2,0.1778496026992798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,8,power_law_1.2,0.0549888014793396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,32,power_law_1.2,0.023020799458026885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,32,power_law_1.2,0.02705279886722565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,32,power_law_1.2,0.023001599311828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,32,power_law_1.2,0.023475199937820435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,32,power_law_1.2,0.022963200509548188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,32,power_law_1.2,0.023552000522613525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.2,0.2376703977584839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,32,power_law_1.2,0.025382399559020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,32,power_law_1.2,0.026137599349021913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.2,0.06033279895782471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,32,power_law_1.2,0.027161601185798644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,32,power_law_1.2,0.026707199215888978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,32,power_law_1.2,0.03234559893608093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,32,power_law_1.2,0.03178240060806274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,32,power_law_1.2,0.03240320086479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,32,power_law_1.2,0.052172797918319705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,32,power_law_1.2,0.05145599842071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,32,power_law_1.2,0.0535103976726532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,32,power_law_1.2,0.055980801582336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,32,power_law_1.2,0.06097919940948486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.2,0.06108800172805786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.2,0.08505600094795226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.2,0.09425920248031616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.2,0.07535359859466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.2,0.11637120246887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.2,0.13288320302963258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.2,0.1928447961807251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.2,0.25134079456329345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.2,0.3370752096176147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,1,balanced,0.03591466695070267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.2,0.08828160166740417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,1,balanced,0.03939733405907949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,1,balanced,0.05576533575852712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.2,0.4674047946929932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,1,balanced,0.08381332953770955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,1,balanced,0.13794666528701782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,1,balanced,0.13773866494496664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,1,balanced,0.1378986636797587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,1,balanced,0.1365546683470408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,1,balanced,0.13979732990264893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,1,balanced,0.14620799819628397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,1,balanced,0.1423893372217814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,1,balanced,0.1481013298034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,1,balanced,0.15160533785820007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,1,balanced,0.1592693328857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,1,balanced,0.16775466998418173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,1,balanced,0.17248533169428507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,1,balanced,0.1835413376490275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,1,balanced,0.20494399468104044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,1,balanced,0.23360000054041544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,1,balanced,0.2819253404935201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,1,balanced,0.31541333595911664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,1,balanced,0.446234663327535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,1,balanced,0.5177866617838541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,1,balanced,0.7438133557637533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.2,0.11361279487609863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,1,balanced,0.910154660542806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,1,balanced,1.3427252769470215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.2,0.29513599872589114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,1,balanced,1.7217599550882976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.2,0.40651521682739256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.2,0.1422719955444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.2,0.5171199798583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.2,0.19582719802856446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.2,0.2378943920135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.2,0.31816959381103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.2,0.7522111892700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,2,balanced,0.018085333208243053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,2,balanced,0.0179626668492953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,2,balanced,0.017210666090250015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,2,balanced,0.01897066707412402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,2,balanced,0.019258666783571243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,2,balanced,0.01926933353145917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,2,balanced,0.021317332983016968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,2,balanced,0.021082667013009388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,2,balanced,0.02126399924357732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,2,balanced,0.021488000949223835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,2,balanced,0.023082666099071503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,2,balanced,0.024005333582560223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,2,balanced,0.02922666569550832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,2,balanced,0.027221334477265675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,2,balanced,0.02739733209212621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,2,balanced,0.02932800104220708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,2,balanced,0.03522133330504099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,2,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,2,balanced,0.04181866844495138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,2,balanced,0.055445333321889244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,2,balanced,0.061679999033610024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,2,balanced,0.08072533210118611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,2,balanced,0.09909866253534953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,2,balanced,0.1276586651802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,2,balanced,0.16082666317621866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,2,balanced,0.2244159976641337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,2,balanced,0.2890186707178752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.2,0.43351039886474607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.2,0.9776703834533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,2,power_law_1.01,0.08503040075302123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,2,power_law_1.01,0.13043199777603148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,2,power_law_1.01,0.18991999626159667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,2,power_law_1.01,0.24705278873443604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,2,power_law_1.01,0.24431359767913818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,2,power_law_1.01,0.2458944082260132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,2,power_law_1.01,0.24227840900421144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,2,power_law_1.01,0.2539263963699341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,2,power_law_1.01,0.2630784034729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,2,power_law_1.01,0.26746881008148193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,2,power_law_1.01,0.27658240795135497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,2,power_law_1.01,0.2793215990066528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,2,power_law_1.01,0.28512001037597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,2,power_law_1.01,0.33863039016723634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,2,power_law_1.01,0.37944960594177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,2,power_law_1.01,0.376582407951355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,2,power_law_1.01,0.4034304141998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,2,power_law_1.01,0.4321023941040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,2,power_law_1.01,0.5172800064086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,2,power_law_1.01,0.6218495845794678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,2,power_law_1.01,0.7958208084106445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,2,power_law_1.01,0.9910207748413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,2,power_law_1.01,1.1942336082458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,2,power_law_1.01,1.740243148803711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,2,power_law_1.01,2.3999935150146485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,32,power_law_1.2,0.017433600127696992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,32,power_law_1.2,0.016518400609493257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,32,power_law_1.2,0.016601599752902985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,32,power_law_1.2,0.01777919977903366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,32,power_law_1.2,0.01854719966650009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,32,power_law_1.2,0.02200320065021515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,32,power_law_1.2,0.03210879862308502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,32,power_law_1.2,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,32,power_law_1.2,0.03253760039806366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,32,power_law_1.2,0.03207040131092072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,32,power_law_1.2,0.032179200649261476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,32,power_law_1.2,0.031071999669075014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,32,power_law_1.2,0.031744000315666196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,32,power_law_1.2,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,32,power_law_1.2,0.030092799663543703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,32,power_law_1.2,0.030636799335479737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,32,power_law_1.2,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,32,power_law_1.2,0.03386879861354828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,32,power_law_1.2,0.03889279961585999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,32,power_law_1.2,0.04734080135822296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,32,power_law_1.2,0.0490880012512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,32,power_law_1.2,0.06047359704971313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,32,power_law_1.2,0.07765759825706482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,2,power_law_1.01,3.6846336364746093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,2,power_law_1.2,0.02094080001115799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,2,power_law_1.2,0.025356799364089966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,2,power_law_1.2,0.03917439877986908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,2,power_law_1.2,0.052262401580810545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,2,power_law_1.2,0.06584960222244263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,2,power_law_1.2,0.08265600204467774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,2,power_law_1.2,0.09440000057220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,2,power_law_1.2,0.10028799772262573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,2,power_law_1.2,0.1019711971282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,2,power_law_1.2,0.11185280084609986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,2,power_law_1.2,0.11534080505371094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,2,power_law_1.2,0.11736960411071777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,2,power_law_1.2,0.11811840534210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,32,power_law_1.2,0.1074560046195984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,2,power_law_1.2,0.1259392023086548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,2,power_law_1.2,0.12592639923095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,2,power_law_1.2,0.1314239978790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,2,power_law_1.2,0.13600640296936034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,2,power_law_1.2,0.14847359657287598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,2,power_law_1.2,0.16412160396575928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,2,power_law_1.2,0.3347775936126709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,2,power_law_1.2,0.37319040298461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,2,power_law_1.2,0.2969088077545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,2,power_law_1.2,0.33297278881073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,2,power_law_1.2,0.41590399742126466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,2,power_law_1.01,4.549817657470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,2,power_law_1.2,0.49838080406188967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,2,power_law_1.2,0.654643201828003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,2,power_law_1.2,0.8273792266845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,32,power_law_1.2,0.1251520037651062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,32,power_law_1.2,0.17870719432830812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,32,power_law_1.2,0.2334399938583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,2,power_law_1.2,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,2,power_law_1.2,0.06455680131912231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,2,power_law_1.2,0.07647359967231751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,2,power_law_1.2,0.10119680166244507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,2,power_law_1.2,0.11230720281600952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,2,power_law_1.2,0.1496384024620056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,2,power_law_1.2,0.1578559994697571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,2,power_law_1.2,0.15962239503860473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,2,power_law_1.2,0.16229759454727172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,2,power_law_1.2,0.17235840559005738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,2,power_law_1.2,0.1710911989212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,2,power_law_1.2,0.17744640111923218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,2,power_law_1.2,0.1728127956390381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,2,power_law_1.2,0.18188159465789794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,2,power_law_1.2,0.18384640216827391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,2,power_law_1.2,0.19101439714431762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,2,power_law_1.2,0.19868799448013305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,2,power_law_1.2,0.2176896095275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,2,power_law_1.2,0.23311998844146728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,2,power_law_1.2,0.27797760963439944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,2,power_law_1.2,0.31256320476531985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,2,power_law_1.2,0.37512960433959963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,2,power_law_1.2,0.46372480392456056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,64,power_law_1.01,0.05553920269012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,64,power_law_1.01,0.05534719824790955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,64,power_law_1.01,0.05246719717979431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,64,power_law_1.01,0.05744640231132507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,64,power_law_1.01,0.05756800174713135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,64,power_law_1.01,0.05716480016708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,64,power_law_1.01,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,64,power_law_1.01,0.05864959955215454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,64,power_law_1.01,0.05989120006561279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,64,power_law_1.01,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,2,power_law_1.2,0.6651519775390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,64,power_law_1.01,0.0642304003238678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,64,power_law_1.01,0.06914560198783874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,64,power_law_1.01,0.07157760262489318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,64,power_law_1.01,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,64,power_law_1.01,0.08393599987030029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,64,power_law_1.01,0.08519039750099182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,64,power_law_1.01,0.09128959774971009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,64,power_law_1.01,0.10957440137863159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,64,power_law_1.01,0.11311359405517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,64,power_law_1.01,0.1426751971244812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,64,power_law_1.01,0.16389119625091553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,64,power_law_1.01,0.21410560607910156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,2,power_law_1.2,0.8176511764526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,64,power_law_1.01,0.23361918926239014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,2,power_law_1.2,1.1111167907714843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,64,power_law_1.01,0.29788799285888673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,64,power_law_1.01,0.36785919666290284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,2,power_law_1.2,1.4923647880554198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,64,power_law_1.01,0.5245312213897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,64,power_law_1.01,0.6044415950775146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,32,power_law_1.01,0.11992319822311401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,32,power_law_1.01,0.16008319854736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,32,power_law_1.01,0.15629440546035767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,32,power_law_1.01,0.1582208037376404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,32,power_law_1.01,0.15976959466934204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,32,power_law_1.01,0.14391039609909057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,32,power_law_1.01,0.1368127942085266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,32,power_law_1.01,0.1562175989151001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,32,power_law_1.01,0.15834239721298218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,32,power_law_1.01,0.13611520528793336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,32,power_law_1.01,0.14570239782333375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,32,power_law_1.01,0.15467519760131837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,32,power_law_1.01,0.1585536003112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,32,power_law_1.01,0.15897599458694459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,32,power_law_1.01,0.16802560091018676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,32,power_law_1.01,0.16191359758377075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,32,power_law_1.01,0.17564799785614013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,32,power_law_1.01,0.17923200130462646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,32,power_law_1.01,0.18940160274505616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,32,power_law_1.01,0.20850560665130616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,32,power_law_1.01,0.23064959049224854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,32,power_law_1.01,0.2843839883804321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,32,power_law_1.01,0.33346560001373293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,32,power_law_1.01,0.43740158081054686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,32,power_law_1.01,0.5553855895996094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,32,power_law_1.01,0.740006399154663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,32,power_law_1.01,0.9143551826477051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,2,power_law_1.01,0.19163520336151124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,2,power_law_1.01,0.32785921096801757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,2,power_law_1.01,0.39750399589538576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,2,power_law_1.01,0.6365568161010742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,2,power_law_1.01,0.8287167549133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,2,power_law_1.01,1.220902442932129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,2,power_law_1.01,1.3440832138061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,2,power_law_1.01,1.3982080459594726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,2,power_law_1.01,1.3933695793151855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,2,power_law_1.01,1.449625587463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,2,power_law_1.01,1.4659392356872558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,2,power_law_1.01,1.481760025024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,2,power_law_1.01,1.5393152236938477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,2,power_law_1.01,1.5568511962890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,2,power_law_1.01,1.605779266357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,2,power_law_1.01,1.6337791442871095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,2,power_law_1.01,1.7450431823730468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,2,power_law_1.01,1.8860544204711913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,2,power_law_1.01,1.9155839920043944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,2,power_law_1.01,2.2411968231201174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,2,power_law_1.01,2.323398399353027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,2,power_law_1.01,2.7332992553710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,2,power_law_1.01,2.942560005187988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,2,power_law_1.01,3.6886528015136717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,2,power_law_1.01,4.841004943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,2,power_law_1.01,6.269279861450196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,1,power_law_1.01,0.10721280574798583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,2,power_law_1.01,7.47681884765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,1,power_law_1.01,0.11219840049743653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,1,power_law_1.01,0.12396160364151002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,1,power_law_1.01,0.1492416024208069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,1,power_law_1.01,0.19811199903488158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,1,power_law_1.01,0.23280000686645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,1,power_law_1.01,0.2584448099136353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,1,power_law_1.01,0.30697600841522216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,1,power_law_1.01,0.3165760040283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,1,power_law_1.01,0.32321279048919677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,1,power_law_1.01,0.33488640785217283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,1,power_law_1.01,0.3505215883255005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,1,power_law_1.01,0.35946240425109866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,1,power_law_1.01,0.3723455905914307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,1,power_law_1.01,0.3843775987625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,1,power_law_1.01,0.3942015886306763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,1,power_law_1.01,0.4172544002532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,1,power_law_1.01,0.4710080146789551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.01,0.5128064155578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.01,0.617625617980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.01,0.7046527862548828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.01,0.9095680236816406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.01,1.1003328323364259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.01,1.5032320022583008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,1,power_law_1.01,2.035763168334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,1,power_law_1.01,2.8108800888061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,1,power_law_1.01,3.502816009521484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,8,balanced,0.07044800122578938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,8,balanced,0.07674666742483775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,8,balanced,0.08949866890907288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,8,balanced,0.12981333335240683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,8,balanced,0.18010665973027548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,8,balanced,0.2759573260943095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,8,balanced,0.27640000979105633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,8,balanced,0.2770719925562541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,8,balanced,0.2791893283526103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,8,balanced,0.27664534250895184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,8,balanced,0.2797279953956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,8,balanced,0.28225600719451904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,8,balanced,0.28180267413457233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,16,balanced,0.043968002001444496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,32,balanced,0.04685866832733154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,32,balanced,0.04669333497683207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,32,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,32,balanced,0.053674668073654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,32,balanced,0.056176001826922096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,32,balanced,0.06630933284759521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,32,balanced,0.06438399851322174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,32,balanced,0.06429333488146464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,32,balanced,0.06500266492366791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,32,balanced,0.06663999954859416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,32,balanced,0.06429333488146464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,32,balanced,0.06841066479682922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,32,balanced,0.06758933266003926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,32,balanced,0.0662773350874583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,32,balanced,0.07246933380762736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,32,balanced,0.0724533349275589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,32,balanced,0.07379200061162312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,32,balanced,0.07893333335717519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,32,balanced,0.08467732866605122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,32,balanced,0.09497599800427754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,32,balanced,0.10114666819572449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,32,balanced,0.12178666392962138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,32,balanced,0.14403200149536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,32,balanced,0.19241066773732504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,32,balanced,0.2336533268292745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,32,balanced,0.31700267394383747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,32,balanced,0.40321600437164307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,16,balanced,0.04557333389918009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,16,balanced,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,16,balanced,0.07027199864387512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,16,balanced,0.09105066458384196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,16,balanced,0.13662399848302206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,16,balanced,0.13665599624315897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,16,balanced,0.13715199629465738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,16,balanced,0.13803733388582864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,16,balanced,0.13929067055384317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,16,balanced,0.13922666509946188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,16,balanced,0.14126933614412943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,16,balanced,0.1437120040257772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,16,balanced,0.14545067151387533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,16,balanced,0.15089600284894308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,16,balanced,0.15343999862670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,16,balanced,0.156031996011734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,16,balanced,0.16759467124938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,16,balanced,0.17786665757497153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,16,balanced,0.19537599881490073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,16,balanced,0.21783999601999918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,16,balanced,0.2568533420562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,16,balanced,0.2883840004603068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,16,balanced,0.37352001667022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,16,balanced,0.4564106861750285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,16,balanced,0.6219946543375651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,16,balanced,0.7642026742299398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,32,balanced,0.02754133443037669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,32,balanced,0.025722667574882507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,32,balanced,0.02754666656255722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,32,balanced,0.0295413335164388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,32,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,32,balanced,0.03166399896144867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,32,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,32,balanced,0.03368533402681351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,32,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,32,balanced,0.03380800038576126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,32,balanced,0.03375466664632162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,32,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,32,balanced,0.035631999373435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,32,balanced,0.03579200059175491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,32,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,32,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,32,balanced,0.0417546679576238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,32,balanced,0.04203199843565623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,32,balanced,0.047925333182017006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,32,balanced,0.05121066669623057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,32,balanced,0.05313600103060404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,32,balanced,0.06518933176994324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,32,balanced,0.07257066667079926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,32,balanced,0.09499733646710713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,32,balanced,0.11548266808191936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,32,balanced,0.15614400307337442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,32,balanced,0.19896533091862997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,8,balanced,0.041706666350364685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,8,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,8,balanced,0.04391466577847799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,8,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,8,balanced,0.051967998345692955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,8,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,8,balanced,0.05940799911816915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,8,balanced,0.06020266811052958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,8,balanced,0.060645331939061485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,8,balanced,0.06105599800745646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,8,balanced,0.060047999024391174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,8,balanced,0.062080000837643944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,8,balanced,0.06480533381303151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,8,balanced,0.06659733255704244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,8,balanced,0.07232533395290375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,8,balanced,0.07262933254241943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,8,balanced,0.07640000184377034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,8,balanced,0.08898666501045227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,8,balanced,0.09619733691215515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,8,balanced,0.13195199767748514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,8,balanced,0.14289066195487976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,8,balanced,0.1974560022354126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,8,balanced,0.2295893430709839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,8,balanced,0.3285440007845561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,8,balanced,0.40036265055338544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,8,balanced,0.5643519957860311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,8,balanced,0.7243306636810303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,4,power_law_1.2,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,4,power_law_1.2,0.06565120220184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,4,power_law_1.2,0.07157760262489318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,4,power_law_1.2,0.08574079871177673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,4,power_law_1.2,0.09638400077819824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,4,power_law_1.2,0.10841599702835084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,4,power_law_1.2,0.1259392023086548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,4,power_law_1.2,0.12728320360183715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,4,power_law_1.2,0.13422080278396606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,4,power_law_1.2,0.12954879999160768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,4,power_law_1.2,0.13466880321502686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,4,power_law_1.2,0.1397696018218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,4,power_law_1.2,0.14582400321960448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,4,power_law_1.2,0.14778879880905152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,4,power_law_1.2,0.16284799575805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,4,power_law_1.2,0.16403199434280397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,4,power_law_1.2,0.17523839473724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,4,power_law_1.2,0.20042240619659424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,4,power_law_1.2,0.2205888032913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,4,power_law_1.2,0.27781119346618655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,4,power_law_1.2,0.305945611000061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,4,power_law_1.2,0.390662407875061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,4,power_law_1.2,0.47446398735046386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,4,power_law_1.2,0.6543295860290528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,4,power_law_1.2,0.8645119667053223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,4,power_law_1.2,1.1374591827392577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,4,power_law_1.2,1.464038372039795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,1,power_law_1.2,0.09843199849128723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,1,power_law_1.2,0.11046400070190429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,1,power_law_1.2,0.13825279474258423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,1,power_law_1.2,0.1852031946182251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,1,power_law_1.2,0.23193600177764892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,1,power_law_1.2,0.2722368001937866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,1,power_law_1.2,0.37237119674682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,1,power_law_1.2,0.38886399269104005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,1,power_law_1.2,0.40288639068603516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,1,power_law_1.2,0.40129919052124025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,1,power_law_1.2,0.4198847770690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,1,power_law_1.2,0.42560639381408694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,1,power_law_1.2,0.44236159324645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,1,power_law_1.2,0.45041279792785643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,1,power_law_1.2,0.46749439239501955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,1,power_law_1.2,0.4824960231781006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,1,power_law_1.2,0.49848318099975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,1,power_law_1.2,0.5682496070861817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.2,0.6092351913452149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.2,0.7106112003326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.2,0.7974400043487548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.2,1.0148223876953124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.2,1.2048831939697267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.2,1.6130559921264649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,16,balanced,0.03818133225043615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,16,balanced,0.03788800040880839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,16,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,16,balanced,0.04641599953174591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.2,2.0130048751831056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.2,2.8067583084106444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,16,balanced,0.06435200075308482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,16,balanced,0.06346666812896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,16,balanced,0.06512000163396199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,16,balanced,0.06577600042025249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,16,balanced,0.06515733400980632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.2,3.5934783935546877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,16,balanced,0.0661653329928716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,16,balanced,0.0658079981803894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,16,balanced,0.06818133095900218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,16,balanced,0.06836266815662384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,16,balanced,0.06885866820812225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,16,balanced,0.07672533392906189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,16,balanced,0.07842133442560832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,16,balanced,0.07832533121109009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,16,balanced,0.08406933148701985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,16,balanced,0.08661866188049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,16,balanced,0.0944640040397644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,16,balanced,0.1011946698029836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,16,balanced,0.1320799986521403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,16,balanced,0.14568533500035605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,16,balanced,0.19188799460728964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,16,balanced,0.2381653388341268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,16,balanced,0.32683734099070233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,16,balanced,0.4259573221206665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,1,balanced,0.03842133283615112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,1,balanced,0.0543146679798762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,1,balanced,0.0784853349129359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,1,balanced,0.1281013290087382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,1,balanced,0.21548267205556235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,1,balanced,0.2188106576601664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,1,balanced,0.21714133024215698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,1,balanced,0.21873066822687784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,1,balanced,0.2212053338686625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,1,balanced,0.22784000635147095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,1,balanced,0.22498132785161337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,1,balanced,0.2297226587931315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,1,balanced,0.2366186579068502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,1,balanced,0.24101332823435465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,1,balanced,0.25068267186482746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,1,balanced,0.25827733675638836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,1,balanced,0.2714080015818278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,1,balanced,0.2972533305486043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,1,balanced,0.3330559929211934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,1,balanced,0.41063467661539715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,1,balanced,0.4591360092163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,1,balanced,0.6225066582361857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,1,balanced,0.716917355855306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,1,balanced,1.0384213129679363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,1,balanced,1.2762293020884197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,1,balanced,1.856170654296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,1,balanced,2.3939359982808432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,128,balanced,0.060496002435684204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,128,balanced,0.05862933397293091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,128,balanced,0.0599839985370636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,128,balanced,0.062447999914487205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,128,balanced,0.060309335589408875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,128,balanced,0.06817066669464111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,128,balanced,0.0668213317791621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,128,balanced,0.0705813318490982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,128,balanced,0.07041599849859874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,128,balanced,0.06883733471234639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,128,balanced,0.07670400043328603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,128,balanced,0.07018666466077168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,128,balanced,0.07043733199437459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,128,balanced,0.07268266876538594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,128,balanced,0.0763679991165797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,128,balanced,0.0769760012626648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,128,balanced,0.07879999776681264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,128,balanced,0.08509866396586101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,128,balanced,0.08933867017428081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,128,balanced,0.10102933645248413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,128,balanced,0.11322666207949321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,128,balanced,0.1344586710135142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,128,balanced,0.1543786625067393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,128,balanced,0.1978506644566854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,128,balanced,0.23673067490259805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,128,balanced,0.3150720000267029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,128,balanced,0.39239466190338135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,32,power_law_1.01,0.08609920144081115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,32,power_law_1.01,0.08520320057868958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,32,power_law_1.01,0.0846015989780426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,32,power_law_1.01,0.08525440096855164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,32,power_law_1.01,0.0855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,32,power_law_1.01,0.091839998960495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,32,power_law_1.01,0.09182080030441284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,32,power_law_1.01,0.09281920194625855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,32,power_law_1.01,0.09353600144386291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,32,power_law_1.01,0.09525759816169739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,32,power_law_1.01,0.09521920084953309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,32,power_law_1.01,0.09594240188598632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,32,power_law_1.01,0.09758719801902771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,32,power_law_1.01,0.09903360009193421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,32,power_law_1.01,0.10733439922332763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,32,power_law_1.01,0.10839040279388427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,32,power_law_1.01,0.12471040487289428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,32,power_law_1.01,0.14177919626235963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,32,power_law_1.01,0.14839680194854737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,32,power_law_1.01,0.1840831995010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,32,power_law_1.01,0.21902079582214357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,32,power_law_1.01,0.26460800170898435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,32,power_law_1.01,0.3187455892562866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,32,power_law_1.01,0.4801599979400635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,32,power_law_1.01,0.6791039943695069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,32,power_law_1.01,0.9127552032470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,32,power_law_1.01,1.2040255546569825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,1,power_law_1.2,0.08025599718093872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,1,power_law_1.2,0.12082560062408447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,1,power_law_1.2,0.126803195476532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,1,power_law_1.2,0.20176000595092775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,1,power_law_1.2,0.209881591796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,1,power_law_1.2,0.21376640796661378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,1,power_law_1.2,0.21880319118499755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,1,power_law_1.2,0.2201024055480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,1,power_law_1.2,0.22449278831481934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,1,power_law_1.2,0.22602241039276122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,1,power_law_1.2,0.22703359127044678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,1,power_law_1.2,0.24703359603881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,1,power_law_1.2,0.24350080490112305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,1,power_law_1.2,0.25806078910827634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,1,power_law_1.2,0.28908159732818606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,1,power_law_1.2,0.3015104055404663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,1,power_law_1.2,0.29809279441833497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,1,power_law_1.2,0.3393984079360962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,32,power_law_1.01,0.0392767995595932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,32,power_law_1.01,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,32,power_law_1.01,0.029919999837875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,32,power_law_1.01,0.03244799971580505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,32,power_law_1.01,0.03139199912548065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,32,power_law_1.01,0.03287039995193482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,32,power_law_1.01,0.03245440125465393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,32,power_law_1.01,0.03313280045986176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,32,power_law_1.01,0.03391999900341034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,32,power_law_1.01,0.03521920144557953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,32,power_law_1.01,0.03562879860401154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,32,power_law_1.01,0.03687680065631867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,32,power_law_1.01,0.03820799887180328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,32,power_law_1.01,0.03937279880046844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,1,power_law_1.2,0.38919041156768797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,32,power_law_1.01,0.043532800674438474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,32,power_law_1.01,0.04545280039310455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,32,power_law_1.01,0.045849600434303285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,32,power_law_1.01,0.05092480182647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,32,power_law_1.01,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,32,power_law_1.01,0.060140800476074216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,32,power_law_1.01,0.06554880142211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,32,power_law_1.01,0.08566399812698364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,32,power_law_1.01,0.10528639554977418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,1,power_law_1.2,0.4837376117706299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,32,power_law_1.01,0.13686399459838866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,32,power_law_1.01,0.16022399663925171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,32,power_law_1.01,0.21571838855743408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,2,balanced,0.027248000105222065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,2,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,32,power_law_1.01,0.2769536018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,2,balanced,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,2,balanced,0.07623999814192454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,2,balanced,0.12594133615493774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,2,balanced,0.21494932969411215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,2,balanced,0.22140800952911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,2,balanced,0.2268213431040446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,2,balanced,0.23196800549825033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,2,balanced,0.24025066693623862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,2,balanced,0.2515520056088765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,2,balanced,0.23810666799545288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,2,balanced,0.24438933531443277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,2,balanced,0.25704532861709595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,2,balanced,0.2688106695810954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,2,balanced,0.27090134223302204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,2,balanced,0.27582399050394696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,2,balanced,0.37060801188151044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,2,balanced,0.3570773204167684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,2,balanced,0.6308319965998331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,2,balanced,0.6932533582051595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,2,balanced,0.5897226730982462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,2,balanced,0.6192533175150553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,2,balanced,0.6564799944559733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,2,balanced,0.7175892988840739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,2,balanced,1.2171626885732014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,1,power_law_1.2,0.6016575813293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,2,balanced,1.3055040041605632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,1,power_law_1.2,0.7954559803009034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,1,power_law_1.2,0.9958271980285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,1,power_law_1.2,1.3980992317199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,1,power_law_1.2,1.8095359802246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,1,power_law_1.2,2.5887872695922853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,64,power_law_1.2,0.031353598833084105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,64,power_law_1.2,0.029190400242805482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,64,power_law_1.2,0.028352001309394838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,64,power_law_1.2,0.030239999294281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,64,power_law_1.2,0.03084160089492798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,64,power_law_1.2,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,64,power_law_1.2,0.03182080090045929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,64,power_law_1.2,0.03178240060806274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,64,power_law_1.2,0.03192960023880005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,64,power_law_1.2,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,64,power_law_1.2,0.03246079981327057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,64,power_law_1.2,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,64,power_law_1.2,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,64,power_law_1.2,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,64,power_law_1.2,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,64,power_law_1.2,0.03985919952392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,64,power_law_1.2,0.041177600622177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,64,power_law_1.2,0.04725759923458099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,64,power_law_1.2,0.04805760085582733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,64,power_law_1.2,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,64,power_law_1.2,0.06778879761695862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,64,power_law_1.2,0.08711680173873901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,64,power_law_1.2,0.08666880130767822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,64,power_law_1.2,0.12640639543533325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,64,power_law_1.2,0.1626688003540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,64,power_law_1.2,0.2268671989440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,64,power_law_1.2,0.31027200222015383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,1,power_law_1.2,3.328384017944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,1,power_law_1.2,0.0525056004524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,1,power_law_1.2,0.05809919834136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,1,power_law_1.2,0.06540799736976624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,1,power_law_1.2,0.08858240246772767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,1,power_law_1.2,0.11521279811859131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,1,power_law_1.2,0.1356735944747925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,1,power_law_1.2,0.1765887975692749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,1,power_law_1.2,0.1866943955421448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,1,power_law_1.2,0.18874239921569824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,8,power_law_1.01,0.036883199214935304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,1,power_law_1.2,0.19169280529022217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,8,power_law_1.01,0.03519999980926514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,1,power_law_1.2,0.2036288022994995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,8,power_law_1.01,0.032793599367141726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,1,power_law_1.2,0.21050240993499755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,8,power_law_1.01,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,8,power_law_1.01,0.036934399604797365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,8,power_law_1.01,0.038796800374984744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,8,power_law_1.01,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,8,power_law_1.01,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,8,power_law_1.01,0.03809280097484589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,8,power_law_1.01,0.040064001083374025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,8,power_law_1.01,0.04035840034484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,8,power_law_1.01,0.04296320080757141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,1,power_law_1.2,0.2200063943862915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,8,power_law_1.01,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,1,power_law_1.2,0.2441920042037964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,1,power_law_1.2,0.2553728103637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,4,power_law_1.2,0.048742398619651794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,1,power_law_1.2,0.26721279621124266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,4,power_law_1.2,0.052108800411224364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,1,power_law_1.2,0.26915199756622316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,4,power_law_1.2,0.0513152003288269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,1,power_law_1.2,0.32586240768432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,4,power_law_1.2,0.055379199981689456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,1,power_law_1.2,0.3940608024597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,4,power_law_1.2,0.0623744010925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,8,power_law_1.01,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,1,power_law_1.2,0.4657599925994873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,8,power_law_1.01,0.055692797899246214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,8,power_law_1.01,0.057811200618743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,1,power_law_1.2,0.5178112030029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,8,power_law_1.01,0.057126402854919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,1,power_law_1.2,0.6791168212890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,8,power_law_1.01,0.0642687976360321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,1,power_law_1.2,0.8445504188537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,8,power_law_1.01,0.07912319898605347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,4,power_law_1.2,0.07370880246162415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,8,power_law_1.01,0.09473279714584351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,1,power_law_1.2,1.1478591918945313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,8,power_law_1.01,0.10064640045166015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,8,power_law_1.01,0.12613760232925414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,8,power_law_1.01,0.1469823956489563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,8,power_law_1.01,0.22587521076202394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,8,power_law_1.01,0.2721152067184448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,8,power_law_1.01,0.33358719348907473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,4,power_law_1.2,0.08011519908905029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,8,power_law_1.01,0.42170238494873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,4,power_law_1.2,0.07837439775466919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,1,power_law_1.2,1.4865280151367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,1,power_law_1.2,2.101651191711426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,4,power_law_1.2,0.07600640058517456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,4,power_law_1.2,0.0800383985042572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,4,power_law_1.2,0.08360959887504578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,4,power_law_1.2,0.08588799834251404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,4,power_law_1.2,0.08799359798431397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,4,power_law_1.2,0.09638400077819824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,4,power_law_1.2,0.10494719743728638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,4,power_law_1.2,0.10888320207595825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,4,power_law_1.2,0.10426880121231079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,4,power_law_1.2,0.11611520051956177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,4,power_law_1.2,0.12577919960021972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,4,power_law_1.2,0.14684799909591675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,4,power_law_1.2,0.16771199703216552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,4,power_law_1.2,0.20901119709014893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,1,power_law_1.2,2.7586687088012694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,4,power_law_1.2,0.252889609336853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,4,power_law_1.2,0.3397952079772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,4,power_law_1.2,0.43383040428161623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,4,power_law_1.2,0.6154687881469727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,4,power_law_1.2,0.7545472145080566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,8,balanced,0.048058668772379555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,8,balanced,0.049082666635513306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,8,balanced,0.07017066578070323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,8,balanced,0.10475732882817586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,8,balanced,0.15467733144760132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,8,balanced,0.18465600411097208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,8,balanced,0.18540799617767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,8,balanced,0.18593066930770874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,8,balanced,0.18974934021631876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,8,balanced,0.18899200359980264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,8,balanced,0.18759999672571817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,8,balanced,0.19181332985560098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,8,balanced,0.19406400124231973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,8,balanced,0.1934453248977661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,8,balanced,0.2023626764615377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,8,balanced,0.20338133970896402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,8,balanced,0.2082186738650004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,8,balanced,0.2384106715520223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,8,balanced,0.2302666703859965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,8,balanced,0.2869173288345337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,8,balanced,0.27266667286554974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,8,balanced,0.4131253163019816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,8,balanced,0.37937601407368976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,8,balanced,0.5487253268559774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,8,balanced,0.5975893338521322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,8,balanced,0.9270666440327963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,8,balanced,1.0801013310750325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,4,balanced,0.04814399778842926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,4,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,4,balanced,0.05628266433874766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,4,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,4,balanced,0.05940799911816915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,4,balanced,0.060959999759991966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,4,balanced,0.061205332477887474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,4,balanced,0.06003733476003011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,4,balanced,0.060085331400235496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,4,balanced,0.0621066689491272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,4,balanced,0.06069866816202799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,4,balanced,0.06225066880385081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,4,balanced,0.06166933476924896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,4,balanced,0.06552533308664958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,4,balanced,0.07084799806276958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,4,balanced,0.07090133428573608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,4,balanced,0.07800533374150594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,4,balanced,0.0849120020866394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,4,balanced,0.0901759962240855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,4,balanced,0.10917866230010986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,4,balanced,0.11358400185902913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,4,balanced,0.14934933185577393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,4,balanced,0.17677332957585654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,4,balanced,0.23410133520762125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,4,balanced,0.29020800193150836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,4,balanced,0.4010293483734131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,4,balanced,0.4936000108718872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,16,power_law_1.01,0.19393919706344603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,16,power_law_1.01,0.24890239238739015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,16,power_law_1.01,0.2476288080215454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,16,power_law_1.01,0.2530431985855103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,16,power_law_1.01,0.2458944082260132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,16,power_law_1.01,0.22001280784606933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,16,power_law_1.01,0.22613120079040527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,16,power_law_1.01,0.22958719730377197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,16,power_law_1.01,0.23036160469055175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,16,power_law_1.01,0.24222080707550048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,16,power_law_1.01,0.23859200477600098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,16,power_law_1.01,0.24253439903259277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,16,power_law_1.01,0.25576961040496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,16,power_law_1.01,0.2611007928848267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,16,power_law_1.01,0.26721920967102053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,16,power_law_1.01,0.2636096000671387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,16,power_law_1.01,0.28819200992584226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,16,power_law_1.01,0.33537280559539795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,16,power_law_1.01,0.3355456113815308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,16,power_law_1.01,0.40977277755737307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,16,power_law_1.01,0.47258877754211426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,16,power_law_1.01,0.5987328052520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,16,power_law_1.01,0.5719359874725342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,16,power_law_1.01,0.7520256042480469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,16,power_law_1.01,0.9357952117919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,16,power_law_1.01,1.389894390106201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,2,power_law_1.2,0.06787199974060058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,16,power_law_1.01,1.3833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,2,power_law_1.2,0.09018880128860474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,2,power_law_1.2,0.10457600355148315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,2,power_law_1.2,0.14741120338439942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,2,power_law_1.2,0.1948799967765808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,2,power_law_1.2,0.23052799701690674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,2,power_law_1.2,0.29821441173553465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,2,power_law_1.2,0.3174207925796509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,2,power_law_1.2,0.3334912061691284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,2,power_law_1.2,0.3265664100646973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,2,power_law_1.2,0.34151039123535154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,2,power_law_1.2,0.3489919900894165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,2,power_law_1.2,0.3702847957611084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,2,power_law_1.2,0.386246395111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,2,power_law_1.2,0.4004608154296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,2,power_law_1.2,0.4116799831390381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,2,power_law_1.2,0.4311232089996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,2,power_law_1.2,0.4825471878051758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,2,power_law_1.2,0.4988224029541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,2,power_law_1.2,0.5895040035247803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,2,power_law_1.2,0.6444672107696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,2,power_law_1.2,0.7552063941955567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,2,power_law_1.2,0.9420096397399902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,2,power_law_1.2,1.2202816009521484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,2,power_law_1.2,1.5000191688537599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,2,power_law_1.2,1.9639360427856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,2,power_law_1.2,2.4753087997436523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,4,balanced,0.04199466605981191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,4,balanced,0.04185600082079569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,4,balanced,0.04572799801826477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,4,balanced,0.05996799965699514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,4,balanced,0.08113066852092743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,4,balanced,0.1197653313477834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,4,balanced,0.1204159955183665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,4,balanced,0.12078400452931722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,4,balanced,0.12181867162386577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,4,balanced,0.12237866719563802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,4,balanced,0.12174933155377705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,4,balanced,0.0444106658299764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,4,balanced,0.12682132919629416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,4,balanced,0.03364266703526179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,4,balanced,0.12821333607037863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,4,balanced,0.03380800038576126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,4,balanced,0.126991997162501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,4,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,4,balanced,0.035631999373435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,4,balanced,0.13643200198809305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,4,balanced,0.03736000011364619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,4,balanced,0.13766933480898538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,4,balanced,0.03573333223660787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,4,balanced,0.14288533727327982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,4,balanced,0.15850133697191873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,4,balanced,0.1661066710948944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,4,balanced,0.191103994846344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,4,balanced,0.2112320065498352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,4,balanced,0.2611413399378459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,4,balanced,0.3022986650466919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,4,balanced,0.4063839912414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,4,balanced,0.5009440183639526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,4,balanced,0.7071466445922852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,4,balanced,0.03589866558710734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,4,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,4,balanced,0.03774933268626531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,4,balanced,0.8888746897379557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,4,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,4,balanced,0.039434666434923805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,4,balanced,0.03751466671625773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,4,balanced,0.03982933362325033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,4,balanced,0.05132266879081726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,4,balanced,0.04969066878159841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,4,balanced,0.05107733110586802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,4,balanced,0.0621013343334198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,4,balanced,0.07026133437951405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,4,balanced,0.0776693324247996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,4,balanced,0.08090133468310039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,4,balanced,0.10725866754849751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,4,balanced,0.12388799587885539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,4,balanced,0.16033599774042764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,4,balanced,0.18849066893259683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,4,balanced,0.254421333471934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,4,balanced,0.32070932785669964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,8,power_law_1.01,0.026240000128746034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,8,power_law_1.01,0.04485119879245758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,8,power_law_1.01,0.04862079918384552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,8,power_law_1.01,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,8,power_law_1.01,0.0657920002937317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,8,power_law_1.01,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,8,power_law_1.01,0.06929280161857605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,8,power_law_1.01,0.06394240260124207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,8,power_law_1.01,0.07189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,8,power_law_1.01,0.07748479843139648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,8,power_law_1.01,0.07420160174369812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,8,power_law_1.01,0.09836159944534302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,8,power_law_1.01,0.0948032021522522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,8,power_law_1.01,0.09637119770050048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,8,power_law_1.01,0.1401087999343872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,8,power_law_1.01,0.14277119636535646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,8,power_law_1.01,0.14844800233840943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,8,power_law_1.01,0.1493056058883667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,8,power_law_1.01,0.14380160570144654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,8,power_law_1.01,0.15485440492630004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,8,power_law_1.01,0.15043840408325196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,8,power_law_1.01,0.18525439500808716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,8,power_law_1.01,0.20571520328521728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,8,power_law_1.01,0.2827647924423218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,8,power_law_1.01,0.3493119955062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,8,power_law_1.01,0.4850624084472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,8,power_law_1.01,0.6316415786743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,2,power_law_1.2,0.034835198521614076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,2,power_law_1.2,0.04013440012931824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,2,power_law_1.2,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,2,power_law_1.2,0.05455999970436096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,2,power_law_1.2,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,2,power_law_1.2,0.05692160129547119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,2,power_law_1.2,0.059468799829483034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,2,power_law_1.2,0.0599232017993927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,2,power_law_1.2,0.06232960224151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,2,power_law_1.2,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,2,power_law_1.2,0.06343039870262146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,2,power_law_1.2,0.06596480011940002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,2,power_law_1.2,0.06638079881668091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,2,power_law_1.2,0.07217919826507568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,2,power_law_1.2,0.08108159899711609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,2,power_law_1.2,0.08243839740753174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,2,power_law_1.2,0.08380159735679626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,2,power_law_1.2,0.0998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,2,power_law_1.2,0.107532799243927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,2,power_law_1.2,0.12549760341644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,2,power_law_1.2,0.14855680465698243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,2,power_law_1.2,0.1847040057182312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,2,power_law_1.2,0.20566399097442628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,2,power_law_1.2,0.3157952070236206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,32,8,balanced,0.0536106675863266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,32,8,balanced,0.05641066531340281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,32,8,balanced,0.058304001887639366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,32,8,balanced,0.06629333396752675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,32,8,balanced,0.07056533296902974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,32,8,balanced,0.09046933054924011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,32,8,balanced,0.09044800202051799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,32,8,balanced,0.0906826655069987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,32,8,balanced,0.09268266956011455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,32,8,balanced,0.09513066212336223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,32,8,balanced,0.09493866562843323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,32,8,balanced,0.09717866778373718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,32,8,balanced,0.09852266311645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,32,8,balanced,0.09686932961146037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,32,8,balanced,0.10196800033251445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,32,8,balanced,0.10739733775456746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,32,8,balanced,0.10578667124112447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,32,8,balanced,0.11593600114186604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,32,8,balanced,0.12359467148780823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,32,8,balanced,0.14065600434939066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,32,8,balanced,0.16090133786201477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,32,8,balanced,0.19585599501927695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,32,8,balanced,0.22933866580327353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,32,8,balanced,0.3201120098431905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,32,8,balanced,0.3914080063501994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,32,8,balanced,0.5514933268229166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,32,8,balanced,0.7078666687011719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,8,power_law_1.2,0.021376000344753267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,8,power_law_1.2,0.024371199309825897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,8,power_law_1.2,0.02571519911289215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,8,power_law_1.2,0.026700800657272337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,8,power_law_1.2,0.029049599170684816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,8,power_law_1.2,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,16,power_law_1.2,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,8,power_law_1.2,0.03368319869041443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,16,power_law_1.2,0.03079040050506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,8,power_law_1.2,0.03391999900341034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,16,power_law_1.2,0.03054719865322113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,16,power_law_1.2,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,16,power_law_1.2,0.033081600069999696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,16,power_law_1.2,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,8,power_law_1.2,0.03736959993839264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,16,power_law_1.2,0.03434880077838898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,16,power_law_1.2,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,8,power_law_1.2,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,8,power_law_1.2,0.03769600093364715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,8,power_law_1.2,0.03871360123157501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,16,power_law_1.2,0.0347135990858078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,16,power_law_1.2,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,2,power_law_1.2,0.39336960315704345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,16,power_law_1.2,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,16,power_law_1.2,0.036422398686408994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,16,power_law_1.2,0.0371071994304657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,16,power_law_1.2,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,16,power_law_1.2,0.04302079975605011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,8,power_law_1.2,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,16,power_law_1.2,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,16,power_law_1.2,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,16,power_law_1.2,0.05073919892311096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,16,power_law_1.2,0.061791998147964475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,16,power_law_1.2,0.07779840230941773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,8,power_law_1.2,0.048416000604629514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,8,power_law_1.2,0.051948797702789304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,16,power_law_1.2,0.0985152006149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,16,power_law_1.2,0.12186880111694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,8,power_law_1.2,0.055199998617172244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,8,power_law_1.2,0.05811840295791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,8,power_law_1.2,0.06917759776115417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,16,power_law_1.2,0.15302400588989257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.2,0.11132800579071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.2,0.12508800029754638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.2,0.11511679887771606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.2,0.132806396484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.2,0.159552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.2,0.21398398876190186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.2,0.26752638816833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,16,power_law_1.2,0.22659199237823485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.2,0.3699007987976074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,16,power_law_1.2,0.2745408058166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.2,0.46834559440612794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,1,power_law_1.01,0.10998400449752807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,1,power_law_1.01,0.13935359716415405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,1,power_law_1.01,0.1892416000366211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,1,power_law_1.01,0.2761663913726807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,1,power_law_1.01,0.3442944049835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,1,power_law_1.01,0.413702392578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,1,power_law_1.01,0.532371187210083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,16,power_law_1.2,0.41159682273864745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,1,power_law_1.01,0.5552832126617432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,1,power_law_1.01,0.5536640167236329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,1,power_law_1.01,0.5684544086456299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,1,power_law_1.01,0.5926720142364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,1,power_law_1.01,0.6044928073883057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,1,power_law_1.01,0.6200191974639893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,1,power_law_1.01,0.6142784118652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,1,power_law_1.01,0.631763219833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,1,power_law_1.01,0.6345344066619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,1,power_law_1.01,0.665235185623169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,1,power_law_1.01,0.713747215270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,1,power_law_1.01,0.7674496173858643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,1,power_law_1.01,0.874623966217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,1,power_law_1.01,0.9962688446044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,2,power_law_1.2,0.5315392017364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,1,power_law_1.01,1.2047360420227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,1,power_law_1.01,1.408409595489502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,1,power_law_1.01,1.81909122467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,1,power_law_1.01,2.226304054260254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,1,power_law_1.01,3.0407487869262697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,1,power_law_1.01,3.8550464630126955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,16,power_law_1.2,0.5780479907989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,2,power_law_1.2,0.7152703762054443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,128,balanced,0.04414399961630503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,128,balanced,0.05320533116658529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,128,balanced,0.04089066634575526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,128,balanced,0.04393066465854645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,128,balanced,0.04384533564249674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,128,balanced,0.04478933413823446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,128,balanced,0.04632533093293508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,128,balanced,0.047135998805363975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,128,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,128,balanced,0.04606399933497111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,128,balanced,0.04826133449872335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,128,balanced,0.04897066454092661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,128,balanced,0.04794666667779287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,128,balanced,0.05007466673851013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,128,balanced,0.05486933390299479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,128,balanced,0.054229333996772766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,128,balanced,0.054458667834599815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,128,balanced,0.05907199780146281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,128,balanced,0.06029866635799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,128,balanced,0.06840533514817555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,128,balanced,0.07468800246715546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,128,balanced,0.09116266171137492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,128,balanced,0.1016373336315155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,128,balanced,0.13215466340382895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,128,balanced,0.1623360017935435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,128,balanced,0.21899199485778809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,128,balanced,0.27609066168467206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,8,power_law_1.01,0.04599039852619171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,8,power_law_1.01,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,8,power_law_1.01,0.045817598700523376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,8,power_law_1.01,0.05015680193901062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,8,power_law_1.01,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,8,power_law_1.01,0.05394560098648071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,8,power_law_1.01,0.055430400371551516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,8,power_law_1.01,0.05596799850463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,8,power_law_1.01,0.05719040036201477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,8,power_law_1.01,0.05676159858703613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,8,power_law_1.01,0.058432000875473025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,8,power_law_1.01,0.062041598558425906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,8,power_law_1.01,0.06312959790229797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,8,power_law_1.01,0.0664896011352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,8,power_law_1.01,0.07434239983558655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,8,power_law_1.01,0.07921280264854431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,8,power_law_1.01,0.0796671986579895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,8,power_law_1.01,0.0939840018749237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,8,power_law_1.01,0.09745280146598816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,8,power_law_1.01,0.12646399736404418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,8,power_law_1.01,0.1533184051513672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,8,power_law_1.01,0.18696320056915283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,8,power_law_1.01,0.22304000854492187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,8,power_law_1.01,0.31000959873199463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,8,power_law_1.01,0.40316162109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,8,power_law_1.01,0.5210815906524658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,8,power_law_1.01,0.6985536098480225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,8,balanced,0.04224533339341482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,8,balanced,0.04366933306058248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,8,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,8,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,8,balanced,0.05946666498978933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,8,balanced,0.08378133177757263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,8,balanced,0.10017066200574239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,8,balanced,0.09893332918485005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,8,balanced,0.10080533226331075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,8,balanced,0.10193600257237752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,8,balanced,0.10051733255386353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,8,balanced,0.10389332969983418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,8,balanced,0.10638399918874104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,8,balanced,0.10972799857457478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,8,balanced,0.11857600013415019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,8,balanced,0.11969066659609477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,8,balanced,0.1262506643931071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,8,balanced,0.14136000474294028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,8,balanced,0.15852266550064087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,8,balanced,0.18486400445302328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,8,balanced,0.20957867304484049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,8,balanced,0.26206399997075397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,8,balanced,0.3302239974339803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,8,balanced,0.4336373408635457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,8,balanced,0.5597333510716757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,8,balanced,0.778112014134725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,8,balanced,1.0031466484069824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,64,power_law_1.01,0.08191360235214233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,64,power_law_1.01,0.06568319797515869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,64,power_law_1.01,0.067084801197052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,64,power_law_1.01,0.07110400199890136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,64,power_law_1.01,0.06686080098152161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,64,power_law_1.01,0.0709119975566864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,64,power_law_1.01,0.07372159957885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,64,power_law_1.01,0.07167360186576843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,64,power_law_1.01,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,64,power_law_1.01,0.073471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,64,power_law_1.01,0.07512320280075073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,64,power_law_1.01,0.07592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,64,power_law_1.01,0.07687039971351624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,64,power_law_1.01,0.07879679799079894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,64,power_law_1.01,0.08165760040283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,64,power_law_1.01,0.08826239705085755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,64,power_law_1.01,0.08824319839477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,64,power_law_1.01,0.10594559907913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,64,power_law_1.01,0.10616960525512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,64,power_law_1.01,0.1324031949043274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,64,power_law_1.01,0.14087680578231812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,64,power_law_1.01,0.1894144058227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,64,power_law_1.01,0.2097536087036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,64,power_law_1.01,0.28351359367370604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,64,power_law_1.01,0.34244480133056643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,64,power_law_1.01,0.49852800369262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,64,power_law_1.01,0.6219711780548096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,8,balanced,0.01926400015751521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,8,balanced,0.019152000546455383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,8,balanced,0.020848001043001812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,8,balanced,0.021141332884629566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,8,balanced,0.023306667804718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,8,balanced,0.02770666778087616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,8,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,8,balanced,0.02980799973011017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,8,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,32,1,balanced,0.021856000026067097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,32,1,balanced,0.023013333479563396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,32,1,balanced,0.02492266645034154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,32,1,balanced,0.029215998947620392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,32,1,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,32,1,balanced,0.06214933097362518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,32,1,balanced,0.07010133564472198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,32,1,balanced,0.07376533250013988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,32,1,balanced,0.07654400169849396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,32,1,balanced,0.07819200058778127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,32,1,balanced,0.07956799864768982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,32,1,balanced,0.08160000046094258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,32,1,balanced,0.0846666693687439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,32,1,balanced,0.08543999989827473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,32,1,balanced,0.08917333682378133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,32,1,balanced,0.09101333220799764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,32,1,balanced,0.09810133775075276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,8,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,32,1,balanced,0.11168000102043152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,8,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,32,1,balanced,0.12710400422414145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,8,balanced,0.04529066880544027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,8,balanced,0.04427733520666758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,8,balanced,0.045642669002215065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,8,balanced,0.06218666831652323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,8,balanced,0.0606826643149058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,8,balanced,0.06193066636721293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,8,balanced,0.07208533088366191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,8,balanced,0.07829866806666057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,8,balanced,0.0860746701558431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,8,balanced,0.09403199950853984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,8,balanced,0.11012267072995503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,8,balanced,0.1252959966659546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,8,balanced,0.15654399991035461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,8,balanced,0.18364266554514566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,8,balanced,0.27958933512369794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,8,balanced,0.3354826768239339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,32,1,balanced,0.1707786719004313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,32,1,balanced,0.19810134172439575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,32,1,balanced,0.27191466093063354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,32,1,balanced,0.30930133660634357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,32,1,balanced,0.35443735122680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,32,1,balanced,0.40490134557088214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,32,1,balanced,0.6345119873682658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,32,1,balanced,0.7462080319722494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,32,power_law_1.2,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,32,power_law_1.2,0.03091199994087219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,32,power_law_1.2,0.031564798951148984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,32,power_law_1.2,0.03221760094165802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,32,power_law_1.2,0.03255679905414581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,32,power_law_1.2,0.03392640054225922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,32,power_law_1.2,0.034227201342582704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,32,power_law_1.2,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,32,power_law_1.2,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,32,power_law_1.2,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,32,power_law_1.2,0.0379584014415741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,32,power_law_1.2,0.04015359878540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,32,power_law_1.2,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,32,power_law_1.2,0.0427264004945755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,32,power_law_1.2,0.04776319861412048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,32,power_law_1.2,0.049804800748825075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,32,power_law_1.2,0.05263360142707825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,32,power_law_1.2,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,32,power_law_1.2,0.06554880142211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,32,power_law_1.2,0.08447999954223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,32,power_law_1.2,0.10304000377655029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,32,power_law_1.2,0.1415295958518982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,32,power_law_1.2,0.1645248055458069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,32,power_law_1.2,0.22695679664611818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,32,power_law_1.2,0.27268478870391843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,2,power_law_1.2,0.045158401131629944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,2,power_law_1.2,0.046649599075317384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,2,power_law_1.2,0.04633600115776062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,2,power_law_1.2,0.05214080214500427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,2,power_law_1.2,0.057785600423812866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,2,power_law_1.2,0.06435199975967407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,2,power_law_1.2,0.07207679748535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,2,power_law_1.2,0.07630720138549804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,2,power_law_1.2,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,2,power_law_1.2,0.076665598154068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,2,power_law_1.2,0.08117759823799134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,2,power_law_1.2,0.08278399705886841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,2,power_law_1.2,0.08600959777832032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,2,power_law_1.2,0.09070079922676086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,2,power_law_1.2,0.09837440252304078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,2,power_law_1.2,0.10318080186843873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,2,power_law_1.2,0.1094208002090454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,2,power_law_1.2,0.12757760286331177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,2,power_law_1.2,0.13645440340042114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,2,power_law_1.2,0.16357760429382323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,2,power_law_1.2,0.18938239812850952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,2,power_law_1.2,0.254150390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,2,power_law_1.2,0.32077438831329347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,2,power_law_1.2,0.4521791934967041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,2,power_law_1.2,0.5539775848388672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,2,power_law_1.2,0.7798655986785888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,2,power_law_1.2,1.0127424240112304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,1,power_law_1.2,0.20789120197296143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,1,power_law_1.2,0.35447039604187014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,1,power_law_1.2,0.42070398330688474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,1,power_law_1.2,0.7119487762451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,1,power_law_1.2,0.72325119972229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,1,power_law_1.2,0.7287744045257568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,1,power_law_1.2,0.7338880062103271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,1,power_law_1.2,0.7541376113891601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,1,power_law_1.2,0.7678016185760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,1,power_law_1.2,0.7605760097503662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,1,power_law_1.2,0.7833343982696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,1,power_law_1.2,0.8126784324645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,1,power_law_1.2,0.8294719696044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,1,power_law_1.2,0.8666879653930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,1,power_law_1.2,0.8968000411987305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,1,power_law_1.2,0.9018495559692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,1,power_law_1.2,1.0591744422912597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,1,power_law_1.2,1.194752025604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,32,power_law_1.2,0.4683584213256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,1,power_law_1.2,1.2936384201049804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,1,power_law_1.2,1.5866559982299804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,1,power_law_1.2,1.8579647064208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,8,power_law_1.2,0.05866879820823669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,8,power_law_1.2,0.07400959730148315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,8,power_law_1.2,0.07713279724121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,8,power_law_1.2,0.08658559918403626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,8,power_law_1.2,0.09970560073852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,8,power_law_1.2,0.1043328046798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,8,power_law_1.2,0.10255359411239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,8,power_law_1.2,0.10375039577484131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,8,power_law_1.2,0.10756479501724243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,8,power_law_1.2,0.1098688006401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,8,power_law_1.2,0.10570240020751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,8,power_law_1.2,0.11369600296020507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,1,power_law_1.2,2.466156768798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,8,power_law_1.2,0.115174400806427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,8,power_law_1.2,0.11431039571762085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,8,power_law_1.2,0.1260800004005432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,8,power_law_1.2,0.12725119590759276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,8,power_law_1.2,0.13777920007705688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,8,power_law_1.2,0.15864319801330568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,8,power_law_1.2,0.15774719715118407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,8,power_law_1.2,0.1883455991744995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,8,power_law_1.2,0.2120192050933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,8,power_law_1.2,0.26391680240631105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,8,power_law_1.2,0.3254784107208252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,8,power_law_1.2,0.4448703765869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,8,power_law_1.2,0.5410111904144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,8,power_law_1.2,0.751635217666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,8,power_law_1.2,0.9464832305908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,64,balanced,0.02362666775782903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,64,balanced,0.02550400048494339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,64,balanced,0.02517866591612498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,64,balanced,0.0271573339899381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,64,balanced,0.02719466636578242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,64,balanced,0.02717866748571396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,64,balanced,0.028277332584063213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,64,balanced,0.02926933268706004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,64,balanced,0.02940800040960312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,64,balanced,0.029461334149042766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,64,balanced,0.03218133250872294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,64,balanced,0.0317546675602595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,64,balanced,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,64,balanced,0.033743999898433685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,64,balanced,0.03527999917666117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,64,balanced,0.035887998839219414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,64,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,64,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,64,balanced,0.03718933214743932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,64,balanced,0.038021333515644073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,64,balanced,0.039818666875362396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,64,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,64,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,64,balanced,0.05269333223501841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,64,balanced,0.05738666653633118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,64,balanced,0.06968533496061961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,64,balanced,0.082805335521698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,1,power_law_1.2,3.4508735656738283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,1,power_law_1.2,4.771686553955078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,32,power_law_1.2,0.6219456195831299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,1,power_law_1.2,6.256671905517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,1,power_law_1.2,8.653401947021484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,4,balanced,0.05946666498978933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,4,balanced,0.06071466704209646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,4,balanced,0.06485333542029063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,4,balanced,0.07283733288447063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,4,balanced,0.08316266536712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,4,balanced,0.11406399806340535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,4,balanced,0.14481600125630698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,4,balanced,0.14305599530537924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,4,balanced,0.14272000392278036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,4,balanced,0.14261333147684732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,4,balanced,0.1433013379573822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,4,balanced,0.146314670642217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,4,balanced,0.15122666954994202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,4,balanced,0.15036267042160034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,4,balanced,0.15948800245920816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,4,balanced,0.16080533464749655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,4,balanced,0.16768000523249307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,4,balanced,0.18307199080785116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,4,balanced,0.19698667526245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,4,balanced,0.22569066286087036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,4,balanced,0.25227200984954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,4,balanced,0.31488533814748126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,4,balanced,0.3792320092519124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,4,balanced,0.5078453222910563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,4,balanced,0.682965358098348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,4,balanced,0.9369173049926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,4,balanced,1.2430880069732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,1,balanced,0.053039997816085815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,1,balanced,0.055685331424077354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,1,balanced,0.06657066444555919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,1,balanced,0.09518399834632874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,1,balanced,0.1456000010172526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,1,balanced,0.14633066455523172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,1,balanced,0.14774399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,1,balanced,0.1511573294798533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,1,balanced,0.1518400013446808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,1,balanced,0.15318933129310608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,1,balanced,0.15870933731396994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,1,balanced,0.15853866934776306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,1,balanced,0.1644426683584849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,1,balanced,0.16880534092585245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,1,balanced,0.17628266414006552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,1,balanced,0.18336000045140585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,1,balanced,0.19839467604955038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,1,balanced,0.23257599274317423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,1,balanced,0.2595733404159546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,1,balanced,0.3274826606114705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,1,balanced,0.3753653367360433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,1,balanced,0.5495520035425822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,1,balanced,0.6614880164464315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,1,balanced,0.9756746292114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,1,balanced,1.247711976369222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,1,balanced,1.8189973831176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,1,balanced,2.404010613759359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,1,power_law_1.2,11.986860656738282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,64,balanced,0.03277866790692011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,64,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,64,balanced,0.03165333221356074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,64,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,64,balanced,0.03581333408753077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,64,balanced,0.03586133321126302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,64,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,64,balanced,0.03806933263937632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,64,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,64,balanced,0.03811733424663544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,64,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,64,balanced,0.041749333341916404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,64,balanced,0.04197333256403605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,64,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,64,balanced,0.04994133114814758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,64,balanced,0.048026666045188904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,64,balanced,0.050053333242734276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,64,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,64,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,64,balanced,0.06257600088914235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,64,balanced,0.06629866858323415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,64,balanced,0.08541333675384521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,64,balanced,0.09754133224487305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,64,balanced,0.12677866220474243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,64,balanced,0.15457066893577576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,64,balanced,0.20256000757217407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,64,balanced,0.2550026575724284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,2,power_law_1.01,0.02261119931936264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,2,power_law_1.01,0.02812800109386444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,2,power_law_1.01,0.04377599954605103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,2,power_law_1.01,0.060575997829437254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,2,power_law_1.01,0.08355200290679932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,2,power_law_1.01,0.0923520028591156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,2,power_law_1.01,0.10851839780807496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,2,power_law_1.01,0.11457279920578003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,2,power_law_1.01,0.12340480089187622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,2,power_law_1.01,0.12125439643859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,2,power_law_1.01,0.1244928002357483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,2,power_law_1.01,0.12777600288391114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,2,power_law_1.01,0.13545600175857545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,2,power_law_1.01,0.1351359963417053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,2,power_law_1.01,0.1392832040786743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,2,power_law_1.01,0.14485119581222533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,2,power_law_1.01,0.16389119625091553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,2,power_law_1.01,0.18736640214920045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.01,0.19741439819335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.01,0.2491136074066162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.01,0.24154880046844482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.01,0.3084223985671997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.01,0.3681152105331421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.01,0.46017918586730955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.01,0.5667967796325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.01,0.7317440032958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.01,0.9186944007873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,8,power_law_1.01,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,8,power_law_1.01,0.04588800072669983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,8,power_law_1.01,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,8,power_law_1.01,0.05283839702606201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,8,power_law_1.01,0.05898879766464234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,8,power_law_1.01,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,8,power_law_1.01,0.06183680295944214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,8,power_law_1.01,0.060089600086212155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,8,power_law_1.01,0.06313599944114685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,8,power_law_1.01,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,8,power_law_1.01,0.06380159854888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,8,power_law_1.01,0.06647679805755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,8,power_law_1.01,0.0677183985710144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,8,power_law_1.01,0.07084159851074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,8,power_law_1.01,0.07915520071983337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,8,power_law_1.01,0.08132479786872863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,8,power_law_1.01,0.08734719753265381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,8,power_law_1.01,0.09761279821395874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,8,power_law_1.01,0.10542080402374268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,8,power_law_1.01,0.12583039999008178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,8,power_law_1.01,0.1462272047996521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,8,power_law_1.01,0.18279680013656616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,8,power_law_1.01,0.23289599418640136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,8,power_law_1.01,0.3023616075515747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,8,power_law_1.01,0.3765376091003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,8,power_law_1.01,0.5604800224304199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,8,power_law_1.01,0.7349376201629638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,8,power_law_1.01,0.04023039937019348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,8,power_law_1.01,0.043110400438308716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,8,power_law_1.01,0.03989120125770569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,8,power_law_1.01,0.049465599656105044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,8,power_law_1.01,0.046598398685455324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,8,power_law_1.01,0.051507198810577394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,8,power_law_1.01,0.05187839865684509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,8,power_law_1.01,0.053907197713851926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,8,power_law_1.01,0.05457280278205871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,8,power_law_1.01,0.05370879769325256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,8,power_law_1.01,0.054579198360443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,8,power_law_1.01,0.05763840079307556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,8,power_law_1.01,0.05858560204505921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,8,power_law_1.01,0.060652798414230345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,8,power_law_1.01,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,8,power_law_1.01,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,8,power_law_1.01,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,8,power_law_1.01,0.07807360291481018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,8,power_law_1.01,0.09476479887962341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,8,power_law_1.01,0.11511039733886719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,8,power_law_1.01,0.13365119695663452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,8,power_law_1.01,0.17770240306854249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,8,power_law_1.01,0.2277888059616089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,8,power_law_1.01,0.3445888042449951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,8,power_law_1.01,0.4430016040802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,8,power_law_1.01,0.6694655895233155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,8,power_law_1.01,0.7948544025421143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,8,power_law_1.01,0.022668799757957457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,8,power_law_1.01,0.027590399980545043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,8,power_law_1.01,0.03856000006198883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,8,power_law_1.01,0.04266240000724793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,8,power_law_1.01,0.046342399716377256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,8,power_law_1.01,0.04938240051269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,8,power_law_1.01,0.05427839756011963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,8,power_law_1.01,0.049235200881958006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,8,power_law_1.01,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,8,power_law_1.01,0.05448960065841675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,8,power_law_1.01,0.056934398412704465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,8,power_law_1.01,0.06920959949493408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,8,power_law_1.01,0.06646400094032287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,8,power_law_1.01,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,8,power_law_1.01,0.10081919431686401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,8,power_law_1.01,0.1001471996307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,8,power_law_1.01,0.10046720504760742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,8,power_law_1.01,0.11523840427398682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.01,0.113811194896698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.01,0.12984319925308227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.01,0.13041919469833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.01,0.15864319801330568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.01,0.18597760200500488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.01,0.24498560428619384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.01,0.2972480058670044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.01,0.4268352031707764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.01,0.5259520053863526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,32,power_law_1.01,0.08227199912071229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,32,power_law_1.01,0.05511040091514587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,32,power_law_1.01,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,32,power_law_1.01,0.05572479963302612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,32,power_law_1.01,0.05735039710998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,32,power_law_1.01,0.06037759780883789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,32,power_law_1.01,0.05863680243492127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,32,power_law_1.01,0.058259201049804685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,32,power_law_1.01,0.06164479851722717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,32,power_law_1.01,0.060729598999023436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,32,power_law_1.01,0.06558719873428345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,32,power_law_1.01,0.0664896011352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,32,power_law_1.01,0.06620799899101257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,32,power_law_1.01,0.0701312005519867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,32,power_law_1.01,0.07800319790840149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,32,power_law_1.01,0.07905920147895813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,32,power_law_1.01,0.0804095983505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,32,power_law_1.01,0.09943040013313294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,32,power_law_1.01,0.09565439820289612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,32,power_law_1.01,0.11248639822006226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,32,power_law_1.01,0.12405760288238525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,32,power_law_1.01,0.15651839971542358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,32,power_law_1.01,0.1946176052093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,32,power_law_1.01,0.26638720035552976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,16,power_law_1.01,0.053504002094268796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,32,power_law_1.01,0.31676158905029295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,16,power_law_1.01,0.07214720249176025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,16,power_law_1.01,0.06517120003700257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,16,power_law_1.01,0.06938880085945129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,16,power_law_1.01,0.07009919881820678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,32,power_law_1.01,0.45080318450927737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,16,power_law_1.01,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,16,power_law_1.01,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,32,power_law_1.01,0.5551167964935303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,16,power_law_1.01,0.07033600211143494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,16,power_law_1.01,0.06817280054092408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,16,power_law_1.01,0.07221119999885559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,16,power_law_1.01,0.07496320009231568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,16,power_law_1.01,0.0763647973537445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,16,power_law_1.01,0.07923840284347534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,16,power_law_1.01,0.08127359747886657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,16,power_law_1.01,0.09147520065307617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,16,power_law_1.01,0.09356799721717834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,16,power_law_1.01,0.09895679950714112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,16,power_law_1.01,0.1106112003326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,16,power_law_1.01,0.11909760236740112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,16,power_law_1.01,0.14434560537338256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,16,power_law_1.01,0.16590720415115356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,16,power_law_1.01,0.21167359352111817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,16,power_law_1.01,0.2661439895629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,16,power_law_1.01,0.3461888074874878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,16,power_law_1.01,0.4580543994903564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,16,power_law_1.01,0.6399231910705566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,16,power_law_1.01,0.8220352172851563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,4,power_law_1.2,0.018668800592422485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,4,power_law_1.2,0.01880960017442703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,4,power_law_1.2,0.020076799392700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,4,power_law_1.2,0.02072319984436035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,4,power_law_1.2,0.021516799926757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,4,power_law_1.2,0.02385919988155365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,4,power_law_1.2,0.025542399287223815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,4,power_law_1.2,0.026745599508285523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,4,power_law_1.2,0.02805759906768799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,4,power_law_1.2,0.027558401226997375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,4,power_law_1.2,0.0282368004322052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,4,power_law_1.2,0.03488639891147614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,4,power_law_1.2,0.03676159977912903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,4,power_law_1.2,0.03457919955253601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,4,power_law_1.2,0.05186560153961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,4,power_law_1.2,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,4,power_law_1.2,0.06592000126838685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,4,power_law_1.2,0.06258559823036194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,4,power_law_1.2,0.06556159853935242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,4,power_law_1.2,0.07608960270881653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,4,power_law_1.2,0.08209919929504395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,4,power_law_1.2,0.10419199466705323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,4,power_law_1.2,0.12015999555587768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,4,power_law_1.2,0.16072959899902345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,4,power_law_1.2,0.1960576057434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,4,power_law_1.2,0.2796351909637451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,16,power_law_1.01,0.0726144015789032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,16,power_law_1.01,0.0717248022556305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,16,power_law_1.01,0.06524800062179566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,16,power_law_1.01,0.0717631995677948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,16,power_law_1.01,0.06837120056152343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,16,power_law_1.01,0.07288960218429566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,16,power_law_1.01,0.07370240092277527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,16,power_law_1.01,0.07531520128250122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,16,power_law_1.01,0.07475839853286743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,16,power_law_1.01,0.07637119889259339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,16,power_law_1.01,0.07678719758987426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,16,power_law_1.01,0.07663360238075256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,16,power_law_1.01,0.07985919713973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,16,power_law_1.01,0.08197759985923767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,16,power_law_1.01,0.08799359798431397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,16,power_law_1.01,0.08999680280685425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,16,power_law_1.01,0.09719039797782898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,16,power_law_1.01,0.10855679512023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.01,0.12007679939270019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.01,0.1380288004875183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.01,0.1624511957168579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.01,0.20061440467834474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.01,0.2523008108139038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.01,0.32934401035308836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.01,0.3945280075073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,4,power_law_1.2,0.048230400681495665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.01,0.5394688129425049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,4,power_law_1.2,0.05246719717979431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.01,0.754201602935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,4,power_law_1.2,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,4,power_law_1.2,0.07720320224761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,4,power_law_1.2,0.08315520286560059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,4,power_law_1.2,0.08503040075302123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,4,power_law_1.2,0.10734080076217652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,4,power_law_1.2,0.1041856050491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,4,power_law_1.2,0.10297600030899048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,4,power_law_1.2,0.10670080184936523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,4,power_law_1.2,0.11072640419006348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,4,power_law_1.2,0.11934720277786255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,4,power_law_1.2,0.12190719842910766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,4,power_law_1.2,0.1339903950691223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,4,power_law_1.2,0.14324480295181274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,4,power_law_1.2,0.15244799852371216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,4,power_law_1.2,0.15600639581680298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,4,power_law_1.2,0.17381759881973266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,4,power_law_1.2,0.19503359794616698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,4,power_law_1.2,0.23619840145111085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,4,power_law_1.2,0.27905280590057374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,4,power_law_1.2,0.37610878944396975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,4,power_law_1.2,0.4440320014953613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,4,power_law_1.2,0.6546304225921631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,4,power_law_1.2,0.8171520233154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,4,power_law_1.2,1.1973055839538573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,4,power_law_1.2,0.349183988571167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,4,power_law_1.2,1.5249024391174317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,32,balanced,0.052709331115086876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,32,balanced,0.053823997577031456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,32,balanced,0.05378133555253347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,32,balanced,0.06057600180308024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,32,balanced,0.0664160003264745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,32,balanced,0.07694933315118153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,32,balanced,0.07693866888682048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,32,balanced,0.07624533275763194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,32,balanced,0.0763733337322871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,32,balanced,0.0761706680059433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,32,balanced,0.0746666689713796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,32,balanced,0.07750399907430013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,32,balanced,0.07937066753705342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,32,balanced,0.07444266478220622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,32,balanced,0.08412266770998637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,32,balanced,0.0825493335723877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,32,balanced,0.08509866396586101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,32,balanced,0.0969493289788564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,32,balanced,0.10114133358001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,32,balanced,0.11588799953460693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,32,balanced,0.12989333271980286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,32,balanced,0.15781866510709128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,32,balanced,0.18288000424702963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,32,balanced,0.2457866668701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,32,balanced,0.2996480067571004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,32,balanced,0.4124373197555542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,32,balanced,0.5281173388163248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,16,balanced,0.04164800047874451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,16,balanced,0.04194133480389913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,16,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,16,balanced,0.04764799773693085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,16,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,16,balanced,0.05388266841570536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,16,balanced,0.053914666175842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,16,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,16,balanced,0.056186666091283165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,16,balanced,0.05793599784374237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,16,balanced,0.05797866483529409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,16,balanced,0.058933332562446594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,16,balanced,0.06190933287143707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,16,balanced,0.06062399844328562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,16,balanced,0.06634666522343953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,16,balanced,0.06823466718196869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,16,balanced,0.07117866476376851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,16,balanced,0.081194669008255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,16,balanced,0.08866666754086812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,16,balanced,0.11550399661064148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,16,balanced,0.13567999998728433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,16,balanced,0.18085867166519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,16,balanced,0.21556266148885092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,16,balanced,0.29445334275563556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,16,balanced,0.3760053316752116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,16,balanced,0.5265973409016927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,16,balanced,0.6787467002868652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,8,balanced,0.03314133236805598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,8,balanced,0.03221333275238673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,8,balanced,0.029343999922275543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,8,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,8,balanced,0.029530666768550873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,8,balanced,0.031557333966096245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,8,balanced,0.031541332602500916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,8,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,8,balanced,0.03162133445342382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,8,balanced,0.031632001201311745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,8,balanced,0.03149333347876867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,8,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,8,balanced,0.031685332457224526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,8,balanced,0.03294399877389272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,8,balanced,0.03764266769091288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,8,balanced,0.03736000011364619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,8,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,8,balanced,0.04005866746107737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,8,balanced,0.04145599901676178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,8,balanced,0.047557334105173744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,8,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,8,balanced,0.060565332571665444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,8,balanced,0.0636106679836909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,8,balanced,0.08793600400288899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,8,balanced,0.1011306643486023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,8,balanced,0.13222400347391763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,8,balanced,0.14941333731015524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,8,power_law_1.01,0.056953597068786624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,8,power_law_1.01,0.05799679756164551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,8,power_law_1.01,0.05606399774551392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,8,power_law_1.01,0.05715199708938599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,8,power_law_1.01,0.05936639904975891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,8,power_law_1.01,0.06669440269470214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,8,power_law_1.01,0.07949439883232116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,8,power_law_1.01,0.07299200296401978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,8,power_law_1.01,0.0763264000415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,8,power_law_1.01,0.2778559923171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,8,power_law_1.01,0.06904320120811462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,8,power_law_1.01,0.07461760044097901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,8,power_law_1.01,0.07454720139503479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,8,power_law_1.01,0.08579840064048767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,8,power_law_1.01,0.10119680166244507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,8,power_law_1.01,0.11691520214080811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,8,power_law_1.01,0.11559040546417236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,8,power_law_1.01,0.1388543963432312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,8,power_law_1.01,0.1643903970718384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,8,power_law_1.01,0.2244800090789795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,8,power_law_1.01,0.2371392011642456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,4,power_law_1.2,0.04896639883518219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,4,power_law_1.2,0.05141119956970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,4,power_law_1.2,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,4,power_law_1.2,0.07147520184516906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,4,power_law_1.2,0.08156160116195679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,4,power_law_1.2,0.08892800211906433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,4,power_law_1.2,0.10533119440078735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,4,power_law_1.2,0.10294400453567505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,4,power_law_1.2,0.10520960092544555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,4,power_law_1.2,0.10535039901733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,4,power_law_1.2,0.11202559471130372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,1,power_law_1.2,0.05270400047302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,4,power_law_1.2,0.11034239530563354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,1,power_law_1.2,0.054201602935791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,8,power_law_1.01,0.37880959510803225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,4,power_law_1.2,0.1158911943435669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,4,power_law_1.2,0.12015999555587768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,4,power_law_1.2,0.13066240549087524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,4,power_law_1.2,0.13939839601516724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,4,power_law_1.2,0.15178879499435424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,4,power_law_1.2,0.17609599828720093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,4,power_law_1.2,0.18137600421905517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,4,power_law_1.2,0.21935999393463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,4,power_law_1.2,0.26264960765838624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,8,power_law_1.01,0.4366591930389404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,1,power_law_1.2,0.05783680081367493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,1,power_law_1.2,0.07277439832687378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,4,power_law_1.2,0.32364161014556886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,1,power_law_1.2,0.08689919710159302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,4,power_law_1.2,0.39818880558013914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,1,power_law_1.2,0.09942399859428405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,4,power_law_1.2,0.5463488101959229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,1,power_law_1.2,0.1211967945098877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,1,power_law_1.2,0.12496639490127563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,4,power_law_1.2,0.683289623260498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,1,power_law_1.2,0.12914559841156006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,1,power_law_1.2,0.13270399570465088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,4,power_law_1.2,1.0255295753479003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,1,power_law_1.2,0.14048000574111938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,1,power_law_1.2,0.14935679435729982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,4,power_law_1.2,1.217414379119873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,1,power_law_1.2,0.15522559881210327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,1,power_law_1.2,0.16305279731750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,1,power_law_1.2,0.17649279832839965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,1,power_law_1.2,0.18595839738845826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,1,power_law_1.2,0.19071359634399415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,1,power_law_1.2,0.22311038970947267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,1,power_law_1.2,0.25257599353790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,1,power_law_1.2,0.31905279159545896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,8,power_law_1.01,0.6688960075378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,1,power_law_1.2,0.3731967926025391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,1,power_law_1.2,0.4953919887542725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,1,power_law_1.2,0.6204544067382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,8,power_law_1.01,0.8805312156677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,1,power_law_1.2,0.8478976249694824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,1,power_law_1.2,1.0870976448059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,1,power_law_1.2,1.5528191566467284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,1,power_law_1.2,2.0299583435058595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,8,power_law_1.01,1.356761646270752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,8,power_law_1.01,1.614784049987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,32,power_law_1.2,0.03324800133705139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,32,power_law_1.2,0.030457600951194763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,32,power_law_1.2,0.03182719945907593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,32,power_law_1.2,0.03182719945907593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,32,power_law_1.2,0.03229439854621887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,32,power_law_1.2,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,32,power_law_1.2,0.03423359990119934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,32,power_law_1.2,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,32,power_law_1.2,0.03549439907073974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,32,power_law_1.2,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,32,power_law_1.2,0.0371071994304657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,32,power_law_1.2,0.03848319947719574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,32,power_law_1.2,0.03880319893360138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,32,power_law_1.2,0.03991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,32,power_law_1.2,0.04467839896678925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,32,power_law_1.2,0.047865599393844604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,32,power_law_1.2,0.04895359873771667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,32,power_law_1.2,0.05314559936523437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,32,power_law_1.2,0.05928959846496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,32,power_law_1.2,0.06859520077705383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,32,power_law_1.2,0.08737919926643371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,32,power_law_1.2,0.11948800086975098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,32,power_law_1.2,0.15178240537643434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,32,power_law_1.2,0.19495680332183837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,16,power_law_1.01,0.06516479849815368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,16,power_law_1.01,0.08152959942817688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,16,power_law_1.01,0.0791808009147644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,16,power_law_1.01,0.08275840282440186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,16,power_law_1.01,0.08384640216827392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,16,power_law_1.01,0.07680000066757202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,16,power_law_1.01,0.08320639729499817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,16,power_law_1.01,0.08605440258979798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,16,power_law_1.01,0.08491520285606384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,32,power_law_1.2,0.295743989944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,16,power_law_1.01,0.08487039804458618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,16,power_law_1.01,0.08680959939956664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,16,power_law_1.01,0.08729599714279175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,16,power_law_1.01,0.08743680119514466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,16,power_law_1.01,0.09040639996528625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,16,power_law_1.01,0.09838719964027405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,16,power_law_1.01,0.09904639720916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,16,power_law_1.01,0.10246399641036988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,16,power_law_1.01,0.11907199621200562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,16,power_law_1.01,0.12632319927215577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,16,power_law_1.01,0.14711040258407593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,16,power_law_1.01,0.16874239444732667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,16,power_law_1.01,0.22651519775390624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,16,power_law_1.01,0.26640000343322756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,16,power_law_1.01,0.3926336050033569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,16,power_law_1.01,0.4921728134155273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,16,power_law_1.01,0.7187200069427491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,16,power_law_1.01,0.9328448295593261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,32,power_law_1.2,0.3830591917037964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,32,power_law_1.2,0.5188159942626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,8,balanced,0.037871999045213066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,8,balanced,0.04154133299986521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,8,balanced,0.04141333450873693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,8,balanced,0.045824001232783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,8,balanced,0.04762666424115499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,8,balanced,0.06729066868623097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,8,balanced,0.06843733290831248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,8,balanced,0.0683840016523997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,8,balanced,0.06845333178838094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,8,balanced,0.0687360018491745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,8,balanced,0.069733331600825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,8,balanced,0.07155199845631917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,8,balanced,0.0732479989528656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,8,balanced,0.0734986662864685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,8,balanced,0.08192533254623413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,8,balanced,0.08225599924723308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,8,balanced,0.08515200018882751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,8,balanced,0.0949173370997111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,8,balanced,0.10142933328946431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,8,balanced,0.1178986628850301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,8,balanced,0.13101866841316223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,8,balanced,0.1711039940516154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,8,balanced,0.19918400049209595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,8,balanced,0.27079466978708905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,8,balanced,0.33084799846013385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,8,balanced,0.462010661760966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,8,balanced,0.5940373341242472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,2,balanced,0.06331199904282887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,2,balanced,0.08545600374539693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,2,balanced,0.13121599952379862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,2,balanced,0.21869333585103354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,2,balanced,0.38470399379730225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,2,balanced,0.7269919713338217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,2,balanced,0.7282506624857584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,2,balanced,0.7275893688201904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,2,balanced,0.7271573543548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,2,balanced,0.7282826900482178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,2,balanced,0.7374560038248698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,2,balanced,0.7375466823577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,2,balanced,0.7413653532663981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,2,balanced,0.7476373513539633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,2,balanced,0.7530773480733236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,2,balanced,0.7600906689961752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,2,balanced,0.7705173492431641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,2,balanced,0.7927573521931967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,2,balanced,0.8269546826680502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,2,balanced,0.8614346981048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,2,balanced,0.9112106959025065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,2,balanced,0.983184019724528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,2,balanced,1.083957354227702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,2,balanced,1.2752479712168376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,2,balanced,1.3985439936319988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,2,balanced,2.12390931447347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,2,balanced,2.251413345336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,16,power_law_1.01,0.05233280062675476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,16,power_law_1.01,0.03868800103664398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,16,power_law_1.01,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,16,power_law_1.01,0.03741439878940582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,16,power_law_1.01,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,16,power_law_1.01,0.037273600697517395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,16,power_law_1.01,0.037567999958992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,16,power_law_1.01,0.03688960075378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,16,power_law_1.01,0.03711360096931458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,16,power_law_1.01,0.03856000006198883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,16,power_law_1.01,0.04023039937019348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,16,power_law_1.01,0.04223999977111816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,16,power_law_1.01,0.0430400013923645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,16,power_law_1.01,0.04590719938278198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,16,power_law_1.01,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,16,power_law_1.01,0.05374079942703247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,16,power_law_1.01,0.053855997323989865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,16,power_law_1.01,0.061286401748657224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,16,power_law_1.01,0.06762239933013917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,16,power_law_1.01,0.07664639949798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,16,power_law_1.01,0.0820032000541687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,16,power_law_1.01,0.10547200441360474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,16,power_law_1.01,0.11458560228347778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,16,power_law_1.01,0.061286401748657224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,16,power_law_1.01,0.14878720045089722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,16,power_law_1.01,0.06929919719696045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,16,power_law_1.01,0.07958400249481201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,16,power_law_1.01,0.08065919876098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,16,power_law_1.01,0.08156800270080566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,16,power_law_1.01,0.08177279829978942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,16,power_law_1.01,0.07939199805259704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,16,power_law_1.01,0.08369280099868774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,16,power_law_1.01,0.08232960104942322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,16,power_law_1.01,0.1890176057815552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,16,power_law_1.01,0.0812928020954132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,16,power_law_1.01,0.26431999206542967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,16,power_law_1.01,0.08179200291633607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,16,power_law_1.01,0.08723199963569642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,16,power_law_1.01,0.08566399812698364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,16,power_law_1.01,0.09036160111427308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,16,power_law_1.01,0.09963520169258118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,16,power_law_1.01,0.09915519952774048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,16,power_law_1.01,0.32118399143218995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,16,power_law_1.01,0.10780160427093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,16,power_law_1.01,0.117958402633667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,16,power_law_1.01,0.1254207968711853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,16,power_law_1.01,0.14901119470596313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,16,power_law_1.01,0.17658239603042603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,16,power_law_1.01,0.2205888032913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,16,power_law_1.01,0.2834815979003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,16,power_law_1.01,0.39080960750579835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,16,power_law_1.01,0.4867839813232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,16,power_law_1.01,0.7708864212036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,16,power_law_1.01,1.0368255615234374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,8,power_law_1.01,0.026796799898147584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,8,power_law_1.01,0.027539199590682982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,8,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,8,power_law_1.01,0.04159359931945801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,8,power_law_1.01,0.042131200432777405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,8,power_law_1.01,0.045407998561859134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,8,power_law_1.01,0.045433598756790164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,8,power_law_1.01,0.04517120122909546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,8,power_law_1.01,0.047443199157714847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,8,power_law_1.01,0.047040000557899475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,8,power_law_1.01,0.053311997652053834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,8,power_law_1.01,0.05436800122261047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,8,power_law_1.01,0.056908798217773435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,8,power_law_1.01,0.05498239994049072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,8,power_law_1.01,0.05726720094680786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,8,power_law_1.01,0.06230400204658508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,8,power_law_1.01,0.06382719874382019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,8,power_law_1.01,0.07219200134277344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,8,power_law_1.01,0.1254464030265808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,8,power_law_1.01,0.13850239515304566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,8,power_law_1.01,0.12503039836883545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,8,power_law_1.01,0.1435520052909851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,8,power_law_1.01,0.1664639949798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,8,power_law_1.01,0.21920640468597413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,8,power_law_1.01,0.2661504030227661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,8,power_law_1.01,0.3620032072067261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,8,power_law_1.01,0.481171178817749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,256,power_law_1.2,0.04549759924411774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,256,power_law_1.2,0.045875200629234315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,256,power_law_1.2,0.045433598756790164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,256,power_law_1.2,0.04842880070209503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,256,power_law_1.2,0.049516800045967105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,256,power_law_1.2,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,8,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,8,balanced,0.04445866743723551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,8,balanced,0.05367999772230784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,8,balanced,0.0650186687707901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,8,balanced,0.08948799967765808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,8,balanced,0.10409067074457805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,8,balanced,0.12246400117874146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,8,balanced,0.1181066632270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,8,balanced,0.12437333663304646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,8,balanced,0.17832533518473306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,8,balanced,0.1616159975528717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,8,balanced,0.17320533593495688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,8,balanced,0.22339733441670737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,8,balanced,0.23014400402704874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,256,power_law_1.2,0.05324159860610962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,8,balanced,0.25838400920232135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,8,balanced,0.2362826665242513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,8,balanced,0.2241226633389791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,8,balanced,0.23923200368881226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,8,balanced,0.26577067375183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,8,balanced,0.30669333537419635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,8,balanced,0.3468213478724162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,8,balanced,0.4155306816101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,8,balanced,0.4737173318862915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,8,balanced,0.7432746887207031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,8,balanced,0.8989439805348715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,256,power_law_1.2,0.05780479907989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,8,balanced,1.3233439922332764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,8,balanced,1.7246185938517253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,256,power_law_1.2,0.05939199924468994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,64,power_law_1.2,0.02250239998102188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,64,power_law_1.2,0.021907199919223786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,64,power_law_1.2,0.02218240052461624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,256,power_law_1.2,0.060844802856445314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,64,power_law_1.2,0.023680000007152556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,256,power_law_1.2,0.05813760161399841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,256,power_law_1.2,0.06146559715270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,256,power_law_1.2,0.06382719874382019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,256,power_law_1.2,0.06218240261077881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,256,power_law_1.2,0.07521920204162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,256,power_law_1.2,0.07537279725074768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,64,power_law_1.2,0.024736000597476958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,64,power_law_1.2,0.025497600436210632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,256,power_law_1.2,0.08058879971504211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,64,power_law_1.2,0.025433599948883057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,64,power_law_1.2,0.026156800985336303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,64,power_law_1.2,0.02928000092506409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,64,power_law_1.2,0.029158401489257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,64,power_law_1.2,0.030937600135803222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,64,power_law_1.2,0.03162240087985992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,64,power_law_1.2,0.031699201464653014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,64,power_law_1.2,0.030483201146125793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,64,power_law_1.2,0.03478400111198425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,64,power_law_1.2,0.034406399726867674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,64,power_law_1.2,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,64,power_law_1.2,0.03925760090351105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,64,power_law_1.2,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,64,power_law_1.2,0.050483202934265135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,256,power_law_1.2,0.09148160219192505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,64,power_law_1.2,0.053958398103713986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,256,power_law_1.2,0.10546560287475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,64,power_law_1.2,0.07468799948692321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,64,power_law_1.2,0.07312639951705932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,256,power_law_1.2,0.13590400218963622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,256,power_law_1.2,0.15761280059814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,256,power_law_1.2,0.24098560810089112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,256,power_law_1.2,0.28378241062164306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,256,power_law_1.2,0.4375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,256,power_law_1.2,0.5971392154693603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,64,power_law_1.2,0.0979200005531311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,256,power_law_1.2,0.9554368019104004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,256,power_law_1.2,1.3848447799682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,64,power_law_1.2,0.13299200534820557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,64,power_law_1.2,0.17623039484024047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,32,balanced,0.03352533280849457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,32,balanced,0.029616000751654308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,32,balanced,0.03176533430814743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,32,balanced,0.03456533451875051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,32,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,32,balanced,0.036015999813874565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,32,balanced,0.03536533315976461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,32,balanced,0.038133333126703896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,32,balanced,0.03740799923737844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,32,balanced,0.037130666275819145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,32,balanced,0.03738133360942205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,32,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,32,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,32,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,32,balanced,0.041434665520985924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,32,balanced,0.04363200068473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,32,balanced,0.044549331068992615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,32,balanced,0.046037331223487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,32,balanced,0.047781333327293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,32,balanced,0.05793599784374237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,32,balanced,0.06487466891606648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,32,balanced,0.08821866909662883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,32,balanced,0.10473066568374634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,32,balanced,0.13820800185203552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,32,balanced,0.17081065972646078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,32,balanced,0.2264853318532308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,32,balanced,0.2855839927991231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,16,power_law_1.2,0.05565440058708191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,16,power_law_1.2,0.07642880082130432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,16,power_law_1.2,0.06327679753303528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,16,power_law_1.2,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,16,power_law_1.2,0.06712960004806519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,16,power_law_1.2,0.06967039704322815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,16,power_law_1.2,0.06773120164871216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,16,power_law_1.2,0.0695680022239685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,16,power_law_1.2,0.07158399820327759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,16,power_law_1.2,0.07074559926986694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,16,power_law_1.2,0.07190399765968322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,64,power_law_1.2,0.20609920024871825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,16,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,16,power_law_1.2,0.07507839798927307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,16,power_law_1.2,0.07971199750900268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,16,power_law_1.2,0.08666880130767822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,16,power_law_1.2,0.08606719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,16,power_law_1.2,0.09765120148658753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,16,power_law_1.2,0.11074559688568116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,16,power_law_1.2,0.12621439695358277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,16,power_law_1.2,0.15585919618606567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,16,power_law_1.2,0.17990399599075318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,16,power_law_1.2,0.2434623956680298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,16,power_law_1.2,0.2667135953903198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,16,power_law_1.2,0.35917439460754397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,16,power_law_1.2,0.4604479789733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,16,power_law_1.2,0.6661503791809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,16,power_law_1.2,0.9044351577758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,4,power_law_1.01,0.08364160060882568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,4,power_law_1.01,0.09874560236930847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,4,power_law_1.01,0.10951039791107178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,4,power_law_1.01,0.12892800569534302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,4,power_law_1.01,0.13598719835281373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,4,power_law_1.01,0.15374720096588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,4,power_law_1.01,0.1707584023475647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,4,power_law_1.01,0.17366399765014648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,4,power_law_1.01,0.17114239931106567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,4,power_law_1.01,0.16917120218276976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,4,power_law_1.01,0.18362879753112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,4,power_law_1.01,0.17816959619522094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,4,power_law_1.01,0.18819199800491332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,4,power_law_1.01,0.18932479619979858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,4,power_law_1.01,0.19219839572906494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,4,power_law_1.01,0.19921280145645143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,4,power_law_1.01,0.2089855909347534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,4,power_law_1.01,0.22553598880767822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,4,power_law_1.01,0.24006400108337403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,4,power_law_1.01,0.2771008014678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,4,power_law_1.01,0.31958398818969724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,4,power_law_1.01,0.41208958625793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,4,power_law_1.01,0.5262784004211426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,4,power_law_1.01,0.6933440208435059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,4,power_law_1.01,0.8940095901489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,4,power_law_1.01,1.2723008155822755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,4,power_law_1.01,1.664352035522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,64,balanced,0.15266666809717813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,64,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,64,balanced,0.04437866806983948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,64,balanced,0.04797333478927612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,64,balanced,0.050213331977526345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,64,balanced,0.07629866898059845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,64,balanced,0.09061333537101746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,64,balanced,0.09118400017420451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,64,balanced,0.0925546685854594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,64,balanced,0.09354133407274882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,64,balanced,0.09268266956011455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,64,balanced,0.09477866689364116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,64,balanced,0.09723200400670369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,64,balanced,0.0981173316637675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,64,balanced,0.10633066296577454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,64,balanced,0.10622400045394897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,64,balanced,0.11314133803049724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,64,balanced,0.12238400181134541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,64,balanced,0.1317866643269857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,64,balanced,0.15270400047302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,64,balanced,0.1704960068066915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,64,balanced,0.21041067441304526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,64,balanced,0.250325342019399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,64,balanced,0.3253386616706848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,64,balanced,0.41333333651224774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,64,balanced,0.5613280137379965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,64,balanced,0.7204693158467611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,4,power_law_1.01,0.061791998147964475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,4,power_law_1.01,0.06663039922714234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,4,power_law_1.01,0.08087040185928344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,4,power_law_1.01,0.084307199716568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,4,power_law_1.01,0.09749119877815246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,4,power_law_1.01,0.10748800039291381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,4,power_law_1.01,0.12045439481735229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,4,power_law_1.01,0.1292672038078308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,4,power_law_1.01,0.12355200052261353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,4,power_law_1.01,0.12345600128173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,4,power_law_1.01,0.1304128050804138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,4,power_law_1.01,0.13674880266189576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,4,power_law_1.01,0.13725440502166747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,4,power_law_1.01,0.13705600500106813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,4,power_law_1.01,0.15094399452209473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,4,power_law_1.01,0.14998400211334229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,4,power_law_1.01,0.15952639579772948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,4,power_law_1.01,0.17660160064697267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,4,power_law_1.01,0.1958400011062622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,4,power_law_1.01,0.23369600772857665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,4,power_law_1.01,0.27014400959014895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,4,power_law_1.01,0.3380863904953003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,4,power_law_1.01,0.41937918663024903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,4,power_law_1.01,0.5826752185821533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,4,power_law_1.01,0.7500224113464355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,4,power_law_1.01,1.0305536270141602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,4,power_law_1.01,1.3667008399963378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,128,power_law_1.01,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,128,power_law_1.01,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,128,power_law_1.01,0.057132798433303836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,128,power_law_1.01,0.05862399935722351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,128,power_law_1.01,0.05950719714164734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,128,power_law_1.01,0.05982720255851746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,128,power_law_1.01,0.0599232017993927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,128,power_law_1.01,0.06027519702911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,128,power_law_1.01,0.06149119734764099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,128,power_law_1.01,0.06103039979934692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,128,power_law_1.01,0.06269440054893494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,128,power_law_1.01,0.06229760050773621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,128,power_law_1.01,0.06295040249824524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,128,power_law_1.01,0.06391680240631104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,128,power_law_1.01,0.06970239877700805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,128,power_law_1.01,0.06968320012092591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,128,power_law_1.01,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,128,power_law_1.01,0.08287360072135926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,128,power_law_1.01,0.09233279824256897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,128,power_law_1.01,0.1156607985496521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,128,power_law_1.01,0.1360640048980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,128,power_law_1.01,0.17585279941558837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,128,power_law_1.01,0.21966080665588378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,128,power_law_1.01,0.2904191970825195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,128,power_law_1.01,0.3895872116088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,128,power_law_1.01,0.543839979171753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,128,power_law_1.01,0.7464320182800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,8,power_law_1.01,0.04292480051517487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,8,power_law_1.01,0.046214398741722104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,8,power_law_1.01,0.04648320078849792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,8,power_law_1.01,0.050432002544403075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,8,power_law_1.01,0.05381119847297668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,8,power_law_1.01,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,8,power_law_1.01,0.05545600056648255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,8,power_law_1.01,0.05424000024795532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,8,power_law_1.01,0.05484799742698669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,8,power_law_1.01,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,8,power_law_1.01,0.05680000185966492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,8,power_law_1.01,0.06048640012741089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,8,power_law_1.01,0.06119679808616638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,8,power_law_1.01,0.06521599888801574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,8,power_law_1.01,0.07219200134277344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,8,power_law_1.01,0.07318400144577027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,8,power_law_1.01,0.07520639896392822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,8,power_law_1.01,0.08541439771652222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,8,power_law_1.01,0.09354239702224731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,8,power_law_1.01,0.11492480039596557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,4,balanced,0.036159999668598175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,8,power_law_1.01,0.13107839822769166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,4,balanced,0.036090667049090065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,4,balanced,0.039877332746982574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,8,power_law_1.01,0.16879359483718873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,4,balanced,0.040591999888420105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,4,balanced,0.040821333726247154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,4,balanced,0.04115733255942663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,4,balanced,0.040149333576361336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,4,balanced,0.0421066681543986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,4,balanced,0.0420959989229838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,4,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,4,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,4,balanced,0.04165333261092504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,4,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,4,balanced,0.041477332512537636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,4,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,4,balanced,0.04594666759173075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,4,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,4,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,4,balanced,0.05407466491063436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,4,balanced,0.06586666901906331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,4,balanced,0.07039466500282288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,4,balanced,0.08825066685676575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,4,balanced,0.10317867000897725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,4,balanced,0.13566399614016214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,4,balanced,0.166703999042511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,4,balanced,0.22860799233118692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,4,balanced,0.2929333249727885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,8,power_law_1.01,0.20863358974456786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,8,power_law_1.01,0.30040318965911866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,8,power_law_1.01,0.3866624116897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,8,power_law_1.01,0.5101696014404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,8,power_law_1.01,0.6463679790496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,4,power_law_1.2,0.021158400177955627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,4,power_law_1.2,0.02385919988155365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,4,power_law_1.2,0.026502400636672974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,4,power_law_1.2,0.033292800188064575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,4,power_law_1.2,0.04119040071964264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,4,power_law_1.2,0.04450559914112091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,4,power_law_1.2,0.052160000801086424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,4,power_law_1.2,0.052665597200393675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,4,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,4,power_law_1.2,0.05381119847297668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,4,power_law_1.2,0.05625600218772888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,4,power_law_1.2,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,4,power_law_1.2,0.05779839754104614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,4,power_law_1.2,0.06110720038414001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,4,power_law_1.2,0.06646400094032287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,4,power_law_1.2,0.07093759775161743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,4,power_law_1.2,0.07431679964065552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,4,power_law_1.2,0.09427199959754944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.2,0.09849600195884704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.2,0.12233599424362182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.2,0.1579967975616455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.2,0.17851519584655762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.2,0.20624001026153566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.2,0.2662911891937256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.2,0.3323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.2,0.4547904014587402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.2,0.5951744079589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,2,power_law_1.2,0.02054399996995926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,2,power_law_1.2,0.021830399334430695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,2,power_law_1.2,0.024160000681877136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,2,power_law_1.2,0.029043200612068176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,2,power_law_1.2,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,2,power_law_1.2,0.04216960072517395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,2,power_law_1.2,0.05121920108795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,2,power_law_1.2,0.0537280023097992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,2,power_law_1.2,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,2,power_law_1.2,0.05857920050621033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,2,power_law_1.2,0.0667136013507843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,2,power_law_1.2,0.062745600938797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,2,power_law_1.2,0.0655680000782013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,2,power_law_1.2,0.0711296021938324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,2,power_law_1.2,0.07048320174217224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,2,power_law_1.2,0.07293440103530884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,2,power_law_1.2,0.07941120266914367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,2,power_law_1.2,0.09816960096359253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,2,power_law_1.2,0.10667519569396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,2,power_law_1.2,0.16244479417800903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,2,power_law_1.2,0.1772160053253174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,2,power_law_1.2,0.20131199359893798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,2,power_law_1.2,0.22294399738311768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,2,power_law_1.2,0.2869055986404419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,2,power_law_1.2,0.33974399566650393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,2,power_law_1.2,0.4664576053619385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,2,power_law_1.2,0.5975615978240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,16,balanced,0.02348266790310542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,16,balanced,0.023130667706330616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,16,balanced,0.02309333284695943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,16,balanced,0.025120000044504803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,16,balanced,0.02722666660944621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,16,balanced,0.03909866760174433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,16,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,16,balanced,0.04067733387152354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,16,balanced,0.05909866591294607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,16,balanced,0.054773335655530296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,16,balanced,0.0544106662273407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,16,balanced,0.09368000427881877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,16,balanced,0.09297600388526917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,16,balanced,0.09300800164540608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,16,balanced,0.09948266545931499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,16,balanced,0.09289066990216573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,16,balanced,0.08784000078837077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,16,balanced,0.09557333588600159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,256,power_law_1.2,0.07847679853439331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,16,balanced,0.10934399565060933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,16,balanced,0.11966400345166524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,16,balanced,0.14148799578348795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,16,balanced,0.179967999458313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,16,balanced,0.21861867109934488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,16,balanced,0.28379732370376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,16,balanced,0.35236799716949463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,16,balanced,0.5349066654841105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,16,balanced,0.6942293643951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,256,power_law_1.2,0.05675519704818725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,256,power_law_1.2,0.04784640073776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,256,power_law_1.2,0.05259519815444946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,256,power_law_1.2,0.05268480181694031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,256,power_law_1.2,0.05256959795951843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,256,power_law_1.2,0.05381760001182556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,256,power_law_1.2,0.05397760272026062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,256,power_law_1.2,0.053907197713851926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,256,power_law_1.2,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,256,power_law_1.2,0.0574783980846405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,256,power_law_1.2,0.06277120113372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,256,power_law_1.2,0.06355839967727661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,256,power_law_1.2,0.06770560145378113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,256,power_law_1.2,0.07472000122070313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,256,power_law_1.2,0.07760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,256,power_law_1.2,0.07955200076103211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,256,power_law_1.2,0.10030720233917237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,256,power_law_1.2,0.1077888011932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,256,power_law_1.2,0.1468224048614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,256,power_law_1.2,0.16894079446792604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,256,power_law_1.2,0.22542719841003417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,256,power_law_1.2,0.24725120067596434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,256,power_law_1.2,0.31020159721374513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,256,power_law_1.2,0.416703987121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,256,power_law_1.2,0.5975039958953857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,256,power_law_1.2,0.6856959819793701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,4,power_law_1.2,0.031692799925804135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,4,power_law_1.2,0.03472639918327332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,4,power_law_1.2,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,4,power_law_1.2,0.03612799942493439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,4,power_law_1.2,0.03672960102558136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,4,power_law_1.2,0.04311679899692535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,4,power_law_1.2,0.04498560130596161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,4,power_law_1.2,0.04693120121955872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,4,power_law_1.2,0.04754559993743897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,4,power_law_1.2,0.04853120148181915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,4,power_law_1.2,0.05134720206260681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,4,power_law_1.2,0.0531711995601654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,4,power_law_1.2,0.05399680137634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,4,power_law_1.2,0.058924800157546996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,4,power_law_1.2,0.06842880249023438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,4,power_law_1.2,0.07019519805908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,4,power_law_1.2,0.08115199804306031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,4,power_law_1.2,0.09063040018081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,1,power_law_1.01,0.08172159790992736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,4,power_law_1.2,0.10777599811553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,1,power_law_1.01,0.10660480260848999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,1,power_law_1.01,0.1567296028137207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,1,power_law_1.01,0.25043840408325196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,1,power_law_1.01,0.42064638137817384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,4,power_law_1.2,0.12349439859390259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,1,power_law_1.01,0.5627264022827149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,1,power_law_1.01,0.6586751937866211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,1,power_law_1.01,0.8353216171264648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,1,power_law_1.01,0.8909695625305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,1,power_law_1.01,0.9195967674255371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,1,power_law_1.01,0.9528256416320801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,1,power_law_1.01,0.9819519996643067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,4,power_law_1.2,0.14384640455245973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,1,power_law_1.2,0.09838079810142517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,1,power_law_1.2,0.10718719959259033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,1,power_law_1.01,1.0011199951171874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,1,power_law_1.2,0.12755199670791625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,1,power_law_1.2,0.1642688035964966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,1,power_law_1.01,1.0436672210693358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,4,power_law_1.2,0.17281919717788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,1,power_law_1.2,0.1963711977005005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,1,power_law_1.2,0.23645439147949218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,1,power_law_1.01,1.090559959411621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,1,power_law_1.2,0.30127999782562254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,1,power_law_1.2,0.3163583993911743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,1,power_law_1.2,0.3230720043182373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,1,power_law_1.2,0.33213438987731936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,1,power_law_1.2,0.3397952079772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,1,power_law_1.2,0.3491199970245361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,1,power_law_1.2,0.3519615888595581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,1,power_law_1.01,1.1118271827697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,1,power_law_1.2,0.3629823923110962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,1,power_law_1.2,0.3761728048324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,1,power_law_1.01,1.170361614227295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,1,power_law_1.2,0.38480639457702637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,1,power_law_1.2,0.4141823768615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,1,power_law_1.01,1.2585151672363282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,1,power_law_1.2,0.4670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,1,power_law_1.01,1.3644415855407714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,1,power_law_1.01,1.546560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,1,power_law_1.01,1.6109312057495118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,1,power_law_1.2,0.5115520000457764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,1,power_law_1.01,1.9440576553344726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,1,power_law_1.2,0.6065792083740235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,1,power_law_1.2,0.6878143787384033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,1,power_law_1.01,2.0421695709228516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,1,power_law_1.2,0.876153564453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,1,power_law_1.2,1.0671808242797851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,1,power_law_1.01,2.5619647979736326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,1,power_law_1.2,1.4271488189697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,1,power_law_1.2,1.8026239395141601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,1,power_law_1.01,3.1441408157348634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,4,power_law_1.2,0.23402879238128663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,1,power_law_1.01,4.078124618530273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,1,power_law_1.2,2.5231103897094727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,1,power_law_1.01,5.135692977905274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,1,power_law_1.2,3.2548992156982424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,4,power_law_1.2,0.347110390663147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,4,power_law_1.2,0.49826560020446775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,4,power_law_1.2,0.5915391921997071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,2,balanced,0.07261866827805837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,2,balanced,0.0988159974416097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,2,balanced,0.15198399623235068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,2,balanced,0.2619306643803914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,2,balanced,0.4713386694590251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,2,balanced,0.584389328956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,2,balanced,0.589136004447937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,2,balanced,0.5860426823298136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,2,balanced,0.5900053183237711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,2,balanced,0.5909333229064941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,2,balanced,0.5946133136749268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,2,balanced,0.6018720070521036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,2,balanced,0.602944016456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,2,balanced,0.6094506581624349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,2,balanced,0.6198133230209351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,2,balanced,0.6249013344446818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,2,balanced,0.6356159845987955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,2,balanced,0.736197312672933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,2,balanced,0.6850293477376302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,2,balanced,0.836581309636434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,2,balanced,0.7907360394795736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,2,balanced,1.084170659383138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,2,balanced,1.0187626679738362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,2,balanced,1.5960586865743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,2,balanced,1.6507360140482585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,2,balanced,2.8015626271565757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,2,balanced,3.012645403544108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,4,power_law_1.2,0.7208703994750977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,4,power_law_1.2,0.044659200310707095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,4,power_law_1.2,0.046342399716377256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,4,power_law_1.2,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,4,power_law_1.2,0.05011199712753296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,4,power_law_1.2,0.05588480234146118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,4,power_law_1.2,0.05856000185012818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,4,power_law_1.2,0.06381440162658691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,4,power_law_1.2,0.06888319849967957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,4,power_law_1.2,0.06705920100212097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,4,power_law_1.2,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,4,power_law_1.2,0.07085440158843995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,4,power_law_1.2,0.07333120107650756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,4,power_law_1.2,0.07185919880867005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,4,power_law_1.2,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,4,power_law_1.2,0.08318079710006714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,4,power_law_1.2,0.08654080033302307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,4,power_law_1.2,0.09467520117759705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,4,power_law_1.2,0.10776319503784179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,4,power_law_1.2,0.1130560040473938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,4,power_law_1.2,0.1346943974494934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,4,power_law_1.2,0.15438079833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,4,power_law_1.2,0.19886720180511475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,4,power_law_1.2,0.2401792049407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,4,power_law_1.2,0.3270143985748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,4,power_law_1.2,0.4552320003509521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,4,power_law_1.2,0.6492544174194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,4,power_law_1.2,0.7413824081420899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,1,balanced,0.06746133168538411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,1,balanced,0.09090133508046468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,1,balanced,0.13315733273824057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,1,balanced,0.22462934255599976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,1,balanced,0.3917440176010132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,1,balanced,0.7321866353352865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,1,balanced,0.7411519686381022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,1,balanced,0.7437067031860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,1,balanced,0.7447893619537354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,1,balanced,0.7513866424560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,1,balanced,0.7537546952565511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,1,balanced,0.7627786795298258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,1,balanced,0.767136017481486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,1,balanced,0.7767093181610107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,1,balanced,0.7900319894154867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,1,balanced,0.7987573146820068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,1,balanced,0.8214720090230306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,1,balanced,0.8743573029836019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,1,balanced,0.9094666639963785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,1,balanced,0.9798666636149088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,1,balanced,1.0925172964731853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,1,balanced,1.2262132962544758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,1,balanced,1.3881920178731282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,1,balanced,1.895125389099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,1,balanced,2.1196373303731284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,1,balanced,3.0567092895507812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,1,balanced,3.8105014165242515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,256,power_law_1.2,0.07817599773406983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,256,power_law_1.2,0.07704960107803345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,256,power_law_1.2,0.07823359966278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,256,power_law_1.2,0.08081279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,256,power_law_1.2,0.08097919821739197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,256,power_law_1.2,0.08236799836158752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,256,power_law_1.2,0.08124160170555114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,256,power_law_1.2,0.08241919875144958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,256,power_law_1.2,0.08128640055656433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,256,power_law_1.2,0.08072959780693054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,256,power_law_1.2,0.08110079765319825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,256,power_law_1.2,0.08111360073089599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,256,power_law_1.2,0.08046079874038696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,256,power_law_1.2,0.0804032027721405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,256,power_law_1.2,0.08405759930610657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,256,power_law_1.2,0.08346239924430847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,256,power_law_1.2,0.09915519952774048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,256,power_law_1.2,0.10436480045318604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,256,power_law_1.2,0.11621760129928589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,256,power_law_1.2,0.14325120449066162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,256,power_law_1.2,0.17004159688949586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,256,power_law_1.2,0.22040319442749023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,256,power_law_1.2,0.2936896085739136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,256,power_law_1.2,0.45879039764404295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,256,power_law_1.2,0.5377855777740479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,256,power_law_1.2,0.7637951850891114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,256,power_law_1.2,1.1236991882324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,16,balanced,0.04866666595141093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,16,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,16,balanced,0.04974400003751119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,16,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,16,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,16,balanced,0.0641546646753947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,16,balanced,0.06921599805355072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,16,balanced,0.07023466626803081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,16,balanced,0.0705386648575465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,16,balanced,0.07239999870459239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,16,balanced,0.07234666744867961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,16,balanced,0.07656000057856242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,16,balanced,0.07657599945863088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,16,balanced,0.07640533149242401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,16,balanced,0.08273066580295563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,16,balanced,0.08450667063395183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,16,balanced,0.090938667456309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,16,balanced,0.10296000043551128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,16,balanced,0.11327999830245972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,16,balanced,0.13424000144004822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,16,balanced,0.15010666847229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,16,balanced,0.1876266598701477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,16,balanced,0.2251360019048055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,16,balanced,0.29129600524902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,16,balanced,0.3792373339335124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,16,balanced,0.5101066827774048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,16,balanced,0.662448008855184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,2,balanced,0.03499733408292135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,2,balanced,0.03373866776625315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,2,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,2,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,2,balanced,0.03963200002908707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,2,balanced,0.039349332451820374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,2,balanced,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,2,balanced,0.03941866755485535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,2,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,2,balanced,0.03961066653331121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,2,balanced,0.03963200002908707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,2,balanced,0.03976000100374222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,2,balanced,0.04182399809360504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,2,balanced,0.04557333389918009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,2,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,2,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,2,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,2,balanced,0.06699733436107635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,2,balanced,0.06851199766000111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,2,balanced,0.0881706674893697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,2,balanced,0.10111467043558757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,2,balanced,0.05172266562779745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,2,balanced,0.12980799873669943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,2,balanced,0.1562399963537852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,2,balanced,0.20587199926376343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,2,balanced,0.24498667319615683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,2,balanced,0.34587732950846356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,2,balanced,0.44993066787719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,2,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,2,balanced,0.06247999767462412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,2,balanced,0.0827893316745758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,2,balanced,0.11456533273061116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,2,balanced,0.17634665966033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,2,balanced,0.23663999636967978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,2,balanced,0.22621333599090576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,2,balanced,0.23004267613093057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,2,balanced,0.2309066653251648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,2,balanced,0.23322133223215738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,2,balanced,0.2382240096728007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,2,balanced,0.24209600687026978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,2,balanced,0.24732265869776407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,2,balanced,0.2565386692682902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,2,balanced,0.2618880073229472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,2,balanced,0.27507734298706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,2,balanced,0.30667199691136676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,2,balanced,0.3404373327891032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,2,balanced,0.39872535069783527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,2,balanced,0.448202649752299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,2,balanced,0.5571306546529134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,2,balanced,0.6775893370310465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,2,balanced,0.880949338277181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,2,balanced,1.1634559631347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,2,balanced,1.56769593556722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,2,balanced,2.100821336110433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,64,balanced,0.07236266632874806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,64,balanced,0.041936000188191734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,64,balanced,0.04178666571776072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,64,balanced,0.045834665497144066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,64,balanced,0.04428799947102865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,64,balanced,0.04962133367856344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,64,balanced,0.049653331438700356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,64,balanced,0.04965866605440775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,64,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,64,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,64,balanced,0.05173333485921224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,64,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,64,balanced,0.05401599903901418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,64,balanced,0.054042667150497437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,64,balanced,0.05994133154551188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,64,balanced,0.06019733349482218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,64,balanced,0.06233066817124685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,64,balanced,0.07038400073846181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,64,balanced,0.07448000212510426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,64,balanced,0.0909546713034312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,64,balanced,0.11272533734639485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,64,balanced,0.1479200025399526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,64,balanced,0.17627733945846558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,64,balanced,0.23600000143051147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,64,balanced,0.29575467109680176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,64,balanced,0.4075839916865031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,64,balanced,0.5280693372090658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,2,power_law_1.2,0.02968960106372833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,2,power_law_1.2,0.02803199887275696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,2,power_law_1.2,0.029369598627090453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,2,power_law_1.2,0.03189760148525238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,2,power_law_1.2,0.03155199885368347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,2,power_law_1.2,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,2,power_law_1.2,0.03315199911594391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,2,power_law_1.2,0.0335999995470047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,2,power_law_1.2,0.03441919982433319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,2,power_law_1.2,0.03573119938373566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,2,power_law_1.2,0.03537920117378235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,2,power_law_1.2,0.03687039911746979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,2,power_law_1.2,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,2,power_law_1.2,0.037196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,2,power_law_1.2,0.04267520010471344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,2,power_law_1.2,0.045542401075363156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,2,power_law_1.2,0.05017600059509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,2,power_law_1.2,0.06526079773902893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,2,power_law_1.2,0.06174719929695129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,2,power_law_1.2,0.08108159899711609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,2,power_law_1.2,0.09082239866256714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,2,power_law_1.2,0.11468160152435303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,2,power_law_1.2,0.12806400060653686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,2,power_law_1.2,0.16707199811935425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,2,power_law_1.01,0.07869439721107482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,2,power_law_1.01,0.09414399862289428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,2,power_law_1.01,0.10329600572586059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,2,power_law_1.01,0.1296447992324829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,2,power_law_1.01,0.14954240322113038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,2,power_law_1.01,0.1754688024520874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,2,power_law_1.01,0.2050175905227661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,2,power_law_1.01,0.20986878871917725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,2,power_law_1.01,0.21644799709320067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,2,power_law_1.01,0.22758400440216064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,2,power_law_1.01,0.22250878810882568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,2,power_law_1.01,0.2268160104751587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,2,power_law_1.01,0.2375488042831421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,2,power_law_1.01,0.24342401027679444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,2,power_law_1.01,0.2446336030960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,2,power_law_1.2,0.22649600505828857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,2,power_law_1.01,0.2452608108520508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,2,power_law_1.01,0.259500789642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,2,power_law_1.01,0.2859519958496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,2,power_law_1.01,0.31388800144195556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,2,power_law_1.01,0.3641664028167725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,2,power_law_1.01,0.4236288070678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,2,power_law_1.01,0.5157631874084473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,2,power_law_1.01,0.6204800128936767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,2,power_law_1.01,0.8741375923156738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,2,power_law_1.01,1.0463744163513184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,2,power_law_1.01,1.49487361907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,2,power_law_1.01,1.9394048690795898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,2,power_law_1.2,0.3235136032104492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,8,power_law_1.01,0.1575935959815979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,8,power_law_1.01,0.23107199668884276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,8,power_law_1.01,0.22216320037841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,8,power_law_1.01,0.2672384023666382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,8,power_law_1.01,0.24919679164886474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,8,power_law_1.01,0.27645440101623536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,8,power_law_1.01,0.2957247972488403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,8,power_law_1.01,0.2807807922363281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,8,power_law_1.01,0.28859519958496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,8,power_law_1.01,0.28786559104919435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,8,power_law_1.01,0.3017280101776123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,8,power_law_1.01,0.29811840057373046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,8,power_law_1.01,0.3057791948318481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,8,power_law_1.01,0.3248447895050049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,8,power_law_1.01,0.33427839279174804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,8,power_law_1.01,0.33829760551452637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,8,power_law_1.01,0.3857343912124634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,8,power_law_1.01,0.4201791763305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,8,power_law_1.01,0.41384320259094237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,8,power_law_1.01,0.5121024131774903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,8,power_law_1.01,0.5665728092193604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,8,power_law_1.01,0.7864128112792969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,8,power_law_1.01,0.7033408164978028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,8,power_law_1.01,0.9405823707580566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,8,power_law_1.01,1.2324159622192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,8,power_law_1.01,1.8031871795654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,8,power_law_1.01,2.0055871963500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,32,power_law_1.01,0.0741312026977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,32,power_law_1.01,0.08039039969444275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,32,power_law_1.01,0.07701759934425353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,32,power_law_1.01,0.08172159790992736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,32,power_law_1.01,0.08086400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,32,power_law_1.01,0.078438401222229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,32,power_law_1.01,0.07916160225868225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,32,power_law_1.01,0.07964159846305847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,2,power_law_1.2,0.40705280303955077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,32,power_law_1.01,0.08046720027923585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,32,power_law_1.01,0.08310400247573853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,32,power_law_1.01,0.08573439717292786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,32,power_law_1.01,0.08700159788131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,32,power_law_1.01,0.08790400028228759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,32,power_law_1.01,0.09676160216331482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,32,power_law_1.01,0.1056704044342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,32,power_law_1.01,0.10712319612503052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,32,power_law_1.01,0.10417280197143555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,32,power_law_1.01,0.12137600183486938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,32,power_law_1.01,0.12828799486160278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,32,power_law_1.01,0.15262720584869385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,32,power_law_1.01,0.17333760261535644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,32,power_law_1.01,0.21809918880462648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,32,power_law_1.01,0.2643775939941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,32,power_law_1.01,0.3641983985900879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,32,power_law_1.01,0.47772798538208006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,32,power_law_1.01,0.7024064064025879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,32,power_law_1.01,0.9114944458007812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,16,power_law_1.2,0.14750080108642577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,16,power_law_1.2,0.1927616000175476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,16,power_law_1.2,0.2016511917114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,16,power_law_1.2,0.19717119932174682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,16,power_law_1.2,0.21470720767974855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,16,power_law_1.2,0.17908480167388915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,16,power_law_1.2,0.17964160442352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,16,power_law_1.2,0.18247040510177612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,16,power_law_1.2,0.17081600427627563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,16,power_law_1.2,0.16914559602737428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,16,power_law_1.2,0.18261120319366456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,16,power_law_1.2,0.1831168055534363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,16,power_law_1.2,0.18426239490509033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,16,power_law_1.2,0.1812991976737976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,16,power_law_1.2,0.19288320541381837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,16,power_law_1.2,0.19182080030441284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,16,power_law_1.2,0.19738240242004396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,16,power_law_1.2,0.21687679290771483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.2,0.2540031909942627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.2,0.2617151975631714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.2,0.3128511905670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.2,0.38881919384002683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.2,0.44318079948425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.2,0.6259520053863525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.2,0.857049560546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.2,1.104915237426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.2,1.2844223976135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,2,balanced,0.03236266722281774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,2,balanced,0.033359999457995095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,2,balanced,0.03787733366092046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,2,balanced,0.059861332178115845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,2,balanced,0.0860533316930135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,2,balanced,0.08935466408729553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,2,balanced,0.08844266335169475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,2,balanced,0.09055999914805095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,2,balanced,0.09222933650016785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,2,balanced,0.0928053359190623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,2,balanced,0.0942133367061615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,2,balanced,0.0977226694424947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,2,balanced,0.09800533453623454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,2,balanced,0.10081600149472554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,2,balanced,0.10900800426801045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,2,balanced,0.10967466235160828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,2,balanced,0.1146399974822998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,2,balanced,0.13381333152453104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,2,balanced,0.13461333513259888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,2,balanced,0.17154133319854736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,2,balanced,0.1744640072186788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,2,balanced,0.24854934215545654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,2,balanced,0.28220800558725995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,2,balanced,0.40885333220163983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,2,balanced,0.5087413390477499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,2,balanced,0.7259413401285807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,2,balanced,0.913429339726766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,8,power_law_1.01,0.0757695972919464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,8,power_law_1.01,0.10149760246276855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,8,power_law_1.01,0.10215679407119752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,8,power_law_1.01,0.12760319709777831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,8,power_law_1.01,0.1287616014480591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,8,power_law_1.01,0.14424959421157837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,8,power_law_1.01,0.14391039609909057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,8,power_law_1.01,0.1313088059425354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,8,power_law_1.01,0.13332480192184448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,8,power_law_1.01,0.13565440177917482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,8,power_law_1.01,0.12664320468902587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,8,power_law_1.01,0.1269376039505005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,8,power_law_1.01,0.1344383955001831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,8,power_law_1.01,0.13612159490585327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,8,power_law_1.01,0.1477504014968872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,8,power_law_1.01,0.1460736036300659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,8,power_law_1.01,0.15281920433044432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,8,power_law_1.01,0.164409601688385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,8,power_law_1.01,0.17402880191802977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,8,power_law_1.01,0.19589760303497314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,8,power_law_1.01,0.2219775915145874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,8,power_law_1.01,0.27015678882598876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,8,power_law_1.01,0.32600960731506345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,8,power_law_1.01,0.4227007865905762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,8,power_law_1.01,0.5309887886047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,8,power_law_1.01,0.7561535835266113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,8,power_law_1.01,0.9917695999145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,1,power_law_1.2,0.020851199328899384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,1,power_law_1.2,0.022592000663280487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,1,power_law_1.2,0.026867198944091796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,1,power_law_1.2,0.03433600068092346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,1,power_law_1.2,0.04816640019416809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,1,power_law_1.2,0.07039999961853027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,1,power_law_1.2,0.08484479784965515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,1,power_law_1.2,0.08711680173873901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,1,power_law_1.2,0.09096959829330445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,1,power_law_1.2,0.09380480051040649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,1,power_law_1.2,0.09807360172271729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,1,power_law_1.2,0.101363205909729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,1,power_law_1.2,0.10700800418853759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,1,power_law_1.2,0.10935039520263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,1,power_law_1.2,0.1130944013595581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,1,power_law_1.2,0.11792000532150268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,1,power_law_1.2,0.12772480249404908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,1,power_law_1.2,0.1463039994239807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.2,0.1632256031036377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.2,0.20113279819488525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.2,0.2340991973876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.2,0.3065920114517212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.2,0.3395136117935181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.2,0.5040832042694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.2,0.5736576080322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.2,0.8105152130126954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.2,1.0501888275146485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,1,power_law_1.2,0.08091520071029663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,1,power_law_1.2,0.09479680061340331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,1,power_law_1.2,0.12924799919128419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,1,power_law_1.2,0.1799936056137085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,1,power_law_1.2,0.22584960460662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,1,power_law_1.2,0.36559998989105225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,1,power_law_1.2,0.39009919166564944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,1,power_law_1.2,0.4050559997558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,1,power_law_1.2,0.4173567771911621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,1,power_law_1.2,0.4270463943481445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,1,power_law_1.2,0.4457727909088135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,1,power_law_1.2,0.45516161918640136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,1,power_law_1.2,0.47121920585632326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,1,power_law_1.2,0.4902656078338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,1,power_law_1.2,0.5170112133026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,1,power_law_1.2,0.537504005432129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,1,power_law_1.2,0.5633600234985352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,1,power_law_1.2,0.6414847850799561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,1,power_law_1.2,0.7206975936889648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,1,power_law_1.2,0.8821311950683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,1,power_law_1.2,1.0677568435668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,1,power_law_1.2,1.3384063720703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,1,power_law_1.2,1.5592384338378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,1,power_law_1.2,2.121504020690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,1,power_law_1.2,2.597804832458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,1,power_law_1.2,3.6028480529785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,1,power_law_1.2,4.756576156616211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,4,power_law_1.01,0.048147198557853696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,4,power_law_1.01,0.05170559883117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,4,power_law_1.01,0.056505602598190305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,4,power_law_1.01,0.06302080154418946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,4,power_law_1.01,0.06711680293083191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,4,power_law_1.01,0.07459840178489685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,4,power_law_1.01,0.07660160064697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,4,power_law_1.01,0.07783039808273315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,4,power_law_1.01,0.0804032027721405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,4,power_law_1.01,0.07978879809379577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,4,power_law_1.01,0.08044160008430482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,4,power_law_1.01,0.0824512004852295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,4,power_law_1.01,0.08525440096855164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,4,power_law_1.01,0.08685439825057983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,4,power_law_1.01,0.0895359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,4,power_law_1.01,0.09091200232505799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,4,power_law_1.01,0.0977728009223938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,4,power_law_1.01,0.10904959440231324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,4,power_law_1.01,0.1262336015701294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,4,power_law_1.01,0.15205760002136232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,4,power_law_1.01,0.18069759607315064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,4,power_law_1.01,0.2588479995727539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,4,power_law_1.01,0.3112639904022217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,4,power_law_1.01,0.42849278450012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,4,power_law_1.01,0.5610816001892089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,4,power_law_1.01,0.8687744140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,4,power_law_1.01,1.1677760124206542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,2,power_law_1.2,0.09981439709663391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,2,power_law_1.2,0.13434239625930786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,2,power_law_1.2,0.1710592031478882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,2,power_law_1.2,0.22897279262542725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,2,power_law_1.2,0.3129535913467407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,2,power_law_1.2,0.4133440017700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,2,power_law_1.2,0.46077442169189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,2,power_law_1.2,0.5845632076263427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,2,power_law_1.2,0.5685696125030517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,2,power_law_1.2,0.6052735805511474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,2,power_law_1.2,0.6022655963897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,2,power_law_1.2,0.6577151775360107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,2,power_law_1.2,0.6492800235748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,2,power_law_1.2,0.6661312103271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,2,power_law_1.2,0.7264704227447509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,2,power_law_1.2,0.703545618057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,2,power_law_1.2,0.746611213684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,2,power_law_1.2,0.7801407814025879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,2,power_law_1.2,0.855571174621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,2,power_law_1.2,0.9529343605041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,2,power_law_1.2,1.063430404663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,2,power_law_1.2,1.3076288223266601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,2,power_law_1.2,1.4125247955322267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,2,power_law_1.2,1.7684928894042968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,2,power_law_1.2,2.229555130004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,2,power_law_1.2,2.8936384201049803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,2,power_law_1.2,3.5141185760498046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,2,power_law_1.01,0.05225600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,2,power_law_1.01,0.058169597387313844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,2,power_law_1.01,0.06677119731903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,2,power_law_1.01,0.07820799946784973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,2,power_law_1.01,0.09804800152778625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,2,power_law_1.01,0.11986559629440308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,2,power_law_1.01,0.1310464024543762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,2,power_law_1.01,0.13767679929733276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,2,power_law_1.01,0.14855680465698243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,2,power_law_1.01,0.15675519704818724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,2,power_law_1.01,0.15238399505615235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,2,power_law_1.01,0.16179840564727782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,2,power_law_1.01,0.15986560583114623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,2,power_law_1.01,0.1729151964187622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,2,power_law_1.01,0.18353919982910155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,2,power_law_1.01,0.1853503942489624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,2,power_law_1.01,0.20066559314727783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,2,power_law_1.01,0.2230976104736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,2,power_law_1.01,0.2563584089279175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,2,power_law_1.01,0.3084671974182129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,2,power_law_1.01,0.3600895881652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,2,power_law_1.01,0.46404480934143066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,2,power_law_1.01,0.5700287818908691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,2,power_law_1.01,0.7662591934204102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,2,power_law_1.01,1.0012479782104493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,2,power_law_1.01,1.341267204284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,2,power_law_1.01,1.7749311447143554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,32,power_law_1.01,0.050809597969055174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,32,power_law_1.01,0.04990079998970032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,32,power_law_1.01,0.04910080134868622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,32,power_law_1.01,0.052716797590255736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,32,power_law_1.01,0.05260159969329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,32,power_law_1.01,0.051583999395370485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,32,power_law_1.01,0.05432320237159729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,32,power_law_1.01,0.05475839972496033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,32,power_law_1.01,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,32,power_law_1.01,0.055340802669525145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,32,power_law_1.01,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,32,power_law_1.01,0.0574400007724762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,32,power_law_1.01,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,32,power_law_1.01,0.0605184018611908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,32,power_law_1.01,0.06568319797515869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,32,power_law_1.01,0.0675711989402771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,32,power_law_1.01,0.07354879975318909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,32,power_law_1.01,0.08127359747886657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,32,power_law_1.01,0.08773760199546814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,16,balanced,0.05422399938106537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,16,balanced,0.05398933092753092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,16,balanced,0.05789333085219065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,16,balanced,0.07061333457628886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,16,balanced,0.08674666285514832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,16,balanced,0.12122666835784912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,16,balanced,0.12246400117874146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,16,balanced,0.12175466616948445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,16,balanced,0.12441066900889079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,16,balanced,0.12298666437466939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,32,power_law_1.01,0.1067199945449829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,32,power_law_1.01,0.12707200050354003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,32,power_law_1.01,0.15136640071868895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,32,power_law_1.01,0.1808895945549011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,16,balanced,0.12366933623949687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,16,balanced,0.12614933649698892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,32,power_law_1.01,0.22697598934173585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,16,balanced,0.1258026659488678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,16,balanced,0.12808533509572348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,32,power_law_1.01,0.2714560031890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,32,power_law_1.01,0.3938175916671753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,32,power_law_1.01,0.5059711933135986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,16,balanced,0.14004266262054443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,16,balanced,0.14028267065684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,16,balanced,0.1456160048643748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,16,balanced,0.16470932960510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,16,balanced,0.1688906749089559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,16,balanced,0.18555200099945068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,16,balanced,0.20152533054351807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,16,balanced,0.23483200867970785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,16,balanced,0.265994668006897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,16,balanced,0.35462931791941327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,16,balanced,0.4222400188446045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,16,balanced,0.5953493515650431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,16,balanced,0.7672800223032633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,64,power_law_1.01,0.03949440121650696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,64,power_law_1.01,0.03806079924106598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,64,power_law_1.01,0.03883520066738129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,64,power_law_1.01,0.03984639942646027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,64,power_law_1.01,0.040889599919319154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,64,power_law_1.01,0.04185599982738495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,64,power_law_1.01,0.041894400119781496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,64,power_law_1.01,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,64,power_law_1.01,0.041196799278259276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,64,power_law_1.01,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,64,power_law_1.01,0.04211840033531189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,64,power_law_1.01,0.04206719994544983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,64,power_law_1.01,0.04287999868392944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,64,power_law_1.01,0.04390400052070618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,64,power_law_1.01,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,64,power_law_1.01,0.04890879988670349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,64,power_law_1.01,0.051712000370025636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,64,power_law_1.01,0.0551360011100769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.01,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.01,0.06875519752502442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.01,0.07475200295448303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.01,0.09552639722824097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.01,0.11950080394744873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.01,0.15628160238265992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.01,0.1960576057434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.01,0.27426559925079347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.01,0.3169663906097412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,16,power_law_1.01,0.14615679979324342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,16,power_law_1.01,0.2519999980926514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,16,power_law_1.01,0.3425535917282104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,16,power_law_1.01,0.3235071897506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,16,power_law_1.01,0.33892478942871096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,16,power_law_1.01,0.30364160537719725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,16,power_law_1.01,0.3239487886428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,16,power_law_1.01,0.3384320020675659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,16,power_law_1.01,0.28136320114135743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,16,power_law_1.01,0.29416959285736083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,16,power_law_1.01,0.3012671947479248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,16,power_law_1.01,0.2887615919113159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,16,power_law_1.01,0.29754879474639895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,16,power_law_1.01,0.29539198875427247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,16,power_law_1.01,0.32503039836883546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,16,power_law_1.01,0.3360447883605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,16,power_law_1.01,0.3376319885253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,16,power_law_1.01,0.3568320035934448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,16,power_law_1.01,0.3969856023788452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,16,power_law_1.01,0.47646079063415525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,16,power_law_1.01,0.45586562156677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,16,power_law_1.01,0.5780223846435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,16,power_law_1.01,0.637395191192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,16,power_law_1.01,0.7979328155517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,16,power_law_1.01,0.8621631622314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,16,power_law_1.01,1.0963775634765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,16,power_law_1.01,1.3869695663452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,16,power_law_1.01,0.04310399889945984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,16,power_law_1.01,0.04056960046291351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,16,power_law_1.01,0.03909119963645935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,16,power_law_1.01,0.042771199345588685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,16,power_law_1.01,0.04135040044784546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,16,power_law_1.01,0.04197759926319122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,16,power_law_1.01,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,16,power_law_1.01,0.04555520117282867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,16,power_law_1.01,0.0438400000333786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,16,power_law_1.01,0.04588800072669983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,16,power_law_1.01,0.04429439902305603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,16,power_law_1.01,0.05392640233039856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,16,power_law_1.01,0.054630398750305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,16,power_law_1.01,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,64,power_law_1.01,0.07003520131111145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,64,power_law_1.01,0.06928640007972717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,64,power_law_1.01,0.06883199810981751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,16,power_law_1.01,0.05868160128593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,64,power_law_1.01,0.07395840287208558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,16,power_law_1.01,0.062080001831054686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,64,power_law_1.01,0.07496320009231568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,16,power_law_1.01,0.06783360242843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,16,power_law_1.01,0.07759360074996949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,16,power_law_1.01,0.08377599716186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,16,power_law_1.01,0.1024448037147522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,16,power_law_1.01,0.12684799432754518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,16,power_law_1.01,0.15921920537948608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,64,power_law_1.01,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,16,power_law_1.01,0.18743679523468018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,64,power_law_1.01,0.07835519909858704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,64,power_law_1.01,0.07723519802093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,16,power_law_1.01,0.26853759288787843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,64,power_law_1.01,0.0788096010684967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,16,power_law_1.01,0.37141759395599366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,16,power_law_1.01,0.5644800186157226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,16,power_law_1.01,0.7955711841583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,2,balanced,0.034234667817751564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,2,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,2,balanced,0.033514666060606636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,2,balanced,0.03859733293453852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,2,balanced,0.04713066418965658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,64,power_law_1.01,0.07910400032997131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,2,balanced,0.04739200075467428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,2,balanced,0.046367997924486794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,2,balanced,0.04837866624196371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,64,power_law_1.01,0.07909759879112244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,2,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,2,balanced,0.048842668533325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,2,balanced,0.05050133168697357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,2,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,2,balanced,0.054287999868392944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,2,balanced,0.05853333572546641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,2,balanced,0.062394668658574425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,2,balanced,0.06655466556549072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,2,balanced,0.0710506687561671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,2,balanced,0.084906667470932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,2,balanced,0.09299199779828389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,2,balanced,0.121370663245519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,64,power_law_1.01,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,2,balanced,0.1341493328412374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,64,power_law_1.01,0.08699520230293274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,2,balanced,0.1835413376490275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,2,balanced,0.21570666631062826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,2,balanced,0.3134453296661377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,64,power_law_1.01,0.08465279936790467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,64,power_law_1.01,0.09669119715690613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,64,power_law_1.01,0.09459199905395507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,64,power_law_1.01,0.09845759868621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,64,power_law_1.01,0.11063679456710815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,64,power_law_1.01,0.11855360269546508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,2,balanced,0.41436266899108887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,64,power_law_1.01,0.14359040260314943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,64,power_law_1.01,0.17011200189590453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,64,power_law_1.01,0.22867839336395263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,64,power_law_1.01,0.26306560039520266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,64,power_law_1.01,0.36762878894805906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,2,balanced,0.611407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,2,balanced,0.8188640276590983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,64,power_law_1.01,0.4682303905487061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,64,power_law_1.01,0.6672639846801758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,64,power_law_1.01,0.9161151885986328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,1,power_law_1.01,0.028652799129486085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,1,power_law_1.01,0.030630400776863097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,1,power_law_1.01,0.031436800956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,1,power_law_1.01,0.03506560027599335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,1,power_law_1.01,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,1,power_law_1.01,0.036396801471710205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,1,power_law_1.01,0.039129599928855896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,1,power_law_1.01,0.041791999340057374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,1,power_law_1.01,0.04490880072116852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,1,power_law_1.01,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,1,power_law_1.01,0.0453247994184494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,1,power_law_1.01,0.04635519981384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,1,power_law_1.01,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,1,power_law_1.01,0.05264000296592712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,1,power_law_1.01,0.06285439729690552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,1,power_law_1.01,0.06453760266304016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,1,power_law_1.01,0.06312959790229797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,1,power_law_1.01,0.07896959781646729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,1,power_law_1.01,0.08113920092582702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,1,power_law_1.01,0.10439039468765259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,1,power_law_1.01,0.11610879898071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,1,power_law_1.01,0.1549056053161621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,8,balanced,0.027552001178264618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,8,balanced,0.027514666318893433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,8,balanced,0.029215998947620392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,8,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,8,balanced,0.0346666673819224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,8,balanced,0.035605333745479584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,8,balanced,0.03572800010442734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,8,balanced,0.035546667873859406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,8,balanced,0.037087999284267426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,8,balanced,0.035946667194366455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,8,balanced,0.03729599962631861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,8,balanced,0.036805334190527596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,8,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,8,balanced,0.03844800094763438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,8,balanced,0.04170133173465729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,8,balanced,0.04376000165939331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,8,balanced,0.04378133515516917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,8,balanced,0.04764799773693085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,8,balanced,0.05150400102138519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,8,balanced,0.05830933153629303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,8,balanced,0.062421331803003945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,8,balanced,0.07568533221880595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,8,balanced,0.09034666419029236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,8,balanced,0.12982400258382162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,8,balanced,0.1715466578801473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,8,balanced,0.23226666450500488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,8,balanced,0.2893120050430298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,1,power_law_1.01,0.17488640546798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,1,power_law_1.01,0.23796479701995848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,1,power_law_1.01,0.30792961120605467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,1,power_law_1.01,0.4287680149078369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,1,power_law_1.01,0.554041576385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,16,power_law_1.2,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,16,power_law_1.2,0.08422399759292602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,16,power_law_1.2,0.08113279938697815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,16,power_law_1.2,0.08600320219993592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,16,power_law_1.2,0.08729599714279175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,16,power_law_1.2,0.07143679857254029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,16,power_law_1.2,0.0840831995010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,16,power_law_1.2,0.08536959886550903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,16,power_law_1.2,0.08004480004310607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,16,power_law_1.2,0.08599680066108703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,16,power_law_1.2,0.08931840062141419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,16,power_law_1.2,0.08824319839477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,16,power_law_1.2,0.09025279879570007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,16,power_law_1.2,0.09170560240745544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,16,power_law_1.2,0.10144640207290649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,16,power_law_1.2,0.10095360279083251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,16,power_law_1.2,0.11226880550384521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,16,power_law_1.2,0.13153280019760133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,16,power_law_1.2,0.1300287961959839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,16,power_law_1.2,0.15599360466003417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,16,power_law_1.2,0.17175040245056153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,16,power_law_1.2,0.20728960037231445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,16,power_law_1.2,0.2547647953033447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,16,power_law_1.2,0.3442496061325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,16,power_law_1.2,0.4574143886566162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,16,power_law_1.2,0.6792895793914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,16,power_law_1.2,0.7680448055267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,16,power_law_1.2,0.06409599781036376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,16,power_law_1.2,0.05937280058860779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,16,power_law_1.2,0.0639360010623932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,16,power_law_1.2,0.06295679807662964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,16,power_law_1.2,0.053932797908782956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,16,power_law_1.2,0.060031998157501223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,16,power_law_1.2,0.06341760158538819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,16,power_law_1.2,0.06093440055847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,16,power_law_1.2,0.06636160016059875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,16,power_law_1.2,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,16,power_law_1.2,0.07025920152664185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,16,power_law_1.2,0.07338240146636962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,16,power_law_1.2,0.07720959782600403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,16,power_law_1.2,0.07621759772300721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,16,power_law_1.2,0.08326399922370911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,16,power_law_1.2,0.08727040290832519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,16,power_law_1.2,0.09548159837722778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,16,power_law_1.2,0.11484160423278808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,16,power_law_1.2,0.11932799816131592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,16,power_law_1.2,0.1451840043067932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,16,power_law_1.2,0.1673472046852112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,16,power_law_1.2,0.2213887929916382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,16,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,16,balanced,0.048154667019844055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,16,balanced,0.05044800043106079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,16,balanced,0.05550399919350942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,16,balanced,0.05421866476535797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,16,balanced,0.06200533111890157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,16,balanced,0.06198399762312571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,16,balanced,0.06403199831644694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,16,balanced,0.06404800216356914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,16,balanced,0.06452799836794536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,16,balanced,0.06403199831644694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,16,balanced,0.0663679987192154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,16,balanced,0.06831466654936473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,16,balanced,0.06714666883150737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,16,balanced,0.07262399792671204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,16,balanced,0.07257066667079926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,16,balanced,0.07654400169849396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,32,power_law_1.01,0.1441215991973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,16,balanced,0.08296533425649007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,16,balanced,0.08996267120043437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,16,balanced,0.10403199990590413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,16,balanced,0.12893333037694296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,16,balanced,0.15654399991035461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,16,balanced,0.18258132537206015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,16,balanced,0.24229333798090616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,16,balanced,0.29573333263397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,16,balanced,0.4007519880930583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,16,balanced,0.510693351427714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,32,power_law_1.01,0.1435647964477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,32,power_law_1.01,0.14101120233535766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,32,power_law_1.01,0.14393600225448608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,32,power_law_1.01,0.13528319597244262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,16,power_law_1.2,0.32648959159851076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,32,power_law_1.01,0.12282240390777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,16,power_law_1.2,0.3983360052108765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,32,power_law_1.01,0.1431040048599243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,16,power_law_1.2,0.5188352108001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,32,power_law_1.01,0.14232959747314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,32,power_law_1.01,0.13770240545272827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,32,power_law_1.01,0.14343039989471434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,16,power_law_1.2,0.7360832214355468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,32,power_law_1.01,0.14635519981384276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,32,power_law_1.01,0.14188159704208375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,32,power_law_1.01,0.15359359979629517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,32,power_law_1.01,0.14594559669494628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,32,power_law_1.01,0.15864959955215455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,32,power_law_1.01,0.15774719715118407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,32,power_law_1.01,0.17955199480056763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,32,power_law_1.01,0.19115519523620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,32,power_law_1.01,0.19211519956588746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,32,power_law_1.01,0.2255295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,32,power_law_1.01,0.2367039918899536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,32,power_law_1.01,0.3005631923675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,32,power_law_1.01,0.3429120063781738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,32,power_law_1.01,0.4820864200592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,32,power_law_1.01,0.5556863784790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,32,power_law_1.01,0.8032959938049317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,32,power_law_1.01,0.9644031524658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,16,power_law_1.2,0.9102911949157715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,2,power_law_1.01,0.045126399397850035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,2,power_law_1.01,0.04711039960384369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,2,power_law_1.01,0.04764159917831421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,2,power_law_1.01,0.054067200422286986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,2,power_law_1.01,0.05960959792137146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,2,power_law_1.01,0.06520320177078247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,2,power_law_1.01,0.07355520129203796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,2,power_law_1.01,0.0750656008720398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,2,power_law_1.01,0.07693439722061157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,2,power_law_1.01,0.0766592025756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,2,power_law_1.01,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,2,power_law_1.01,0.08217599987983704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,2,power_law_1.01,0.08507519960403442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,2,power_law_1.01,0.08893439769744874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,2,power_law_1.01,0.09581440091133117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,2,power_law_1.01,0.10080640316009522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,2,power_law_1.01,0.10584959983825684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,2,power_law_1.01,0.12738560438156127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,2,power_law_1.01,0.1334272027015686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,2,power_law_1.01,0.1627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,2,power_law_1.01,0.190937602519989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,2,power_law_1.01,0.25822720527648924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,2,power_law_1.01,0.3284032106399536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,2,power_law_1.01,0.43444480895996096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,2,power_law_1.01,0.5425856113433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,2,power_law_1.01,0.7659647941589356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,2,power_law_1.01,0.97259521484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,32,power_law_1.01,0.1916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,8,balanced,0.023541333774725597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,8,balanced,0.024154665569464367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,8,balanced,0.025072000920772552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,8,balanced,0.025834667185942333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,8,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,8,balanced,0.045567999283472695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,8,balanced,0.05341866612434387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,8,balanced,0.05030933519204458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,8,balanced,0.05372266471385956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,8,balanced,0.06609066824118297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,8,balanced,0.06491733094056447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,8,balanced,0.06708799799283345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,8,balanced,0.08886933326721191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,8,balanced,0.08917867143948872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,8,balanced,0.08912533521652222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,8,balanced,0.10077333450317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,8,balanced,0.09949866930643718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,8,balanced,0.11153599619865417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,8,balanced,0.12573867042859396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,8,balanced,0.1332480013370514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,8,balanced,0.15589333573977152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,8,balanced,0.18650132417678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,8,balanced,0.21105066935221353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,8,balanced,0.3144479990005493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,8,balanced,0.38365864753723145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,8,balanced,0.5454560120900472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,8,balanced,0.7231733004252116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,32,power_law_1.01,0.2558655977249146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,32,power_law_1.01,0.2456063985824585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,32,power_law_1.01,0.2520639896392822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,32,power_law_1.01,0.2505408048629761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,32,power_law_1.01,0.23566079139709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,32,power_law_1.01,0.2259200096130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,32,power_law_1.01,0.22581119537353517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,32,power_law_1.01,0.2431936025619507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,32,power_law_1.01,0.23036160469055175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,32,power_law_1.01,0.2535552024841309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,32,power_law_1.01,0.2444607973098755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,32,power_law_1.01,0.25537919998168945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,32,power_law_1.01,0.23484160900115966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,32,power_law_1.01,0.24425599575042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,32,power_law_1.01,0.24414079189300536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,32,power_law_1.01,0.24350080490112305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,32,power_law_1.01,0.2442624092102051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,32,power_law_1.01,0.2733439922332764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,32,power_law_1.01,0.3184767961502075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,32,power_law_1.01,0.3479167938232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,32,power_law_1.01,0.4138688087463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,32,power_law_1.01,0.47388157844543455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,32,power_law_1.01,0.6195648193359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,32,power_law_1.01,0.8180352210998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,32,power_law_1.01,1.0762368202209474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,32,power_law_1.01,1.3840895652770997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,8,power_law_1.2,0.13500800132751464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,8,power_law_1.2,0.21278719902038573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,8,power_law_1.2,0.2083967924118042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,8,power_law_1.2,0.2466815948486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,8,power_law_1.2,0.30763518810272217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,8,power_law_1.2,0.3167680025100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,8,power_law_1.2,0.35047039985656736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,1,balanced,0.03793066740036011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,8,power_law_1.2,0.3710848093032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,1,balanced,0.04562133550643921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,1,balanced,0.06669866542021434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,1,balanced,0.07007466753323872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,1,balanced,0.07148266832033794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,1,balanced,0.07146133482456207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,1,balanced,0.07044266661008199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,1,balanced,0.07431999842325847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,1,balanced,0.0731573353211085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,1,balanced,0.07217066486676534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,1,balanced,0.07457600037256877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,1,balanced,0.07549866537253062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,1,balanced,0.0795360008875529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,1,balanced,0.08333333333333333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,1,balanced,0.08897599577903748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,1,balanced,0.09314666191736858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,1,balanced,0.0963253378868103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,1,balanced,0.11769599715868632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,1,balanced,0.12657066186269125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,1,balanced,0.1609173317750295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,1,balanced,0.1992266575495402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,8,power_law_1.2,0.35697920322418214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,1,balanced,0.26130133867263794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,1,balanced,0.31621332963307697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,8,power_law_1.2,0.34257280826568604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,1,balanced,0.43881066640218097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,8,power_law_1.2,0.33856639862060545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,8,power_law_1.2,0.3562304019927979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,8,power_law_1.2,0.33924479484558107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,8,power_law_1.2,0.37048959732055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,8,power_law_1.2,0.3837696075439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,8,power_law_1.2,0.38689279556274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,1,balanced,0.5606773296991984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,8,power_law_1.2,0.4341119766235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,1,balanced,0.7960159778594971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,8,power_law_1.2,0.47080321311950685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,1,balanced,1.0483787059783936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,8,power_law_1.2,0.5001920223236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,8,power_law_1.2,0.5671103954315185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,8,power_law_1.2,0.5413055896759034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,8,power_law_1.2,0.6486527919769287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,8,power_law_1.2,0.7394495964050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,8,power_law_1.2,0.9861248016357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,8,power_law_1.2,1.0454336166381837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,8,power_law_1.2,1.4180992126464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,8,power_law_1.2,1.792518424987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,2,power_law_1.2,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,2,power_law_1.2,0.12923519611358641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,2,power_law_1.2,0.15693440437316894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,2,power_law_1.2,0.22922239303588868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,2,power_law_1.2,0.29600000381469727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,2,power_law_1.2,0.4297215938568115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,2,power_law_1.2,0.5916672229766846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,2,power_law_1.2,0.5998079776763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,2,power_law_1.2,0.5801663875579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,2,power_law_1.2,0.6210432052612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,2,power_law_1.2,0.6280576229095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,2,power_law_1.2,0.6283391952514649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,2,power_law_1.2,0.6630527973175049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,2,power_law_1.2,0.6832128047943116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,2,power_law_1.2,0.6938176155090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,2,power_law_1.2,0.7370431900024415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,2,power_law_1.2,0.7531839847564697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,2,power_law_1.2,0.8440832138061524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,2,power_law_1.2,0.8613632202148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,2,power_law_1.2,1.0170368194580077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,2,power_law_1.2,0.9971199989318847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,2,power_law_1.2,1.1863871574401856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,2,power_law_1.2,1.3552767753601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,2,power_law_1.2,1.7371456146240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,2,power_law_1.2,2.1073535919189452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,2,power_law_1.2,2.693894386291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,2,power_law_1.2,3.260569763183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,4,power_law_1.2,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,4,power_law_1.2,0.05299839973449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,4,power_law_1.2,0.055936002731323244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,4,power_law_1.2,0.060601598024368285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,4,power_law_1.2,0.06954879760742187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,4,power_law_1.2,0.07217280268669128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,4,power_law_1.2,0.09203199744224548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,4,power_law_1.2,0.08650239706039428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,4,power_law_1.2,0.08869119882583618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,4,power_law_1.2,0.0924351990222931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,4,power_law_1.2,0.0960640013217926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,4,power_law_1.2,0.10305919647216796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,4,power_law_1.2,0.10837759971618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,4,power_law_1.2,0.11549439430236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,4,power_law_1.2,0.12272000312805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,4,power_law_1.2,0.1327296018600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,4,power_law_1.2,0.1349503993988037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,4,power_law_1.2,0.15467519760131837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,4,power_law_1.2,0.17912319898605347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,4,power_law_1.2,0.21812479496002196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,4,power_law_1.2,0.2545344114303589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,4,power_law_1.2,0.3167167901992798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,4,power_law_1.2,0.40577921867370603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,4,power_law_1.2,0.5551424026489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,4,power_law_1.2,0.7383808135986328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,4,power_law_1.2,1.01146240234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,4,power_law_1.2,1.3476160049438477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,1,balanced,0.023578666150569916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,1,balanced,0.025472000241279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,1,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,1,balanced,0.0554613322019577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,1,balanced,0.08658666412035625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,1,balanced,0.15019733707110086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,1,balanced,0.15210666259129843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,1,balanced,0.1566933294137319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,1,balanced,0.1623146633307139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,1,balanced,0.1649279991785685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,1,balanced,0.17521067460378012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,1,balanced,0.1835413376490275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,1,balanced,0.19402132431666055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,1,balanced,0.21069333950678507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,1,balanced,0.1965279976526896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,1,balanced,0.20458666483561197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,1,balanced,0.22409067551294962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,1,balanced,0.22394132614135742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,1,balanced,0.24654932816823324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,1,balanced,0.2572266658147176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,1,balanced,0.2804800073305766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,1,balanced,0.5594186782836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,1,balanced,0.6367306709289551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,1,balanced,0.5255573193232218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,1,balanced,0.5766133467356364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,1,balanced,0.9480640093485514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,1,balanced,1.085098663965861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,8,balanced,0.021525333325068157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,8,balanced,0.020037333170572918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,8,balanced,0.019029332945744198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,8,balanced,0.0210506667693456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,8,balanced,0.023503998915354412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,8,balanced,0.029066666960716248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,8,balanced,0.03160000095764796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,8,balanced,0.029850666721661884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,8,balanced,0.03146666785081228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,8,balanced,0.03349866718053818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,8,balanced,0.03287466615438461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,8,balanced,0.03536533315976461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,8,balanced,0.03713600089152654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,8,balanced,0.03934400031963984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,8,balanced,0.039135999977588654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,8,balanced,0.04139200101296107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,8,balanced,0.06567466755708058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,8,balanced,0.07085333267847697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,8,balanced,0.07835733393828075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,8,balanced,0.10506133238474528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,8,balanced,0.090938667456309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,8,balanced,0.11100266377131145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,8,balanced,0.13052800297737122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,8,balanced,0.18796799580256143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,8,balanced,0.2218773365020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,8,balanced,0.313914676507314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,8,balanced,0.40757866700490314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,16,power_law_1.2,0.04689280092716217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,16,power_law_1.2,0.06445440053939819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,16,power_law_1.2,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,16,power_law_1.2,0.07087360024452209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,16,power_law_1.2,0.07207679748535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,16,power_law_1.2,0.07289599776268005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,16,power_law_1.2,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,16,power_law_1.2,0.0707264006137848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,16,power_law_1.2,0.06794880032539367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,16,power_law_1.2,0.06876159906387329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,16,power_law_1.2,0.07210879921913146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,16,power_law_1.2,0.07747840285301208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,16,power_law_1.2,0.0794368028640747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,16,power_law_1.2,0.08276479840278625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,16,power_law_1.2,0.08984959721565247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,16,power_law_1.2,0.09543039798736572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,16,power_law_1.2,0.10620160102844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,2,power_law_1.2,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,16,power_law_1.2,0.12350720167160034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,2,power_law_1.2,0.05552639961242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,16,power_law_1.2,0.13087999820709229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,2,power_law_1.2,0.0636352002620697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,16,power_law_1.2,0.16405760049819945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,2,power_law_1.2,0.07383040189743043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,16,power_law_1.2,0.1876736044883728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,2,power_law_1.2,0.08163840174674988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,16,power_law_1.2,0.2585088014602661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,2,power_law_1.2,0.10227199792861938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,2,power_law_1.2,0.10490880012512208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,16,power_law_1.2,0.317792010307312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,2,power_law_1.2,0.10888960361480712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,16,power_law_1.2,0.43389439582824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,2,power_law_1.2,0.10856959819793702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,16,power_law_1.2,0.5548096179962159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,2,power_law_1.2,0.10968960523605346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,2,power_law_1.2,0.11074559688568116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,16,power_law_1.2,0.7777984142303467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,2,power_law_1.2,0.1103551983833313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,2,power_law_1.2,0.11187200546264649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,2,power_law_1.2,0.11498880386352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,2,power_law_1.2,0.1212224006652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,2,power_law_1.2,0.12889599800109863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,2,power_law_1.2,0.13075200319290162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,2,power_law_1.2,0.1481727957725525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,2,power_law_1.2,0.16383999586105347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,2,power_law_1.2,0.19688960313796997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,16,power_law_1.2,1.0344767570495605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,2,power_law_1.2,0.24099199771881102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,2,power_law_1.2,0.2881216049194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,32,balanced,0.049973333875338234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,32,balanced,0.044069334864616394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,32,balanced,0.04353600243727366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,32,balanced,0.04629333317279816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,32,balanced,0.048842668533325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,32,balanced,0.06897066533565521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,32,balanced,0.07050666709740956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,32,balanced,0.07039999961853027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,32,balanced,0.07017066578070323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,32,balanced,0.0721973329782486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,32,balanced,0.07203199962774913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,32,balanced,0.07281066477298737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,2,power_law_1.2,0.38226559162139895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,2,power_law_1.2,0.545849609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,32,balanced,0.07462933162848155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,32,balanced,0.0763733337322871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,32,balanced,0.07943999767303467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,32,balanced,0.08106666803359985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,32,balanced,0.08247466882069905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,32,balanced,0.08793600400288899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,32,balanced,0.09207466244697571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,32,balanced,0.10342933734258015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,32,balanced,0.11115200320879619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,32,balanced,0.12958932916323343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,32,balanced,0.1497813363869985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,32,balanced,0.19211200873057047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,32,balanced,0.22959999243418375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,32,balanced,0.3091786702473958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,2,power_law_1.2,0.6685440063476562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,32,balanced,0.38949867089589435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,2,power_law_1.2,0.9484543800354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,128,power_law_1.2,0.05916159749031067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,128,power_law_1.2,0.06183040142059326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,128,power_law_1.2,0.060659199953079224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,128,power_law_1.2,0.06096640229225159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,128,power_law_1.2,0.059084802865982056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,128,power_law_1.2,0.058790397644042966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,128,power_law_1.2,0.06042879819869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,128,power_law_1.2,0.060127997398376466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,128,power_law_1.2,0.061324799060821535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,128,power_law_1.2,0.061324799060821535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,128,power_law_1.2,0.06130560040473938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,128,power_law_1.2,0.06462720036506653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,128,power_law_1.2,0.0648256003856659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,128,power_law_1.2,0.06711040139198303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,128,power_law_1.2,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,128,power_law_1.2,0.0759872019290924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,128,power_law_1.2,0.08391680121421814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,128,power_law_1.2,0.09845759868621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,128,power_law_1.2,0.11281919479370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,128,power_law_1.2,0.14689919948577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,128,power_law_1.2,0.1655295968055725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,128,power_law_1.2,0.2267712116241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,128,power_law_1.2,0.22929279804229735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,128,power_law_1.2,0.32624640464782717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,128,power_law_1.2,0.39819519519805907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,1,power_law_1.01,0.055244797468185426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,128,power_law_1.2,0.5648064136505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,1,power_law_1.01,0.05916159749031067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,1,power_law_1.01,0.06673280000686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,1,power_law_1.01,0.09028480052947999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,1,power_law_1.01,0.11230720281600952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,1,power_law_1.01,0.1290112018585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,1,power_law_1.01,0.1634559988975525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,1,power_law_1.01,0.1686079978942871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,1,power_law_1.01,0.17452800273895264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,1,power_law_1.01,0.17891199588775636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,1,power_law_1.01,0.18659199476242067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,128,power_law_1.2,0.7301055908203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,1,power_law_1.01,0.20007679462432862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,16,power_law_1.2,0.07051519751548767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,1,power_law_1.01,0.20787839889526366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,16,power_law_1.2,0.10887680053710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,16,power_law_1.2,0.0995199978351593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,16,power_law_1.2,0.11219840049743653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,16,power_law_1.2,0.09836800098419189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,16,power_law_1.2,0.08220800161361694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,16,power_law_1.2,0.10201599597930908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,1,power_law_1.01,0.22875521183013917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,1,power_law_1.01,0.24260480403900148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,1,power_law_1.01,0.25679359436035154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,1,power_law_1.01,0.2486272096633911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,1,power_law_1.01,0.29910399913787844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,1,power_law_1.01,0.3337791919708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,16,power_law_1.2,0.10648319721221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,1,power_law_1.01,0.4230783939361572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,16,power_law_1.2,0.10106240510940552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,1,power_law_1.01,0.500435209274292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,1,power_law_1.01,0.6672255992889404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,1,power_law_1.01,0.8294464111328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,1,power_law_1.01,1.1422143936157227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,16,power_law_1.2,0.10353920459747315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,1,power_law_1.01,1.4438079833984374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,16,power_law_1.2,0.10675200223922729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,16,power_law_1.2,0.10962560176849365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,16,power_law_1.2,0.11103359460830689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,1,power_law_1.01,2.0558528900146484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,16,power_law_1.2,0.11294080018997192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,16,power_law_1.2,0.11715199947357177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,16,power_law_1.2,0.12383359670639038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,16,power_law_1.2,0.1283455967903137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,1,power_law_1.01,2.6814720153808596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,16,power_law_1.2,0.15010559558868408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,16,power_law_1.2,0.1475648045539856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,16,power_law_1.2,0.17452800273895264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,16,power_law_1.2,0.18343039751052856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,16,power_law_1.2,0.22490239143371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,16,power_law_1.2,0.24598400592803954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,16,power_law_1.2,0.33018879890441893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,16,power_law_1.2,0.3998784065246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,16,power_law_1.2,0.6359295845031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,16,power_law_1.2,0.713478422164917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,2,power_law_1.2,1.2140288352966309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,4,power_law_1.01,0.0451200008392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,4,power_law_1.01,0.048588800430297854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,4,power_law_1.01,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,4,power_law_1.01,0.055327999591827395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,4,power_law_1.01,0.06058239936828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,4,power_law_1.01,0.06817280054092408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,4,power_law_1.01,0.07122560143470764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,4,power_law_1.01,0.07674239873886109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,4,power_law_1.01,0.07553279995918274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,4,power_law_1.01,0.07438079714775085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,4,power_law_1.01,0.5279295921325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,4,power_law_1.01,0.08078719973564148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,4,power_law_1.01,0.08198400139808655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,4,power_law_1.01,0.08412799835205079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,4,power_law_1.01,0.09266560077667237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,4,power_law_1.01,0.09685760140419006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,4,power_law_1.01,0.10084480047225952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,4,power_law_1.01,0.116428804397583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,4,power_law_1.01,0.13150080442428588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,4,power_law_1.01,0.16074880361557006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,4,power_law_1.01,0.21103360652923583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,4,power_law_1.01,0.26866559982299804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,4,power_law_1.01,0.3032448053359985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,4,power_law_1.01,0.43503360748291015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,4,power_law_1.01,0.555244779586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,4,power_law_1.01,0.7687424182891845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,4,power_law_1.01,0.9693120002746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,4,power_law_1.01,0.037625598907470706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,4,power_law_1.01,0.037011200189590455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,4,power_law_1.01,0.036057600378990175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,4,power_law_1.01,0.03969280123710632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,4,power_law_1.01,0.04017280042171478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,4,power_law_1.01,0.04009599983692169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,4,power_law_1.01,0.04074240028858185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,4,power_law_1.01,0.04077439904212952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,4,power_law_1.01,0.04095999896526337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,4,power_law_1.01,0.04089600145816803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,4,power_law_1.01,0.0411327987909317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,4,power_law_1.01,0.04216319918632507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,4,power_law_1.01,0.04356479942798615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,4,power_law_1.01,0.043673598766326906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,4,power_law_1.01,0.049798399209976196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,4,power_law_1.01,0.04997760057449341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,4,power_law_1.01,0.052288001775741576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,4,power_law_1.01,0.05911679863929749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.01,0.06577919721603394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.01,0.07973120212554932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.01,0.08752639889717102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.01,0.11694079637527466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.01,0.14608639478683472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.01,0.1936511993408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.01,0.2451711893081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,32,8,power_law_1.01,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,8,power_law_1.2,0.0945472002029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,32,8,power_law_1.01,0.04596480131149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,8,power_law_1.2,0.143340802192688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,8,power_law_1.2,0.1393280029296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,32,8,power_law_1.01,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,8,power_law_1.2,0.1724671959877014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,32,8,power_law_1.01,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,8,power_law_1.2,0.1863808035850525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,32,8,power_law_1.01,0.053286397457122804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,8,power_law_1.2,0.19680639505386352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,32,8,power_law_1.01,0.05724800229072571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,8,power_law_1.2,0.20955519676208495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,8,power_law_1.2,0.20158720016479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,8,power_law_1.2,0.22098560333251954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,8,power_law_1.2,0.2398848056793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,8,power_law_1.2,0.221068811416626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,8,power_law_1.2,0.2125312089920044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,32,8,power_law_1.01,0.05663359761238098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,32,8,power_law_1.01,0.05976960062980652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,32,8,power_law_1.01,0.058297598361968996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,32,8,power_law_1.01,0.06030719876289368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,32,8,power_law_1.01,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,32,8,power_law_1.01,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,32,8,power_law_1.01,0.06442239880561829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,8,power_law_1.2,0.23319680690765382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,32,8,power_law_1.01,0.06723840236663818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,32,8,power_law_1.01,0.07433599829673768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,32,8,power_law_1.01,0.0766592025756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,32,8,power_law_1.01,0.08294399976730346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,32,8,power_law_1.01,0.0964031994342804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,32,8,power_law_1.01,0.1034432053565979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,8,power_law_1.2,0.23276801109313966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,32,8,power_law_1.01,0.1251520037651062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,32,8,power_law_1.01,0.14108799695968627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,32,8,power_law_1.01,0.19213440418243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,32,8,power_law_1.01,0.2374720096588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,32,8,power_law_1.01,0.30719358921051027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,32,8,power_law_1.01,0.3846911907196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,8,power_law_1.2,0.24507520198822022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,8,power_law_1.2,0.2502847909927368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,32,8,power_law_1.01,0.5807936191558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,8,power_law_1.2,0.2613312005996704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,32,8,power_law_1.01,0.7376063823699951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,8,power_law_1.2,0.3149696111679077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,8,power_law_1.2,0.32921600341796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,8,power_law_1.2,0.3977407932281494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,8,power_law_1.2,0.41608319282531736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,8,power_law_1.2,0.502291202545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,8,power_law_1.2,0.5186495780944824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,8,power_law_1.2,0.7477952003479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,8,power_law_1.2,0.8200639724731446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,8,power_law_1.2,1.0712384223937987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,8,power_law_1.2,1.3880703926086426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.01,0.38128640651702883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.01,0.47133440971374513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,32,power_law_1.01,0.08010879755020142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,32,power_law_1.01,0.088755202293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,32,power_law_1.01,0.08613119721412658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,32,power_law_1.01,0.08814079761505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,32,power_law_1.01,0.08852480053901672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,32,power_law_1.01,0.08618239760398864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,32,power_law_1.01,0.08432000279426574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,32,power_law_1.01,0.08740479946136474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,32,power_law_1.01,0.0899392008781433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,32,power_law_1.01,0.08695679903030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,32,power_law_1.01,0.09048320055007934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,32,power_law_1.01,0.08867200016975403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,32,power_law_1.01,0.09157760143280029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,32,power_law_1.01,0.09077119827270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,32,power_law_1.01,0.09902719855308532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,32,power_law_1.01,0.09874560236930847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,32,power_law_1.01,0.10362240076065063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,32,power_law_1.01,0.11264640092849731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,32,power_law_1.01,0.121343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,32,power_law_1.01,0.13854080438613892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,32,power_law_1.01,0.1577407956123352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,32,power_law_1.01,0.19392000436782836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,32,power_law_1.01,0.22791678905487062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,32,power_law_1.01,0.2945280075073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,1,power_law_1.01,0.07713919878005981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,32,power_law_1.01,0.3802175998687744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,1,power_law_1.01,0.10271999835968018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,32,power_law_1.01,0.5283904075622559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,1,power_law_1.01,0.14650239944458007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,32,power_law_1.01,0.7001471996307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,1,power_law_1.01,0.24970879554748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,1,power_law_1.01,0.35387520790100097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,1,power_law_1.01,0.4575039863586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,1,power_law_1.01,0.6338624000549317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,1,power_law_1.01,0.6529727935791015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,1,power_law_1.01,0.6805568218231202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,32,balanced,0.05377600093682607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,32,balanced,0.04390400151411692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,1,power_law_1.01,0.7039040088653564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,32,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,32,balanced,0.048986668388048805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,32,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,32,balanced,0.07468800246715546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,1,power_law_1.01,0.7287680149078369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,1,power_law_1.01,0.7485504150390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,1,power_law_1.01,0.7618815898895264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,1,power_law_1.01,0.7808703899383544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,1,power_law_1.01,0.8285440444946289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,32,balanced,0.09193066755930583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,32,balanced,0.09152000149091084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,32,balanced,0.09096533060073853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,32,balanced,0.0918293297290802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,1,power_law_1.01,0.8295680046081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,32,balanced,0.09291199843088786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,32,balanced,0.09194667140642802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,32,balanced,0.0926026701927185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,32,balanced,0.09469333291053772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,32,balanced,0.10022399822870891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,1,power_law_1.01,0.8886079788208008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,32,balanced,0.1011946698029836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,32,balanced,0.1060746709505717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,32,balanced,0.11272000273068745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,1,power_law_1.01,0.9715071678161621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,32,balanced,0.1181066632270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,32,balanced,0.13270399967829385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,1,power_law_1.01,1.0095808029174804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,32,balanced,0.14321600397427878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,32,balanced,0.17034133275349936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,1,power_law_1.01,1.155174446105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,1,power_law_1.01,1.225875186920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,1,power_law_1.01,1.4673727989196776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,32,balanced,0.2007466753323873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,32,balanced,0.2550666729609172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,32,balanced,0.3105280001958211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,1,power_law_1.01,1.7039615631103515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,32,balanced,0.4118613402048747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,32,balanced,0.5250720183054606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,1,power_law_1.01,2.2130111694335937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,1,power_law_1.01,2.683193588256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,1,power_law_1.01,3.647372817993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,1,power_law_1.01,4.61409912109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,4,power_law_1.2,0.0884607970714569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,4,power_law_1.2,0.10488959550857543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,4,power_law_1.2,0.11636480093002319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,4,power_law_1.2,0.13085440397262574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,4,power_law_1.2,0.14606720209121704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,4,power_law_1.2,0.16625280380249025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,4,power_law_1.2,0.17072639465332032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,4,power_law_1.2,0.18647040128707887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,4,power_law_1.2,0.19298559427261353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,4,power_law_1.2,0.1907263994216919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,4,power_law_1.2,0.18754559755325317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,4,power_law_1.2,0.18829439878463744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,4,power_law_1.2,0.19349759817123413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,4,power_law_1.2,0.19808000326156616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,4,power_law_1.2,0.21426560878753662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,4,power_law_1.2,0.21059200763702393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,4,power_law_1.2,0.22111999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,4,power_law_1.2,0.24368638992309571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,4,power_law_1.2,0.2644479990005493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,4,power_law_1.2,0.3190783977508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,4,power_law_1.2,0.38204801082611084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,4,power_law_1.2,0.4799808025360107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,4,power_law_1.2,0.6073535919189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,4,power_law_1.2,0.8409279823303223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,4,power_law_1.2,1.0657855987548828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,64,balanced,0.019194666296243668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,64,balanced,0.01836266616980235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,64,balanced,0.01887999971707662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,64,balanced,0.01903466631968816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,4,power_law_1.2,1.513702392578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,64,balanced,0.020981334149837494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,64,balanced,0.02276800076166789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,64,balanced,0.03739733248949051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,64,balanced,0.037317333122094475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,64,balanced,0.031221332649389904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,64,balanced,0.031221332649389904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,64,balanced,0.03125333289305369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,4,power_law_1.2,1.8527551651000977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,64,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,64,balanced,0.031173333525657654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,64,balanced,0.031370667119820915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,64,balanced,0.0377813329299291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,64,balanced,0.03700799991687139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,64,balanced,0.03739733248949051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,64,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,64,balanced,0.04286933441956838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,64,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,64,balanced,0.062037333846092224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,64,balanced,0.07630933324495952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,64,balanced,0.09478400150934856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,64,balanced,0.12652800480524698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,64,balanced,0.16247999668121338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,64,balanced,0.22612265745798746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,64,balanced,0.29496532678604126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,32,balanced,0.03201599915822347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,32,balanced,0.027402666707833607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,32,balanced,0.029493334392706554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,32,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,32,balanced,0.02951466788848241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,32,balanced,0.03160533308982849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,32,balanced,0.03142933299144109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,256,balanced,0.053157334526379905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,256,balanced,0.0644160012404124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,256,balanced,0.04205333193143209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,256,balanced,0.043882668018341064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,256,balanced,0.0414986660083135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,256,balanced,0.04428799947102865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,256,balanced,0.04390933116277059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,256,balanced,0.04554666578769684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,256,balanced,0.045610666275024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,256,balanced,0.04604800045490265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,256,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,256,balanced,0.046800002455711365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,256,balanced,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,256,balanced,0.04808533191680908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,256,balanced,0.0525493323802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,32,balanced,0.02974933385848999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,256,balanced,0.05389333268006643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,32,balanced,0.031189332405726116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,256,balanced,0.05605333546797434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,32,balanced,0.03296533226966858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,256,balanced,0.06204266846179962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,256,balanced,0.0653599997361501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,256,balanced,0.07426666716734569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,256,balanced,0.07876800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,256,balanced,0.0971999963124593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,64,power_law_1.2,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,256,balanced,0.11942399541536967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,256,balanced,0.17021334171295166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,256,balanced,0.2128480076789856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,256,balanced,0.29134400685628253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,256,balanced,0.3720800081888835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,32,balanced,0.03143999973932902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,32,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,32,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,32,balanced,0.03146666785081228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,32,balanced,0.03728000074625015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,32,balanced,0.03551999976237615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,32,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,32,balanced,0.039861333866914116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,32,balanced,0.039690665900707245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,32,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,32,balanced,0.04112533231576284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,32,balanced,0.047610665361086525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,32,balanced,0.04784533381462097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,32,balanced,0.055674667159716286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,32,balanced,0.06399466594060262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,32,balanced,0.08072000245253245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,32,balanced,0.09679466485977173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,64,power_law_1.2,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,64,power_law_1.2,0.030988800525665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,64,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,64,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,64,power_law_1.2,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,64,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,64,power_law_1.2,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,64,power_law_1.2,0.03548159897327423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,64,power_law_1.2,0.03529599905014038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,64,power_law_1.2,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,64,power_law_1.2,0.036473599076271054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,64,power_law_1.2,0.037567999958992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,64,power_law_1.2,0.03774079978466034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,64,power_law_1.2,0.041536000370979306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,64,power_law_1.2,0.04140160083770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,64,power_law_1.2,0.042694398760795595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,64,power_law_1.2,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,64,power_law_1.2,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,64,power_law_1.2,0.05331839919090271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,64,power_law_1.2,0.06275200247764587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,64,power_law_1.2,0.07223039865493774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,64,power_law_1.2,0.08513280153274536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,64,power_law_1.2,0.11413760185241699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,64,power_law_1.2,0.14730240106582643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,64,power_law_1.2,0.18727680444717407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,2,power_law_1.01,0.03160960078239441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,2,power_law_1.01,0.03091199994087219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,2,power_law_1.01,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,2,power_law_1.01,0.032441601157188416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,2,power_law_1.01,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,2,power_law_1.01,0.03898879885673523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,2,power_law_1.01,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,2,power_law_1.01,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,2,power_law_1.01,0.06314240097999572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,2,power_law_1.01,0.05172479748725891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,2,power_law_1.01,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,2,power_law_1.01,0.06955519914627076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,2,power_law_1.01,0.07627519965171814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,2,power_law_1.01,0.07375360131263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,32,power_law_1.01,0.06846719980239868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,32,power_law_1.01,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,2,power_law_1.01,0.1065600037574768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,2,power_law_1.01,0.08462719917297364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,32,power_law_1.01,0.06645119786262513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,2,power_law_1.01,0.09880319833755494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,32,power_law_1.01,0.06694399714469909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,2,power_law_1.01,0.13818880319595336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.01,0.16272000074386597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.01,0.23059840202331544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,32,power_law_1.01,0.06762880086898804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,32,power_law_1.01,0.0692351996898651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,32,power_law_1.01,0.06890239715576171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,32,power_law_1.01,0.06779519915580749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.01,0.2918528079986572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,32,power_law_1.01,0.07034879922866821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,128,power_law_1.2,0.022163200378417968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,32,power_law_1.01,0.07208319902420043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,128,power_law_1.2,0.021484799683094025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,32,power_law_1.01,0.07185279726982116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,128,power_law_1.2,0.022137600183486938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,32,power_law_1.01,0.07278079986572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,128,power_law_1.2,0.023801599442958832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,32,power_law_1.01,0.07448319792747497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,128,power_law_1.2,0.0310591995716095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,32,power_law_1.01,0.07717120051383972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.01,0.4189311981201172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,128,power_law_1.2,0.0506816029548645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,128,power_law_1.2,0.048870399594306946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,128,power_law_1.2,0.049497601389884946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,128,power_law_1.2,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,128,power_law_1.2,0.04912639856338501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,128,power_law_1.2,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,32,power_law_1.01,0.08179200291633607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,128,power_law_1.2,0.049670401215553286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,32,power_law_1.01,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,128,power_law_1.2,0.047814399003982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.01,0.5304831981658935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,32,power_law_1.01,0.09190400242805481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,128,power_law_1.2,0.04362240135669708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,32,power_law_1.01,0.10665600299835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,128,power_law_1.2,0.0427839994430542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,32,power_law_1.01,0.11949440240859985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,32,power_law_1.01,0.14876799583435057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,32,power_law_1.01,0.17310719490051268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,32,power_law_1.01,0.2203903913497925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,32,power_law_1.01,0.26834559440612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,32,power_law_1.01,0.37544960975646974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,128,power_law_1.2,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,32,power_law_1.01,0.4864704132080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,128,power_law_1.2,0.05023360252380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,128,power_law_1.2,0.056454402208328244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,32,power_law_1.01,0.6836671829223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,128,power_law_1.2,0.06714239716529846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,32,power_law_1.01,0.9024831771850585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,128,power_law_1.2,0.07765759825706482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,128,power_law_1.2,0.09619200229644775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,64,power_law_1.2,0.26796159744262693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,128,power_law_1.2,0.12588160037994384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,128,power_law_1.2,0.16291199922561644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,128,power_law_1.2,0.22501759529113768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,128,power_law_1.2,0.2934720039367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,128,power_law_1.2,0.42163839340209963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,128,power_law_1.2,0.5521535873413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.01,0.782585620880127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.01,1.0201536178588868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,4,power_law_1.01,0.020108799636363982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,4,power_law_1.01,0.019091199338436126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,4,power_law_1.01,0.021433599293231964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,4,power_law_1.01,0.024255999922752382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,4,power_law_1.01,0.024748800694942473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,4,power_law_1.01,0.027980801463127137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.01,1.5096256256103515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,2,balanced,0.048986668388048805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,2,balanced,0.058373332023620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,2,balanced,0.07204799850781758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,2,balanced,0.11359467109044392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,2,balanced,0.1660426656405131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,2,balanced,0.19234667221705118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,2,balanced,0.1744746764500936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,2,balanced,0.17312000195185342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,2,balanced,0.166703999042511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,2,balanced,0.1663093368212382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,2,balanced,0.17051732540130615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,4,power_law_1.01,0.029555198550224305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,2,balanced,0.16959466536839804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,2,balanced,0.17146132389704385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,2,balanced,0.17353065808614096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,2,balanced,0.18069332838058472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,2,balanced,0.18343466520309448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,2,balanced,0.1786293387413025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,2,balanced,0.18708266814549765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,2,balanced,0.19607466459274292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,2,balanced,0.212991992632548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,2,balanced,0.23460266987482706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,2,balanced,0.37141867478688556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,2,balanced,0.38756267229715985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,4,power_law_1.01,0.030828800797462464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,2,balanced,0.572154680887858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,4,power_law_1.01,0.03139199912548065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,2,balanced,0.6527946790059408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,2,balanced,0.9600213368733724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,2,balanced,1.1896159648895264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,4,power_law_1.01,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,4,power_law_1.01,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,4,power_law_1.01,0.0377023994922638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,4,power_law_1.01,0.04035840034484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,4,power_law_1.01,0.038227200508117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,4,power_law_1.01,0.04232319891452789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,4,power_law_1.01,0.04519039988517761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,4,power_law_1.01,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,4,power_law_1.01,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.01,0.09437440037727356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.01,0.10417920351028442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.01,0.08817279934883118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.01,0.10847359895706177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.01,0.1295232057571411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.01,0.17076480388641357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.01,0.20811519622802735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.01,0.291430401802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.01,0.37635838985443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.01,1.9986560821533204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,16,balanced,0.02995733420054118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,16,balanced,0.027701333165168762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,16,balanced,0.0276853342851003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,16,balanced,0.029674666623274486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,16,balanced,0.03275199979543686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,16,balanced,0.033359999457995095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,16,balanced,0.03310933212439219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,16,balanced,0.034117333590984344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,16,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,16,balanced,0.03375466664632162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,16,balanced,0.03548266738653183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,16,balanced,0.033728001018365227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,16,balanced,0.035930665830771126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,16,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,16,balanced,0.03937600056330363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,16,balanced,0.04146666576464971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,16,balanced,0.04156800111134847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,16,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,16,balanced,0.043578664461771645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,16,balanced,0.05012799799442291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,16,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,16,power_law_1.2,0.18995840549468995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,16,balanced,0.06413866579532623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,16,power_law_1.2,0.2368704080581665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,16,balanced,0.070592001080513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,16,power_law_1.2,0.24876160621643068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,16,balanced,0.10466667016347249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,16,balanced,0.13153599699338278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,16,balanced,0.1750453313191732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,16,balanced,0.22073066234588623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,16,power_law_1.2,0.25614080429077146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,16,power_law_1.2,0.2339008092880249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,16,power_law_1.2,0.2398400068283081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,16,power_law_1.2,0.2294464111328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,16,power_law_1.2,0.24584319591522216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,16,power_law_1.2,0.23204479217529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,16,power_law_1.2,0.23473920822143554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,16,power_law_1.2,0.23466880321502687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,16,power_law_1.2,0.24458880424499513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,16,power_law_1.2,0.24630401134490967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,16,power_law_1.2,0.2557568073272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,16,power_law_1.2,0.27181439399719237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,16,power_law_1.2,0.279417610168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,16,power_law_1.2,0.3000447988510132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,16,power_law_1.2,0.3419264078140259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,16,power_law_1.2,0.35364480018615724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,16,power_law_1.2,0.43237757682800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,16,power_law_1.2,0.5112192153930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,16,power_law_1.2,0.6621568202972412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,16,power_law_1.2,0.6405375957489013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,16,power_law_1.2,0.7948160171508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,16,power_law_1.2,1.0592255592346191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,16,power_law_1.2,1.5802495956420899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,4,balanced,0.06445866823196411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,4,balanced,0.0869813362757365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,4,balanced,0.12752532958984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,4,balanced,0.2087200085322062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,4,balanced,0.3415733178456624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,4,balanced,0.45954132080078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,4,balanced,0.4615679979324341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,4,balanced,0.45746668179829914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,4,balanced,0.46053866545359295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,4,balanced,0.4594080050786336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,4,balanced,0.4652426640192668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,4,balanced,0.4691093365351359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,4,balanced,0.4697546561559041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,4,balanced,0.4768213431040446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,4,balanced,0.48552000522613525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,4,balanced,0.48767467339833576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,4,balanced,0.5003519852956136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,4,balanced,0.5709973176320394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,4,balanced,0.5432053407033285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,4,balanced,0.6939786275227865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,4,balanced,0.6380000114440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,4,balanced,1.0261867046356201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,4,balanced,0.8092106978098551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,4,balanced,1.2650293509165447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,4,balanced,1.2377173105875652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,4,balanced,1.9948533376057942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,4,balanced,2.048543930053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,16,power_law_1.2,1.6366847991943358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,8,power_law_1.01,0.022316800057888032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,8,power_law_1.01,0.024031999707221984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,8,power_law_1.01,0.025785601139068602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,8,power_law_1.01,0.0274944007396698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,8,power_law_1.01,0.027430400252342224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,8,power_law_1.01,0.030220800638198854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,8,power_law_1.01,0.03487359881401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,8,power_law_1.01,0.034169599413871765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,8,power_law_1.01,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,8,power_law_1.01,0.044607999920845035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,8,power_law_1.01,0.043244799971580504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,8,power_law_1.01,0.06490880250930786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,8,power_law_1.01,0.06444159746170045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,8,power_law_1.01,0.06417919993400574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,8,power_law_1.01,0.07073280215263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,8,power_law_1.01,0.07185919880867005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,64,power_law_1.01,0.061900800466537474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,64,power_law_1.01,0.05103359818458557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,64,power_law_1.01,0.049395200610160825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,64,power_law_1.01,0.05261440277099609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,64,power_law_1.01,0.05204480290412903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,64,power_law_1.01,0.05186560153961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,8,power_law_1.01,0.06535040140151978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,64,power_law_1.01,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,64,power_law_1.01,0.05429760217666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,64,power_law_1.01,0.054636800289154054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,64,power_law_1.01,0.05425919890403748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,64,power_law_1.01,0.05472000241279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,64,power_law_1.01,0.05633919835090637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,64,power_law_1.01,0.05687680244445801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,8,power_law_1.01,0.06415359973907471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,64,power_law_1.01,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.01,0.07123839855194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,64,power_law_1.01,0.06358399987220764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,64,power_law_1.01,0.06405119895935059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,64,power_law_1.01,0.07061759829521179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,64,power_law_1.01,0.08295680284500122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,64,power_law_1.01,0.08902400135993957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,64,power_law_1.01,0.10840320587158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,64,power_law_1.01,0.1217919945716858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.01,0.08083840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,64,power_law_1.01,0.1476863980293274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.01,0.08620160222053527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,64,power_law_1.01,0.17912319898605347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.01,0.10814080238342286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,64,power_law_1.01,0.23829119205474852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,64,power_law_1.01,0.30192639827728274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.01,0.12614400386810304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,64,power_law_1.01,0.4203775882720947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.01,0.17427200078964233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.01,0.20884480476379394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,64,power_law_1.01,0.5155903816223144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.01,0.3031872034072876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.01,0.37363200187683104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,128,power_law_1.2,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,128,power_law_1.2,0.05134080052375793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,128,power_law_1.2,0.05315840244293213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,128,power_law_1.2,0.0553600013256073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,128,power_law_1.2,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,128,power_law_1.2,0.05642240047454834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,64,power_law_1.01,0.03207040131092072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,64,power_law_1.01,0.03147520124912262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,64,power_law_1.01,0.03141759932041168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,64,power_law_1.01,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,64,power_law_1.01,0.034246399998664856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,64,power_law_1.01,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,64,power_law_1.01,0.03540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,64,power_law_1.01,0.03619840145111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,64,power_law_1.01,0.03640320003032684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,64,power_law_1.01,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,128,power_law_1.2,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,64,power_law_1.01,0.03712640106678009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,128,power_law_1.2,0.05660799741744995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,128,power_law_1.2,0.057196801900863646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,128,power_law_1.2,0.057196801900863646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,128,power_law_1.2,0.058348798751831056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,64,power_law_1.01,0.03779839873313904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,64,power_law_1.01,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,64,power_law_1.01,0.03828479945659637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,64,power_law_1.01,0.04082559943199158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,128,power_law_1.2,0.06215680241584778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,64,power_law_1.01,0.04156799912452698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,64,power_law_1.01,0.042508798837661746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,64,power_law_1.01,0.0445248007774353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,64,power_law_1.01,0.04805760085582733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,64,power_law_1.01,0.0526528000831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,64,power_law_1.01,0.05676159858703613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,64,power_law_1.01,0.06453120112419128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,64,power_law_1.01,0.073990398645401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,64,power_law_1.01,0.1050879955291748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,64,power_law_1.01,0.12078720331192017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,128,power_law_1.2,0.06266239881515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,64,power_law_1.01,0.18326400518417357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,128,power_law_1.2,0.06450560092926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,64,power_law_1.01,0.20920319557189943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,128,power_law_1.2,0.07073919773101807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,128,power_law_1.2,0.07294719815254211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,128,power_law_1.2,0.07845759987831116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,128,power_law_1.2,0.092467200756073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,128,power_law_1.2,0.1059008002281189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,128,power_law_1.2,0.13446400165557862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,128,power_law_1.2,0.15368319749832154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,128,power_law_1.2,0.21489920616149902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,128,power_law_1.2,0.21410560607910156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,128,power_law_1.2,0.2971839904785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,128,power_law_1.2,0.3379519939422607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,128,power_law_1.2,0.5239744186401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,128,power_law_1.2,0.6650688171386718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,32,power_law_1.2,0.04410879909992218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,32,power_law_1.2,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,32,power_law_1.2,0.04408319890499115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,32,power_law_1.2,0.04555520117282867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,32,power_law_1.2,0.04581120014190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,32,power_law_1.2,0.04572800099849701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,32,power_law_1.2,0.0472896009683609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,32,power_law_1.2,0.04793600142002106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,32,power_law_1.2,0.04888960123062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,32,power_law_1.2,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,32,power_law_1.2,0.05062400102615357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,32,power_law_1.2,0.05167359709739685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,32,power_law_1.2,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,32,power_law_1.2,0.05539199709892273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,32,power_law_1.2,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,32,power_law_1.2,0.06071680188179016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,32,power_law_1.2,0.06529920101165772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,32,power_law_1.2,0.07425280213356018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,32,power_law_1.2,0.07720959782600403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,32,power_law_1.2,0.09227520227432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,32,power_law_1.2,0.10379519462585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,32,power_law_1.2,0.14088319540023803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,32,power_law_1.2,0.17756799459457398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,32,power_law_1.2,0.24664320945739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,32,power_law_1.2,0.30531198978424073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,32,power_law_1.2,0.44209918975830076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,32,power_law_1.2,0.6120704174041748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,8,power_law_1.2,0.026233598589897156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,8,power_law_1.2,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,8,power_law_1.2,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,8,power_law_1.2,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,8,power_law_1.2,0.0634112000465393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,8,power_law_1.2,0.0664192020893097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,8,power_law_1.2,0.06789119839668274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,8,power_law_1.2,0.06696320176124573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,8,power_law_1.2,0.0732096016407013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,8,power_law_1.2,0.0760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,8,power_law_1.2,0.07279360294342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,8,power_law_1.2,0.09086080193519593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,8,power_law_1.2,0.09217919707298279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,8,power_law_1.2,0.08696960210800171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,8,power_law_1.2,0.14096640348434447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,8,power_law_1.2,0.14251519441604615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,8,power_law_1.2,0.1471743941307068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,8,power_law_1.2,0.14929920434951782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,8,power_law_1.2,0.14320000410079955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,8,power_law_1.2,0.15294079780578612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,8,power_law_1.2,0.1698815941810608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,8,power_law_1.2,0.19356800317764283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,8,power_law_1.2,0.23669760227203368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,8,power_law_1.2,0.30380160808563234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,8,power_law_1.2,0.36376960277557374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,8,power_law_1.2,0.5558976173400879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,8,power_law_1.2,0.6797567844390869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,1,balanced,0.0782773345708847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,1,balanced,0.12184533476829529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,1,balanced,0.20934400955835977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,1,balanced,0.2114773392677307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,1,balanced,0.21276267369588217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,1,balanced,0.21211200952529907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,1,balanced,0.21313599745432535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,1,power_law_1.01,0.040582400560379026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,1,balanced,0.21650133530298868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,1,power_law_1.01,0.05170559883117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,1,power_law_1.01,0.06614400148391723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,1,power_law_1.01,0.08275840282440186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,1,power_law_1.01,0.08379520177841186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,1,power_law_1.01,0.08499199748039246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,1,power_law_1.01,0.08753920197486878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,1,power_law_1.01,0.08840320110321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,1,power_law_1.01,0.08836479783058167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,1,power_law_1.01,0.09078400135040283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,1,balanced,0.21770666042963663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,1,power_law_1.01,0.09174399971961975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,1,power_law_1.01,0.09514240026474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,1,power_law_1.01,0.09928320050239563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,1,power_law_1.01,0.10128639936447144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,1,power_law_1.01,0.11057920455932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,1,power_law_1.01,0.11638400554656983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,1,balanced,0.21783467133839926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,1,power_law_1.01,0.12076799869537354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,1,balanced,0.22428800662358603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,1,balanced,0.22805333137512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,1,balanced,0.23152534166971842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,1,balanced,0.24330133199691772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,1,balanced,0.25884799162546795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,1,balanced,0.2624586621920268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,1,balanced,0.2665440042813619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,1,balanced,0.32415999968846637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,1,balanced,0.3433973391850789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,1,balanced,0.43670932451883954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,1,balanced,0.5010879834493002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,1,balanced,0.693615992863973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,1,balanced,0.8964266777038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,1,balanced,1.2953706582387288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,1,power_law_1.01,0.1411967992782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,1,balanced,1.6789439519246419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,1,power_law_1.01,0.16079360246658325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,1,balanced,2.473466714223226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,1,power_law_1.01,0.19752320051193237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,1,balanced,3.2025814056396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,2,balanced,0.05388266841570536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,2,balanced,0.06826133529345195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,2,balanced,0.09435199697812398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,2,balanced,0.1518400013446808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,2,balanced,0.2467893362045288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,2,balanced,0.444485346476237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,2,balanced,0.4456319808959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,2,balanced,0.4459253152211507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,2,balanced,0.4467466672261556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,2,balanced,0.4503999948501587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,2,balanced,0.4519413312276204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,2,balanced,0.45954668521881104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,2,balanced,0.4629280169804891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,2,balanced,0.46828265984853107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,2,balanced,0.47888000806172687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,1,power_law_1.01,0.23522560596466063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,2,balanced,0.4819360176722209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,2,balanced,0.49217065175374347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,2,balanced,0.514138658841451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,2,balanced,0.5393120050430298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,2,balanced,0.5846399863560995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,2,balanced,0.6311786572138468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,2,balanced,0.7092639605204264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,4,power_law_1.01,0.04834559857845307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,4,power_law_1.01,0.05014399886131286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,4,power_law_1.01,0.06440320014953613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,4,power_law_1.01,0.07057920098304749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,4,power_law_1.01,0.07881600260734559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,1,power_law_1.01,0.3128895998001099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,2,balanced,0.8053279717763265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,2,balanced,1.021504004796346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,2,balanced,1.1678400039672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,1,power_law_1.01,0.38593919277191163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,2,balanced,1.7318399747212727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,4,power_law_1.01,0.09170560240745544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,4,power_law_1.01,0.09442560076713562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,4,power_law_1.01,0.09431040287017822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,2,balanced,2.0647786458333335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,4,power_law_1.01,0.09557120203971863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,4,power_law_1.01,0.09833599925041199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,4,power_law_1.01,0.09916800260543823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,4,power_law_1.01,0.10206719636917114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,4,power_law_1.01,0.10323200225830079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,4,power_law_1.01,0.11073280572891235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,4,power_law_1.01,0.11612800359725953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,4,power_law_1.01,0.12063360214233398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,1,power_law_1.01,0.5249919891357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,4,power_law_1.01,0.12405120134353638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,4,power_law_1.01,0.14679039716720582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.01,0.15931520462036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.01,0.181222403049469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.01,0.21485440731048583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.01,0.2756927967071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.01,0.3414207935333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.01,0.4691455841064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,1,power_law_1.01,0.6644351959228516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.01,0.5848959922790528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.01,0.8585023880004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.01,0.9841407775878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,1,power_law_1.01,0.9500288009643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,1,power_law_1.01,0.08701440095901489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,1,power_law_1.01,0.1094208002090454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,1,power_law_1.01,0.140774405002594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,1,power_law_1.01,0.17854080200195313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,1,power_law_1.01,0.21208319664001465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,1,power_law_1.01,0.2895359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,1,power_law_1.01,0.3107392072677612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,1,power_law_1.01,0.3018111944198608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,1,power_law_1.01,0.31113600730895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,1,power_law_1.01,0.3052864074707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,1,power_law_1.01,0.3216063976287842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,1,power_law_1.01,0.32673919200897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,32,1,power_law_1.2,0.055225598812103274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,32,1,power_law_1.2,0.05857920050621033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,32,1,power_law_1.2,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,32,1,power_law_1.2,0.08435199856758117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,32,1,power_law_1.2,0.10589439868927002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,32,1,power_law_1.2,0.12439680099487305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,1,power_law_1.01,0.3269439935684204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,1,power_law_1.01,0.3392895936965942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,1,power_law_1.01,0.341759991645813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,1,power_law_1.01,0.3520319938659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,1,power_law_1.01,0.37574400901794436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,1,power_law_1.01,0.429369592666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,32,1,power_law_1.2,0.1603968024253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.01,0.47694082260131837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,32,1,power_law_1.2,0.1650431990623474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.01,0.5685056209564209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,32,1,power_law_1.2,0.16815359592437745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,32,1,power_law_1.2,0.17447680234909058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.01,0.6665728092193604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,32,1,power_law_1.2,0.18422399759292601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,32,1,power_law_1.2,0.19134080410003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,32,1,power_law_1.2,0.19952640533447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,32,1,power_law_1.2,0.2075968027114868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,32,1,power_law_1.2,0.21753599643707275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.01,0.8572223663330079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.01,1.050291156768799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.01,1.435923194885254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,16,16,power_law_1.01,0.05374720096588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,32,1,power_law_1.2,0.22690560817718505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,16,16,power_law_1.01,0.058233600854873654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.01,1.8194047927856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,16,16,power_law_1.01,0.055206400156021115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,16,16,power_law_1.01,0.0565887987613678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.01,2.586086463928223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,16,16,power_law_1.01,0.058169597387313844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,16,16,power_law_1.01,0.06172159910202026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,32,1,power_law_1.2,0.27816319465637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.01,3.3462398529052733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,16,16,power_law_1.01,0.0616703987121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,32,1,power_law_1.2,0.3160128116607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,16,16,power_law_1.01,0.061612802743911746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,32,1,power_law_1.2,0.35034880638122556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,16,16,power_law_1.01,0.06285439729690552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,32,1,power_law_1.2,0.4166463851928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,16,16,power_law_1.01,0.06401280164718628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,32,1,power_law_1.2,0.4612800121307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,16,16,power_law_1.01,0.06283519864082336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,32,1,power_law_1.2,0.6044352054595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,16,16,power_law_1.01,0.06760960221290588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,16,16,power_law_1.01,0.067603200674057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,16,16,power_law_1.01,0.06928640007972717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,32,1,power_law_1.2,0.7528128147125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,32,1,power_law_1.2,1.0255040168762206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,16,16,power_law_1.01,0.07895039916038513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,32,1,power_law_1.2,1.3028927803039552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,16,16,power_law_1.01,0.08044800162315369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,16,16,power_law_1.01,0.08435199856758117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,32,1,power_law_1.2,1.8259584426879882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,32,1,power_law_1.2,2.3842687606811523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,16,16,power_law_1.01,0.09514880180358887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,16,16,power_law_1.01,0.10821759700775146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,16,16,power_law_1.01,0.12533119916915894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,16,16,power_law_1.01,0.14380160570144654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,16,16,power_law_1.01,0.18802560567855836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,16,16,power_law_1.01,0.2224639892578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,16,16,power_law_1.01,0.31434879302978513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,16,16,power_law_1.01,0.4276927947998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,1,power_law_1.01,1.2235584259033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,16,16,power_law_1.01,0.6140607833862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,16,16,power_law_1.01,0.8291456222534179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,2,power_law_1.01,0.050860798358917235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,2,power_law_1.01,0.05882880091667175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,2,power_law_1.01,0.0673792004585266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,2,power_law_1.01,0.07598080039024353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,2,power_law_1.01,0.07323520183563233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,2,power_law_1.01,0.07597439885139465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,2,power_law_1.01,0.07209600210189819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,2,power_law_1.01,0.07587839961051941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,2,power_law_1.01,0.07525759935379028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,2,power_law_1.01,0.07263360023498536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,2,power_law_1.01,0.07454079985618592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,2,power_law_1.01,0.07923840284347534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,2,power_law_1.01,0.08119040131568908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,2,power_law_1.01,0.07933440208435058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,2,power_law_1.01,0.08813440203666686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,2,power_law_1.01,0.09402880072593689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,2,power_law_1.01,0.10928640365600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,2,power_law_1.01,0.10901119709014892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.01,0.14358400106430053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.01,0.17384320497512817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.01,0.1853824019432068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.01,0.23683199882507325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.01,0.3084480047225952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.01,0.41561598777770997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.01,0.5517439842224121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,1,balanced,0.03994666785001755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,1,balanced,0.049312000473340355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,1,balanced,0.07502933343251546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,1,balanced,0.12115200360616048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,1,balanced,0.20576000213623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,1,balanced,0.20789867639541626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,1,balanced,0.20779200394948324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,1,balanced,0.2075786590576172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,1,balanced,0.21130132675170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,1,balanced,0.21215999126434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,1,balanced,0.21167999505996704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,1,balanced,0.21619200706481934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,1,balanced,0.21597866217295328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,1,balanced,0.21900800863901773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,1,balanced,0.22645332415898642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,1,balanced,0.22902933756510416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,1,balanced,0.23573867479960123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,4,power_law_1.2,0.08133760094642639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,1,balanced,0.25170133511225384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,4,power_law_1.2,0.09003520011901855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,4,power_law_1.2,0.1004863977432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,1,balanced,0.27432000637054443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,1,balanced,0.3086400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,4,power_law_1.2,0.1296895980834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,1,balanced,0.32947200536727905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,4,power_law_1.2,0.13005440235137938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,1,balanced,0.41923201084136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,4,power_law_1.2,0.1358016014099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,1,balanced,0.47516266504923504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,4,power_law_1.2,0.1403839945793152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,1,balanced,0.6605173349380493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,4,power_law_1.2,0.14447360038757323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,1,balanced,0.7927412986755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,4,power_law_1.2,0.14318079948425294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,1,balanced,1.1302239894866943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,1,balanced,1.4803412755330403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,4,power_law_1.2,0.1553536057472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,4,power_law_1.2,0.16124160289764405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,4,power_law_1.2,0.16951040029525757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,4,power_law_1.2,0.17310080528259278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,4,power_law_1.2,0.16554880142211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,4,power_law_1.2,0.19880319833755494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,4,power_law_1.2,0.20821759700775147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,4,power_law_1.2,0.22160639762878417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,4,power_law_1.2,0.2922624111175537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,4,power_law_1.2,0.3324863910675049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,4,power_law_1.2,0.480620813369751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.01,0.809171199798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,4,power_law_1.2,0.5880703926086426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,4,power_law_1.2,0.7409215927124023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,32,balanced,0.048021331429481506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,32,balanced,0.02741333345572154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,32,balanced,0.029365333418051403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,32,balanced,0.031199999153614044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,32,balanced,0.031514666974544525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,32,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,32,balanced,0.03180799881617228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,32,balanced,0.03320533285538355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,32,balanced,0.03365333378314972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,32,balanced,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,32,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,32,balanced,0.03321066747109095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,32,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,32,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,32,balanced,0.037674665451049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,32,balanced,0.03957333415746689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,32,balanced,0.039861333866914116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,32,balanced,0.04353066782156626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,32,balanced,0.041690667470296226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,32,balanced,0.04971733192602793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,32,balanced,0.047930667797724404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,32,balanced,0.05797333518664042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,32,balanced,0.06525866687297821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,32,balanced,0.09649067123730977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,32,balanced,0.1122773289680481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,32,balanced,0.15054933230082193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,32,balanced,0.18816532691319784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,4,power_law_1.2,0.9463871955871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,2,power_law_1.01,0.05971199870109558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,2,power_law_1.01,0.07125120162963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,2,power_law_1.01,0.07652480006217957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,2,power_law_1.01,0.10930559635162354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,2,power_law_1.01,0.1189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,2,power_law_1.01,0.16135040521621705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,2,power_law_1.01,0.15360000133514404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,2,power_law_1.01,0.16654080152511597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,2,power_law_1.01,0.15864959955215455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,2,power_law_1.01,0.16414079666137696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,2,power_law_1.01,0.17081600427627563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,4,power_law_1.01,0.04663040041923523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,2,power_law_1.01,0.17422720193862914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,4,power_law_1.01,0.04798719882965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,4,power_law_1.2,1.3397439956665038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,2,power_law_1.01,0.17405439615249635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,2,power_law_1.01,0.17788159847259521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,4,power_law_1.01,0.05003520250320435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,2,power_law_1.01,0.1781823992729187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,2,power_law_1.01,0.18392319679260255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,2,power_law_1.01,0.1876479983329773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,2,power_law_1.01,0.2007296085357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,2,power_law_1.01,0.22435200214385986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,2,power_law_1.01,0.2643455982208252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,2,power_law_1.01,0.3018687963485718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.01,1.102284812927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,4,power_law_1.01,0.06167680025100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,4,power_law_1.01,0.06268799901008607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,4,power_law_1.01,0.07175040245056152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,4,power_law_1.01,0.08431360125541687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,4,power_law_1.01,0.0766975998878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,2,power_law_1.01,0.36798079013824464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,4,power_law_1.01,0.07559679746627808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,2,power_law_1.01,0.4461952209472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,4,power_law_1.01,0.07865599989891052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,2,power_law_1.01,0.5977471828460693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,2,power_law_1.01,0.7917376041412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,2,power_law_1.01,1.0445311546325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,2,power_law_1.01,1.358572769165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,4,power_law_1.01,0.08209279775619507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,4,power_law_1.01,0.08519039750099182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,4,power_law_1.01,0.0265855997800827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,4,power_law_1.01,0.08524159789085388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,4,power_law_1.01,0.036774399876594546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,4,power_law_1.01,0.092467200756073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,4,power_law_1.01,0.045023998618125914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,4,power_law_1.01,0.1012992024421692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,4,power_law_1.01,0.05047680139541626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,4,power_law_1.01,0.10772479772567749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,4,power_law_1.01,0.056576001644134524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,4,power_law_1.01,0.10480639934539795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,4,power_law_1.01,0.06499199867248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,4,power_law_1.01,0.11446399688720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,4,power_law_1.01,0.07084800004959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,4,power_law_1.01,0.12343679666519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,4,power_law_1.01,0.06830080151557923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,4,power_law_1.01,0.14101760387420653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,4,power_law_1.01,0.16446080207824706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,4,power_law_1.01,0.20661120414733886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,4,power_law_1.01,0.24992640018463136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,4,power_law_1.01,0.3233920097351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,4,power_law_1.01,0.41539201736450193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,4,power_law_1.01,0.6105792045593261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,4,power_law_1.01,0.07559679746627808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,4,power_law_1.01,0.07360000014305115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,4,power_law_1.01,0.7151167869567872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,4,power_law_1.01,0.07451519966125489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,4,power_law_1.01,0.1078336000442505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,4,power_law_1.01,0.08817279934883118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,4,power_law_1.01,0.08491520285606384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,4,power_law_1.2,1.9672000885009766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,4,power_law_1.01,0.14102400541305543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,4,power_law_1.01,0.14095360040664673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,4,power_law_1.01,0.15690879821777343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,4,power_law_1.01,0.1334272027015686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.01,0.13087999820709229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.01,0.1466048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.01,0.1563007950782776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.01,0.18903039693832396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.01,0.22531840801239014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.01,0.2999295949935913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.01,0.389849591255188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.01,0.540499210357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.01,0.6801599979400634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,4,power_law_1.2,2.743404769897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,4,power_law_1.2,3.5489791870117187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,128,power_law_1.2,0.06817280054092408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,128,power_law_1.2,0.05382400155067444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,128,power_law_1.2,0.05920640230178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,128,power_law_1.2,0.052160000801086424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,128,power_law_1.2,0.0522816002368927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,128,power_law_1.2,0.05292159914970398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,128,power_law_1.2,0.053446400165557864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,128,power_law_1.2,0.05426560044288635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,128,power_law_1.2,0.055078399181365964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,128,power_law_1.2,0.055353599786758426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,128,power_law_1.2,0.05480960011482239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,128,power_law_1.2,0.0572160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,128,power_law_1.2,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,128,power_law_1.2,0.060089600086212155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,128,power_law_1.2,0.06576640009880066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,128,power_law_1.2,0.06605439782142639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,128,power_law_1.2,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,128,power_law_1.2,0.08579199910163879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,128,power_law_1.2,0.09313279986381531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,128,power_law_1.2,0.11654399633407593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,128,power_law_1.2,0.13825279474258423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,128,power_law_1.2,0.17596800327301027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,128,power_law_1.2,0.21719679832458497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,128,power_law_1.2,0.2888767957687378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,128,power_law_1.2,0.3494208097457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,128,power_law_1.2,0.5199039936065674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,128,power_law_1.2,0.6840896129608154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,64,balanced,0.05649599929650625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,64,balanced,0.057946667075157166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,64,balanced,0.057248001297314964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,64,balanced,0.06203199923038483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,64,balanced,0.06609066824118297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,64,balanced,0.09281599521636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,64,balanced,0.08653866251309712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,64,balanced,0.08685866991678874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,64,balanced,0.08678932984670003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,64,balanced,0.08674666285514832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,64,balanced,0.08859733740488689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,64,balanced,0.09065066774686177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,64,balanced,0.0902880032857259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,64,balanced,0.08573333422342937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,64,balanced,0.09706667065620422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,64,balanced,0.09398933251698811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,64,balanced,0.09693333506584167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,64,balanced,0.10531199971834819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,64,balanced,0.10719999670982361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,64,balanced,0.12133866548538208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,64,balanced,0.12997866670290628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,64,balanced,0.15430399775505066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,64,balanced,0.1779359976450602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,64,balanced,0.24235733350118002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,64,balanced,0.2884213328361511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,64,balanced,0.39345065752665204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,64,balanced,0.49941333134969074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,4,balanced,0.04773866633574168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,4,balanced,0.04762666424115499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,4,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,4,balanced,0.04869333406289419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,4,balanced,0.051589335004488625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,4,balanced,0.05448000133037567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,4,balanced,0.0767146646976471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,4,balanced,0.06971199810504913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,4,balanced,0.09218666950861613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,4,balanced,0.08168533444404602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,4,balanced,0.07733866572380066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,4,balanced,0.13515200217564902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,4,balanced,0.13076800107955933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,4,balanced,0.10346133510271709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,4,balanced,0.1527733306090037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,4,balanced,0.1269493301709493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,4,balanced,0.1630506714185079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,4,balanced,0.21423467000325522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,4,balanced,0.27737067143122357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,4,balanced,0.39425067106882733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,4,balanced,0.5226240158081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,4,balanced,0.7469546794891357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,4,balanced,1.0030720233917236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,4,balanced,1.4590346018473308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,4,balanced,1.922309398651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,4,balanced,2.8775412241617837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,4,balanced,3.8105332056681314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,128,balanced,0.05189333359400431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,128,balanced,0.05295466880003611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,128,balanced,0.052655999859174095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,128,balanced,0.06021333237489065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,128,balanced,0.05983999868233999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,128,balanced,0.06017066538333893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,128,balanced,0.0620000014702479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,128,balanced,0.06192000210285187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,128,balanced,0.062133332093556724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,128,balanced,0.0620000014702479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,128,balanced,0.06190933287143707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,128,balanced,0.06493866443634033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,128,balanced,0.06468800206979115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,128,balanced,0.06625600159168243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,128,balanced,0.0703893353541692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,128,balanced,0.07045866549015045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,128,balanced,0.07520000139872234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,128,balanced,0.08482666810353597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,128,balanced,0.0909440020720164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,128,balanced,0.11715733011563619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,128,balanced,0.13205867012341818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,128,balanced,0.18094400564829508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,128,balanced,0.17290133237838745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,128,balanced,0.21979733308156332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,128,balanced,0.24405866861343384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,128,balanced,0.32021866242090863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,128,balanced,0.3859626849492391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,128,balanced,0.060778667529424034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,128,balanced,0.060138667623202004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,128,balanced,0.06226666768391927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,128,balanced,0.06470400094985962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,128,balanced,0.061477333307266235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,128,balanced,0.07258666555086772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,128,balanced,0.07049599786599477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,128,balanced,0.07257066667079926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,128,balanced,0.07871999839941661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,128,balanced,0.07656533519426982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,128,balanced,0.0743999977906545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,128,balanced,0.07858133316040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,128,balanced,0.0828000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,128,balanced,0.08100800216197968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,128,balanced,0.08734933535257976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,128,balanced,0.09321066737174988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,128,balanced,0.0913866659005483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,128,balanced,0.10309333602587382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,128,balanced,0.11362133423487346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,128,balanced,0.12917866309483847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,128,balanced,0.1457866628964742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,128,balanced,0.1835093299547831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,128,balanced,0.21740265687306723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,128,balanced,0.2914026578267415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,128,balanced,0.3609120051066081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,128,balanced,0.5072906812032064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,128,balanced,0.6470453341801962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,16,power_law_1.01,0.05711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,16,power_law_1.01,0.06247040033340454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,16,power_law_1.01,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,16,power_law_1.01,0.06069759726524353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,16,power_law_1.01,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,16,power_law_1.01,0.0586624026298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,16,power_law_1.01,0.06231679916381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,16,power_law_1.01,0.06220800280570984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,16,power_law_1.01,0.06244480013847351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,16,power_law_1.01,0.06297600269317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,16,power_law_1.01,0.06368640065193176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,16,power_law_1.01,0.0654591977596283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,16,power_law_1.01,0.066457599401474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,16,power_law_1.01,0.06659839749336242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,16,power_law_1.01,0.07174400091171265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,16,power_law_1.01,0.07129600048065185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,16,power_law_1.01,0.074508798122406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,16,power_law_1.01,0.08297600150108338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,16,power_law_1.01,0.09007999897003174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,16,power_law_1.01,0.10555520057678222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,16,power_law_1.01,0.118886399269104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,16,power_law_1.01,0.14970879554748534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,16,power_law_1.01,0.18102400302886962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,16,power_law_1.01,0.2415231943130493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,16,power_law_1.01,0.29040639400482177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,16,power_law_1.01,0.42531838417053225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,16,power_law_1.01,0.5475071907043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,4,balanced,0.06067200005054474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,4,balanced,0.0603413333495458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,4,balanced,0.07628266513347626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,4,balanced,0.08316799998283386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,4,balanced,0.07854933540026347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,4,balanced,0.07640533149242401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,4,balanced,0.0849120020866394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,4,balanced,0.07874133189519246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,4,balanced,0.07940266529719035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,4,balanced,0.08260799944400787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,4,balanced,0.07255466779073079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,4,balanced,0.07447466750939687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,4,balanced,0.0746666689713796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,4,balanced,0.07225066423416138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,4,balanced,0.0729066679875056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,4,balanced,0.07416533430417378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,4,balanced,0.07357333103815715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,4,balanced,0.10503466924031575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,4,balanced,0.09476799766222636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,4,balanced,0.11956800023714702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,4,balanced,0.13844799995422363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,4,balanced,0.1834239959716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,4,balanced,0.22447999318440756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,4,balanced,0.31121599674224854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,4,balanced,0.39405866463979083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,4,balanced,0.5625973145167033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,4,balanced,0.7377173105875651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,8,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,8,balanced,0.03668266783157984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,8,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,8,balanced,0.04166933397452036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,8,balanced,0.04340266684691111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,8,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,8,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,8,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,8,balanced,0.047685335079828896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,8,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,8,balanced,0.05115733544031779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,8,balanced,0.05009600023428599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,8,balanced,0.049957334995269775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,8,balanced,0.05188799897829691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,8,balanced,0.055871998270352684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,8,balanced,0.05725333094596863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,8,balanced,0.05874133110046387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,8,balanced,0.06633066634337108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,8,balanced,0.07189866900444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,8,balanced,0.08239999910195668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,8,balanced,0.0906880001227061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,8,balanced,0.1176479955514272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,8,balanced,0.14470932881037393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,8,balanced,0.20151466131210327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,8,balanced,0.27724266052246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,8,balanced,0.40351998805999756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,8,balanced,0.4850826660792033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,2,balanced,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,2,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,2,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,2,balanced,0.049728001157442726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,2,balanced,0.07216000060240428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,2,balanced,0.08170666793982188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,2,balanced,0.08055999875068665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,2,balanced,0.0803466687599818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,2,balanced,0.0803306649128596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,2,balanced,0.08021333316961925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,2,balanced,0.0810346653064092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,2,balanced,0.08276799817879994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,2,balanced,0.0820000022649765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,2,balanced,0.08275733391443889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,2,balanced,0.08659733335177104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,2,balanced,0.0883840024471283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,2,balanced,0.09293867150942485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,2,balanced,0.10847466190656026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,2,balanced,0.10621866583824158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,2,balanced,0.14356266458829245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,2,balanced,0.13945066928863525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,2,balanced,0.18237332503000894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,2,balanced,0.19140799840291342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,2,balanced,0.2711679935455322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,2,balanced,0.30640532573064166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,2,balanced,0.4346933364868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,2,balanced,0.5262720187505087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,16,power_law_1.01,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,16,power_law_1.01,0.0591871976852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,16,power_law_1.01,0.04917759895324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,16,power_law_1.01,0.05178239941596985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,16,power_law_1.01,0.052108800411224364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,16,power_law_1.01,0.05414400100708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,16,power_law_1.01,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,16,power_law_1.01,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,16,power_law_1.01,0.05655679702758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,16,power_law_1.01,0.05551360249519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,16,power_law_1.01,0.05878400206565857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,16,power_law_1.01,0.057606399059295654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,16,power_law_1.01,0.06044800281524658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,16,power_law_1.01,0.06360960006713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,16,power_law_1.01,0.06789759993553161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,16,power_law_1.01,0.0719488024711609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,16,power_law_1.01,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,16,power_law_1.01,0.08628479838371277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,16,power_law_1.01,0.08942080140113831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,16,power_law_1.01,0.10858240127563476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,16,power_law_1.01,0.1263167977333069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,16,power_law_1.01,0.16197119951248168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,16,power_law_1.01,0.18698879480361938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,16,power_law_1.01,0.256441593170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,16,power_law_1.01,0.3178368091583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,16,power_law_1.01,0.44341120719909666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,16,power_law_1.01,0.5672575950622558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,32,8,power_law_1.01,0.01738879978656769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,32,8,power_law_1.01,0.017215999960899352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,32,8,power_law_1.01,0.017183999717235564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,32,8,power_law_1.01,0.017030400037765504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,32,8,power_law_1.01,0.01713919937610626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,32,8,power_law_1.01,0.017606399953365326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,32,8,power_law_1.01,0.01796479970216751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,32,8,power_law_1.01,0.020339199900627138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,32,8,power_law_1.01,0.021254399418830873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,32,8,power_law_1.01,0.020691199600696562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,32,8,power_law_1.01,0.02945919930934906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,32,8,power_law_1.01,0.030144000053405763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,32,8,power_law_1.01,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,32,8,power_law_1.01,0.03023360073566437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,32,8,power_law_1.01,0.03091199994087219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,32,8,power_law_1.01,0.03108479976654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,32,8,power_law_1.01,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,32,8,power_law_1.01,0.03640959858894348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,32,8,power_law_1.01,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,32,8,power_law_1.01,0.046214398741722104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,32,8,power_law_1.01,0.05006080269813538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,32,8,power_law_1.01,0.06931840181350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,32,8,power_law_1.01,0.07897599935531616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,32,8,power_law_1.01,0.10695040225982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,32,8,power_law_1.01,0.13418240547180177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,32,power_law_1.01,0.01825920045375824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,32,power_law_1.01,0.018105599284172057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,32,power_law_1.01,0.01777919977903366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,32,power_law_1.01,0.018624000251293182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,32,power_law_1.01,0.019859200716018675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,32,power_law_1.01,0.022124800086021423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,32,power_law_1.01,0.03966079950332642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,32,power_law_1.01,0.03943679928779602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,32,power_law_1.01,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,32,power_law_1.01,0.03343360126018524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,32,power_law_1.01,0.033452799916267394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,32,power_law_1.01,0.03311359882354736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,32,power_law_1.01,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,32,power_law_1.01,0.03208959996700287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,32,power_law_1.01,0.033004799485206605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,32,power_law_1.01,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,32,power_law_1.01,0.03787519931793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,32,power_law_1.01,0.04072319865226746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.01,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.01,0.05056639909744263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.01,0.05619840025901794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.01,0.07597439885139465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.01,0.08934400081634522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.01,0.11224960088729859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.01,0.14162559509277345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,32,8,power_law_1.01,0.18824959993362428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.01,0.20465919971466065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.01,0.2715008020401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,4,power_law_1.01,0.025094398856163026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,4,power_law_1.01,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,4,power_law_1.01,0.043942400813102724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,4,power_law_1.01,0.05754240155220032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,4,power_law_1.01,0.06892799735069274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,4,power_law_1.01,0.06821759939193725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,4,power_law_1.01,0.08019199967384338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,4,power_law_1.01,0.08622080087661743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,4,power_law_1.01,0.0866047978401184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,4,power_law_1.01,0.08335360288619995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,4,power_law_1.01,0.08834559917449951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,4,power_law_1.01,0.10696959495544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,4,power_law_1.01,0.09359359741210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,4,power_law_1.01,0.0964352011680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,4,power_law_1.01,0.12740479707717894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,4,power_law_1.01,0.11274880170822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,4,power_law_1.01,0.10856319665908813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,4,power_law_1.01,0.19549440145492553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,4,power_law_1.01,0.21842560768127442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,4,power_law_1.01,0.19896960258483887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,4,power_law_1.01,0.1978943943977356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,4,power_law_1.01,0.19560960531234742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,4,power_law_1.01,0.2196671962738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,4,power_law_1.01,0.2671488046646118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,4,power_law_1.01,0.34168961048126223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,4,power_law_1.01,0.46564478874206544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,4,power_law_1.01,0.6134912014007569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,32,8,power_law_1.01,0.23561599254608154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,2,power_law_1.2,0.04832639992237091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,2,power_law_1.2,0.05242239832878113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,2,power_law_1.2,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,2,power_law_1.2,0.07313920259475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,2,power_law_1.2,0.08565120100975036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,2,power_law_1.2,0.09936000108718872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,2,power_law_1.2,0.12341120243072509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,2,power_law_1.2,0.12476160526275634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,2,power_law_1.2,0.1322816014289856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,2,power_law_1.2,0.12879999876022338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,2,power_law_1.2,0.13827840089797974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,2,power_law_1.2,0.14629759788513183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,2,power_law_1.2,0.15063680410385133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,2,power_law_1.2,0.15962239503860473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,2,power_law_1.2,0.1757375955581665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,2,power_law_1.2,0.18347519636154175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,2,power_law_1.2,0.1847808003425598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,2,power_law_1.2,0.21619839668273927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,2,power_law_1.2,0.24438400268554689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,2,power_law_1.2,0.3027071952819824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,2,power_law_1.2,0.3622591972351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,2,power_law_1.2,0.475110387802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,2,power_law_1.2,0.5863679885864258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,2,power_law_1.2,0.8004927635192871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,2,power_law_1.2,1.073580837249756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,2,power_law_1.2,1.4660672187805175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,2,power_law_1.2,1.9205440521240233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,1,balanced,0.05285866558551788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,1,balanced,0.06436266501744588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,1,balanced,0.09572266538937886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,1,balanced,0.0997973382472992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,1,balanced,0.10142933328946431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,1,balanced,0.10075733065605164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,1,balanced,0.10151466727256775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,1,balanced,0.10348266363143921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,1,balanced,0.10116266210873921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,1,balanced,0.10353599985440572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,1,balanced,0.10269866387049358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,1,balanced,0.10626666744550069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,1,balanced,0.10620799660682678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,1,balanced,0.10828266541163127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,1,balanced,0.11558933059374492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,1,balanced,0.1190880040327708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,1,balanced,0.11946666240692139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,1,balanced,0.15495999654134116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,1,balanced,0.164901336034139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,1,balanced,0.2204373280207316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,1,balanced,0.26313066482543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,1,balanced,0.3731573422749837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,1,balanced,0.475328008333842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,1,balanced,0.6814186573028564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,1,balanced,0.8874666690826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,1,balanced,1.279520034790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,1,balanced,1.6875306765238445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,16,balanced,0.04367466767628988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,16,balanced,0.046069333950678505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,16,balanced,0.04391466577847799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,16,balanced,0.04905066887537638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,16,balanced,0.07627733548482259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,16,balanced,0.08306133250395457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,16,balanced,0.08241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,16,balanced,0.08404800295829773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,16,balanced,0.08282666901747386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,16,balanced,0.08281599978605907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,16,balanced,0.08390933275222778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,16,balanced,0.08586666981379192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,16,balanced,0.08431466420491536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,16,balanced,0.08689066767692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,16,balanced,0.09195199608802795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,16,balanced,0.0922933320204417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,16,balanced,0.09474666913350423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,16,balanced,0.10139733552932739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,16,balanced,0.10357866684595744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,16,balanced,0.12293866276741028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,16,balanced,0.12743999560674033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,16,balanced,0.15923200050989786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,16,balanced,0.1694613297780355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,16,balanced,0.23666133483250937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,16,balanced,0.2712106704711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,16,balanced,0.35252801577250165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,16,balanced,0.4368373155593872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,8,balanced,0.08227733274300893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,8,balanced,0.08196266492207845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,8,balanced,0.08067200084527333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,8,balanced,0.08355200290679932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,8,balanced,0.08653866251309712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,8,balanced,0.08661866188049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,8,balanced,0.0849226713180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,8,balanced,0.08675199747085571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,8,balanced,0.08784533540407817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,8,balanced,0.0879146655400594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,8,balanced,0.0909493366877238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,8,balanced,0.09050666292508443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,8,balanced,0.09289600451787312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,8,balanced,0.09858666857083638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,8,balanced,0.10157333811124165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,8,balanced,0.10493333141009013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,8,balanced,0.11000532905260722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,8,balanced,0.13452800114949545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,8,balanced,0.13502933581670126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,8,balanced,0.1813066601753235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,8,balanced,0.1998080015182495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,8,balanced,0.25358933210372925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,8,balanced,0.31086933612823486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,8,balanced,0.43668266137441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,8,balanced,0.5521173477172852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,8,balanced,0.7891146341959635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,8,balanced,1.0492533047993977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,4,power_law_1.01,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,4,power_law_1.01,0.055225598812103274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,4,power_law_1.01,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,4,power_law_1.01,0.0692799985408783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,4,power_law_1.01,0.0818943977355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,4,power_law_1.01,0.09144960045814514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,4,power_law_1.01,0.09472640156745911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,4,power_law_1.01,0.10710400342941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,4,power_law_1.01,0.1047935962677002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,4,power_law_1.01,0.1124351978302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,4,power_law_1.01,0.11062400341033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,4,power_law_1.01,0.11228159666061402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,4,power_law_1.01,0.11125119924545288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,4,power_law_1.01,0.1206015944480896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,4,power_law_1.01,0.12842880487442015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,4,power_law_1.01,0.12495360374450684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,4,power_law_1.01,0.139136004447937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,4,power_law_1.01,0.15365760326385497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,4,power_law_1.01,0.17066240310668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,4,power_law_1.01,0.21111040115356444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,4,power_law_1.01,0.23625600337982178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,4,power_law_1.01,0.30039680004119873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,4,power_law_1.01,0.3805311918258667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,4,power_law_1.01,0.5069375991821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,4,power_law_1.01,0.667577600479126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,2,power_law_1.01,0.04692479968070984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,2,power_law_1.01,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,2,power_law_1.01,0.045414400100708005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,4,power_law_1.01,0.9397695541381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,2,power_law_1.01,0.04774399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,2,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,2,power_law_1.01,0.05403519868850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,2,power_law_1.01,0.0710591971874237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,2,power_law_1.01,0.0674560010433197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,2,power_law_1.01,0.08862720131874084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,2,power_law_1.01,0.08006399869918823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,2,power_law_1.01,0.0760703980922699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,2,power_law_1.01,0.1362496018409729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,2,power_law_1.01,0.1280959963798523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,4,power_law_1.01,1.2682623863220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,2,power_law_1.01,0.10088319778442383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,2,power_law_1.01,0.1460096001625061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,2,power_law_1.01,0.12539520263671874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,2,power_law_1.01,0.1610751986503601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,2,power_law_1.01,0.21098880767822265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.01,0.27418880462646483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.01,0.3891200065612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.01,0.5213376045227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.01,0.7439231872558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.01,0.9882368087768555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.01,1.4545599937438964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.01,1.9162111282348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.01,2.873708724975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,8,balanced,0.025648000339667004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,8,balanced,0.02311466634273529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,8,balanced,0.02311466634273529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,8,balanced,0.025173333783944447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,8,balanced,0.02722666660944621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,8,balanced,0.027488000690937042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.01,3.7999935150146484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,8,balanced,0.02921066681543986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,8,balanced,0.029792000850041706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,8,balanced,0.02826133370399475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,8,balanced,0.029706666866938274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,8,balanced,0.029893333713213604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,8,balanced,0.03165333221356074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,8,balanced,0.030261332790056866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,8,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,8,balanced,0.03718400001525879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,8,balanced,0.035775999228159584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,8,balanced,0.03986666599909464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,8,balanced,0.04162666698296865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,8,balanced,0.045893331368764244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,8,balanced,0.049045334259668984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,8,balanced,0.05381333331267039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,8,balanced,0.061893333991368614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,8,balanced,0.07461866736412048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,8,balanced,0.09860799709955852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,8,balanced,0.12339199582735698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,8,balanced,0.15618667006492615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,8,balanced,0.19126399358113608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,4,balanced,0.055957332253456116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,4,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,4,balanced,0.0603359987338384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,4,balanced,0.07016533116499583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,4,balanced,0.08488000432650249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,4,balanced,0.1256160040696462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,4,balanced,0.1572426656881968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,4,balanced,0.1601599951585134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,4,balanced,0.16154133280118307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,4,balanced,0.16241600116093954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,4,balanced,0.15620799859364828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,4,balanced,0.15958933035532633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,4,balanced,0.16288533806800842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,4,balanced,0.16125333309173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,4,balanced,0.17146666844685873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,4,balanced,0.17218667268753052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,4,balanced,0.18056533734003702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,4,balanced,0.1956640084584554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,4,balanced,0.2096853256225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,4,balanced,0.23837333917617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,4,balanced,0.26502933104832965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,4,balanced,0.32620267073313397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,4,balanced,0.38702932993570965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,4,balanced,0.5153173208236694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,4,balanced,0.6891573270161947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,4,balanced,0.9338080088297526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,4,balanced,1.2437653541564941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,8,power_law_1.01,0.08963840007781983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,8,power_law_1.01,0.10309760570526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,8,power_law_1.01,0.1449023962020874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,8,power_law_1.01,0.14295040369033812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,8,power_law_1.01,0.14565119743347169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,8,power_law_1.01,0.16273280382156372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,8,power_law_1.01,0.15890560150146485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,8,power_law_1.01,0.16516480445861817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,8,power_law_1.01,0.16442879438400268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,8,power_law_1.01,0.15612159967422484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,8,power_law_1.01,0.16240639686584474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,8,power_law_1.01,0.16647679805755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,8,power_law_1.01,0.16312320232391359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,8,power_law_1.01,0.16488959789276122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,8,power_law_1.01,0.18142080307006836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,8,power_law_1.01,0.17505919933319092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,8,power_law_1.01,0.18396159410476684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,8,power_law_1.01,0.19201279878616334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,8,power_law_1.01,0.20940160751342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,8,power_law_1.01,0.2529727935791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,8,power_law_1.01,0.28132479190826415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,8,power_law_1.01,0.366048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,8,power_law_1.01,0.44463357925415037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,8,power_law_1.01,0.5725503921508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,128,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,8,power_law_1.01,0.7537024021148682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,128,power_law_1.2,0.053324800729751584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,8,power_law_1.01,1.0871808052062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,128,power_law_1.2,0.05255039930343628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,128,power_law_1.2,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,8,power_law_1.01,1.446553611755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,128,power_law_1.2,0.05752320289611816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,128,power_law_1.2,0.05928320288658142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,128,power_law_1.2,0.059648001194000246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,128,power_law_1.2,0.06099839806556702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,128,power_law_1.2,0.061312001943588254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,128,power_law_1.2,0.06296319961547851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,128,power_law_1.2,0.06755200028419495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,128,power_law_1.2,0.07427840232849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,128,power_law_1.2,0.07653759717941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,128,power_law_1.2,0.0705024003982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,4,power_law_1.2,0.018163199722766876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,128,power_law_1.2,0.08275200128555298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,4,power_law_1.2,0.019020800292491914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,4,power_law_1.2,0.020000000298023225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,128,power_law_1.2,0.08773760199546814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,4,power_law_1.2,0.022284799814224245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,128,power_law_1.2,0.09017599821090698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,4,power_law_1.2,0.023929600417613984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,128,power_law_1.2,0.10780160427093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,4,power_law_1.2,0.024249599874019624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,128,power_law_1.2,0.11919360160827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,128,power_law_1.2,0.14614399671554565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,128,power_law_1.2,0.16927360296249389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,128,power_law_1.2,0.22286720275878907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,4,power_law_1.2,0.027167999744415285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,128,power_law_1.2,0.2703039884567261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,4,power_law_1.2,0.027968001365661622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,4,power_law_1.2,0.028889599442481994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,128,power_law_1.2,0.423033618927002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,4,power_law_1.2,0.030048000812530517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,4,power_law_1.2,0.031641599535942075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,128,power_law_1.2,0.5191936016082763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,4,power_law_1.2,0.03392640054225922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,4,power_law_1.2,0.03371520042419433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,4,power_law_1.2,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,128,power_law_1.2,0.8220735549926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,4,power_law_1.2,0.03763200044631958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,4,power_law_1.2,0.039059200882911684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,4,power_law_1.2,0.04653440117835998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,4,power_law_1.2,0.050918400287628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,4,power_law_1.2,0.05493119955062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,4,power_law_1.2,0.06791039705276489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,4,power_law_1.2,0.08691200017929077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,4,power_law_1.2,0.09768959879875183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,4,power_law_1.2,0.1106495976448059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,128,power_law_1.2,1.1208383560180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,4,power_law_1.2,0.13694080114364623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,4,power_law_1.2,0.166048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,4,power_law_1.2,0.22796800136566162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,4,power_law_1.2,0.2800704002380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,8,balanced,0.041946664452552795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,8,balanced,0.043434664607048035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,8,balanced,0.043663998444875084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,8,balanced,0.05002133548259735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,8,balanced,0.0533493310213089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,8,balanced,0.07682133217652638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,8,balanced,0.07872533301512401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,8,balanced,0.07999999821186066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,8,balanced,0.08041066428025563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,8,balanced,0.08176533381144206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,8,balanced,0.08244266609350841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,8,balanced,0.0848533312479655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,8,balanced,0.08676266670227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,8,balanced,0.08718933661778767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,8,balanced,0.09607999523480733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,8,balanced,0.09703466296195984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,8,balanced,0.10173333684603374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,8,balanced,0.11486933628718059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,8,balanced,0.12416533629099528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,8,balanced,0.14429333806037903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,8,balanced,0.16341867049535116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,8,balanced,0.21516267458597818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,8,balanced,0.25498666365941364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,8,balanced,0.34629865487416583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,8,balanced,0.4472320079803467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,8,balanced,0.6226666768391927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,8,balanced,0.8068319956461588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,2,balanced,0.055973331133524575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,2,balanced,0.0670826683441798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,2,balanced,0.09321600198745728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,2,balanced,0.09742400050163269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,2,balanced,0.09879466891288757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,2,balanced,0.09846400221188863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,2,balanced,0.09915199875831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,2,balanced,0.1009386678536733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,2,balanced,0.1002400020758311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,2,balanced,0.10035733381907146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,2,balanced,0.09915733337402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,2,balanced,0.10301333665847778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,2,balanced,0.10214400291442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,2,balanced,0.103493332862854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,2,balanced,0.10974933703740437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,2,balanced,0.11319466431935628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,2,balanced,0.11308800180753072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,2,balanced,0.16031466921170553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,2,balanced,0.15728533267974854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,2,balanced,0.2015786568323771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,2,balanced,0.2479520042737325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,2,balanced,0.34068799018859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,2,balanced,0.420629342397054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,2,balanced,0.593946655591329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,2,balanced,0.7668000062306722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,2,balanced,1.1365066369374592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,2,balanced,1.4907840092976887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,8,power_law_1.2,0.04501760005950928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,8,power_law_1.2,0.06822400093078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,8,power_law_1.2,0.07331200242042542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,8,power_law_1.2,0.07902719974517822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,8,power_law_1.2,0.10533119440078735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,8,power_law_1.2,0.08739839792251587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,8,power_law_1.2,0.10177279710769653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,8,power_law_1.2,0.11766400337219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,8,power_law_1.2,0.12174079418182374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,8,power_law_1.2,0.12088320255279542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,8,power_law_1.2,0.11290880441665649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,8,power_law_1.2,0.11479040384292602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,8,power_law_1.2,0.11924480199813843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,8,power_law_1.2,0.13243520259857178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,8,power_law_1.2,0.1373952031135559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,8,power_law_1.2,0.14970240592956544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,8,power_law_1.2,0.14944640398025513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,8,power_law_1.2,0.16429439783096314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,8,power_law_1.2,0.32907519340515134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,8,power_law_1.2,0.4179520130157471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,8,power_law_1.2,0.2844928026199341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,8,power_law_1.2,0.2945791959762573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,8,power_law_1.2,0.3540287971496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,8,power_law_1.2,0.4739583969116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,8,power_law_1.2,0.5813504219055176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,8,power_law_1.2,0.7986495971679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,8,power_law_1.2,1.0187071800231933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,16,power_law_1.2,0.1589184045791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,16,power_law_1.2,0.23239679336547853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,16,power_law_1.2,0.22663040161132814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,16,power_law_1.2,0.23935999870300292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,16,power_law_1.2,0.24309759140014647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,16,power_law_1.2,0.19950720071792602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,16,power_law_1.2,0.20188798904418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,16,power_law_1.2,0.21882240772247313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,16,power_law_1.2,0.21368319988250734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,16,power_law_1.2,0.2066431999206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,16,power_law_1.2,0.22188160419464112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,16,power_law_1.2,0.22157440185546876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,16,power_law_1.2,0.22532479763031005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,16,power_law_1.2,0.25043840408325196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,16,power_law_1.2,0.25125761032104493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,16,power_law_1.2,0.26095359325408934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,16,power_law_1.2,0.2679935932159424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,16,power_law_1.2,0.3241919994354248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,1,power_law_1.01,0.10245120525360107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,16,power_law_1.2,0.3339263916015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,1,power_law_1.01,0.15383039712905883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,16,power_law_1.2,0.4158783912658691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,1,power_law_1.01,0.24272000789642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,16,power_law_1.2,0.37528319358825685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,1,power_law_1.01,0.44220800399780275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,16,power_law_1.2,0.5024703979492188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,1,power_law_1.01,0.6454912185668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,16,power_law_1.2,0.5766848087310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,1,power_law_1.01,0.8493375778198242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,16,power_law_1.2,0.7544960021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,1,power_law_1.01,1.1785408020019532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,1,power_law_1.01,1.2286399841308593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,1,power_law_1.01,1.28853759765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,1,power_law_1.01,1.331174373626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,1,power_law_1.01,1.3541312217712402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,1,power_law_1.01,1.4052672386169434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,16,power_law_1.2,0.8780096054077149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,1,power_law_1.01,1.4325311660766602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,16,power_law_1.2,1.2940352439880372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,1,power_law_1.01,1.4596735954284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,16,power_law_1.2,1.8189311981201173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,1,power_law_1.01,1.4873536109924317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,1,power_law_1.01,1.5454784393310548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,1,power_law_1.01,1.607417678833008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,1,power_law_1.01,1.7678207397460937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,1,power_law_1.01,1.723321533203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,1,power_law_1.01,1.9235008239746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,1,power_law_1.01,2.0222591400146483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,1,power_law_1.01,2.326252746582031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,1,power_law_1.01,2.5593727111816404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,1,power_law_1.01,3.1261695861816405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,1,power_law_1.01,3.6004096984863283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,1,power_law_1.01,4.743027114868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,1,power_law_1.01,5.950956726074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,2,power_law_1.2,0.03150720000267029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,2,power_law_1.2,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,2,power_law_1.2,0.03883520066738129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,2,power_law_1.2,0.048870399594306946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,2,power_law_1.2,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,2,power_law_1.2,0.06944000124931335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,2,power_law_1.2,0.07306879758834839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,2,power_law_1.2,0.0730239987373352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,2,power_law_1.2,0.07352960109710693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,2,power_law_1.2,0.0733568012714386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,2,power_law_1.2,0.077920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,2,power_law_1.2,0.07831680178642272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,2,power_law_1.2,0.08083199858665466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,2,power_law_1.2,0.08520960211753845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,2,power_law_1.2,0.09543039798736572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,2,power_law_1.2,0.0991424024105072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,2,power_law_1.2,0.10327680110931396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,2,power_law_1.2,0.11265920400619507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,2,power_law_1.2,0.12405120134353638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,2,power_law_1.2,0.1442944049835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,2,power_law_1.2,0.15614080429077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,2,power_law_1.2,0.19185279607772826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,2,power_law_1.2,0.2202752113342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,2,power_law_1.2,0.28499200344085696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,2,power_law_1.2,0.3491199970245361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,2,power_law_1.2,0.514521598815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,2,power_law_1.01,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,2,power_law_1.01,0.0582144021987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,2,power_law_1.01,0.08005120158195496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,2,power_law_1.01,0.09415680170059204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,2,power_law_1.01,0.12314879894256592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,2,power_law_1.01,0.14277119636535646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,2,power_law_1.01,0.15718400478363037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,2,power_law_1.01,0.18744959831237792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,2,power_law_1.01,0.197324800491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,2,power_law_1.01,0.19738880395889283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,2,power_law_1.01,0.2085439920425415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,2,power_law_1.01,0.20896000862121583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,2,power_law_1.01,0.21729280948638915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,2,power_law_1.01,0.22325758934020995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,2,power_law_1.01,0.23772799968719482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,2,power_law_1.01,0.237939190864563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,2,power_law_1.01,0.25394558906555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,2,power_law_1.01,0.2881727933883667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,2,power_law_1.01,0.3101567983627319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,2,power_law_1.01,0.37271039485931395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,2,power_law_1.01,0.44401278495788576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,2,power_law_1.01,0.5609663963317871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,2,power_law_1.01,0.6412352085113525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,2,power_law_1.01,0.8595392227172851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,2,power_law_1.01,1.122758388519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,2,power_law_1.01,1.501523208618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,2,power_law_1.01,1.88787841796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,2,power_law_1.2,0.6308159828186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,16,power_law_1.01,0.04688639938831329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,16,power_law_1.01,0.052537602186203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,16,power_law_1.01,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,16,power_law_1.01,0.04460160136222839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,16,power_law_1.01,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,16,power_law_1.01,0.04380159974098206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,16,power_law_1.01,0.04593920111656189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,16,power_law_1.01,0.04615040123462677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,16,power_law_1.01,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,16,power_law_1.01,0.04657280147075653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,16,power_law_1.01,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,16,power_law_1.01,0.04835839867591858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,16,power_law_1.01,0.049395200610160825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,16,power_law_1.01,0.04959999918937683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,16,power_law_1.01,0.054739201068878175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,16,power_law_1.01,0.05552639961242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,16,power_law_1.01,0.058310401439666745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,16,power_law_1.01,0.0680895984172821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,16,power_law_1.01,0.07158399820327759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,16,power_law_1.01,0.08790400028228759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,16,power_law_1.01,0.09207040071487427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,16,power_law_1.01,0.118886399269104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,16,power_law_1.01,0.15210880041122438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,16,power_law_1.01,0.2004863977432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,16,power_law_1.01,0.23374719619750978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,16,power_law_1.01,0.3414720058441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,16,power_law_1.01,0.413862419128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,1,power_law_1.2,0.08811519742012024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,1,power_law_1.2,0.1248128056526184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,1,power_law_1.2,0.1448639988899231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,1,power_law_1.2,0.2389888048171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,1,power_law_1.2,0.24083199501037597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,1,power_law_1.2,0.2392575979232788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,1,power_law_1.2,0.24910080432891846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,1,power_law_1.2,0.24992640018463136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,1,power_law_1.2,0.25259521007537844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,1,power_law_1.2,0.26403839588165284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,1,power_law_1.2,0.2639039993286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,1,power_law_1.2,0.2775871992111206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,1,power_law_1.2,0.2874367952346802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,1,power_law_1.2,0.3063807964324951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,1,power_law_1.2,0.32797439098358155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,1,power_law_1.2,0.33854079246520996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,8,balanced,0.023434666295846302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,8,balanced,0.023226665953795116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,8,balanced,0.021402666966120403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,8,balanced,0.021418665846188862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,8,balanced,0.023584000766277313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,8,balanced,0.0262773334980011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,8,balanced,0.034703999757766724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,8,balanced,0.029872000217437744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,8,balanced,0.03902400036652883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,8,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,8,balanced,0.03792533278465271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,8,balanced,0.05787733197212219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,8,balanced,0.059343998630841575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,8,balanced,0.04747733473777771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,8,balanced,0.06612800061702728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,8,balanced,0.05620799958705902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,8,balanced,0.0639466643333435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,8,balanced,0.08521599570910136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,8,balanced,0.1009386678536733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,8,balanced,0.1404213309288025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,8,balanced,0.17577066024144491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,8,balanced,0.24361066023508707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,8,balanced,0.303056001663208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,8,balanced,0.4384853442509969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,8,balanced,0.5669973293940226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,8,balanced,0.8321173191070557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,8,balanced,1.0943040053049724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,1,power_law_1.2,0.3628096103668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,1,power_law_1.2,0.41291518211364747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,1,power_law_1.2,0.4766143798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,1,power_law_1.2,0.5924928188323975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,1,power_law_1.2,0.6914688110351562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,1,power_law_1.2,0.954201602935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,32,power_law_1.2,0.044409599900245664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,32,power_law_1.2,0.04416640102863312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,32,power_law_1.2,0.04402559995651245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,32,power_law_1.2,0.04639999866485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,32,power_law_1.2,0.0470335990190506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,32,power_law_1.2,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,32,power_law_1.2,0.0487744003534317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,32,power_law_1.2,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,32,power_law_1.2,0.04963839948177338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,32,power_law_1.2,0.05023999810218811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,32,power_law_1.2,0.05237119793891907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,32,power_law_1.2,0.05316479802131653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,32,power_law_1.2,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,1,power_law_1.2,1.1929535865783691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,32,power_law_1.2,0.05671679973602295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,32,power_law_1.2,0.06416640281677247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,32,power_law_1.2,0.06664320230484008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,32,power_law_1.2,0.07073280215263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,32,power_law_1.2,0.08247039914131164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,32,power_law_1.2,0.08921599984169007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,32,power_law_1.2,0.11779839992523193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,32,power_law_1.2,0.14274560213088988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,32,power_law_1.2,0.18944000005722045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,32,power_law_1.2,0.23370881080627443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,32,power_law_1.2,0.3258944034576416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,32,power_law_1.2,0.4215871810913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,32,power_law_1.2,0.6296192169189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,32,power_law_1.2,0.7956352233886719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,8,power_law_1.2,0.039238399267196654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,8,power_law_1.2,0.039052799344062805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,8,power_law_1.2,0.03958399891853333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,8,power_law_1.2,0.04095999896526337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,8,power_law_1.2,0.041510400176048276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,8,power_law_1.2,0.041484799981117246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,8,power_law_1.2,0.04110080003738403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,8,power_law_1.2,0.040659201145172116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,8,power_law_1.2,0.039987200498580934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,8,power_law_1.2,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,8,power_law_1.2,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,8,power_law_1.2,0.043987199664115906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,8,power_law_1.2,0.04273920059204102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,8,power_law_1.2,0.043910399079322815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,8,power_law_1.2,0.04979200065135956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,8,power_law_1.2,0.050483202934265135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,8,power_law_1.2,0.05319679975509643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,8,power_law_1.2,0.06322559714317322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.2,0.06887680292129517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,1,power_law_1.2,1.7023168563842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.2,0.08539519906044006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.2,0.1022528052330017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.2,0.13023359775543214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,1,power_law_1.2,2.1742015838623048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.2,0.15569280385971068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,1,power_law_1.2,3.1519744873046873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.2,0.22639360427856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.2,0.2883392095565796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,1,power_law_1.2,4.146579360961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,4,power_law_1.01,0.1211583971977234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,4,power_law_1.01,0.1889855980873108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,4,power_law_1.01,0.18833919763565063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,4,power_law_1.01,0.24591999053955077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,4,power_law_1.01,0.3051775932312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,4,power_law_1.01,0.41594882011413575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,4,power_law_1.01,0.40064640045166017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,4,power_law_1.01,0.40926079750061034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.2,0.40650238990783694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,4,power_law_1.01,0.3990015983581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,4,power_law_1.01,0.40508160591125486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,4,power_law_1.01,0.4268671989440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,4,power_law_1.01,0.4216576099395752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,4,power_law_1.01,0.43116159439086915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,4,power_law_1.01,0.44640002250671384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,4,power_law_1.01,0.45873279571533204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,2,balanced,0.03200533241033554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,2,balanced,0.03603733330965042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,4,power_law_1.01,0.47912960052490233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,2,balanced,0.057674666245778404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,2,balanced,0.05880533158779144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,2,balanced,0.06198933223883311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,2,balanced,0.06186666587988535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,2,balanced,0.06020799775918325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,2,balanced,0.06317333380381267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,2,balanced,0.06426666676998138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,2,balanced,0.06486399968465169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,2,balanced,0.06632533172766368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,2,balanced,0.06645333270231883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,2,balanced,0.06850666801134746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,2,balanced,0.07048533360163371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,2,balanced,0.08077866832415263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,2,balanced,0.08303999900817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,2,balanced,0.08102400104204814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,2,balanced,0.09745066364606221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,2,balanced,0.09977066516876221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,2,balanced,0.12346133589744568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,2,balanced,0.14381866653760275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,2,balanced,0.18862932920455933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,2,balanced,0.23184533913930258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,2,balanced,0.3174453377723694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,2,balanced,0.39504532019297284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,2,balanced,0.559002677599589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,4,power_law_1.01,0.4950528144836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,2,balanced,0.7192160288492838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,4,power_law_1.01,0.5601791858673095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,4,power_law_1.01,0.5867519855499268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,4,power_law_1.01,0.6843967914581299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,4,power_law_1.01,0.7810175895690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,4,power_law_1.01,0.9450816154479981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,4,power_law_1.01,0.9780672073364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,4,power_law_1.01,1.182252788543701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,4,power_law_1.01,1.6511360168457032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,4,power_law_1.01,2.122604751586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,4,power_law_1.01,2.431283187866211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,16,balanced,0.02349333216746648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,16,balanced,0.02310933421055476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,16,balanced,0.022474666436513264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,16,balanced,0.023317334552605946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,16,balanced,0.02749866743882497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,16,balanced,0.03307733436425527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,16,balanced,0.04331733286380768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,16,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,16,balanced,0.0637066662311554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,16,balanced,0.06352533400058746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,16,balanced,0.062090665102005005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,16,balanced,0.06423999865849812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,16,balanced,0.06193066636721293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,16,balanced,0.059205333391825356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,16,balanced,0.07924800117810567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,16,balanced,0.077674667040507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,16,balanced,0.06938666601975758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,16,balanced,0.07991999884446462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,16,balanced,0.09262933333714803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,16,balanced,0.10845866799354553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,16,balanced,0.13104533155759177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,16,balanced,0.1649493376413981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,16,balanced,0.19853333632151285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,16,balanced,0.2935733397801717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,16,power_law_1.01,0.026425600051879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,16,power_law_1.01,0.02587519884109497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,16,power_law_1.01,0.025964799523353576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,16,balanced,0.36214931805928546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,16,power_law_1.01,0.023347200453281404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,16,balanced,0.5266613165537516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,16,power_law_1.01,0.02510719895362854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,16,balanced,0.6956266562143961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,16,power_law_1.01,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,16,power_law_1.01,0.043033599853515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,16,power_law_1.01,0.04264959990978241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,16,power_law_1.01,0.062438398599624634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,16,power_law_1.01,0.06232960224151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,16,power_law_1.01,0.06332799792289734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.2,0.5395967960357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,16,power_law_1.01,0.06554880142211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,16,power_law_1.01,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,16,power_law_1.01,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,16,power_law_1.01,0.06430720090866089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,16,power_law_1.01,0.06265599727630615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,16,power_law_1.01,0.0655232012271881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,16,power_law_1.01,0.06391040086746216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.01,0.06936320066452026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.01,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.01,0.09101439714431762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.01,0.12074879407882691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.01,0.1461184024810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.01,0.20558080673217774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.01,0.25369598865509035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.01,0.34497919082641604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.01,0.4600959777832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,4,balanced,0.03175999969244003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,4,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,4,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,4,balanced,0.04996266464392344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,4,balanced,0.07023466626803081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,4,balanced,0.08330666522185008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,4,balanced,0.08098666866620381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,4,balanced,0.08366933465003967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,4,balanced,0.0844053328037262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,4,balanced,0.08411733309427898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,4,balanced,0.0851200024286906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,4,balanced,0.08424533406893413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,4,balanced,0.08661333719889323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,4,balanced,0.08793600400288899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,4,balanced,0.09063999851544698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,4,balanced,0.09304533402125041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,4,balanced,0.09476266304651897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,4,balanced,0.102101335922877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,4,balanced,0.10361066460609436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,4,balanced,0.1244053343931834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,4,balanced,0.12364799777666728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,4,balanced,0.16430399815241495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,4,balanced,0.17033066352208456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,4,balanced,0.23899734020233154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,4,balanced,0.25752532482147217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,4,balanced,0.3591200113296509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,4,balanced,0.42670400937398273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,128,power_law_1.2,0.01672320067882538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,128,power_law_1.2,0.016633599996566772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,128,power_law_1.2,0.017023999989032746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,128,power_law_1.2,0.019232000410556793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,128,power_law_1.2,0.030668801069259642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,128,power_law_1.2,0.026099199056625368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,128,power_law_1.2,0.02595840096473694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,128,power_law_1.2,0.025836798548698425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,128,power_law_1.2,0.02602880001068115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,128,power_law_1.2,0.026470398902893065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,128,power_law_1.2,0.026790401339530943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,128,power_law_1.2,0.027263998985290527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,128,power_law_1.2,0.027027198672294618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,128,power_law_1.2,0.027603200078010558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,128,power_law_1.2,0.029260799288749695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,128,power_law_1.2,0.029318401217460634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,128,power_law_1.2,0.030112001299858093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,128,power_law_1.2,0.032518398761749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,128,power_law_1.2,0.03437440097332001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,128,power_law_1.2,0.038380798697471616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,128,power_law_1.2,0.0398719996213913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,128,power_law_1.2,0.04723840057849884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,128,power_law_1.2,0.057011198997497556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,128,power_law_1.2,0.07267839908599853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,64,power_law_1.01,0.03397760093212128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,64,power_law_1.01,0.038553598523139956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,64,power_law_1.01,0.03289600014686585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,64,power_law_1.01,0.035078400373458864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,64,power_law_1.01,0.03489919900894165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,64,power_law_1.01,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,64,power_law_1.01,0.03555839955806732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,64,power_law_1.01,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,64,power_law_1.01,0.035750401020050046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,64,power_law_1.01,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,64,power_law_1.01,0.03992320001125336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,64,power_law_1.01,0.04202240109443665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,64,power_law_1.01,0.04296959936618805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,64,power_law_1.01,0.047839999198913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,64,power_law_1.01,0.05397760272026062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,64,power_law_1.01,0.05799679756164551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,64,power_law_1.01,0.055955201387405396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,64,power_law_1.01,0.0655232012271881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,64,power_law_1.01,0.0688256025314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,64,power_law_1.01,0.08267520070075988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,64,power_law_1.01,0.09034240245819092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,64,power_law_1.01,0.11644799709320068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,128,power_law_1.2,0.09294080138206481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,64,power_law_1.01,0.13193600177764891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,64,power_law_1.01,0.16680320501327514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,32,power_law_1.2,0.0723904013633728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,64,power_law_1.01,0.19875839948654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,32,power_law_1.2,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,64,power_law_1.01,0.31413118839263915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,32,power_law_1.2,0.06292480230331421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,64,power_law_1.01,0.47144317626953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,32,power_law_1.2,0.06706560254096985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,32,power_law_1.2,0.06528639793395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,32,power_law_1.2,0.061267197132110596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,32,power_law_1.2,0.06475520133972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,32,power_law_1.2,0.06569600105285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,32,power_law_1.2,0.06659200191497802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,32,power_law_1.2,0.06556159853935242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,32,power_law_1.2,0.06941440105438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,32,power_law_1.2,0.07080960273742676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,32,power_law_1.2,0.07189760208129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,32,power_law_1.2,0.07464320063591004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,32,power_law_1.2,0.08313599824905396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,32,power_law_1.2,0.08244479894638061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,32,power_law_1.2,0.08821120262145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,32,power_law_1.2,0.10297600030899048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,32,power_law_1.2,0.10383360385894776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,32,power_law_1.2,0.13690240383148194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,32,power_law_1.2,0.14097919464111328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,32,power_law_1.2,0.1844607949256897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,32,power_law_1.2,0.21397759914398193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,32,power_law_1.2,0.29788799285888673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,32,power_law_1.2,0.3585536003112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,32,power_law_1.2,0.5902656078338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,32,power_law_1.2,0.628384017944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,128,power_law_1.2,0.12846720218658447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,128,balanced,0.04461866617202759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,128,balanced,0.047055999437967934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,128,balanced,0.039488000174363456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,128,balanced,0.042208001017570496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,128,balanced,0.04231466849644979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,128,balanced,0.04624533156553904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,128,balanced,0.04638933142026266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,128,balanced,0.04635733366012573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,128,balanced,0.046351999044418335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,128,balanced,0.047397335370381675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,128,balanced,0.046298667788505554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,128,balanced,0.047653332352638245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,128,balanced,0.04971200227737427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,128,balanced,0.048656001687049866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,128,balanced,0.053946668903032936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,128,balanced,0.05436266462008158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,128,balanced,0.05608533322811127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,128,balanced,0.061978667974472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,128,balanced,0.06504000226656596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,128,balanced,0.07426133255163829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,128,balanced,0.0807360013326009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,128,balanced,0.09875200192133586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,128,balanced,0.12158933281898499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,128,balanced,0.1755680044492086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,128,balanced,0.21628799041112265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,128,balanced,0.29550933837890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,128,balanced,0.3758026758829753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,128,power_law_1.2,0.1620800018310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,16,power_law_1.01,0.058271998167037965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,16,power_law_1.01,0.07580800056457519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,16,power_law_1.01,0.07368959784507752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,16,power_law_1.01,0.07786239981651306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,16,power_law_1.01,0.07714560031890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,16,power_law_1.01,0.07915520071983337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,16,power_law_1.01,0.07731199860572815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,16,power_law_1.01,0.0754688024520874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,16,power_law_1.01,0.07801600098609925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,16,power_law_1.01,0.07992320060729981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,16,power_law_1.01,0.08046079874038696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,16,power_law_1.01,0.07930880188941955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,16,power_law_1.01,0.08160640001296997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,16,power_law_1.01,0.08303999900817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,16,power_law_1.01,0.09195520281791687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,16,power_law_1.01,0.09395840167999267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,16,power_law_1.01,0.10140800476074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,16,power_law_1.01,0.11208319664001465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,16,power_law_1.01,0.11559040546417236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,16,power_law_1.01,0.13870079517364503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,16,power_law_1.01,0.14958720207214354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,16,power_law_1.01,0.18649599552154542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,16,power_law_1.01,0.23822081089019775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,16,power_law_1.01,0.289574408531189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,16,power_law_1.01,0.3698944091796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,16,power_law_1.01,0.5643904209136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,16,power_law_1.01,0.6910655975341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,2,balanced,0.057061334451039634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,2,balanced,0.062074666221936546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,2,balanced,0.07524799803892772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,2,balanced,0.10506133238474528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,2,balanced,0.1470026671886444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,2,balanced,0.2346400022506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,2,balanced,0.23610132932662964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,2,balanced,0.23705067237218222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,2,balanced,0.237119992574056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,2,balanced,0.240831991036733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,2,balanced,0.24152000745137533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,2,balanced,0.2462559938430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,2,balanced,0.24802666902542114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,2,balanced,0.2511306603749593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,2,balanced,0.2614720066388448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,2,balanced,0.26311999559402466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,2,balanced,0.27131734291712445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,2,balanced,0.2909440000851949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,2,balanced,0.3076373338699341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,2,balanced,0.3484479983647664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,2,balanced,0.390064001083374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,2,balanced,0.4672000010808309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,2,balanced,0.5372426509857178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,2,balanced,0.7594133218129476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,2,balanced,0.9087839921315511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,2,balanced,1.296346664428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,2,balanced,1.6541333198547363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,16,power_law_1.01,0.021542400121688843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,16,power_law_1.01,0.026092800498008727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,16,power_law_1.01,0.026681599020957947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,16,power_law_1.01,0.026528000831604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,16,power_law_1.01,0.024831999838352204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,16,power_law_1.01,0.024480000138282776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,16,power_law_1.01,0.02481279969215393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,16,power_law_1.01,0.02537600100040436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,16,power_law_1.01,0.026950401067733765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,16,power_law_1.01,0.027340799570083618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,16,power_law_1.01,0.02967039942741394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,16,power_law_1.01,0.031769600510597226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,16,power_law_1.01,0.030739200115203858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,16,power_law_1.01,0.03823359906673431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,16,power_law_1.01,0.04077439904212952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,16,power_law_1.01,0.0426367998123169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,16,power_law_1.01,0.056531202793121335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,16,power_law_1.01,0.05980160236358643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,16,power_law_1.01,0.06848000288009644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,16,power_law_1.01,0.08344320058822632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,16,power_law_1.01,0.06823679804801941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,16,power_law_1.01,0.08127359747886657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,16,power_law_1.01,0.0939840018749237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,16,power_law_1.01,0.11863679885864258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,16,power_law_1.01,0.14855040311813356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,16,power_law_1.01,0.21543679237365723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,16,power_law_1.01,0.2615936040878296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,8,power_law_1.2,0.06949120163917541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,8,power_law_1.2,0.07928959727287292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,8,power_law_1.2,0.08481919765472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,8,power_law_1.2,0.09726719856262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,8,power_law_1.2,0.10607359409332276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,8,power_law_1.2,0.10547840595245361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,8,power_law_1.2,0.10210560560226441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,8,power_law_1.2,0.1033087968826294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,8,power_law_1.2,0.10236799716949463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,8,power_law_1.2,0.10375679731369018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,8,power_law_1.2,0.10736000537872314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,8,power_law_1.2,0.11472640037536622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,8,power_law_1.2,0.11440000534057618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,8,power_law_1.2,0.12065279483795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,8,power_law_1.2,0.1265727996826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,8,power_law_1.2,0.1320255994796753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,8,power_law_1.2,0.14156160354614258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,8,power_law_1.2,0.16362240314483642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,8,power_law_1.2,0.17935999631881713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,8,power_law_1.2,0.21969919204711913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,8,power_law_1.2,0.2707135915756226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,8,power_law_1.2,0.3520512104034424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,8,power_law_1.2,0.401043176651001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,8,power_law_1.2,0.5381247997283936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,8,power_law_1.2,0.6995327949523926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,8,power_law_1.2,0.9103551864624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,16,power_law_1.2,0.0722495973110199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,16,power_law_1.2,0.0712768018245697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,16,power_law_1.2,0.07459840178489685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,16,power_law_1.2,0.07736960053443909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,16,power_law_1.2,0.07121919989585876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,16,power_law_1.2,0.07789440155029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,16,power_law_1.2,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,16,power_law_1.2,0.07895039916038513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,16,power_law_1.2,0.07889919877052307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,16,power_law_1.2,0.08120959997177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,16,power_law_1.2,0.08247680068016053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,16,power_law_1.2,0.08199040293693542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,16,power_law_1.2,0.08966400027275086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,8,power_law_1.01,0.03144319951534271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,16,power_law_1.2,0.08828799724578858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,8,power_law_1.01,0.030579200387001036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,16,power_law_1.2,0.09457280039787293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,8,power_law_1.2,1.1385279655456544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,16,power_law_1.2,0.09481599926948547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,16,power_law_1.2,0.10641920566558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,16,power_law_1.2,0.11839359998703003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,16,power_law_1.2,0.12348799705505371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,16,power_law_1.2,0.15537279844284058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,16,power_law_1.2,0.16547839641571044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,16,power_law_1.2,0.21724801063537597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,8,power_law_1.01,0.03154560029506683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,8,power_law_1.01,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,16,power_law_1.2,0.2849024057388306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,8,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,8,power_law_1.01,0.03427839875221252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,16,power_law_1.2,0.3538815975189209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,8,power_law_1.01,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,8,power_law_1.01,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,16,power_law_1.2,0.5100480079650879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,32,1,balanced,0.08268266419569652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,32,1,balanced,0.08725866675376892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,32,1,balanced,0.09877866506576538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,32,1,balanced,0.12211199601491292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,32,1,balanced,0.15988266468048096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,32,1,balanced,0.23838933308919272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,32,1,balanced,0.2428213357925415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,8,power_law_1.01,0.03460479974746704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,32,1,balanced,0.244869331518809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,32,1,balanced,0.24570665756861368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,8,power_law_1.01,0.035334399342536925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,8,power_law_1.01,0.03637759983539581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,32,1,balanced,0.24887466430664062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,32,1,balanced,0.24942400058110556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,32,1,balanced,0.25913067658742267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,32,1,balanced,0.2608853379885356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,32,1,balanced,0.2630666693051656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,32,1,balanced,0.27530133724212646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,32,1,balanced,0.2802559932072957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,32,1,balanced,0.29289066791534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,32,1,balanced,0.337007999420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,32,1,balanced,0.36658668518066406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,32,1,balanced,0.445093313852946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,32,1,balanced,0.5028640031814575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,32,1,balanced,0.663429339726766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,8,power_law_1.01,0.03735679984092712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,32,1,balanced,0.8171839714050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,8,power_law_1.01,0.03953279852867127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,32,1,balanced,1.2103466987609863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,32,1,balanced,1.5402240753173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,32,1,balanced,2.255530675252279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,8,power_law_1.01,0.04097920060157776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,32,1,balanced,2.9785385131835938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,16,power_law_1.2,0.6029056072235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,8,power_law_1.01,0.047788798809051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,16,power_law_1.2,0.8060223579406738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,8,power_law_1.01,0.05278080105781555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,8,power_law_1.01,0.056377601623535153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,8,power_law_1.01,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,8,power_law_1.01,0.08147839903831482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,8,power_law_1.01,0.10734720230102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,8,power_law_1.01,0.1415488004684448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,128,power_law_1.01,0.26924159526824953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,128,power_law_1.01,0.3060096025466919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,8,power_law_1.01,0.19540480375289918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,128,power_law_1.01,0.06207360029220581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,8,power_law_1.01,0.2034048080444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,128,power_law_1.01,0.06626560091972351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,128,power_law_1.01,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,128,power_law_1.01,0.06542080044746398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,128,power_law_1.01,0.0707264006137848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,128,power_law_1.01,0.06886399984359741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,128,power_law_1.01,0.06904320120811462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,128,power_law_1.01,0.06960639953613282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,128,power_law_1.01,0.07155200242996215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,128,power_law_1.01,0.07348480224609374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,128,power_law_1.01,0.07567999958992004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,8,power_law_1.01,0.34371199607849123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,128,power_law_1.01,0.08039039969444275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,128,power_law_1.01,0.08727040290832519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,128,power_law_1.01,0.08826239705085755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,128,power_law_1.01,0.0914687991142273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,128,power_law_1.01,0.10755200386047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,128,power_law_1.01,0.1198464035987854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,128,power_law_1.01,0.15947519540786742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,128,power_law_1.01,0.16919039487838744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,128,power_law_1.01,0.21851520538330077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,128,power_law_1.01,0.2632512092590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,128,power_law_1.01,0.3655359983444214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,128,power_law_1.01,0.4644480228424072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,128,power_law_1.01,0.656108808517456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,128,power_law_1.01,0.8277119636535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,16,power_law_1.2,0.06389120221138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,16,power_law_1.2,0.08279039859771728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,16,power_law_1.2,0.0849727988243103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,16,power_law_1.2,0.09088000059127807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,16,power_law_1.2,0.09202560186386108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,16,power_law_1.2,0.09399679899215699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,16,power_law_1.2,0.08067839741706848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,16,power_law_1.2,0.09162880182266235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,16,power_law_1.2,0.08981119990348815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,16,power_law_1.2,0.09336959719657897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,16,power_law_1.2,0.09397119879722596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,16,power_law_1.2,0.09382399916648865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,16,power_law_1.2,0.09291520118713378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,16,power_law_1.2,0.0977728009223938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,16,power_law_1.2,0.10645760297775268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,16,power_law_1.2,0.11144319772720337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,16,power_law_1.2,0.12247040271759033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,16,power_law_1.2,0.13652479648590088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,16,power_law_1.2,0.1400320053100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,16,power_law_1.2,0.16966400146484376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,16,power_law_1.2,0.190937602519989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,16,power_law_1.2,0.25654399394989014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,16,power_law_1.2,0.3347584009170532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,16,power_law_1.2,0.460371208190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,16,power_law_1.2,0.5770048141479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,8,power_law_1.01,0.4525951862335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,16,power_law_1.2,0.9067968368530274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,16,power_law_1.2,1.1099967956542969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,8,power_law_1.01,0.6992256164550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,8,power_law_1.01,1.1137663841247558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,128,power_law_1.2,0.037651199102401736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,128,power_law_1.2,0.03738879859447479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,128,power_law_1.2,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,128,power_law_1.2,0.040031999349594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,128,power_law_1.2,0.040428799390792844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,128,power_law_1.2,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,128,power_law_1.2,0.04154239892959595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,128,power_law_1.2,0.04073599874973297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,128,power_law_1.2,0.040531200170516965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,128,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,128,power_law_1.2,0.040608000755310056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,128,power_law_1.2,0.04102399945259094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,128,power_law_1.2,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,128,power_law_1.2,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,128,power_law_1.2,0.04604800045490265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,128,power_law_1.2,0.046003198623657225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,128,power_law_1.2,0.045875200629234315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,128,power_law_1.2,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,128,power_law_1.2,0.05256320238113403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,128,power_law_1.2,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,128,power_law_1.2,0.06524800062179566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,128,power_law_1.2,0.08617600202560424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,128,power_law_1.2,0.11041280031204223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,128,power_law_1.2,0.14138879776000976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,128,power_law_1.2,0.17761919498443604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,128,power_law_1.2,0.26732161045074465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,128,power_law_1.2,0.374182391166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,4,balanced,0.019386666516462963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,4,balanced,0.019039999693632126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,4,balanced,0.019248000035683315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,4,balanced,0.019354666272799175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,4,balanced,0.02130666623512904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,4,balanced,0.021029333273569744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,4,balanced,0.02718399961789449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,4,balanced,0.02516799916823705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,4,balanced,0.029120000700155895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,4,balanced,0.027258666853109997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,4,balanced,0.029189333319664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,4,balanced,0.03133866687615713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,4,balanced,0.03397866586844126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,4,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,4,balanced,0.04552533229192098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,4,balanced,0.043840001026789345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,4,balanced,0.045893331368764244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,4,balanced,0.059248000383377075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,4,balanced,0.06714666883150737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,4,balanced,0.08683199683825175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,4,balanced,0.10945600271224976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,4,balanced,0.14867732922236124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,4,balanced,0.18172266085942587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,4,balanced,0.25752532482147217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,4,balanced,0.3290346662203471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,256,power_law_1.01,0.0795199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,4,balanced,0.4766240119934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,4,balanced,0.6247733434041342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,256,power_law_1.01,0.07997440099716187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,256,power_law_1.01,0.07909119725227357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,256,power_law_1.01,0.08127999901771546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,256,power_law_1.01,0.0825215995311737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,256,power_law_1.01,0.08568320274353028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,256,power_law_1.01,0.08314239978790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,256,power_law_1.01,0.08312960267066956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,256,power_law_1.01,0.08392320275306701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,256,power_law_1.01,0.08384000062942505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,256,power_law_1.01,0.0830784022808075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,256,power_law_1.01,0.08311039805412293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,256,power_law_1.01,0.08406400084495544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,256,power_law_1.01,0.08181759715080261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,256,power_law_1.01,0.08456959724426269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,256,power_law_1.01,0.08327680230140685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,256,power_law_1.01,0.08863360285758973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,256,power_law_1.01,0.1045632004737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,256,power_law_1.01,0.10566400289535523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,256,power_law_1.01,0.12551679611206054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,256,power_law_1.01,0.15575040578842164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,256,power_law_1.01,0.19237120151519777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,256,power_law_1.01,0.21383039951324462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,256,power_law_1.01,0.3090431928634644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,256,power_law_1.01,0.39996159076690674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,256,power_law_1.01,0.5594367980957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,256,power_law_1.01,0.6939775943756104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,256,power_law_1.01,0.07348480224609374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,256,power_law_1.01,0.07191680073738098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,256,power_law_1.01,0.07209600210189819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,256,power_law_1.01,0.07528960108757018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,256,power_law_1.01,0.077920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,256,power_law_1.01,0.07980800271034241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,256,power_law_1.01,0.07736960053443909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,256,power_law_1.01,0.07794560194015503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,256,power_law_1.01,0.07793920040130616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,256,power_law_1.01,0.07742080092430115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,256,power_law_1.01,0.07643520236015319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,256,power_law_1.01,0.07628160119056701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,256,power_law_1.01,0.07601919770240784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,256,power_law_1.01,0.0754047989845276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,256,power_law_1.01,0.07640320062637329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,256,power_law_1.01,0.07557759881019592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,256,power_law_1.01,0.07643520236015319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,256,power_law_1.01,0.09345279932022095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,256,power_law_1.01,0.09553279876708984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,256,power_law_1.01,0.11080319881439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,256,power_law_1.01,0.1382143974304199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,256,power_law_1.01,0.16851199865341188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,256,power_law_1.01,0.19837440252304078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,256,power_law_1.01,0.2782912015914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,256,power_law_1.01,0.3425856113433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,256,power_law_1.01,0.4868288040161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,256,power_law_1.01,0.6498176097869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,2,balanced,0.03196266790231069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,2,balanced,0.03346133232116699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,2,balanced,0.052101333936055504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,2,balanced,0.05509333312511444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,2,balanced,0.057962665955225624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,2,balanced,0.058037335673967995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,2,balanced,0.05830933153629303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,2,balanced,0.06002666552861532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,2,balanced,0.06011733412742615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,2,balanced,0.060138667623202004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,2,balanced,0.06227200229962667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,2,balanced,0.06198399762312571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,2,balanced,0.06576533118883769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,2,balanced,0.06713066498438518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,2,balanced,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,2,balanced,0.07713066538174947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,2,balanced,0.08291199803352356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,2,balanced,0.0927946666876475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,2,balanced,0.10225600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,2,balanced,0.12782399853070578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,2,balanced,0.1514133314291636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,2,balanced,0.19869865973790488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,2,balanced,0.24486400683720908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,2,balanced,0.3375519911448161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,2,balanced,0.42028268178304035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,2,balanced,0.6042933464050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,2,balanced,0.775328000386556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,32,power_law_1.2,0.060198402404785155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,32,power_law_1.2,0.06259840130805969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,32,power_law_1.2,0.057766401767730714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,32,power_law_1.2,0.060153597593307497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,32,power_law_1.2,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,32,power_law_1.2,0.062636798620224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,32,power_law_1.2,0.062028801441192626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,32,power_law_1.2,0.06356480121612548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,32,power_law_1.2,0.06472319960594178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,32,power_law_1.2,0.06616960167884826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,32,power_law_1.2,0.06737279891967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,32,power_law_1.2,0.07442560195922851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,32,power_law_1.2,0.07589759826660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,32,power_law_1.2,0.07613440155982971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,32,power_law_1.2,0.0834176003932953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,32,power_law_1.2,0.08770560026168824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,32,power_law_1.2,0.09166719913482665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,32,power_law_1.2,0.10804480314254761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,32,power_law_1.2,0.12028160095214843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,32,power_law_1.2,0.15623040199279786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,32,power_law_1.2,0.18294399976730347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,32,power_law_1.2,0.2412992000579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,32,power_law_1.2,0.25137279033660886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,32,power_law_1.2,0.3529344081878662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,32,balanced,0.04634666442871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,32,balanced,0.04632533093293508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,32,power_law_1.2,0.4065855979919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,32,power_law_1.2,0.5943808078765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,32,power_law_1.2,0.7538496017456054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,32,balanced,0.047226667404174805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,32,balanced,0.051882664362589516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,32,balanced,0.05202133456865946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,32,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,32,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,32,balanced,0.055813332398732506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,32,balanced,0.05489066739877065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,32,balanced,0.05598400036493937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,32,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,32,balanced,0.05827199916044871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,32,balanced,0.05850133299827576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,32,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,32,balanced,0.0645066648721695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,32,balanced,0.06436266501744588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,32,balanced,0.06508799890677135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,32,balanced,0.07050133248170216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,32,balanced,0.0765173335870107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,32,balanced,0.09158933162689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,32,balanced,0.09942400455474854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,32,balanced,0.12844799955685934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,32,balanced,0.14733866850535074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,32,balanced,0.1908373236656189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,32,balanced,0.22220265865325928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,32,balanced,0.3139946659406026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,32,balanced,0.3749866485595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,4,power_law_1.01,0.06261759996414185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,4,power_law_1.01,0.0671231985092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,4,power_law_1.01,0.06629760265350342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,4,power_law_1.01,0.07404159903526306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,4,power_law_1.01,0.07411199808120728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,4,power_law_1.01,0.07329919934272766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,4,power_law_1.01,0.07226240038871765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,4,power_law_1.01,0.07251840233802795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,4,power_law_1.01,0.07747840285301208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,4,power_law_1.01,0.07698559761047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,4,power_law_1.01,0.07352960109710693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,4,power_law_1.01,0.07686399817466735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,4,power_law_1.01,0.07241600155830383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,4,power_law_1.01,0.07500159740447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,4,power_law_1.01,0.09080320000648498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,4,power_law_1.01,0.09154559969902039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,4,power_law_1.01,0.10722559690475464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,4,power_law_1.01,0.11422719955444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,4,power_law_1.01,0.13258880376815796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,4,power_law_1.01,0.16263680458068847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,4,power_law_1.01,0.21834878921508788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,4,power_law_1.01,0.30456318855285647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,4,power_law_1.01,0.31653759479522703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,32,balanced,0.05036800106366476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,32,balanced,0.049914668003718056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,32,balanced,0.050106664498647056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,32,balanced,0.05494399865468343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,32,balanced,0.06488533318042755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,32,balanced,0.07429333527882893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,32,balanced,0.07424533367156982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,32,balanced,0.07348266740640004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,32,balanced,0.07613866527875264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,32,balanced,0.08101866642634074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,32,balanced,0.08193066716194153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,32,balanced,0.08476799726486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,32,balanced,0.08053866525491078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,32,balanced,0.08648533622423808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,32,balanced,0.09392533699671428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,32,balanced,0.09052800138791402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,32,balanced,0.09798933068911235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,32,balanced,0.1032426655292511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,32,balanced,0.10764267047246297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,4,power_law_1.01,0.49187841415405276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,32,balanced,0.12004799644152324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,32,balanced,0.13326399525006613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,32,balanced,0.16285333037376404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,32,balanced,0.18316266934076944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,32,balanced,0.24374399582544962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,32,balanced,0.2930826743443807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,32,balanced,0.3991520007451375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,32,balanced,0.5099786520004272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,1,power_law_1.2,0.10398080348968505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,1,power_law_1.2,0.1536128044128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,1,power_law_1.2,0.24157440662384033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,1,power_law_1.2,0.44423680305480956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,1,power_law_1.2,0.6953472137451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,1,power_law_1.2,0.9746303558349609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,1,power_law_1.2,1.1573568344116212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,1,power_law_1.2,1.5485504150390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,1,power_law_1.2,1.648966407775879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,1,power_law_1.2,1.6810943603515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,1,power_law_1.2,1.7622655868530273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,4,power_law_1.01,0.602732801437378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,1,power_law_1.2,1.8757375717163085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,1,power_law_1.2,1.8937088012695313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,1,power_law_1.2,1.9727680206298828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,1,power_law_1.2,2.040838432312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,1,power_law_1.2,2.0782144546508787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,1,balanced,0.08506666620572408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,1,balanced,0.12109333276748657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,1,power_law_1.2,2.1871871948242188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,1,power_law_1.2,2.314489555358887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,1,balanced,0.18901334206263223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,1,balanced,0.3333653211593628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,1,power_law_1.2,2.492780876159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,1,balanced,0.6060106754302979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,1,balanced,0.8666346867879232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,1,balanced,0.8706933657328287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,1,power_law_1.2,2.872435188293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,1,balanced,0.8747413158416748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,1,balanced,0.8819733460744222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,1,balanced,0.8832746346791586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,1,power_law_1.2,2.715750312805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,1,balanced,0.891541322072347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,1,balanced,0.8990560372670492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,1,power_law_1.2,3.0692928314208983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,1,balanced,0.9047893683115641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,1,balanced,0.9167146682739258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,1,balanced,0.9347573121388754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,1,power_law_1.2,3.7549633026123046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,1,balanced,0.9461812973022461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,1,balanced,0.9661813577016195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,1,balanced,1.1157173315684001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,1,balanced,1.0762560367584229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,1,power_law_1.2,4.637945556640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,1,balanced,1.4257653554280598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,1,power_law_1.2,4.690278244018555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,1,balanced,1.277898629506429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,1,balanced,1.9762399991353352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,1,balanced,1.8016586303710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,1,power_law_1.2,5.831027221679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,1,balanced,2.7788852055867515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,1,balanced,3.0241705576578775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,1,power_law_1.2,7.822112274169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,1,balanced,4.877621332804362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,1,balanced,5.358714421590169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,4,power_law_1.01,1.0187583923339845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,8,balanced,0.019797333826621372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,8,balanced,0.0210506667693456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,8,balanced,0.021183999876181286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,8,balanced,0.023039999107519787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,8,balanced,0.027717334528764088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,8,balanced,0.04241600135962168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,8,balanced,0.04481600224971771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,8,balanced,0.047482664386431374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,8,balanced,0.05387733379999796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,8,balanced,0.05307200054327647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,8,balanced,0.05373866856098175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,8,balanced,0.07116266588370006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,8,balanced,0.06807999809583028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,8,balanced,0.06970666845639546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,8,balanced,0.08136533200740814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,8,balanced,0.0813920001188914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,8,balanced,0.08689066767692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,8,balanced,0.09460266431172688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,8,balanced,0.09946667154630025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,8,balanced,0.10963732997576396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,8,balanced,0.12417599558830261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,8,balanced,0.14015466968218485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,8,balanced,0.1613920032978058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,8,balanced,0.19298666715621948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,8,balanced,0.2244053284327189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,8,balanced,0.3510293165842692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,8,balanced,0.4157866636912028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,4,power_law_1.2,0.04258559942245484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,4,power_law_1.2,0.040908798575401306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,4,power_law_1.2,0.046988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,4,power_law_1.2,0.05287039875984192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,4,power_law_1.2,0.054739201068878175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,4,power_law_1.2,0.07091839909553528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,4,power_law_1.01,1.1265664100646973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,4,power_law_1.2,0.07331839799880982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,4,power_law_1.2,0.07253760099411011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,4,power_law_1.2,0.07516160011291503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,4,power_law_1.2,0.07570559978485107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,4,power_law_1.2,0.07607679963111877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,4,power_law_1.2,0.07912960052490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,4,power_law_1.2,0.08205440044403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,4,power_law_1.2,0.08694400191307068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,4,power_law_1.2,0.09221119880676269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,4,power_law_1.2,0.098854398727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,4,power_law_1.2,0.10535039901733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,4,power_law_1.2,0.11928319931030273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.2,0.1248128056526184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.2,0.15545599460601806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.2,0.18391679525375365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.2,0.22043519020080565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.2,0.31040639877319337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.2,0.39260799884796144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,1,power_law_1.2,0.04873600006103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,1,power_law_1.2,0.0643455982208252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,1,power_law_1.2,0.09543039798736572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,1,power_law_1.2,0.14014079570770263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,1,power_law_1.2,0.1856384038925171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,1,power_law_1.2,0.3087296009063721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,1,power_law_1.2,0.31668479442596437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,1,power_law_1.2,0.34335999488830565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,1,power_law_1.2,0.34997758865356443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,1,power_law_1.2,0.3568128108978271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,1,power_law_1.2,0.361843204498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,1,power_law_1.2,0.3764031887054443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,1,power_law_1.2,0.3746623992919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,1,power_law_1.2,0.38620800971984864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,1,power_law_1.2,0.39915521144866944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,1,power_law_1.2,0.40718722343444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,1,power_law_1.2,0.43695998191833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,1,power_law_1.2,0.4818687915802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,1,power_law_1.2,0.4925568103790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,1,power_law_1.2,0.5595327854156494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,1,power_law_1.2,0.5999360084533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.2,0.49811201095581054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,1,power_law_1.2,0.7123456001281738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,1,power_law_1.2,0.7807744026184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,1,power_law_1.2,0.9482751846313476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,1,power_law_1.2,1.2329471588134766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,1,power_law_1.2,1.5678208351135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,1,power_law_1.2,1.8022272109985351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.2,0.6811327934265137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,1,balanced,0.07165866593519847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,1,balanced,0.09698133667310078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,1,balanced,0.15680533647537231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,1,balanced,0.1622773309548696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,1,balanced,0.1646773318449656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,1,balanced,0.16427733500798544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,1,balanced,0.16478932897249857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,1,balanced,0.16635200381278992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,1,balanced,0.16501866777737936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,1,balanced,0.165994664033254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,1,balanced,0.1653600037097931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,1,balanced,0.16952532529830933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,1,balanced,0.16959466536839804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,1,balanced,0.17116800944010416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,1,balanced,0.18013866742451987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,1,balanced,0.18384534120559692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,1,balanced,0.18607999881108603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,1,balanced,0.23407999674479166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,1,balanced,0.2477173407872518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.2,0.9606656074523926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,1,balanced,0.34857598940531415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,1,balanced,0.430245320002238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,1,balanced,0.6101280053456625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,1,balanced,0.7811306317647299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,1,balanced,1.1161119937896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,1,balanced,1.456725279490153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,1,balanced,2.16373872756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,1,balanced,2.843056042989095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,1,power_law_1.2,0.021990400552749634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,1,power_law_1.2,0.020640000700950623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,1,power_law_1.2,0.02099200040102005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,1,power_law_1.2,0.02033279985189438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,1,power_law_1.2,0.020960000157356263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,1,power_law_1.2,0.02479359954595566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,1,power_law_1.2,0.030771198868751525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,1,power_law_1.2,0.02693760097026825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,1,power_law_1.2,0.03425920009613037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,1,power_law_1.2,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,1,power_law_1.2,0.03559040129184723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,1,power_law_1.2,0.03961600065231323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,1,power_law_1.2,0.04056960046291351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,1,power_law_1.2,0.0406464010477066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,1,power_law_1.2,0.058739197254180905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,1,power_law_1.2,0.05260800123214722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,1,power_law_1.2,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,1,power_law_1.2,0.07297279834747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.2,0.09050880074501037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.2,0.12561919689178466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.2,0.14968960285186766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.2,0.21098239421844484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.2,0.26739840507507323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.2,0.3807807922363281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.2,0.49763197898864747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,8,power_law_1.2,0.01865600049495697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,8,power_law_1.2,0.018246400356292724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,8,power_law_1.2,0.019276799261569978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,8,power_law_1.2,0.019891199469566346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,8,power_law_1.2,0.021382400393486024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,8,power_law_1.2,0.022515200078487396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,8,power_law_1.2,0.02380799949169159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,8,power_law_1.2,0.024582399427890776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,8,power_law_1.2,0.025843200087547303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,8,power_law_1.2,0.026399999856948853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,8,power_law_1.2,0.02632319927215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,8,power_law_1.2,0.03294079899787903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,8,power_law_1.2,0.032287999987602234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,8,power_law_1.2,0.0331712007522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,8,power_law_1.2,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,8,power_law_1.2,0.04576640129089356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,8,power_law_1.2,0.04663040041923523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,8,power_law_1.2,0.05400959849357605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,8,power_law_1.2,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,8,power_law_1.2,0.06280959844589233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,8,power_law_1.2,0.0669376015663147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,8,power_law_1.2,0.078847998380661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,8,power_law_1.2,0.09306880235671997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,8,power_law_1.2,0.11667840480804444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,8,power_law_1.2,0.13865599632263184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,8,power_law_1.2,0.2007551908493042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,8,power_law_1.2,0.23850879669189454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.2,0.7294015884399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,64,balanced,0.03381866713364919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,64,balanced,0.03336533407370249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,64,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,64,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,64,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,64,balanced,0.03749333322048187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,64,balanced,0.03978666663169861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,64,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,64,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,64,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,64,balanced,0.043552001317342125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,64,balanced,0.04603200157483419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,64,balanced,0.042591998974482216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,64,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,64,balanced,0.051685333251953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,64,balanced,0.05161599814891815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,64,balanced,0.051114668448766075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,64,balanced,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,64,balanced,0.057722667853037514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,64,balanced,0.06648000081380208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,64,balanced,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,64,balanced,0.08944533268610637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,64,balanced,0.10293333729108174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,64,balanced,0.13403200109799704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,64,balanced,0.16061333815256754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,64,balanced,0.22018667062123617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,64,balanced,0.2749653259913127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,2,power_law_1.2,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,2,power_law_1.2,0.03429119884967804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,2,power_law_1.2,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,2,power_law_1.2,0.03841919898986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,2,power_law_1.2,0.03888640105724335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,2,power_law_1.2,0.03943679928779602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,2,power_law_1.2,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,2,power_law_1.2,0.04046719968318939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,2,power_law_1.2,0.04023039937019348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,2,power_law_1.2,0.04030719995498657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,2,power_law_1.2,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,2,power_law_1.2,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,2,power_law_1.2,0.05064319968223572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,2,power_law_1.2,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.2,0.955840015411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,2,power_law_1.2,0.05711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,2,power_law_1.2,0.06108160018920898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,2,power_law_1.2,0.06636160016059875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,2,power_law_1.2,0.07728000283241272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,2,power_law_1.2,0.09266560077667237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,2,power_law_1.2,0.12258559465408325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,2,power_law_1.2,0.15395840406417846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,8,power_law_1.2,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,8,power_law_1.2,0.07612159848213196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,8,power_law_1.2,0.08731520175933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,8,power_law_1.2,0.11103999614715576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,16,balanced,0.04543999830881754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,8,power_law_1.2,0.11539839506149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,16,balanced,0.04634666442871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,16,balanced,0.046037331223487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,8,power_law_1.2,0.10640000104904175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,16,balanced,0.05611733098824819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,16,balanced,0.07433066765467326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,8,power_law_1.2,0.1125440001487732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,8,power_law_1.2,0.1141759991645813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,8,power_law_1.2,0.11358079910278321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,8,power_law_1.2,0.11623680591583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,8,power_law_1.2,0.11712000370025635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,8,power_law_1.2,0.12343039512634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,16,balanced,0.10299733281135559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,16,balanced,0.12693333625793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,8,power_law_1.2,0.12558079957962037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,16,balanced,0.12993066509564719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,16,balanced,0.12913599610328674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,16,balanced,0.13077867031097412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,16,balanced,0.1322826643784841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,16,balanced,0.13404800494511923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,16,balanced,0.1368160049120585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,16,balanced,0.13773866494496664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,16,balanced,0.1462453305721283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,16,balanced,0.14800000190734863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,16,balanced,0.15268799662590027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,16,balanced,0.16568000117937723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,16,balanced,0.17595734198888144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,16,balanced,0.19958933194478354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,16,balanced,0.22312533855438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,8,power_law_1.2,0.13180160522460938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,8,power_law_1.2,0.14179199934005737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,8,power_law_1.2,0.1514623999595642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,8,power_law_1.2,0.15619200468063354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,8,power_law_1.2,0.18167680501937866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,2,power_law_1.2,0.20851199626922606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,8,power_law_1.2,0.19720319509506226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,8,power_law_1.2,0.23624958992004394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,8,power_law_1.2,0.2607232093811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,8,power_law_1.2,0.3122688055038452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,16,balanced,0.271232008934021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,16,balanced,0.32818132638931274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,16,balanced,0.41997333367665607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,16,balanced,0.5115946531295776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,16,balanced,0.6947680314381918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,16,balanced,0.8995893001556396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,8,power_law_1.2,0.43527679443359374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,8,power_law_1.2,0.557696008682251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,8,power_law_1.2,0.6807360172271728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,8,power_law_1.2,0.9861503601074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,2,power_law_1.2,0.2790208101272583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,8,power_law_1.2,1.4455295562744142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,2,power_law_1.2,0.40760321617126466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,2,power_law_1.2,0.5047743797302247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,2,power_law_1.2,0.7819456100463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,2,balanced,0.08497066299120586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,2,balanced,0.13004266222318014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,2,balanced,0.21390400330225626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,2,balanced,0.3919573227564494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,2,balanced,0.7202080090840658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,2,balanced,1.3943840662638347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,2,balanced,1.401792049407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,2,balanced,1.4010987281799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,2,balanced,1.3988800048828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,2,balanced,1.3997920354207356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,2,balanced,1.4141119321187336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,2,balanced,1.4149972597757976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,2,balanced,1.4195574124654133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,2,balanced,1.4322880109151204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,2,balanced,1.432458718617757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,2,balanced,1.4501973787943523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,2,balanced,1.4582026799519856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,2,balanced,1.4823625882466633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,2,balanced,1.5236426989237468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,2,balanced,1.5780906677246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,2,balanced,1.6313014030456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,2,balanced,1.7185813585917156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,2,balanced,1.8751200040181477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,2,balanced,2.0861120223999023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,2,balanced,2.210437297821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,2,balanced,3.2912801106770835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,2,balanced,3.6264425913492837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,2,power_law_1.2,1.103775978088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,8,balanced,0.04784533381462097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,8,balanced,0.05093333125114441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,8,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,8,balanced,0.08010133107503255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,8,balanced,0.09796266754468282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,8,balanced,0.11355200409889221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,8,balanced,0.11149332920710246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,8,balanced,0.11180266737937927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,8,balanced,0.1149066686630249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,8,balanced,0.11309867103894551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,8,balanced,0.11527466773986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,8,balanced,0.11831999818483989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,8,balanced,0.11778666575749715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,8,balanced,0.12190399567286174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,8,balanced,0.1291253368059794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,8,balanced,0.12962133685747781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,8,balanced,0.13425599535306296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,8,balanced,0.1547040045261383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,8,balanced,0.1630453368028005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,8,balanced,0.2095680038134257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,8,balanced,0.22861866156260172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,8,balanced,0.32341333230336505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,8,balanced,0.31089067459106445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,8,balanced,0.4527680079142253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,8,balanced,0.5018506844838461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,8,balanced,0.7192053000132242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,8,balanced,0.8721066315968832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,1,balanced,0.03437866767247518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,1,balanced,0.035877334574858345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,1,balanced,0.04656533400217692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,1,balanced,0.07077866792678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,1,balanced,0.1120693286259969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,1,balanced,0.13412800431251526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,1,balanced,0.1336426635583242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,1,balanced,0.13062399625778198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,1,balanced,0.13263466954231262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,1,balanced,0.1334826648235321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,1,balanced,0.13435199856758118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,1,balanced,0.1351146697998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,1,balanced,0.13517333070437113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,1,balanced,0.13865066568056741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,1,balanced,0.1453973352909088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,1,balanced,0.1434986690680186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,1,balanced,0.15178133050600687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,1,balanced,0.1823199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,1,balanced,0.17261866728464761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,1,balanced,0.21221333742141724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,1,balanced,0.22415467103322348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,1,balanced,0.3164106607437134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,1,balanced,0.33554665247599286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,1,balanced,0.5163040161132812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,1,balanced,0.5353493293126425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,1,balanced,0.796229362487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,1,balanced,0.9661493301391602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,8,power_law_1.2,0.033881598711013795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,8,power_law_1.2,0.03261440098285675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,8,power_law_1.2,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,8,power_law_1.2,0.037862399220466615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,8,power_law_1.2,0.0353983998298645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,8,power_law_1.2,0.03950720131397247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,8,power_law_1.2,0.04100480079650879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,8,power_law_1.2,0.04273279905319214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,8,power_law_1.2,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,8,power_law_1.2,0.0438400000333786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,8,power_law_1.2,0.04750719964504242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,8,power_law_1.2,0.052262401580810545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,8,power_law_1.2,0.05563520193099976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,8,power_law_1.2,0.051769602298736575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,8,power_law_1.2,0.060089600086212155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,8,power_law_1.2,0.06543999910354614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,8,power_law_1.2,0.08661760091781616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,8,power_law_1.2,0.06433280110359192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,8,power_law_1.2,0.07635200023651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,8,power_law_1.2,0.1148800015449524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,8,power_law_1.2,0.09539200067520141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,8,power_law_1.2,0.1376255989074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,8,power_law_1.2,0.11868799924850464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,8,power_law_1.2,0.1671231985092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,8,power_law_1.2,0.13197439908981323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,8,power_law_1.2,0.20434560775756835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,8,power_law_1.2,0.22289280891418456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,8,power_law_1.2,0.17742079496383667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,8,power_law_1.2,0.20161919593811034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,8,power_law_1.2,0.2070336103439331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,8,power_law_1.2,0.18772480487823487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,8,power_law_1.2,0.2083712100982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,8,power_law_1.2,0.20101120471954345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,8,power_law_1.2,0.22196478843688966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,8,power_law_1.2,0.2257024049758911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,8,power_law_1.2,0.20629119873046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,8,power_law_1.2,0.23895039558410644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,8,power_law_1.2,0.24003200531005858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,8,power_law_1.2,0.23998720645904542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,8,power_law_1.2,0.2678591966629028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,8,power_law_1.2,0.2894144058227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,8,power_law_1.2,0.22910718917846679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,8,power_law_1.2,0.324563193321228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,8,power_law_1.2,0.3847807884216309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,8,power_law_1.2,0.47177600860595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,8,power_law_1.2,0.5651840209960938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,8,power_law_1.2,0.27418880462646483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,8,power_law_1.2,0.7057472229003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,8,power_law_1.2,0.843564796447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,8,power_law_1.2,1.163980770111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,8,power_law_1.2,0.3862848043441772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,8,power_law_1.2,1.4842944145202637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,8,power_law_1.2,0.5548480033874512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,32,balanced,0.04330666859944662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,32,balanced,0.04228800038496653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,32,balanced,0.043749332427978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,32,balanced,0.043525333205858864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,32,balanced,0.04614933331807455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,32,balanced,0.06591466565926869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,32,balanced,0.06595199803511302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,32,balanced,0.06598400076230367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,32,balanced,0.06630933284759521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,32,balanced,0.06633066634337108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,32,balanced,0.06607466439406078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,32,balanced,0.07011733452479045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,32,balanced,0.06842133402824402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,32,balanced,0.0710346649090449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,32,balanced,0.07720533510049184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,32,balanced,0.07658133407433827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,32,balanced,0.08051200211048126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,32,balanced,0.08590400218963623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,32,balanced,0.09128000338872273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,32,balanced,0.10603732864061992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,32,balanced,0.11617066462834676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,32,balanced,0.13970667123794556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,32,balanced,0.1625440021355947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,32,balanced,0.20751466353734335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,32,balanced,0.2545066674550374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,32,balanced,0.349178671836853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,32,balanced,0.445306658744812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,8,power_law_1.2,0.7076799869537354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,4,power_law_1.2,0.07580800056457519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,4,power_law_1.2,0.08795520067214965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,4,power_law_1.2,0.09056000113487243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,4,power_law_1.2,0.11752959489822387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,4,power_law_1.2,0.12929279804229737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,4,power_law_1.2,0.13540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,4,power_law_1.2,0.15361920595169068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,4,power_law_1.2,0.14743679761886597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,4,power_law_1.2,0.1549504041671753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,4,power_law_1.2,0.15801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,4,power_law_1.2,0.15356800556182862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,4,power_law_1.2,0.15774719715118407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,4,power_law_1.2,0.16839679479598998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,4,power_law_1.2,0.16389119625091553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,4,power_law_1.2,0.1793984055519104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,4,power_law_1.2,0.17916159629821776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,4,power_law_1.2,0.18108799457550048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,4,power_law_1.2,0.2046720027923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.2,0.2263808012008667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.2,0.25922560691833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.2,0.2987071990966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.2,0.4039936065673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.2,0.4799488067626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.2,0.6178815841674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,32,power_law_1.2,0.04247680008411407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.2,0.7909247875213623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,32,power_law_1.2,0.04041599929332733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,32,power_law_1.2,0.04575360119342804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.2,1.0945216178894044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,32,power_law_1.2,0.037913599610328676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,32,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.2,1.4867775917053223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,32,power_law_1.2,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,32,power_law_1.2,0.04356479942798615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,32,power_law_1.2,0.05303679704666138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,32,power_law_1.2,0.05360640287399292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,32,power_law_1.2,0.053446400165557864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,32,power_law_1.2,0.08323839902877808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,32,power_law_1.2,0.08333439826965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,32,power_law_1.2,0.08454399704933166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,32,power_law_1.2,0.08529279828071594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,32,power_law_1.2,0.09564160108566284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,32,power_law_1.2,0.10752639770507813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,32,power_law_1.2,0.11387519836425782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,32,power_law_1.2,0.1615679979324341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.2,0.19130239486694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.2,0.2620352029800415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.2,0.1424512028694153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.2,0.19865599870681763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.2,0.28105599880218507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.2,0.3479423999786377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.2,0.44327678680419924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.2,0.7303359985351563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.2,0.9279935836791993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,2,power_law_1.2,0.021516799926757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,2,power_law_1.2,0.02276480048894882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,2,power_law_1.2,0.028569599986076354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,2,power_law_1.2,0.037299200892448425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,2,power_law_1.2,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,2,power_law_1.2,0.055692797899246214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,2,power_law_1.2,0.0586624026298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,2,power_law_1.2,0.061990398168563846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,2,power_law_1.2,0.06476160287857055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,2,power_law_1.2,0.06598399877548218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,2,power_law_1.2,0.06894720196723939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,2,power_law_1.2,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,2,power_law_1.2,0.07088639736175537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,2,power_law_1.2,0.07588480114936828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,2,power_law_1.2,0.07993599772453308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,2,power_law_1.2,0.08407040238380432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,2,power_law_1.2,0.08843520283699036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,2,power_law_1.2,0.1617408037185669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,2,power_law_1.2,0.19048320055007933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,2,power_law_1.2,0.17582080364227295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,2,power_law_1.2,0.18665599822998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,2,power_law_1.2,0.22050559520721436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,2,power_law_1.2,0.25768959522247314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,2,power_law_1.2,0.32572801113128663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,2,power_law_1.2,0.40485758781433107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,2,power_law_1.2,0.5506048202514648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,2,power_law_1.2,0.6717311859130859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,4,power_law_1.01,0.04256640076637268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,4,power_law_1.01,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,4,power_law_1.01,0.04906879961490631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,4,power_law_1.01,0.0598143994808197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,4,power_law_1.01,0.060627198219299315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,4,power_law_1.01,0.06170240044593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,4,power_law_1.01,0.06428160071372986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,4,power_law_1.01,0.06515840291976929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,4,power_law_1.01,0.06826239824295044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,4,power_law_1.01,0.0687936007976532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,4,power_law_1.01,0.06855040192604064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,4,power_law_1.01,0.07235199809074402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,4,power_law_1.01,0.07206400036811829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,4,power_law_1.01,0.07870079874992371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,4,power_law_1.01,0.08713600039482117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,4,power_law_1.01,0.09406080245971679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,4,power_law_1.01,0.11185280084609986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,4,power_law_1.01,0.13753600120544435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,4,power_law_1.01,0.1447487950325012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,4,power_law_1.01,0.17598079442977904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,4,power_law_1.01,0.20730240345001222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,4,power_law_1.01,0.2659712076187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,4,power_law_1.01,0.3665087938308716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,4,power_law_1.01,0.4934976100921631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,4,power_law_1.01,0.7335360050201416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,32,power_law_1.01,0.09518719911575317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,32,power_law_1.01,0.08549119830131531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,32,power_law_1.01,0.08583040237426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,32,power_law_1.01,0.09285759925842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,32,power_law_1.01,0.09111679792404175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,32,power_law_1.01,0.08887040019035339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,32,power_law_1.01,0.08792960047721862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,32,power_law_1.01,0.09102720022201538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,32,power_law_1.01,0.089683198928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,32,power_law_1.01,0.09502720236778259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,32,power_law_1.01,0.09139840006828308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,32,power_law_1.01,0.09516800045967103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,32,power_law_1.01,0.09800320267677307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,32,power_law_1.01,0.09937279820442199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,32,power_law_1.01,0.10641920566558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,32,power_law_1.01,0.10942720174789429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,32,power_law_1.01,0.11679999828338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,32,power_law_1.01,0.13430399894714357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,32,power_law_1.01,0.1513216018676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,32,power_law_1.01,0.17968640327453614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,32,power_law_1.01,0.20382080078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,32,power_law_1.01,0.2684992074966431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,128,power_law_1.2,0.06416000127792358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,128,power_law_1.2,0.06353920102119445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,32,power_law_1.01,0.30138239860534666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,128,power_law_1.2,0.06209920048713684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,128,power_law_1.2,0.06467199921607972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,32,power_law_1.01,0.3792128086090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,128,power_law_1.2,0.06389759778976441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,32,power_law_1.01,0.509503984451294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,128,power_law_1.2,0.0646016001701355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,32,power_law_1.01,0.6878079891204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,128,power_law_1.2,0.06761599779129028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,32,power_law_1.01,0.8147904396057128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,128,power_law_1.2,0.0662015974521637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,128,power_law_1.2,0.06561279892921448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,128,power_law_1.2,0.06647679805755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,128,power_law_1.2,0.06557440161705017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,128,power_law_1.2,0.06992639899253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,128,power_law_1.2,0.07158399820327759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,128,power_law_1.2,0.07134720087051391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,128,power_law_1.2,0.07775999903678894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,128,power_law_1.2,0.08001279830932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,128,power_law_1.2,0.08520320057868958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,4,power_law_1.01,0.9014080047607422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,128,power_law_1.2,0.10193920135498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,128,power_law_1.2,0.11710079908370971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,128,power_law_1.2,0.14774399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,128,power_law_1.2,0.17087359428405763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,128,power_law_1.2,0.2293247938156128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,128,power_law_1.2,0.2808703899383545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,128,power_law_1.2,0.3946880102157593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,128,power_law_1.2,0.4966464042663574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,128,power_law_1.2,0.7312191963195801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,128,power_law_1.2,0.9970751762390136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,4,power_law_1.01,1.2275903701782227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,4,power_law_1.01,0.03273600041866302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,4,power_law_1.01,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,4,power_law_1.01,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,4,power_law_1.01,0.035488000512123107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,4,power_law_1.01,0.035596799850463864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,4,power_law_1.01,0.04376960098743439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,4,power_law_1.01,0.04679040014743805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,4,power_law_1.01,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,4,power_law_1.01,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,4,power_law_1.01,0.048876801133155824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,4,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,4,power_law_1.01,0.052876800298690796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,4,power_law_1.01,0.05422080159187317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,4,power_law_1.01,0.059436798095703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,4,power_law_1.01,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,4,power_law_1.01,0.07088000178337098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,4,power_law_1.01,0.07324159741401673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,4,power_law_1.01,0.08869760036468506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,4,power_law_1.01,0.1001471996307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,4,power_law_1.01,0.12148480415344239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,4,power_law_1.01,0.14511359930038453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,4,power_law_1.01,0.177183997631073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,4,power_law_1.01,0.22755839824676513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,4,power_law_1.01,0.34503679275512694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,8,balanced,0.08228800197442372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,8,balanced,0.08293866614500682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,8,balanced,0.08258666594823201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,8,balanced,0.08161599934101105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,8,balanced,0.08308266599973042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,8,balanced,0.08266666531562805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,8,balanced,0.08231466511885326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,8,balanced,0.08080533146858215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,8,balanced,0.08247999846935272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,8,balanced,0.08286400139331818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,8,balanced,0.08393067121505737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,8,balanced,0.08451199531555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,8,balanced,0.08346133430798848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,8,balanced,0.08616000413894653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,8,balanced,0.09673066933949788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,8,balanced,0.09545600414276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,8,balanced,0.09569600224494934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,8,balanced,0.1183093289534251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,8,balanced,0.11755200227101643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,8,balanced,0.13408533732096353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,8,balanced,0.1599146624406179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,8,balanced,0.19817600647608438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,8,balanced,0.24517865975697836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,8,balanced,0.33054399490356445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,8,balanced,0.40646398067474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,8,balanced,0.5836426814397176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,8,balanced,0.7721707026163737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,8,balanced,0.07291199763615926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,8,balanced,0.07238399982452393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,8,balanced,0.07770133515199025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,8,balanced,0.09300800164540608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,8,balanced,0.10827199618021648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,8,balanced,0.14645333091417947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,8,balanced,0.1667840083440145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,8,balanced,0.16499732931454977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,8,balanced,0.16756800810496011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,8,balanced,0.16605866948763529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,8,balanced,0.16790932416915894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,8,balanced,0.17390400171279907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,8,balanced,0.17482666174570718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,8,balanced,0.17723733186721802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,8,balanced,0.1871946652730306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,8,balanced,0.1881600022315979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,8,balanced,0.19352533419926962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,8,balanced,0.20618132750193277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,8,balanced,0.2129439910252889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,8,balanced,0.2345973253250122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,8,balanced,0.26123199860254925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,8,balanced,0.3107573390007019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,8,balanced,0.36295998096466064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,8,balanced,0.4731466770172119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,8,balanced,0.6321333249409994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,8,balanced,0.8479519685109457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,8,balanced,1.0997493267059326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,4,power_law_1.01,0.4444736003875732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,4,power_law_1.01,0.6083327770233155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,4,power_law_1.01,0.7033919811248779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,2,power_law_1.01,0.047366398572921756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,2,power_law_1.01,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,2,power_law_1.01,0.056959998607635495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,2,power_law_1.01,0.07184000015258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,2,power_law_1.01,0.08328319787979126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,2,power_law_1.01,0.09735680222511292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,2,power_law_1.01,0.10771839618682862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,2,power_law_1.01,0.11504640579223632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,2,power_law_1.01,0.11734399795532227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,2,power_law_1.01,0.11586560010910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,2,power_law_1.01,0.12584960460662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,2,power_law_1.01,0.1310464024543762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,2,power_law_1.01,0.1347391963005066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,2,power_law_1.01,0.1457152009010315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,2,power_law_1.01,0.15975040197372437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,2,power_law_1.01,0.166048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,2,power_law_1.01,0.16361600160598755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,2,power_law_1.01,0.18804479837417604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,2,power_law_1.01,0.2172032117843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,2,power_law_1.01,0.2655168056488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,2,power_law_1.01,0.31731839179992677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,2,power_law_1.01,0.40945281982421877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,2,power_law_1.01,0.5258560180664062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,2,power_law_1.01,0.7308032035827636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,2,power_law_1.01,0.9137727737426757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,2,power_law_1.01,1.2909184455871583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,2,power_law_1.01,1.7390655517578124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,64,power_law_1.01,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,64,power_law_1.01,0.03054080009460449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,64,power_law_1.01,0.027558401226997375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,64,power_law_1.01,0.030022400617599487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,64,power_law_1.01,0.030297601222991945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,64,power_law_1.01,0.0313728004693985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,64,power_law_1.01,0.031174400448799135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,128,power_law_1.01,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,64,power_law_1.01,0.03131519854068756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,128,power_law_1.01,0.033471998572349546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,64,power_law_1.01,0.03189760148525238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,128,power_law_1.01,0.03345920145511627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,64,power_law_1.01,0.03219200074672699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,128,power_law_1.01,0.03591040074825287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,128,power_law_1.01,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,128,power_law_1.01,0.03781760036945343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,128,power_law_1.01,0.037427198886871335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,128,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,128,power_law_1.01,0.037785598635673524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,128,power_law_1.01,0.03785600066184998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,128,power_law_1.01,0.03811840116977692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,128,power_law_1.01,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,64,power_law_1.01,0.03256320059299469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,128,power_law_1.01,0.03935360014438629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,64,power_law_1.01,0.03325439989566803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,128,power_law_1.01,0.040268799662590025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,64,power_law_1.01,0.03383679986000061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,128,power_law_1.01,0.042719998955726625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,64,power_law_1.01,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,128,power_law_1.01,0.04325760006904602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,64,power_law_1.01,0.03950720131397247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,128,power_law_1.01,0.044659200310707095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,64,power_law_1.01,0.03949440121650696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,128,power_law_1.01,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,64,power_law_1.01,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,64,power_law_1.01,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,64,power_law_1.01,0.047635200619697574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,64,power_law_1.01,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,64,power_law_1.01,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,64,power_law_1.01,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,64,power_law_1.01,0.09624959826469422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,64,power_law_1.01,0.1394368052482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,128,power_law_1.01,0.048172798752784726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,64,power_law_1.01,0.148198401927948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,128,power_law_1.01,0.05303679704666138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,64,power_law_1.01,0.19920639991760253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,128,power_law_1.01,0.05767679810523987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,128,power_law_1.01,0.06883199810981751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,64,power_law_1.01,0.25462400913238525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,128,power_law_1.01,0.08791679739952088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,128,power_law_1.01,0.12290560007095337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,128,power_law_1.01,0.1513983964920044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,128,power_law_1.01,0.20959360599517823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,128,power_law_1.01,0.27242240905761717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,8,power_law_1.01,0.023686400055885314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,8,power_law_1.01,0.023270399868488313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,8,power_law_1.01,0.0226623997092247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,8,power_law_1.01,0.02220800071954727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,8,power_law_1.01,0.02330880016088486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,8,power_law_1.01,0.026022401452064515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,8,power_law_1.01,0.03397760093212128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,8,power_law_1.01,0.029158401489257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,8,power_law_1.01,0.03868800103664398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,8,power_law_1.01,0.033888000249862674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,8,power_law_1.01,0.038022398948669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,8,power_law_1.01,0.0585536003112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,16,power_law_1.2,0.0441536009311676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,8,power_law_1.01,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,8,power_law_1.01,0.047328001260757445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,16,power_law_1.2,0.06097919940948486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,16,power_law_1.2,0.05451520085334778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,16,power_law_1.2,0.059059202671051025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,16,power_law_1.2,0.06407039761543273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,16,power_law_1.2,0.058937597274780276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,16,power_law_1.2,0.0626688003540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,16,power_law_1.2,0.0605184018611908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,16,power_law_1.2,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,16,power_law_1.2,0.05936639904975891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,16,power_law_1.2,0.06296319961547851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,16,power_law_1.2,0.06629120111465454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,16,power_law_1.2,0.06783999800682068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,16,power_law_1.2,0.0709119975566864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,16,power_law_1.2,0.07728639841079712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,8,power_law_1.01,0.06783999800682068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,8,power_law_1.01,0.05493760108947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,8,power_law_1.01,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,16,power_law_1.2,0.07690879702568054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,16,power_law_1.2,0.07931519746780395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,16,power_law_1.2,0.08909440040588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,8,power_law_1.01,0.08398720026016235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,16,power_law_1.2,0.09059200286865235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,16,power_law_1.2,0.10661120414733886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,16,power_law_1.2,0.11958400011062623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,16,power_law_1.2,0.1450495958328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,16,power_law_1.2,0.17788159847259521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,16,power_law_1.2,0.2381376028060913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,16,power_law_1.2,0.28881919384002686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.01,0.09919360280036926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,2,power_law_1.01,0.045337599515914914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,2,power_law_1.01,0.04748800098896026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,2,power_law_1.01,0.06225919723510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,2,power_law_1.01,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,2,power_law_1.01,0.07009279727935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,2,power_law_1.01,0.07180799841880799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,2,power_law_1.01,0.07266560196876526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,2,power_law_1.01,0.07344639897346497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,16,power_law_1.2,0.42011518478393556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,16,power_law_1.2,0.5211584091186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.01,0.1374079942703247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,2,power_law_1.01,0.07715200185775757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,2,power_law_1.01,0.07389439940452576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,2,power_law_1.01,0.0797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,2,power_law_1.01,0.08364800214767457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,2,power_law_1.01,0.08452479839324951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,2,power_law_1.01,0.08683519959449768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,2,power_law_1.01,0.09872639775276185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,2,power_law_1.01,0.10084480047225952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,2,power_law_1.01,0.11013760566711425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,2,power_law_1.01,0.1344256043434143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.01,0.14316799640655517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.01,0.17151999473571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.01,0.1701375961303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.01,0.23916161060333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.01,0.22041599750518798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.01,0.2992511987686157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.01,0.4342912197113037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.01,0.2653439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.01,0.32203519344329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.01,0.5620607852935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.01,0.48738560676574705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,4,balanced,0.04613866905371348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,4,balanced,0.04562133550643921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,4,balanced,0.05019199848175049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,4,balanced,0.05176533261934916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,4,balanced,0.05789866546789805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,4,balanced,0.05420266588528951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,4,balanced,0.05197333296140035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,4,balanced,0.0531626691420873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,4,balanced,0.051957334081331887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,4,balanced,0.054570664962132774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,4,balanced,0.05242133140563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,4,balanced,0.049653331438700356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,4,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,4,balanced,0.05217599868774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,4,balanced,0.05645333230495453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,4,balanced,0.05989866455396017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,4,balanced,0.06228266656398773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,4,balanced,0.07944533228874207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,4,balanced,0.0844586690266927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,4,balanced,0.10556800166765849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,4,balanced,0.12236266334851582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,4,balanced,0.15845866998036703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,4,balanced,0.19091200828552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,4,balanced,0.2667893369992574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,4,balanced,0.3340640068054199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,4,balanced,0.4795946677525838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,4,balanced,0.6221973498662313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.01,0.6095232009887696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.01,0.8286335945129395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.01,0.8714240074157715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.01,1.0936320304870606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.01,1.2848575592041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,8,power_law_1.01,0.01693439930677414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,8,power_law_1.01,0.01674239933490753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,8,power_law_1.01,0.01664000004529953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,8,power_law_1.01,0.017312000691890716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,8,power_law_1.01,0.017843200266361235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,8,power_law_1.01,0.01812479943037033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,8,power_law_1.01,0.018303999304771425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,8,power_law_1.01,0.020351999998092653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,8,power_law_1.01,0.02112639993429184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,8,power_law_1.01,0.021030400693416596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,8,power_law_1.01,0.029267200827598573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,8,power_law_1.01,0.03136639893054962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,8,power_law_1.01,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,8,power_law_1.01,0.030534398555755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,8,power_law_1.01,0.031411200761795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,8,power_law_1.01,0.031699201464653014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,8,power_law_1.01,0.03301759958267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,8,power_law_1.01,0.03826560080051422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.01,0.04028159976005554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.01,0.05386880040168762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.01,0.059680002927780154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,16,balanced,0.02218666672706604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,16,balanced,0.0230880007147789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,16,balanced,0.023120000958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,16,balanced,0.02311466634273529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,16,balanced,0.02516799916823705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,16,balanced,0.02759466568628947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,16,balanced,0.029290666182835896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,16,balanced,0.029301332930723827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,16,balanced,0.03844800094763438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,16,balanced,0.03779733429352442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,16,balanced,0.038015998899936676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,16,balanced,0.06089066465695699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,16,balanced,0.06145066519578298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,16,balanced,0.06237866481145223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,16,balanced,0.059903999169667564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,16,balanced,0.06321066617965698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,16,balanced,0.06702400247255962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,16,balanced,0.07588799794514973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,16,balanced,0.08250133196512859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,16,balanced,0.09488532940546672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,16,balanced,0.11659733454386394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,16,balanced,0.14478933811187744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,16,balanced,0.18035733699798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,16,balanced,0.22957332928975424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,16,balanced,0.2898400028546651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,16,balanced,0.43116267522176105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,1,power_law_1.2,0.03147520124912262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,16,balanced,0.5560373465220133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.01,0.07877119779586791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,1,power_law_1.2,0.03489919900894165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,1,power_law_1.2,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,1,power_law_1.2,0.05435519814491272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,1,power_law_1.2,0.05497599840164184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,1,power_law_1.2,0.055871999263763426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,1,power_law_1.2,0.05813760161399841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,1,power_law_1.2,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,1,power_law_1.2,0.06156799793243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,1,power_law_1.2,0.06232960224151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,1,power_law_1.2,0.06188160181045532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,1,power_law_1.2,0.06738560199737549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,1,power_law_1.2,0.06570879817008972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,1,power_law_1.2,0.07103360295295716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,1,power_law_1.2,0.08330879807472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,1,power_law_1.2,0.08711680173873901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,1,power_law_1.2,0.09936000108718872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,1,power_law_1.2,0.11552000045776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,1,power_law_1.2,0.1255295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.01,0.09564160108566284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,1,power_law_1.2,0.157043194770813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,1,power_law_1.2,0.18291840553283692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.01,0.12483839988708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,1,power_law_1.2,0.2506239891052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,1,power_law_1.2,0.30522239208221436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,32,2,power_law_1.01,0.04694400131702423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,32,2,power_law_1.01,0.0502016007900238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,32,2,power_law_1.01,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,32,2,power_law_1.01,0.06372479796409607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,32,2,power_law_1.01,0.07452800273895263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.01,0.15791360139846802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,32,2,power_law_1.01,0.0852735996246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,32,2,power_law_1.01,0.09931520223617554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,32,2,power_law_1.01,0.10145280361175538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,32,2,power_law_1.01,0.10448640584945679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,32,2,power_law_1.01,0.10659840106964111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,32,2,power_law_1.01,0.1084928035736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,32,2,power_law_1.01,0.11684479713439941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,32,2,power_law_1.01,0.12124799489974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,32,2,power_law_1.01,0.1270848035812378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,32,2,power_law_1.01,0.1406592011451721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,32,2,power_law_1.01,0.14624639749526977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,32,2,power_law_1.01,0.14545919895172119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,32,2,power_law_1.01,0.1715008020401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,32,2,power_law_1.01,0.19949439764022828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,32,2,power_law_1.01,0.24425599575042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,32,2,power_law_1.01,0.29640960693359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,32,2,power_law_1.01,0.4076416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,32,2,power_law_1.01,0.4988607883453369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,1,power_law_1.2,0.4268799781799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,32,2,power_law_1.01,0.7056640148162842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,32,2,power_law_1.01,0.8868927955627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,32,2,power_law_1.01,1.2409215927124024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,32,2,power_law_1.01,1.6595264434814454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,1,power_law_1.2,0.5523263931274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.01,0.22187519073486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,1,power_law_1.2,0.7823232173919678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.01,0.2854399919509888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,16,4,power_law_1.2,0.06935039758682252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,16,4,power_law_1.2,0.07674239873886109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,16,4,power_law_1.2,0.0836031973361969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,16,4,power_law_1.2,0.09573119878768921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,16,4,power_law_1.2,0.10743680000305175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,16,4,power_law_1.2,0.11042560338973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,16,4,power_law_1.2,0.11934080123901367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,16,4,power_law_1.2,0.1174015998840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,16,4,power_law_1.2,0.11994880437850952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,16,4,power_law_1.2,0.12319359779357911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,16,4,power_law_1.2,0.121670401096344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,16,4,power_law_1.2,0.12699520587921143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,16,4,power_law_1.2,0.1292799949645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,16,4,power_law_1.2,0.13354239463806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,16,4,power_law_1.2,0.13848320245742798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,16,4,power_law_1.2,0.14108799695968627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,16,4,power_law_1.2,0.1467136025428772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,16,4,power_law_1.2,0.1662783980369568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,16,4,power_law_1.2,0.18143999576568604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,16,4,power_law_1.2,0.23399679660797118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,16,4,power_law_1.2,0.2819391965866089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,16,4,power_law_1.2,0.36970880031585696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,16,4,power_law_1.2,0.4311552047729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,16,4,power_law_1.2,0.6568831920623779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,16,4,power_law_1.2,0.9041407585144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,16,4,power_law_1.2,1.18088960647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,16,4,power_law_1.2,1.6109888076782226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,1,power_law_1.2,1.0291584014892579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,8,power_law_1.2,0.05116159915924072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,8,power_law_1.2,0.06840959787368775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,8,power_law_1.2,0.07310720086097718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,8,power_law_1.2,0.08303999900817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,8,power_law_1.2,0.08979200124740601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,8,power_law_1.2,0.09165440201759338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,8,power_law_1.2,0.09404799938201905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,8,power_law_1.2,0.10097279548645019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,8,power_law_1.2,0.10183039903640748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,8,power_law_1.2,0.10254080295562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,8,power_law_1.2,0.10246399641036988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,8,power_law_1.2,0.10429439544677735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,8,power_law_1.2,0.10776959657669068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,8,power_law_1.2,0.11856640577316284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,8,power_law_1.2,0.12543359994888306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,8,power_law_1.2,0.12562559843063353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,8,power_law_1.2,0.1375488042831421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,8,power_law_1.2,0.15343999862670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,8,power_law_1.2,0.17159680128097535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,8,power_law_1.2,0.20941441059112548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,8,power_law_1.2,0.23374080657958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,8,power_law_1.2,0.298905611038208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,8,power_law_1.2,0.39704959392547606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,8,power_law_1.2,0.5081600189208985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,8,power_law_1.2,0.7258240222930908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,8,power_law_1.2,1.042246437072754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,8,power_law_1.2,1.269542407989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,16,power_law_1.01,0.12262400388717651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,16,power_law_1.01,0.17875200510025024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,16,power_law_1.01,0.16903680562973022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,16,power_law_1.01,0.1769600033760071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,16,power_law_1.01,0.17798399925231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,16,power_law_1.01,0.150271999835968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,16,power_law_1.01,0.16457600593566896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,16,power_law_1.01,0.1693824052810669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,16,power_law_1.01,0.1721984028816223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,16,power_law_1.01,0.16614400148391723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,16,power_law_1.01,0.17842559814453124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,16,power_law_1.01,0.18241920471191406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,16,power_law_1.01,0.17933440208435059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,16,power_law_1.01,0.1777791976928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,16,power_law_1.01,0.18599679470062255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,16,power_law_1.01,0.20599679946899413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,16,power_law_1.01,0.20575358867645263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,16,power_law_1.01,0.22295680046081542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,16,power_law_1.01,0.23376638889312745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,16,power_law_1.01,0.2561856031417847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,16,power_law_1.01,0.3033407926559448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,16,power_law_1.01,0.35557119846343993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,16,power_law_1.01,0.4216447830200195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,16,power_law_1.01,0.6047167778015137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,16,power_law_1.01,0.7605055809020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,16,power_law_1.01,1.0903039932250977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,16,power_law_1.01,1.5542400360107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,32,power_law_1.2,0.025536000728607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,32,power_law_1.2,0.024928000569343568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,32,power_law_1.2,0.026713600754737853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,32,power_law_1.2,0.0239424005150795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,32,power_law_1.2,0.024275200068950654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,32,power_law_1.2,0.0271807998418808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,32,power_law_1.2,0.027526399493217467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,32,power_law_1.2,0.03624959886074066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,32,power_law_1.2,0.0383103996515274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,32,power_law_1.2,0.03695360124111176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,32,power_law_1.2,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,32,power_law_1.2,0.05441920161247253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,32,power_law_1.2,0.05533440113067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,32,power_law_1.2,0.05759360194206238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,32,power_law_1.2,0.06714879870414733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,32,power_law_1.2,0.07726719975471497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,32,power_law_1.2,0.08242560029029847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,32,power_law_1.2,0.09870079755783082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.2,0.1310271978378296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.2,0.15616639852523803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.2,0.0997376024723053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.2,0.12387839555740357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.2,0.16001919507980347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.2,0.2257024049758911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.2,0.30320000648498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.2,0.4535679817199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.2,0.5837759971618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,128,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,128,power_law_1.01,0.047513601183891294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,128,power_law_1.01,0.04775680005550385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,128,power_law_1.01,0.049798399209976196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,128,power_law_1.01,0.050419199466705325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,128,power_law_1.01,0.051072001457214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,128,power_law_1.01,0.05118079781532288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,128,power_law_1.01,0.05137919783592224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,128,power_law_1.01,0.05144960284233093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,128,power_law_1.01,0.05225600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,128,power_law_1.01,0.05242879986763001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,128,power_law_1.01,0.05423359870910645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,128,power_law_1.01,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,128,power_law_1.01,0.05541759729385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,128,power_law_1.01,0.059564799070358276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,128,power_law_1.01,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,128,power_law_1.01,0.06212480068206787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,128,power_law_1.01,0.06899200081825256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,128,power_law_1.01,0.07482879757881164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,128,power_law_1.01,0.08648319840431214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,128,power_law_1.01,0.09800320267677307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,128,power_law_1.01,0.11960320472717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,128,power_law_1.01,0.14174720048904418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,128,power_law_1.01,0.19169919490814208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,128,power_law_1.01,0.23543040752410888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,128,power_law_1.01,0.32277119159698486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,128,power_law_1.01,0.4061312198638916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,16,balanced,0.04784533381462097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,16,balanced,0.050901333491007485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,16,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,16,balanced,0.08131733536720276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,16,balanced,0.09985066453615825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,16,balanced,0.11007466912269592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,16,balanced,0.11075733105341594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,16,balanced,0.10998400052388509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,16,balanced,0.11104533076286316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,16,balanced,0.11184533437093098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,16,balanced,0.11239467064539592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,16,balanced,0.11555199821790059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,16,balanced,0.11702932914098103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,16,balanced,0.11794132987658183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,16,balanced,0.12377066413561504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,16,balanced,0.12611200412114462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,16,balanced,0.13179199894269308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,16,balanced,0.14429333806037903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,16,balanced,0.15422399838765463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,16,balanced,0.19586666425069174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,16,balanced,0.2071946660677592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,16,balanced,0.29076266288757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,16,balanced,0.2783786654472351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,16,balanced,0.3858506679534912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,16,balanced,0.43585066000620526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,16,balanced,0.6136426528294882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,16,balanced,0.7279626528422037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,4,power_law_1.2,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,4,power_law_1.2,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,4,power_law_1.2,0.03403519988059998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,4,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,4,power_law_1.2,0.03633280098438263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,4,power_law_1.2,0.03691520094871521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,4,power_law_1.2,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,4,power_law_1.2,0.03779839873313904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,4,power_law_1.2,0.03826560080051422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,4,power_law_1.2,0.03933440148830414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,4,power_law_1.2,0.03989759981632233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,4,power_law_1.2,0.04099839925765991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,4,power_law_1.2,0.041017600893974306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,4,balanced,0.0740586668252945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,4,balanced,0.09477333227793376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,4,balanced,0.12763733665148416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,4,balanced,0.2115359902381897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,4,balanced,0.33908267815907794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,4,balanced,0.5622400045394897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,4,balanced,0.5485333204269409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,4,balanced,0.5536479949951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,4,balanced,0.5273280143737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,4,balanced,0.5218453407287598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,4,balanced,0.5110079844792684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,4,balanced,0.4980586767196655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,4,power_law_1.2,0.04245760142803192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,4,power_law_1.2,0.04748159945011139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,4,power_law_1.2,0.04958719909191132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,4,balanced,0.5314613183339437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,4,power_law_1.2,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,4,balanced,0.5374506711959839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,4,power_law_1.2,0.05713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,4,balanced,0.5330613454182943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,4,balanced,0.5502986510594686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,4,balanced,0.5106240113576254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,4,balanced,0.5490133364995321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,4,balanced,0.5565546751022339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.2,0.0641215980052948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,4,balanced,0.5602026780446371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,4,balanced,0.5666559934616089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.2,0.07973759770393371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,4,balanced,0.5966773430506388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,4,balanced,0.6470933357874552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,4,balanced,1.090666691462199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,4,balanced,1.0887733300526936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,4,balanced,1.5100053151448567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,4,balanced,1.9684267044067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.2,0.09644799828529357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.2,0.12778879404067994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,1,power_law_1.01,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.2,0.1565951943397522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,1,power_law_1.01,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,1,power_law_1.01,0.049055999517440795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,1,power_law_1.01,0.06902400255203248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,1,power_law_1.01,0.07992320060729981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,1,power_law_1.01,0.11200000047683716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,1,power_law_1.01,0.11521919965744018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,1,power_law_1.01,0.1195072054862976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,1,power_law_1.01,0.12201600074768067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,1,power_law_1.01,0.12585599422454835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,1,power_law_1.01,0.13063039779663085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,1,power_law_1.01,0.13630720376968383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,1,power_law_1.01,0.13738880157470704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,1,power_law_1.01,0.1408192038536072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,1,power_law_1.01,0.14661120176315307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,1,power_law_1.01,0.14856319427490233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,1,power_law_1.01,0.15797760486602783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,1,power_law_1.01,0.17521920204162597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,1,power_law_1.01,0.19267840385437013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,1,power_law_1.01,0.23412480354309081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,1,power_law_1.01,0.2566015958786011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,1,power_law_1.01,0.3324352025985718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,1,power_law_1.01,0.4136064052581787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,1,power_law_1.01,0.5501247882843018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,1,power_law_1.01,0.6888383865356446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,1,power_law_1.01,0.9647295951843262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,1,power_law_1.01,1.2446656227111816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.2,0.20647039413452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,4,power_law_1.2,0.08579840064048767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,4,power_law_1.2,0.089273601770401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,4,power_law_1.2,0.10440319776535034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,4,power_law_1.2,0.11490559577941895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,4,power_law_1.2,0.13614720106124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,4,power_law_1.2,0.15545599460601806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,4,power_law_1.2,0.1624575972557068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,4,power_law_1.2,0.16455039978027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,4,power_law_1.2,0.15689599514007568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,4,power_law_1.2,0.15945600271224974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,4,power_law_1.2,0.15338879823684692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,4,power_law_1.2,0.1637503981590271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,4,power_law_1.2,0.1633471965789795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,4,power_law_1.2,0.16562559604644775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,4,power_law_1.2,0.17523200511932374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,4,power_law_1.2,0.1821887969970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,4,power_law_1.2,0.18619519472122192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,4,power_law_1.2,0.2071615934371948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.2,0.22447359561920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.2,0.2700927972793579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.2,0.32822399139404296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.2,0.4130943775177002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.2,0.478169584274292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.2,0.6276671886444092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.2,0.8652799606323243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.2,0.2363136053085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.2,1.216486358642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.2,1.4920255661010742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.2,0.3309567928314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,2,power_law_1.2,0.12486399412155151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,2,power_law_1.2,0.16308480501174927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,2,power_law_1.2,0.2307647943496704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,2,power_law_1.2,0.30855679512023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,2,power_law_1.2,0.40755839347839357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,2,power_law_1.2,0.671833610534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,2,power_law_1.2,0.7197824001312256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,2,power_law_1.2,0.7234432220458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,2,power_law_1.2,0.7518464088439941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,2,power_law_1.2,0.7543935775756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,2,power_law_1.2,0.7918015956878662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,2,power_law_1.2,0.8242688179016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,2,power_law_1.2,0.8208640098571778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,2,power_law_1.2,0.8534784317016602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,2,power_law_1.2,0.8883328437805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,2,power_law_1.2,0.9254015922546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,2,power_law_1.2,0.9819775581359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,2,power_law_1.2,1.0905728340148926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,2,power_law_1.2,1.0956031799316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,2,power_law_1.2,1.2444095611572266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,2,power_law_1.2,1.3581055641174316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,2,power_law_1.2,1.589247989654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,2,power_law_1.2,1.7357887268066405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,2,power_law_1.2,2.158323287963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,2,power_law_1.2,2.6851648330688476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,2,power_law_1.2,3.5769279479980467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.2,0.4717887878417969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,2,power_law_1.2,4.098239898681641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,32,2,power_law_1.01,0.04744960069656372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,32,2,power_law_1.01,0.051846402883529666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,32,2,power_law_1.01,0.05434240102767944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,32,2,power_law_1.01,0.06669440269470214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,32,2,power_law_1.01,0.07932159900665284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,32,2,power_law_1.01,0.09460480213165283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,32,2,power_law_1.01,0.1109503984451294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,32,2,power_law_1.01,0.1083840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,32,2,power_law_1.01,0.114028799533844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,32,2,power_law_1.01,0.11444480419158935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,32,2,power_law_1.01,0.11764479875564575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,32,2,power_law_1.01,0.11932799816131592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,32,2,power_law_1.01,0.12151039838790893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,32,2,power_law_1.01,0.12988799810409546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,32,2,power_law_1.01,0.13216639757156373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,32,2,power_law_1.01,0.13995519876480103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,32,2,power_law_1.01,0.14954240322113038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,32,2,power_law_1.01,0.17545599937438966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,32,2,power_law_1.01,0.19629440307617188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,32,2,power_law_1.01,0.24428160190582277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,32,2,power_law_1.01,0.2882623910903931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,32,2,power_law_1.01,0.36401278972625734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,32,2,power_law_1.01,0.4628352165222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,32,2,power_law_1.01,0.6210879802703857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,32,2,power_law_1.01,0.8288512229919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,32,2,power_law_1.01,1.1283007621765138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,32,2,power_law_1.01,1.5186688423156738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,32,balanced,0.05046933392683665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,32,balanced,0.050661335388819374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,32,balanced,0.05041066805521647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,32,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,32,balanced,0.05418666700522105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,32,balanced,0.06193066636721293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,32,balanced,0.07044266661008199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,32,balanced,0.0717439999183019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,32,balanced,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,32,balanced,0.07299200197060902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,32,balanced,0.07306666672229767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,32,balanced,0.0766186664501826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,32,balanced,0.0759626676638921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,32,balanced,0.07923200229803722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,32,balanced,0.08900800347328186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,32,balanced,0.08917333682378133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,32,balanced,0.09572800000508626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,32,balanced,0.11243733763694763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,32,balanced,0.11383466919263203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,32,balanced,0.1378720005353292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,32,balanced,0.15442132949829102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,32,balanced,0.19267733891805014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,32,balanced,0.2254719932874044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,32,balanced,0.29053332408269245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,32,balanced,0.3638506730397542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,32,balanced,0.5030666589736938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,32,balanced,0.6487679878870646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,8,1,power_law_1.2,0.03772160112857818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,8,1,power_law_1.2,0.05004159808158874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,8,1,power_law_1.2,0.05384320020675659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,8,1,power_law_1.2,0.07559040188789368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,8,1,power_law_1.2,0.0757695972919464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,8,1,power_law_1.2,0.07993599772453308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,8,1,power_law_1.2,0.08076800107955932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,8,1,power_law_1.2,0.0806335985660553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,8,1,power_law_1.2,0.08234879970550538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,8,1,power_law_1.2,0.08467199802398681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,8,1,power_law_1.2,0.08520320057868958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,8,1,power_law_1.2,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,8,1,power_law_1.2,0.09162880182266235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,8,1,power_law_1.2,0.09324799776077271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,8,1,power_law_1.2,0.10705920457839965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,8,1,power_law_1.2,0.11134719848632812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,8,1,power_law_1.2,0.11975680589675904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,8,1,power_law_1.2,0.13745280504226684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,8,1,power_law_1.2,0.15639040470123292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,8,1,power_law_1.2,0.20174078941345214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,8,1,power_law_1.2,0.24143359661102295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,8,1,power_law_1.2,0.3207040071487427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,8,1,power_law_1.2,0.39079039096832274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,4,power_law_1.2,0.036748799681663516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,4,power_law_1.2,0.034016001224517825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,4,power_law_1.2,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,4,power_law_1.2,0.05427839756011963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,4,power_law_1.2,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,4,power_law_1.2,0.06973440051078797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,4,power_law_1.2,0.07133439779281617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,4,power_law_1.2,0.07331839799880982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,4,power_law_1.2,0.07352319955825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,4,power_law_1.2,0.0725823998451233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,4,power_law_1.2,0.07511039972305297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,4,power_law_1.2,0.07705600261688232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,4,power_law_1.2,0.07872639894485474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,4,power_law_1.2,0.08149759769439698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,4,power_law_1.2,0.08902400135993957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,4,power_law_1.2,0.09163519740104675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,4,power_law_1.2,0.09221760034561158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,4,power_law_1.2,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,4,power_law_1.2,0.11797120571136474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,4,power_law_1.2,0.13968640565872192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,4,power_law_1.2,0.13903360366821288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,4,power_law_1.2,0.17783039808273315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,4,power_law_1.2,0.20638720989227294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,8,1,power_law_1.2,0.5400832176208497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,4,power_law_1.2,0.26571519374847413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,4,power_law_1.2,0.3575999975204468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,8,1,power_law_1.2,0.6859776020050049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,4,power_law_1.2,0.46971521377563474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,4,power_law_1.2,0.6685376167297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,8,balanced,0.04637333254019419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,8,balanced,0.06297599772612254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,8,balanced,0.08688533306121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,8,balanced,0.13586666186650595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,8,balanced,0.21926399072011313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,8,power_law_1.01,0.07523840069770812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,8,balanced,0.3848533233006795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,8,balanced,0.3875253200531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,8,power_law_1.01,0.09670400023460388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,8,power_law_1.01,0.10231679677963257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,8,power_law_1.01,0.12309759855270386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,8,1,power_law_1.2,0.9774911880493165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,8,power_law_1.01,0.12448639869689941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,8,power_law_1.01,0.12447359561920165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,8,balanced,0.39185067017873126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,8,power_law_1.01,0.12985600233078004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,8,balanced,0.3884426752726237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,8,balanced,0.3923786481221517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,8,power_law_1.01,0.12743040323257446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,8,balanced,0.3906666835149129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,8,balanced,0.3946666717529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,8,power_law_1.01,0.12737280130386353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,8,balanced,0.3972426652908325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,8,balanced,0.39988799889882404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,8,power_law_1.01,0.1270591974258423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,8,balanced,0.40694932142893475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,8,balanced,0.4059893290201823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,8,power_law_1.01,0.13534079790115355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,8,power_law_1.01,0.13308160305023192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,8,power_law_1.01,0.13943040370941162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,8,power_law_1.01,0.14315520524978637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,8,power_law_1.01,0.14759680032730102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,8,balanced,0.409498651822408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,8,power_law_1.01,0.148307204246521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,8,balanced,0.4174079895019531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,8,balanced,0.42851734161376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,8,power_law_1.01,0.1522304058074951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,8,balanced,0.4469333489735921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,8,power_law_1.01,0.16078720092773438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,8,balanced,0.47414398193359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,8,balanced,0.5057013432184855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,8,power_law_1.01,0.1710528016090393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,8,balanced,0.5619093179702759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,8,power_law_1.01,0.19328000545501708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,8,power_law_1.01,0.21336960792541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,8,power_law_1.01,0.2661504030227661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,8,power_law_1.01,0.32652161121368406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,8,power_law_1.01,0.4242688179016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,8,power_law_1.01,0.54269437789917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,8,power_law_1.01,0.8202943801879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,8,power_law_1.01,1.065452766418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,8,balanced,0.6339573462804159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,8,balanced,0.6941226323445638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,8,balanced,0.9830719629923502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,8,balanced,1.08515731493632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,8,1,power_law_1.2,1.274675178527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,128,balanced,0.02258666604757309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,128,balanced,0.021354667842388153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,128,balanced,0.02346133440732956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,128,balanced,0.025381334125995636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,128,balanced,0.03376533339420954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,128,balanced,0.052485331892967224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,128,balanced,0.05137600004673004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,128,balanced,0.05019199848175049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,128,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,128,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,128,balanced,0.048453330993652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,128,balanced,0.045781334241231285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,128,balanced,0.04307733476161957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,128,balanced,0.03990400085846583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,128,balanced,0.05667733152707418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,128,balanced,0.05658666789531708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,128,balanced,0.04570133487383524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,128,balanced,0.06135466694831848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,128,balanced,0.06975999971230824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,128,balanced,0.09240000446637471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,128,balanced,0.11315733194351196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,128,balanced,0.15657066305478415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,128,balanced,0.19445333878199259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,128,balanced,0.27774399518966675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,128,balanced,0.36152533690134686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,128,balanced,0.528714656829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,128,balanced,0.6966293652852377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,1,power_law_1.01,0.07929599881172181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,1,power_law_1.01,0.09964159727096558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,1,power_law_1.01,0.1221824049949646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,1,power_law_1.01,0.18533120155334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,1,power_law_1.01,0.2394047975540161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,1,power_law_1.01,0.363155198097229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,1,power_law_1.01,0.38646399974823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,1,power_law_1.01,0.4015552043914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,1,power_law_1.01,0.41324801445007325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,1,power_law_1.01,0.42668800354003905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,1,power_law_1.01,0.4385727882385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,1,power_law_1.01,0.45113601684570315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,1,power_law_1.01,0.46136960983276365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,1,power_law_1.01,0.4772223949432373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,1,power_law_1.01,0.5065216064453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,1,power_law_1.01,0.5239295959472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,1,power_law_1.01,0.5613632202148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,1,power_law_1.01,0.6415359973907471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,1,power_law_1.01,0.7187392234802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,1,power_law_1.01,0.8841792106628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,1,power_law_1.01,1.115065574645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,1,power_law_1.01,1.4136384010314942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,1,power_law_1.01,1.7101823806762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,1,power_law_1.01,2.3231807708740235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,1,power_law_1.01,2.769273567199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,1,power_law_1.01,3.8988094329833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,1,power_law_1.01,5.213151931762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,1,power_law_1.01,0.13587199449539183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,1,power_law_1.01,0.203385591506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,1,power_law_1.01,0.3225152015686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,1,power_law_1.01,0.5380991935729981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,1,power_law_1.01,0.7456511974334716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,1,power_law_1.01,1.2258687973022462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,1,power_law_1.01,1.3009663581848145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,1,power_law_1.01,1.3677311897277833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,1,power_law_1.01,1.407801628112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,1,power_law_1.01,1.4403136253356934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,1,power_law_1.01,1.4687935829162597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,1,power_law_1.01,1.5202431678771973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,1,power_law_1.01,1.5362239837646485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,1,power_law_1.01,1.5852543830871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,1,power_law_1.01,1.6310335159301759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,1,power_law_1.01,1.696678352355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,1,power_law_1.01,1.771072006225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,1,power_law_1.01,1.9316415786743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,1,power_law_1.01,2.1206592559814452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,1,power_law_1.01,2.4830015182495115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,1,power_law_1.01,2.470009613037109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,1,power_law_1.01,3.1166080474853515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,1,power_law_1.01,3.179769515991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,1,power_law_1.01,3.878867340087891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,1,power_law_1.01,5.185990524291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,1,power_law_1.01,6.744972991943359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,1,power_law_1.01,8.037344360351563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,4,power_law_1.2,0.1137984037399292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,4,power_law_1.2,0.14361599683761597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,4,power_law_1.2,0.15135999917984008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,4,power_law_1.2,0.18350720405578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,4,power_law_1.2,0.18415360450744628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,4,power_law_1.2,0.17569279670715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,4,power_law_1.2,0.17313920259475707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,4,power_law_1.2,0.17848960161209107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,4,power_law_1.2,0.16229759454727172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,4,power_law_1.2,0.16388479471206666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,4,power_law_1.2,0.1583295941352844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,4,power_law_1.2,0.1916416049003601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,4,power_law_1.2,0.20225920677185058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,4,power_law_1.2,0.1914944052696228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,4,power_law_1.2,0.2185663938522339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,1,balanced,0.09609599908192952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,4,power_law_1.2,0.21473920345306396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,4,power_law_1.2,0.25068800449371337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,4,power_law_1.2,0.33019518852233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,1,balanced,0.10373866558074951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,1,balanced,0.1272586683432261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,1,balanced,0.17618133624394736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,1,balanced,0.2577280004819234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.2,0.42430720329284666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,1,balanced,0.4248693386713664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,1,balanced,0.5901439984639486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,1,balanced,0.5807520151138306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,1,balanced,0.5853813489278158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,1,balanced,0.585536003112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,1,balanced,0.5879093408584595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,1,balanced,0.6006079912185669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,1,balanced,0.6054933468500773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,1,balanced,0.613589326540629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,1,balanced,0.6252640088399252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,1,balanced,0.6301706631978353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,1,balanced,0.6496586799621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,1,balanced,0.7011840343475342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,1,balanced,0.7356853485107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,1,balanced,0.817909320195516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,1,balanced,0.8889919916788737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,1,balanced,1.0580373605092366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,1,balanced,1.266309340794881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,1,balanced,1.638389269510905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,1,balanced,2.1471360524495444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.2,0.5025856018066406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,1,balanced,2.8924694061279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,1,balanced,3.957845369974772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.2,0.6350848197937011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,4,power_law_1.2,0.02707200050354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,4,power_law_1.2,0.03270399868488312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,4,power_law_1.2,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,4,power_law_1.2,0.05324159860610962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,4,power_law_1.2,0.05766400098800659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,4,power_law_1.2,0.06330239772796631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,4,power_law_1.2,0.0695039987564087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,4,power_law_1.2,0.07019519805908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,4,power_law_1.2,0.0760703980922699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,4,power_law_1.2,0.0763647973537445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,4,power_law_1.2,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,4,power_law_1.2,0.1004480004310608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,4,power_law_1.2,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,4,power_law_1.2,0.09363840222358703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,4,power_law_1.2,0.1419584035873413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,4,power_law_1.2,0.1499392032623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.2,0.9220288276672364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,4,power_law_1.2,0.16062719821929933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,4,power_law_1.2,0.1385472059249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.2,0.1383039951324463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.2,0.14596480131149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.2,0.1668287992477417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.2,0.20120959281921386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.2,0.2395008087158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,16,power_law_1.2,0.05548160076141358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,16,power_law_1.2,0.05912320017814636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.2,0.3151360034942627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,16,power_law_1.2,0.04991360008716583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,16,power_law_1.2,0.05554559826850891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,16,power_law_1.2,0.05432959794998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,16,power_law_1.2,0.05619199872016907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,16,power_law_1.2,0.05856639742851257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,16,power_law_1.2,0.05873280167579651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,16,power_law_1.2,0.05923839807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,16,power_law_1.2,0.06017919778823853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,16,power_law_1.2,0.06030719876289368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,16,power_law_1.2,0.061427199840545656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,16,power_law_1.2,0.06335999965667724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,16,power_law_1.2,0.06629120111465454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,16,power_law_1.2,0.07443199753761291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.2,1.1506175994873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,16,power_law_1.2,0.07867519855499268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,16,power_law_1.2,0.08535040020942689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,16,power_law_1.2,0.09794560074806213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,16,power_law_1.2,0.09995520114898682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,16,power_law_1.2,0.12044800519943237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,16,power_law_1.2,0.14675840139389038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,16,power_law_1.2,0.18610559701919555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,16,power_law_1.2,0.22535040378570556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,16,power_law_1.2,0.3223488092422485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,16,power_law_1.2,0.3821376085281372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,16,power_law_1.2,0.6098688125610352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,16,power_law_1.2,0.6879615783691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.2,1.838547134399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.2,0.424454402923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.2,2.315180778503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.2,0.5752639770507812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.2,3.7658111572265627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,8,power_law_1.01,0.03544960021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,8,power_law_1.01,0.03550080060958862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,8,power_law_1.01,0.03052160143852234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,8,power_law_1.01,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,8,power_law_1.01,0.034643200039863584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,8,power_law_1.01,0.03619840145111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,8,power_law_1.01,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,8,power_law_1.01,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,8,power_law_1.01,0.03663359880447388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,8,power_law_1.01,0.03720319867134094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,8,power_law_1.01,0.03775359988212586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,8,power_law_1.01,0.038124799728393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,8,power_law_1.01,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,8,power_law_1.01,0.04135040044784546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,8,power_law_1.01,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,8,power_law_1.01,0.04872319996356964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,8,power_law_1.01,0.05086719989776611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,8,power_law_1.01,0.05792639851570129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,8,power_law_1.01,0.06798719763755798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,8,power_law_1.01,0.08432639837265014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,8,power_law_1.01,0.08673920035362244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,8,power_law_1.01,0.11445120573043824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,8,power_law_1.01,0.13952640295028687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,8,power_law_1.01,0.19247360229492189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,8,power_law_1.01,0.2551232099533081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,8,power_law_1.01,0.33263359069824217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,8,power_law_1.01,0.4531392097473145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.2,0.7833727836608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,2,balanced,0.04074133435885111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,2,balanced,0.04114133367935816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,2,balanced,0.043237333496411644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,2,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,2,balanced,0.062309334675470986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,2,balanced,0.06239999830722809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,2,balanced,0.06229333579540253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,2,balanced,0.06393066545327504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,2,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,2,balanced,0.06428266565004985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,2,balanced,0.06609066824118297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,2,balanced,0.06785599887371063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,2,balanced,0.07032000025113423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,2,balanced,0.07120533287525177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,2,balanced,0.07461333274841309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,2,balanced,0.07853333155314128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,2,balanced,0.08247999846935272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,2,balanced,0.09462933739026387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,2,balanced,0.10579199592272441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,2,balanced,0.1295093297958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,2,balanced,0.15084266662597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,2,balanced,0.20745599269866943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,2,balanced,0.2485546668370565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,2,balanced,0.3588586648305257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,2,balanced,0.4610026677449544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,2,balanced,0.6780959765116373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,2,balanced,0.8909493287404379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.2,4.202444839477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,1,power_law_1.01,0.07651200294494628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,1,power_law_1.01,0.09987840056419373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,1,power_law_1.01,0.15148799419403075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,1,power_law_1.01,0.22704000473022462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,1,power_law_1.01,0.3116031885147095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,1,power_law_1.01,0.46353921890258787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,1,power_law_1.01,0.49643521308898925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,1,power_law_1.01,0.5133823871612548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,1,power_law_1.01,0.5243135929107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,1,power_law_1.01,0.5336703777313232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,1,power_law_1.01,0.5508416175842286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,1,power_law_1.01,0.5625343799591065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,1,power_law_1.01,0.5726784229278564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,1,power_law_1.01,0.5850111961364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,1,power_law_1.01,0.6166143894195557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,1,power_law_1.01,0.6341504096984864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,1,power_law_1.01,0.6608767986297608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,1,power_law_1.01,0.7557055950164795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,1,power_law_1.01,0.7808703899383544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,1,power_law_1.01,0.9207615852355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,1,power_law_1.01,1.0538559913635255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,1,power_law_1.01,1.3219264030456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,1,power_law_1.01,1.4302720069885253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,1,power_law_1.01,1.8883007049560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,1,power_law_1.01,2.630956840515137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,1,power_law_1.01,3.4778175354003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,1,power_law_1.01,4.3251392364501955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,4,power_law_1.2,0.12383999824523925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,4,power_law_1.2,0.15507839918136596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,4,power_law_1.2,0.1592319965362549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,4,power_law_1.2,0.2093951940536499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,4,power_law_1.2,0.211027193069458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,4,power_law_1.2,0.22371199131011962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,4,power_law_1.2,0.22812159061431886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,4,power_law_1.2,0.2205888032913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,4,power_law_1.2,0.24787840843200684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,4,power_law_1.2,0.2558016061782837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,4,power_law_1.2,0.2456063985824585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,4,power_law_1.2,0.2882496118545532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,4,power_law_1.2,0.2768768072128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,4,power_law_1.2,0.29166719913482664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,4,power_law_1.2,0.33622400760650634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,4,power_law_1.2,0.3875839948654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,4,power_law_1.2,0.43645439147949217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,4,power_law_1.2,0.5501632213592529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,4,power_law_1.2,0.5527232170104981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,4,power_law_1.2,0.7182528018951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,4,power_law_1.2,0.8562111854553223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,4,power_law_1.2,1.2381888389587403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,4,power_law_1.2,1.7990976333618165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,4,power_law_1.2,2.489388847351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,16,power_law_1.01,0.05554559826850891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,16,power_law_1.01,0.06407039761543273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,16,power_law_1.01,0.05922560095787048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,16,power_law_1.01,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,16,power_law_1.01,0.06109439730644226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,4,power_law_1.2,3.410073471069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,16,power_law_1.01,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,16,power_law_1.01,0.0635968029499054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,16,power_law_1.01,0.0664255976676941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,16,power_law_1.01,0.06652160286903382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,16,power_law_1.01,0.06744959950447083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,16,power_law_1.01,0.06563199758529663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,16,power_law_1.01,0.06931840181350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,16,power_law_1.01,0.0704255998134613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,16,power_law_1.01,0.07149440050125122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,16,power_law_1.01,0.08024960160255432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,16,power_law_1.01,0.08037760257720947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,16,power_law_1.01,0.08859519958496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,16,power_law_1.01,0.10107519626617431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,16,power_law_1.01,0.11420799493789673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,16,power_law_1.01,0.13354239463806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,16,power_law_1.01,0.1553279995918274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,16,power_law_1.01,0.1935487985610962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,16,power_law_1.01,0.22875521183013917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,16,power_law_1.01,0.29904000759124755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,16,power_law_1.01,0.381932806968689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,16,power_law_1.01,0.551148796081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,16,power_law_1.01,0.6819519996643066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,2,power_law_1.01,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,2,power_law_1.01,0.09354239702224731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,2,power_law_1.01,0.10216319561004639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,2,power_law_1.01,0.13459839820861816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,2,power_law_1.01,0.1630784034729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,2,power_law_1.01,0.23023359775543212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,2,power_law_1.01,0.23694078922271727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,2,power_law_1.01,0.23306241035461425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,2,power_law_1.01,0.23939199447631837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,2,power_law_1.01,0.24802560806274415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,2,power_law_1.01,0.25040640830993655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,2,power_law_1.01,0.26166400909423826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,2,power_law_1.01,0.2686336040496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,2,power_law_1.01,0.2737663984298706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,2,power_law_1.01,0.2895103931427002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,2,power_law_1.01,0.3040895938873291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,2,power_law_1.01,0.3207616090774536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,2,power_law_1.01,0.37278079986572266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,2,power_law_1.01,0.39702401161193845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,2,power_law_1.01,0.5083136081695556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,2,power_law_1.01,0.6028543949127197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,2,power_law_1.01,0.760038423538208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,2,power_law_1.01,0.8712512016296386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,2,power_law_1.01,1.1852352142333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,2,power_law_1.01,1.4390144348144531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,2,power_law_1.01,2.009747123718262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,2,power_law_1.01,2.692947196960449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,4,power_law_1.2,5.576294326782227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,4,balanced,0.044079999128977455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,4,balanced,0.045642669002215065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,4,balanced,0.04981866478919983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,4,balanced,0.07020799815654755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,4,balanced,0.09512533744176228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,4,balanced,0.1402133305867513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,4,balanced,0.14198399583498636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,4,balanced,0.1431893308957418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,4,balanced,0.1432960033416748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,4,balanced,0.14421866337458292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,4,balanced,0.14794133106867471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,4,balanced,0.14696533481280008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,4,balanced,0.15036799510320029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,4,balanced,0.15337066849072775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,4,balanced,0.15828800201416016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,4,balanced,0.1597813367843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,4,balanced,0.16554666558901468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,4,balanced,0.18107199668884277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,4,balanced,0.1888320048650106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,4,balanced,0.214303990205129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,4,balanced,0.24116800228754678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,4,balanced,0.29450132449467975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,4,balanced,0.3341546853383382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,4,balanced,0.4503413438796997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,4,balanced,0.5343733231226603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,4,balanced,0.7529919942220052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,4,balanced,0.9379626909891764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,256,power_law_1.2,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,256,power_law_1.2,0.06417279839515685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,256,power_law_1.2,0.0549888014793396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,256,power_law_1.2,0.05974400043487549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,256,power_law_1.2,0.05639680027961731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,256,power_law_1.2,0.05803520083427429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,256,power_law_1.2,0.05850239992141724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,256,power_law_1.2,0.0591871976852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,256,power_law_1.2,0.05923200249671936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,256,power_law_1.2,0.06002560257911682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,256,power_law_1.2,0.05924479961395264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,256,power_law_1.2,0.06289920210838318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,256,power_law_1.2,0.06303359866142273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,256,power_law_1.2,0.06484479904174804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,256,power_law_1.2,0.0704576015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,256,power_law_1.2,0.07319039702415467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,256,power_law_1.2,0.0794368028640747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,256,power_law_1.2,0.0920960009098053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,256,power_law_1.2,0.10303360223770142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,256,power_law_1.2,0.12859519720077514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,256,power_law_1.2,0.1570688009262085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,256,power_law_1.2,0.20625920295715333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,256,power_law_1.2,0.21169919967651368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,256,power_law_1.2,0.2959552049636841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,256,power_law_1.2,0.3488640069961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,256,power_law_1.2,0.502784013748169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,256,power_law_1.2,0.6896192073822022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,4,power_law_1.2,5.926240158081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,1,balanced,0.05303466816743215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,1,balanced,0.056048000852266945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,1,balanced,0.0641653339068095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,1,balanced,0.08479467034339905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,1,balanced,0.12986666957537332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,1,balanced,0.12966932853062949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,1,balanced,0.13243732849756876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,1,balanced,0.1339466671148936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,1,balanced,0.13497066497802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,1,balanced,0.13577600320180258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,1,balanced,0.14136000474294028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,1,balanced,0.14129066467285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,1,balanced,0.14669332901636759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,1,balanced,0.15051733454068503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,1,balanced,0.15736533204714456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,1,balanced,0.16369066635767618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,1,balanced,0.175327996412913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,1,balanced,0.2021333376566569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,1,balanced,0.22539732853571573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,1,balanced,0.2908906737963359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,1,balanced,0.33787735303243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,1,balanced,0.503114660580953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,1,balanced,0.6173493464787801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,1,balanced,0.8894826571146647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,1,balanced,1.1484746932983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,1,balanced,1.6873386700948079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,1,balanced,2.234389305114746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,8,balanced,0.04190933207670847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,8,balanced,0.044922664761543274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,8,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,8,balanced,0.049626668294270836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,8,balanced,0.052058666944503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,8,balanced,0.06596266726652782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,8,balanced,0.08181866506735484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,8,balanced,0.082997332016627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,8,balanced,0.08303999900817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,8,balanced,0.08519466718037923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,8,balanced,0.0841919978459676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,8,balanced,0.08685866991678874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,8,balanced,0.088837335507075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,8,balanced,0.08986133337020874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,8,balanced,0.09586133559544881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,8,balanced,0.0990666647752126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,8,balanced,0.10421866178512573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,8,balanced,0.11406933267911275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,8,balanced,0.12405332922935486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,8,balanced,0.14429866274197897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,8,balanced,0.15794666608174643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,8,balanced,0.19190933307011923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,8,balanced,0.25012799104054767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,8,balanced,0.32020799318949383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,8,balanced,0.41437868277231854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,8,balanced,0.589189330736796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,8,balanced,0.7394293149312338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,1,power_law_1.01,0.1323456048965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,1,power_law_1.01,0.20162560939788818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,1,power_law_1.01,0.3250047922134399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,1,power_law_1.01,0.5618495941162109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,1,power_law_1.01,0.7475840091705322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,1,power_law_1.01,1.2268671989440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,1,power_law_1.01,1.3239423751831054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,1,power_law_1.01,1.3749376296997071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,1,power_law_1.01,1.413535976409912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,1,power_law_1.01,1.4344575881958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,1,power_law_1.01,1.4769791603088378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,1,power_law_1.01,1.5152511596679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,1,power_law_1.01,1.552665615081787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,1,power_law_1.01,1.5878144264221192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,1,power_law_1.01,1.6171903610229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,1,power_law_1.01,1.6930944442749023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,1,power_law_1.01,1.7888832092285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,1,power_law_1.01,2.0427583694458007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,1,power_law_1.01,1.973036766052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,1,power_law_1.01,2.265100860595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,1,power_law_1.01,2.2823040008544924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,1,power_law_1.01,2.867136001586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,1,power_law_1.01,3.0827199935913088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,1,power_law_1.01,3.8379329681396483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,1,power_law_1.01,4.631731033325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,1,power_law_1.01,6.322182464599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,1,power_law_1.01,7.593644714355468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,4,balanced,0.04902400076389313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,4,balanced,0.05193600058555603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,4,balanced,0.05392533540725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,4,balanced,0.06309866905212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,4,balanced,0.08233066896597545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,4,balanced,0.09477866689364116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,4,balanced,0.09484799702962239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,4,balanced,0.09638933340708415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,4,balanced,0.09679999947547913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,4,balanced,0.0965173343817393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,4,balanced,0.09709866841634114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,4,balanced,0.09885866443316142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,4,balanced,0.10405332843462627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,4,balanced,0.1036906639734904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,4,balanced,0.11318399508794148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,4,balanced,0.11395200093587239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,4,balanced,0.11761066317558289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,4,balanced,0.13029332955678305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,4,balanced,0.1418506701787313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,4,balanced,0.16674667596817017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,4,balanced,0.18523732821146646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,4,balanced,0.2567039926846822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,4,balanced,0.2974613308906555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,4,balanced,0.4325973192850749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,4,balanced,0.5356213251749674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,4,balanced,0.7806026935577393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,4,balanced,0.9897066752115885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,1,balanced,0.050213331977526345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,1,balanced,0.053130666414896645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,1,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,1,balanced,0.07062933345635732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,1,balanced,0.09917866190274556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,1,balanced,0.14328533411026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,1,balanced,0.14247467120488486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,1,balanced,0.14339733123779297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,1,balanced,0.14723733067512512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,1,balanced,0.1493119994799296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,1,balanced,0.15007999539375305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,1,balanced,0.15550933281580606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,1,balanced,0.15877333283424377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,1,balanced,0.15913599729537964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,1,balanced,0.16421866416931152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,1,balanced,0.16946667432785034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,1,balanced,0.18171733617782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,1,balanced,0.20226667324701944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,1,balanced,0.22389866908391318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,1,balanced,0.2653866608937581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,1,balanced,0.2957386573155721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,1,balanced,0.4375573396682739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,1,power_law_1.01,0.10342400074005127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,1,balanced,0.5063039859135946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,1,power_law_1.01,0.12622079849243165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,1,balanced,0.7030879656473795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,1,balanced,0.8818933169047037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,1,power_law_1.01,0.15996160507202148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,1,balanced,1.3418240547180176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,1,power_law_1.01,0.2183232069015503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,1,power_law_1.01,0.32451839447021485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,1,power_law_1.01,0.41926398277282717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,1,power_law_1.01,0.48450560569763185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,1,power_law_1.01,0.6020351886749268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,1,balanced,1.7052853902180989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,1,power_law_1.01,0.6287936210632324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,1,power_law_1.01,0.6528384208679199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,1,power_law_1.01,0.6816192150115967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,1,power_law_1.01,0.7123839855194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,1,power_law_1.01,0.7231616020202637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,1,power_law_1.01,0.7453887939453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,1,power_law_1.01,0.7690815925598145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,1,power_law_1.01,0.7998720169067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,1,power_law_1.01,0.8332608222961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,1,power_law_1.01,0.9016256332397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.01,1.0113984107971192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.01,1.1450112342834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.01,1.199065589904785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.01,1.4615488052368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.01,1.670035171508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.01,2.1349567413330077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,1,power_law_1.01,2.8072128295898438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,1,power_law_1.01,3.740934371948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,1,power_law_1.01,4.529280090332032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,128,power_law_1.01,0.031699201464653014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,128,power_law_1.01,0.03089280128479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,128,power_law_1.01,0.03147520124912262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,128,power_law_1.01,0.03278079926967621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,128,power_law_1.01,0.033452799916267394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,128,power_law_1.01,0.03367680013179779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,128,power_law_1.01,0.03421440124511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,128,power_law_1.01,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,128,power_law_1.01,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,128,power_law_1.01,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,128,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,128,power_law_1.01,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,128,power_law_1.01,0.03617280125617981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,128,power_law_1.01,0.03617280125617981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,128,power_law_1.01,0.039001598954200745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,128,power_law_1.01,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,128,power_law_1.01,0.0392767995595932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,128,power_law_1.01,0.04092159867286682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,128,power_law_1.01,0.04177919924259186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,128,power_law_1.01,0.04439040124416351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,128,power_law_1.01,0.047654399275779726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,128,power_law_1.01,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,128,power_law_1.01,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,128,power_law_1.01,0.07935360074043274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,128,power_law_1.01,0.09856640100479126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,128,power_law_1.01,0.1296064019203186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,128,power_law_1.01,0.16822400093078613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,1,balanced,0.07214400172233582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,1,balanced,0.08891200025876363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,1,balanced,0.13803199927012125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,1,balanced,0.14396799604098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,1,balanced,0.1442346672217051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,1,balanced,0.14453333616256714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,1,balanced,0.1456000010172526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,1,balanced,0.1465013325214386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,1,balanced,0.14598400394121805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,1,balanced,0.14664533734321594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,1,balanced,0.14860799908638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,1,balanced,0.15169599652290344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,1,balanced,0.15215999881426492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,1,balanced,0.15587199727694193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,1,balanced,0.1639306644598643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,1,balanced,0.16950400670369467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,1,balanced,0.16965866088867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,1,balanced,0.22242667277654013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,1,balanced,0.23693867524464926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,1,balanced,0.3057919939359029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,1,balanced,0.38710931936899823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,1,balanced,0.5479946533838908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,1,balanced,0.7112533251444498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,1,balanced,1.0274559656778972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,1,balanced,1.3325386842091878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,1,balanced,1.9689547220865886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,1,balanced,2.582538604736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,8,balanced,0.0317546675602595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,8,balanced,0.029296000798543293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,8,balanced,0.029546665648619335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,8,balanced,0.031258667508761086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,8,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,8,balanced,0.03232000023126602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,8,balanced,0.03136533250411352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,8,balanced,0.0332640012105306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,8,balanced,0.033674667278925575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,8,balanced,0.03329066683848699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,8,balanced,0.03369066615899404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,8,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,8,balanced,0.03566399961709976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,8,balanced,0.035445332527160645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,8,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,8,balanced,0.04192533095677694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,8,balanced,0.04138666639725367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,8,balanced,0.04769066472848257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,8,balanced,0.04594666759173075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,8,balanced,0.06140799820423126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,8,balanced,0.06205866734186808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,8,balanced,0.09168000022570293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,8,balanced,0.10494933525721233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,8,balanced,0.1438826620578766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,8,balanced,0.1721173326174418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,8,balanced,0.23699732621510824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,8,balanced,0.295141339302063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,32,2,balanced,0.023306667804718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,32,2,balanced,0.023130667706330616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,32,2,balanced,0.02515733242034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,32,2,balanced,0.02958400050799052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,32,2,balanced,0.039450667798519135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,32,2,balanced,0.05171733101209005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,32,2,balanced,0.05593599875768026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,32,2,balanced,0.06022400160630544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,32,2,balanced,0.06258666515350342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,32,2,balanced,0.06529599924882253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,32,2,balanced,0.0702400008837382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,32,2,balanced,0.07250666618347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,32,2,balanced,0.07646400233109792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,32,2,balanced,0.08245866497357686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,32,2,balanced,0.08041066428025563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,32,2,balanced,0.08241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,32,2,balanced,0.09110933542251587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,32,2,balanced,0.1192586620648702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,32,2,balanced,0.12705600261688232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,32,2,balanced,0.18844266732533774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,32,2,balanced,0.2012373407681783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,32,2,balanced,0.2596693237622579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,32,2,balanced,0.288480003674825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,32,2,balanced,0.34397868315378827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,32,2,balanced,0.39162667592366535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,32,2,balanced,0.623797337214152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,32,2,balanced,0.7340906461079916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,8,power_law_1.2,0.05375360250473023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,8,power_law_1.2,0.05947520136833191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,8,power_law_1.2,0.0615231990814209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,8,power_law_1.2,0.06848000288009644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,8,power_law_1.2,0.0747327983379364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,8,power_law_1.2,0.0753920018672943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,8,power_law_1.2,0.0791167974472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,8,power_law_1.2,0.07730559706687927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,8,power_law_1.2,0.07991039752960205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,8,power_law_1.2,0.077183997631073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,8,power_law_1.2,0.08009600043296813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,8,power_law_1.2,0.08273280262947083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,8,power_law_1.2,0.08390399813652039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,8,power_law_1.2,0.08732159733772278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,8,power_law_1.2,0.09068800210952759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,8,power_law_1.2,0.09101439714431762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,16,power_law_1.2,0.0834176003932953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,8,power_law_1.2,0.09876480102539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,16,power_law_1.2,0.09984639883041382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,16,power_law_1.2,0.09896960258483886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,16,power_law_1.2,0.1054144024848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,16,power_law_1.2,0.09924479722976684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,8,power_law_1.2,0.10811519622802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,16,power_law_1.2,0.10389759540557861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,8,power_law_1.2,0.12130559682846069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,16,power_law_1.2,0.09964159727096558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,16,power_law_1.2,0.10375679731369018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,16,power_law_1.2,0.10473599433898925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,16,power_law_1.2,0.10414079427719117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,16,power_law_1.2,0.10748800039291381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,16,power_law_1.2,0.11384960412979125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,2,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,2,balanced,0.07637866834799449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,16,power_law_1.2,0.11463040113449097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,2,balanced,0.12174933155377705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,2,balanced,0.12307733297348022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,8,power_law_1.2,0.14365439414978026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,8,power_law_1.2,0.16935039758682252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,8,power_law_1.2,0.21555840969085693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,8,power_law_1.2,0.26019198894500734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,2,balanced,0.1202826698621114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,2,balanced,0.1218239963054657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,2,balanced,0.12231466174125671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,8,power_law_1.2,0.3662911891937256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,16,power_law_1.2,0.11455359458923339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,2,balanced,0.12168533603350322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,2,balanced,0.12416533629099528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,8,power_law_1.2,0.44403839111328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,8,power_law_1.2,0.6939263820648194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,8,power_law_1.2,0.9728192329406739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,16,power_law_1.2,0.12588160037994384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,2,balanced,0.1246560017267863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,2,balanced,0.12523200114568075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,2,balanced,0.12581866979599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,16,power_law_1.2,0.13147519826889037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,16,power_law_1.2,0.13639039993286134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,16,power_law_1.2,0.15701119899749755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,16,power_law_1.2,0.17219200134277343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,2,balanced,0.12668266892433167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,2,balanced,0.13165866335233053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,16,power_law_1.2,0.2178368091583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,2,balanced,0.14628266294797262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,2,balanced,0.14601600170135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,16,power_law_1.2,0.253606390953064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,2,balanced,0.14756266276041666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,2,balanced,0.18447999159495035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,16,power_law_1.2,0.33676159381866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,2,balanced,0.1869973341623942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,2,balanced,0.23240000009536743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,16,power_law_1.2,0.3493760108947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,2,balanced,0.2684160073598226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,2,balanced,0.35738134384155273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,16,power_law_1.2,0.4458752155303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,2,balanced,0.4161653518676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,2,balanced,0.5949066480000814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,2,balanced,0.7706613540649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,16,power_law_1.2,0.5861440181732178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,2,balanced,1.108031988143921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,2,balanced,1.4373760223388672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,16,power_law_1.2,0.8808639526367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,16,power_law_1.2,1.010969638824463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,128,power_law_1.01,0.06962559819221496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,128,power_law_1.01,0.06747519969940186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,128,power_law_1.01,0.06798719763755798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,128,power_law_1.01,0.07025920152664185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,128,power_law_1.01,0.07247359752655029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,128,power_law_1.01,0.07647359967231751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,128,power_law_1.01,0.07459200024604798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,128,power_law_1.01,0.07568640112876893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,128,power_law_1.01,0.07515519857406616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,128,power_law_1.01,0.07493119835853576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,128,power_law_1.01,0.07504640221595764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,128,power_law_1.01,0.07822719812393189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,128,power_law_1.01,0.07760639786720276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,128,power_law_1.01,0.0798143982887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,128,power_law_1.01,0.08267520070075988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,128,power_law_1.01,0.08761600255966187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,4,balanced,0.03702399879693985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,128,power_law_1.01,0.08919039964675904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,128,power_law_1.01,0.10186879634857178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,128,power_law_1.01,0.10458879470825196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,128,power_law_1.01,0.12357759475708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,128,power_law_1.01,0.13587199449539183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,4,balanced,0.033743999898433685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,4,balanced,0.03336533407370249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,4,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,4,balanced,0.03855466594298681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,4,balanced,0.03745600084463755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,4,balanced,0.03783999880154928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,4,balanced,0.03942399968703588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,4,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,4,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,4,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,128,power_law_1.01,0.17712639570236205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,128,power_law_1.01,0.20171520709991456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,128,power_law_1.01,0.2784895896911621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,128,power_law_1.01,0.35226879119873045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,128,power_law_1.01,0.5240767955780029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,128,power_law_1.01,0.6430463790893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,4,balanced,0.04146133363246918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,4,balanced,0.03966933240493139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,4,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,4,balanced,0.04642133414745331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,4,balanced,0.047839999198913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,4,balanced,0.05197866757710775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,4,balanced,0.05811200042565664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,4,balanced,0.0641653339068095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,4,balanced,0.08131733536720276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,4,balanced,0.08666666348775227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,4,balanced,0.11549333731333415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,4,balanced,0.14071999986966452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,4,balanced,0.18440532684326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,4,balanced,0.24458134174346924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,4,balanced,0.34958934783935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,4,balanced,0.46083199977874756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,4,power_law_1.2,0.03563520014286041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,4,power_law_1.2,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,4,power_law_1.2,0.04506880044937134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,4,power_law_1.2,0.05620480179786682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,4,power_law_1.2,0.06267520189285278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,4,power_law_1.2,0.07394559979438782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,4,power_law_1.2,0.07247999906539918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,4,power_law_1.2,0.07660800218582153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,4,power_law_1.2,0.07440000176429748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,4,power_law_1.2,0.07925119996070862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,4,power_law_1.2,0.08054400086402894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,4,power_law_1.2,0.08166400194168091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,4,power_law_1.2,0.08296959996223449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,4,power_law_1.2,0.08554239869117737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,4,power_law_1.2,0.09236479997634887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,4,power_law_1.2,0.09229440093040467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,4,power_law_1.2,0.09597439765930176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,4,power_law_1.2,0.10833920240402221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,4,power_law_1.2,0.11706880331039429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,4,power_law_1.2,0.13287680149078368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,4,power_law_1.2,0.15130239725112915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,4,power_law_1.2,0.1870144009590149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,4,power_law_1.2,0.2079551935195923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,4,power_law_1.2,0.2630975961685181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,4,power_law_1.2,0.33641600608825684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,4,power_law_1.2,0.42946557998657225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,4,power_law_1.2,0.5163136005401612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,16,power_law_1.2,0.021939200162887574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,16,power_law_1.2,0.021715199947357176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,16,power_law_1.2,0.021580800414085388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,16,power_law_1.2,0.020768000185489653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,16,power_law_1.2,0.02316800057888031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,16,power_law_1.2,0.025254398584365845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,16,power_law_1.2,0.031641599535942075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,16,power_law_1.2,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,16,power_law_1.2,0.04377599954605103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,16,power_law_1.2,0.044352000951766966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,16,power_law_1.2,0.04325760006904602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,16,power_law_1.2,0.049772799015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,16,power_law_1.2,0.04850560128688812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,16,power_law_1.2,0.048147198557853696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,16,power_law_1.2,0.04681600034236908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,16,power_law_1.2,0.04903680086135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,16,power_law_1.2,0.046412798762321475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,16,power_law_1.2,0.05199360251426697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.2,0.05678719878196716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.2,0.06427519917488098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.2,0.07426559925079346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.2,0.0964352011680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.2,0.11834880113601684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.2,0.15575040578842164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.2,0.1915071964263916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,1,power_law_1.2,0.052742397785186766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,1,power_law_1.2,0.05782399773597717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,1,power_law_1.2,0.07297279834747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,1,power_law_1.2,0.10013439655303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,1,power_law_1.2,0.12650879621505737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,1,power_law_1.2,0.15348479747772217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,4,power_law_1.01,0.043110400438308716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,1,power_law_1.2,0.2005631923675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,4,power_law_1.01,0.04593920111656189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,1,power_law_1.2,0.21119999885559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,4,power_law_1.01,0.04916479885578155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,1,power_law_1.2,0.21909120082855224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,4,power_law_1.01,0.06042879819869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,4,power_law_1.01,0.0690559983253479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,4,power_law_1.01,0.07716479897499084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,4,power_law_1.01,0.08347520232200623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,4,power_law_1.01,0.0823743999004364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,4,power_law_1.01,0.08290560245513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,1,power_law_1.2,0.22812159061431886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,4,power_law_1.01,0.0832319974899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,1,power_law_1.2,0.23518080711364747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,4,power_law_1.01,0.08361600041389465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,1,power_law_1.2,0.25002241134643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,4,power_law_1.01,0.08951039910316468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,4,power_law_1.01,0.09269760251045227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,4,power_law_1.01,0.09686400294303894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,4,power_law_1.01,0.10371199846267701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,4,power_law_1.01,0.10748159885406494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,4,power_law_1.01,0.1118783950805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,1,power_law_1.2,0.2629184007644653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,1,power_law_1.2,0.27752320766448973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.2,0.29473280906677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,1,power_law_1.2,0.295743989944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,1,power_law_1.2,0.3132800102233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,4,power_law_1.01,0.12584960460662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,1,power_law_1.2,0.32400639057159425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,1,power_law_1.2,0.38348801136016847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,1,power_law_1.2,0.41683201789855956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,1,power_law_1.2,0.5123839855194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,1,power_law_1.2,0.609606409072876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,4,power_law_1.01,0.1373247981071472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,4,power_law_1.01,0.16518399715423585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,4,power_law_1.01,0.1889024019241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,4,power_law_1.01,0.2393343925476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,4,power_law_1.01,0.29178240299224856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,4,power_law_1.01,0.41013121604919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,4,power_law_1.01,0.497376012802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,1,power_law_1.2,0.7780735969543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,4,power_law_1.01,0.7123136043548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,1,power_law_1.2,0.942899227142334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,4,power_law_1.01,0.9350079536437989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,1,power_law_1.2,1.2687104225158692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,1,power_law_1.2,1.5864704132080079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,1,power_law_1.2,2.2360511779785157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,1,power_law_1.2,2.908537673950195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,64,balanced,0.01934933289885521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,64,balanced,0.017221332838137943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,64,balanced,0.016869333883126576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,64,balanced,0.01904533306757609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,64,balanced,0.019002666076024372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,64,balanced,0.02120000123977661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,64,balanced,0.021194666624069214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,64,balanced,0.029301332930723827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,64,balanced,0.029232000311215717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,64,balanced,0.03038399914900462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,64,balanced,0.029178666571776073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,64,balanced,0.02934933453798294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,64,balanced,0.029253333806991577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,64,balanced,0.029322666426499683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,64,balanced,0.03133866687615713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,64,balanced,0.031397332747777305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,64,balanced,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,64,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,64,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,64,balanced,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,64,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,64,balanced,0.057487999399503074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,64,balanced,0.06878399848937988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,64,balanced,0.09309867024421692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,64,balanced,0.11685867110888164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,64,balanced,0.16126400232315063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,64,balanced,0.2059839963912964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,16,power_law_1.01,0.09690880179405212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,16,power_law_1.01,0.13782399892807007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,16,power_law_1.01,0.1546560049057007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,16,power_law_1.01,0.15449600219726561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,16,power_law_1.01,0.1557952046394348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,16,power_law_1.01,0.15674879550933837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,16,power_law_1.01,0.14045439958572387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,16,power_law_1.01,0.15155199766159058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,16,power_law_1.01,0.14454400539398193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,16,power_law_1.01,0.14746880531311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,16,power_law_1.01,0.14994560480117797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,16,power_law_1.01,0.1568384051322937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,16,power_law_1.01,0.15767040252685546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,16,power_law_1.01,0.15718400478363037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,16,power_law_1.01,0.16911360025405883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,16,power_law_1.01,0.16596479415893556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,16,power_law_1.01,0.1745344042778015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,16,power_law_1.01,0.18698240518569947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,16,power_law_1.01,0.1975808024406433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,16,power_law_1.01,0.21875839233398436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,16,power_law_1.01,0.26686079502105714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,16,power_law_1.01,0.3131648063659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,16,power_law_1.01,0.37272961139678956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,16,power_law_1.01,0.5237247943878174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,16,power_law_1.01,0.6107007980346679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,16,power_law_1.01,0.9603967666625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,16,power_law_1.01,1.1630399703979493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,256,power_law_1.01,0.05015040040016174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,256,power_law_1.01,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,256,power_law_1.01,0.05051519870758057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,256,power_law_1.01,0.05299839973449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,256,power_law_1.01,0.05359359979629517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,256,power_law_1.01,0.05459840297698974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,256,power_law_1.01,0.05434240102767944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.2,0.3681663990020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,256,power_law_1.01,0.05529599785804749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,256,power_law_1.01,0.05448319911956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,256,power_law_1.01,0.05486720204353333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,256,power_law_1.01,0.05546879768371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,256,power_law_1.01,0.05633280277252197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,256,power_law_1.01,0.056979197263717654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,256,power_law_1.01,0.0574783980846405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,256,power_law_1.01,0.06034560203552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,256,power_law_1.01,0.06069759726524353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,256,power_law_1.01,0.062438398599624634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,256,power_law_1.01,0.06968960165977478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,256,power_law_1.01,0.07054719924926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,256,power_law_1.01,0.08130559921264649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,256,power_law_1.01,0.08960639834403991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,256,power_law_1.01,0.11512960195541382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,256,power_law_1.01,0.14007680416107177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,256,power_law_1.01,0.1915071964263916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,256,power_law_1.01,0.23489279747009278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,256,power_law_1.01,0.32814080715179444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,256,power_law_1.01,0.41436161994934084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,64,power_law_1.01,0.022681599855422972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,64,power_law_1.01,0.02220800071954727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,64,power_law_1.01,0.022310400009155275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,64,power_law_1.01,0.023321600258350374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,64,power_law_1.01,0.024748800694942473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,64,power_law_1.01,0.025516799092292784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,64,power_law_1.01,0.02550399899482727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,64,power_law_1.01,0.024153600633144378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,64,power_law_1.01,0.02812800109386444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,64,power_law_1.01,0.029657599329948426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,64,power_law_1.01,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,64,power_law_1.01,0.031302401423454286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,64,power_law_1.01,0.031590399146080014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,64,power_law_1.01,0.02996479868888855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,64,power_law_1.01,0.03367680013179779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,64,power_law_1.01,0.03482879996299744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,64,power_law_1.01,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,64,power_law_1.01,0.03666560053825378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,64,power_law_1.01,0.04167680144309997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,64,power_law_1.01,0.04695680141448975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,64,power_law_1.01,0.04808320105075836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,64,power_law_1.01,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,64,power_law_1.01,0.05743359923362732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,64,power_law_1.01,0.08286719918251037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,64,power_law_1.01,0.09297919869422913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,64,power_law_1.01,0.12675199508666993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,64,power_law_1.01,0.16472320556640624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,16,power_law_1.01,0.020268799364566804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,16,power_law_1.01,0.02054399996995926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,16,power_law_1.01,0.01847680062055588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,16,power_law_1.01,0.019270400702953338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,16,power_law_1.01,0.019705599546432494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,16,power_law_1.01,0.02021760046482086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,16,power_law_1.01,0.020902399718761445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,16,power_law_1.01,0.021299199759960176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,16,power_law_1.01,0.021049599349498748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,16,power_law_1.01,0.021305599808692934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,16,power_law_1.01,0.025516799092292784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,16,power_law_1.01,0.02534399926662445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,16,power_law_1.01,0.025939199328422546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,4,power_law_1.01,0.0297791987657547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,16,power_law_1.01,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,16,power_law_1.01,0.03980799913406372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,16,power_law_1.01,0.04298880100250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,16,power_law_1.01,0.04270719885826111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,16,power_law_1.01,0.04390400052070618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.01,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,4,power_law_1.01,0.02927359938621521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.01,0.05271040201187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,4,power_law_1.01,0.029702401161193846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,4,power_law_1.01,0.033529600501060484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,4,power_law_1.01,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,4,power_law_1.01,0.03325439989566803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,4,power_law_1.01,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.01,0.060115200281143186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,4,power_law_1.01,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,4,power_law_1.01,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.01,0.0809664011001587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,4,power_law_1.01,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,4,power_law_1.01,0.035872000455856326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.01,0.0936896026134491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.01,0.1280128002166748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.01,0.1596992015838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,1,balanced,0.15783466895421347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,1,balanced,0.24183466037114462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,1,balanced,0.36821333567301434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,1,balanced,0.5696586767832438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,1,balanced,1.0006879965464275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,1,balanced,1.8071734110514324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.01,0.21766400337219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,1,balanced,1.7866560618082683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.01,0.2849600076675415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,1,balanced,1.738602638244629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,1,balanced,1.7471359570821126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,1,balanced,1.6827252705891926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,1,balanced,1.67413330078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,1,balanced,1.6758294105529785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,1,balanced,1.7148480415344238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,1,balanced,1.725887934366862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,1,balanced,1.7090346018473308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,4,power_law_1.01,0.03818239867687225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,4,power_law_1.01,0.03923200070858002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,4,power_law_1.01,0.042368000745773314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,1,balanced,1.8129599889119465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,1,balanced,1.7842079798380535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,1,balanced,1.8083945910135906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,1,balanced,1.7980586687723796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,1,balanced,1.8536319732666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,1,balanced,1.9121066729227703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,4,power_law_1.01,0.05167359709739685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,1,balanced,2.0422773361206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,4,power_law_1.01,0.05446400046348572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,16,balanced,0.04606399933497111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,1,balanced,2.2142346700032554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,16,balanced,0.04668800036112467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,16,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,16,balanced,0.07495466868082683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,16,balanced,0.09769066174825032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,16,balanced,0.1520960032939911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,16,balanced,0.15030933419863382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,16,balanced,0.15219733119010925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,16,balanced,0.15266133348147073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,16,balanced,0.15411733587582907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,16,balanced,0.15380799770355225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,16,balanced,0.15615466237068176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,16,balanced,0.15870400269826254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,8,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,16,balanced,0.16099733114242554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,16,balanced,0.1670080025990804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,16,balanced,0.16907199223836264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,8,power_law_1.2,0.05847039818763733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,16,balanced,0.1716746687889099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,1,balanced,3.6634400685628257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,8,power_law_1.2,0.06510080099105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,8,power_law_1.2,0.0751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,1,balanced,3.876223882039388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,4,power_law_1.01,0.05987840294837952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,16,balanced,0.18546134233474731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,16,balanced,0.19510400295257568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,1,balanced,5.329013188680013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,16,balanced,0.21862399578094482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,8,power_law_1.2,0.08394240140914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,8,power_law_1.2,0.09227520227432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,8,power_law_1.2,0.08840320110321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,4,power_law_1.01,0.07585279941558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,8,power_law_1.2,0.09802240133285522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,1,balanced,6.866517384847005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,8,power_law_1.2,0.09464319944381713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,16,balanced,0.239519993464152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,8,power_law_1.2,0.09120640158653259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,8,power_law_1.2,0.09288960099220275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,8,power_law_1.2,0.0966592013835907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,8,power_law_1.2,0.09567999839782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,4,power_law_1.01,0.08977280259132385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,16,balanced,0.2824959953625997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,16,balanced,0.3162613312403361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,16,balanced,0.4112693468729655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,16,balanced,0.487280011177063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,16,balanced,0.6763146718343099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,16,balanced,0.833130677541097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,8,power_law_1.2,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,8,power_law_1.2,0.11174399852752685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,8,power_law_1.2,0.11236480474472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,8,power_law_1.2,0.1210368037223816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,8,power_law_1.2,0.13344000577926635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,8,power_law_1.2,0.15505919456481934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,8,power_law_1.2,0.18195199966430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,4,power_law_1.01,0.12252800464630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,4,power_law_1.01,0.134060800075531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,8,power_law_1.2,0.21113600730895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,8,power_law_1.2,0.26791040897369384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,8,power_law_1.2,0.3280128002166748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,8,power_law_1.2,0.4517632007598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,8,power_law_1.2,0.5461376190185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,8,power_law_1.2,0.7411392211914063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,4,power_law_1.01,0.18672640323638917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,8,power_law_1.2,1.0043583869934083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,4,power_law_1.01,0.24150400161743163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,4,power_law_1.01,0.36124160289764407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,4,power_law_1.01,0.4631487846374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,4,power_law_1.01,0.8274623870849609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,4,power_law_1.01,1.1167872428894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,8,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,8,power_law_1.01,0.05108479857444763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,8,power_law_1.01,0.05262719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,8,power_law_1.01,0.05761920213699341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,8,power_law_1.01,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,8,power_law_1.01,0.06159999966621399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,8,power_law_1.01,0.06267520189285278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,8,power_law_1.01,0.06231679916381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,8,power_law_1.01,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,8,power_law_1.01,0.06425600051879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,8,power_law_1.01,0.06540799736976624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,8,power_law_1.01,0.0667967975139618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,8,power_law_1.01,0.06892799735069274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,8,power_law_1.01,0.07109119892120361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,8,power_law_1.01,0.07719680070877075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,8,power_law_1.01,0.07955840229988098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,8,power_law_1.01,0.08704000115394592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,8,power_law_1.01,0.10015360116958619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,8,power_law_1.01,0.10664960145950317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,8,power_law_1.01,0.1320896029472351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,8,power_law_1.01,0.15176960229873657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,8,power_law_1.01,0.19329919815063476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,8,power_law_1.01,0.22727680206298828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,8,power_law_1.01,0.29767038822174074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,8,power_law_1.01,0.3777983903884888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,8,power_law_1.01,0.5392576217651367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,8,power_law_1.01,0.6730624198913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,1,power_law_1.2,0.027430400252342224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,1,power_law_1.2,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,1,power_law_1.2,0.06348159909248352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,1,power_law_1.2,0.0803391993045807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,1,power_law_1.2,0.10098559856414795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,1,power_law_1.2,0.1655295968055725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,1,power_law_1.2,0.1721791982650757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,1,power_law_1.2,0.18581119775772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,1,power_law_1.2,0.18869119882583618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,1,power_law_1.2,0.19685759544372558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,1,power_law_1.2,0.21064960956573486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,1,power_law_1.2,0.1991744041442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,1,power_law_1.2,0.21263999938964845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,1,power_law_1.2,0.21575040817260743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,1,power_law_1.2,0.20741760730743408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,1,power_law_1.2,0.220467209815979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,1,power_law_1.2,0.23419520854949952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,1,power_law_1.2,0.24751360416412355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.2,0.26680960655212405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.2,0.6303296089172363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.2,0.7217343807220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.2,0.4342463970184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.2,0.48418560028076174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.2,0.6036799907684326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.2,0.7631552219390869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.2,1.0231807708740235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,2,balanced,0.01706133286158244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,2,balanced,0.01757866640885671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,2,balanced,0.01924266666173935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,2,balanced,0.021269333859284718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,2,balanced,0.025594666600227356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,2,balanced,0.026848000784715016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,2,balanced,0.027280000348885853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,2,balanced,0.0271519993742307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,2,balanced,0.029205332199732464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,2,balanced,0.02922666569550832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,2,balanced,0.029535998900731403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,2,balanced,0.031285333136717476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,2,balanced,0.032032000521818794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,2,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,2,balanced,0.04312000175317129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,2,balanced,0.04192000130812327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,2,balanced,0.04364266494909922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,2,balanced,0.06095466514428457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,2,balanced,0.06419733166694641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,2,balanced,0.08381332953770955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,2,balanced,0.08875200152397156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,2,balanced,0.09027199943860371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,2,balanced,0.09887466828028361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,2,balanced,0.15433067083358765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,2,balanced,0.16570132970809937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,2,balanced,0.23627734184265137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,2,balanced,0.3040693402290344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,2,power_law_1.01,0.0466623991727829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,2,power_law_1.01,0.04639999866485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,2,power_law_1.01,0.04832000136375427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,2,power_law_1.01,0.052671998739242554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,2,power_law_1.01,0.05135359764099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,2,power_law_1.01,0.05292159914970398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,2,power_law_1.01,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,2,power_law_1.01,0.05322239995002746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,2,power_law_1.01,0.05271040201187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,2,power_law_1.01,0.05263360142707825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,2,power_law_1.01,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,2,power_law_1.01,0.05614079833030701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,2,power_law_1.01,0.059961599111557004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,2,power_law_1.01,0.060249602794647215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,2,power_law_1.01,0.06718080043792725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,2,power_law_1.01,0.06917759776115417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,2,power_law_1.01,0.07463039755821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,2,power_law_1.01,0.07914239764213563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,2,power_law_1.01,0.09263359904289245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,2,power_law_1.01,0.11352959871292115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,2,power_law_1.01,0.1274880051612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,2,power_law_1.01,0.15081599950790406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,2,power_law_1.01,0.19603840112686158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,2,power_law_1.01,0.2822720050811768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,16,power_law_1.2,0.0702015995979309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,2,power_law_1.01,0.31877760887145995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,16,power_law_1.2,0.09257599711418152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,16,power_law_1.2,0.0807039976119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,16,power_law_1.2,0.09013760089874268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,64,power_law_1.2,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,16,power_law_1.2,0.09120000004768372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,64,power_law_1.2,0.05190399885177612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,64,power_law_1.2,0.05030400156974792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,64,power_law_1.2,0.053862398862838744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,64,power_law_1.2,0.05427839756011963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,64,power_law_1.2,0.05596799850463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,16,power_law_1.2,0.08711040019989014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,64,power_law_1.2,0.05631359815597534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,16,power_law_1.2,0.08426880240440368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,64,power_law_1.2,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,16,power_law_1.2,0.0880128026008606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,16,power_law_1.2,0.0884607970714569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,64,power_law_1.2,0.05916159749031067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,16,power_law_1.2,0.09388800263404846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,64,power_law_1.2,0.0590719997882843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,16,power_law_1.2,0.09264000058174134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,64,power_law_1.2,0.05989120006561279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,16,power_law_1.2,0.09465600252151489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,16,power_law_1.2,0.09823359847068787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,16,power_law_1.2,0.10193920135498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,16,power_law_1.2,0.11245440244674683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,16,power_law_1.2,0.11574399471282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,16,power_law_1.2,0.12043520212173461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,16,power_law_1.2,0.13615360260009765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,64,power_law_1.2,0.062035202980041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,16,power_law_1.2,0.14595839977264405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,2,power_law_1.01,0.48090238571166993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,16,power_law_1.2,0.17244800329208373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,16,power_law_1.2,0.2048192024230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,16,power_law_1.2,0.27029120922088623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,16,power_law_1.2,0.31880319118499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,64,power_law_1.2,0.06216319799423218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,16,power_law_1.2,0.43131518363952637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,16,power_law_1.2,0.561625623703003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,16,power_law_1.2,0.8229120254516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,16,power_law_1.2,1.0315648078918458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,64,power_law_1.2,0.06684160232543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,64,power_law_1.2,0.074726402759552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,64,power_law_1.2,0.0812287986278534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.2,1.2831616401672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,64,power_law_1.2,0.08134400248527526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,64,power_law_1.2,0.10408960580825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,64,power_law_1.2,0.11095679998397827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,64,power_law_1.2,0.1338368058204651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,64,power_law_1.2,0.15485440492630004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,64,power_law_1.2,0.20353920459747316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,64,power_law_1.2,0.2439743995666504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,64,power_law_1.2,0.37908480167388914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,64,power_law_1.2,0.49215359687805177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,64,power_law_1.2,0.7509119987487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,64,power_law_1.2,1.0035391807556153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,2,power_law_1.01,0.6486847877502442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,8,balanced,0.0661653329928716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,8,balanced,0.0928000013033549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,8,balanced,0.13089600205421448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,8,balanced,0.22254933913548788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,8,balanced,0.351637323697408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,8,balanced,0.45950933297475177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,8,balanced,0.45849064985911053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,8,balanced,0.4540640115737915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,8,balanced,0.457205335299174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,8,balanced,0.4564799865086873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,8,balanced,0.45786134401957196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,8,balanced,0.465989351272583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,8,balanced,0.4656053384145101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,8,balanced,0.46776000658671063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,8,balanced,0.47596800327301025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,8,balanced,0.478005329767863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,8,balanced,0.48581866423288983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,8,balanced,0.5762720108032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,8,balanced,0.5238720178604126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,8,balanced,0.6664160092671713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,8,balanced,0.5981119871139526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,8,balanced,0.8669013182322184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,8,balanced,0.7208639780680338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,8,balanced,1.136138677597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,8,balanced,1.0615519682566326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,8,balanced,1.7250827153523762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,8,balanced,1.6432852745056152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,64,power_law_1.01,0.5533504009246826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,64,power_law_1.01,0.08120319843292237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,64,power_law_1.01,0.07963520288467407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,64,power_law_1.01,0.08405759930610657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,64,power_law_1.01,0.07753599882125854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,64,power_law_1.01,0.07768319845199585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,64,power_law_1.01,0.08310400247573853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,64,power_law_1.01,0.0880128026008606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,64,power_law_1.01,0.08865280151367187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,64,power_law_1.01,0.08659200072288513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,64,power_law_1.01,0.08858240246772767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,64,power_law_1.01,0.09103999733924865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,64,power_law_1.01,0.0913919985294342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,64,power_law_1.01,0.09587200284004212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,64,power_law_1.01,0.10295039415359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,64,power_law_1.01,0.10671360492706299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,64,power_law_1.01,0.11192320585250855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,64,power_law_1.01,0.12713600397109986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,64,power_law_1.01,0.1388800024986267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,64,power_law_1.01,0.181222403049469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,64,power_law_1.01,0.20151040554046631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,64,power_law_1.01,0.2602880001068115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,64,power_law_1.01,0.3181056022644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,64,power_law_1.01,0.43182082176208497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,64,power_law_1.01,0.513804817199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,64,power_law_1.01,0.7059967994689942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,64,power_law_1.01,0.943718433380127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,8,power_law_1.01,0.0646336019039154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,8,power_law_1.01,0.06446080207824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,8,power_law_1.01,0.06509439945220948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,8,power_law_1.01,0.06514559984207154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,8,power_law_1.01,0.06515840291976929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,8,power_law_1.01,0.06587520241737366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,8,power_law_1.01,0.06805760264396668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,8,power_law_1.01,0.07256960272789001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,8,power_law_1.01,0.07336320281028748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,8,power_law_1.01,0.07916160225868225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,8,power_law_1.01,0.07702400088310242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,8,power_law_1.01,0.08288000226020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,8,power_law_1.01,0.08885759711265565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,8,power_law_1.01,0.09060479998588562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,8,power_law_1.01,0.11266560554504394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,8,power_law_1.01,0.1250496029853821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,8,power_law_1.01,0.12175999879837036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,8,power_law_1.01,0.15971839427947998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,8,power_law_1.01,0.18922239542007446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,8,power_law_1.01,0.24477438926696776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,8,power_law_1.01,0.28792319297790525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,8,power_law_1.01,0.459065580368042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,8,power_law_1.01,0.5268415927886962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,8,power_law_1.01,0.8575743675231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,8,balanced,0.05991466840108236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,8,balanced,0.05902933577696482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,8,balanced,0.05991466840108236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,8,balanced,0.06268799801667531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,8,balanced,0.06584000090758006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,8,balanced,0.06434666613737743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,8,balanced,0.06504533191521962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,8,balanced,0.0666720022757848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,8,balanced,0.07049066821734111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,8,power_law_1.01,1.0146112442016602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,8,balanced,0.07187733550866444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,8,balanced,0.07036800185839336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,8,power_law_1.01,1.446771240234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,8,balanced,0.07431999842325847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,8,balanced,0.07459733386834462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,8,balanced,0.08309866487979889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,8,balanced,0.09077866872151692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,8,balanced,0.08452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,8,balanced,0.09500799576441447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,8,balanced,0.10091200470924377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,8,balanced,0.10505066315333049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,8,balanced,0.13177067041397095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,8,balanced,0.13890133301417032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,8,balanced,0.20066134134928384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,8,balanced,0.2357920010884603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,8,balanced,0.338485320409139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,8,balanced,0.42052265008290607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,8,balanced,0.6217919985453287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,8,balanced,0.7970720132191976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,2,power_law_1.01,0.050886398553848265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,2,power_law_1.01,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,2,power_law_1.01,0.06813439726829529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,2,power_law_1.01,0.08053759932518005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,2,power_law_1.01,0.08323839902877808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,2,power_law_1.01,0.08440960049629212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,2,power_law_1.01,0.08540800213813782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,2,power_law_1.01,0.08947200179100037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,2,power_law_1.01,0.09895679950714112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,2,power_law_1.01,0.10456960201263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,2,power_law_1.01,0.09384959936141968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,2,power_law_1.01,0.09537919759750366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,2,power_law_1.01,0.09617919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,2,power_law_1.01,0.09917439818382263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,2,power_law_1.01,0.11322879791259766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,2,power_law_1.01,0.13080320358276368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,2,power_law_1.01,0.13155200481414794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,2,power_law_1.01,0.14708479642868041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,2,power_law_1.01,0.15903359651565552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,2,power_law_1.01,0.19896960258483887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,8,power_law_1.01,2.029363250732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,2,power_law_1.01,0.24261760711669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,2,power_law_1.01,0.33651199340820315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,16,power_law_1.2,0.044863998889923096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,16,power_law_1.2,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,16,power_law_1.2,0.04602240025997162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,16,power_law_1.2,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,16,power_law_1.2,0.04514560103416443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,2,power_law_1.01,0.43496317863464357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,2,power_law_1.01,0.051123201847076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,16,power_law_1.2,0.044435200095176694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,16,power_law_1.2,0.04568319916725159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,16,power_law_1.2,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,16,power_law_1.2,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,16,power_law_1.2,0.04689919948577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,16,power_law_1.2,0.046758401393890384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,32,power_law_1.2,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,16,power_law_1.2,0.048876801133155824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,16,power_law_1.2,0.05022720098495483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,32,power_law_1.2,0.04292480051517487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,16,power_law_1.2,0.05128960013389587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,32,power_law_1.2,0.041945600509643556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,32,power_law_1.2,0.04421760141849518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,32,power_law_1.2,0.04559360146522522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,32,power_law_1.2,0.046649599075317384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,2,power_law_1.01,0.05644800066947937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,16,power_law_1.2,0.05600000023841858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,32,power_law_1.2,0.047353601455688475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,2,power_law_1.01,0.07980800271034241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,32,power_law_1.2,0.0486847996711731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,2,power_law_1.01,0.09176959991455078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,32,power_law_1.2,0.04894079864025116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,2,power_law_1.01,0.12002559900283813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,32,power_law_1.2,0.049702399969100954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,2,power_law_1.01,0.14582400321960448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,32,power_law_1.2,0.04986880123615265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,2,power_law_1.01,0.1515776038169861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,16,power_law_1.2,0.05592319965362549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,16,power_law_1.2,0.06004480123519897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,2,power_law_1.01,0.5493375778198242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,16,power_law_1.2,0.06766719818115234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,16,power_law_1.2,0.07484800219535828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,16,power_law_1.2,0.09234560132026673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,32,power_law_1.2,0.0514303982257843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,32,power_law_1.2,0.05214080214500427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,2,power_law_1.01,0.18959360122680663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,32,power_law_1.2,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,2,power_law_1.01,0.1913472056388855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,2,power_law_1.01,0.20005760192871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,16,power_law_1.2,0.09952639937400817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,2,power_law_1.01,0.1991744041442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,16,power_law_1.2,0.12405120134353638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,2,power_law_1.01,0.21018240451812745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,16,power_law_1.2,0.15978879928588868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,32,power_law_1.2,0.058284801244735715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,2,power_law_1.01,0.21350400447845458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,32,power_law_1.2,0.05909759998321533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,32,power_law_1.2,0.06529920101165772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,32,power_law_1.2,0.07182080149650574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,32,power_law_1.2,0.08346239924430847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,32,power_law_1.2,0.10376319885253907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,16,power_law_1.2,0.20856959819793702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,2,power_law_1.01,0.22926080226898193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,32,power_law_1.2,0.11932159662246704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,16,power_law_1.2,0.2631360054016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,16,power_law_1.2,0.36720640659332277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,2,power_law_1.01,0.2431999921798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,16,power_law_1.2,0.49784321784973146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,2,power_law_1.01,0.2553215980529785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,32,power_law_1.2,0.1499008059501648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,32,power_law_1.2,0.189900803565979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,32,power_law_1.2,0.2694144010543823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,32,power_law_1.2,0.34596478939056396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,2,power_law_1.01,0.27577600479125974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,32,power_law_1.2,0.5128960132598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,2,power_law_1.01,0.32787199020385743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,32,power_law_1.2,0.8004351615905761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,2,power_law_1.01,0.33605759143829345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,2,power_law_1.01,0.41292800903320315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,2,power_law_1.01,0.471065616607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,2,power_law_1.01,0.6093952178955078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,2,power_law_1.01,0.7162816047668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,2,power_law_1.01,0.980742359161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,2,power_law_1.01,1.3118592262268067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,2,power_law_1.01,1.765964889526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,2,power_law_1.01,2.1810752868652346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,128,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,128,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,128,balanced,0.03385599950949351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,128,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,128,balanced,0.03751466671625773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,128,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,128,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,128,balanced,0.0394400010506312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,128,balanced,0.039488000174363456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,128,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,128,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,128,balanced,0.03871466716130575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,128,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,128,balanced,0.039733332892258964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,128,balanced,0.04351999859015147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,128,balanced,0.043920000394185386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,128,balanced,0.04524800181388855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,128,balanced,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,128,balanced,0.04781333108743032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,128,balanced,0.052853330969810486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,128,balanced,0.054714664816856384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,128,balanced,0.06065066655476888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,128,balanced,0.06633600095907848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,128,balanced,0.09842133522033691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,128,balanced,0.12100799878438313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,128,balanced,0.1572106679280599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,128,balanced,0.19421867529551187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,2,power_law_1.01,0.6448512077331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,2,power_law_1.01,0.8792832374572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,2,power_law_1.01,1.1836095809936524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,4,balanced,0.02126399924357732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,4,balanced,0.023152001202106476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,4,balanced,0.025013332565625507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,4,balanced,0.02958400050799052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,4,balanced,0.04781866570313772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,4,balanced,0.04952533543109894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,4,balanced,0.052576000491778054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,4,balanced,0.054085334142049156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,4,balanced,0.062650665640831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,4,balanced,0.060080001751581825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,4,balanced,0.061520000298817955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,4,balanced,0.08403199911117554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,4,balanced,0.08075733482837677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,4,balanced,0.0820853312810262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,4,balanced,0.1011306643486023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,4,balanced,0.10443199674288432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,4,balanced,0.10899733503659566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,4,balanced,0.11909866333007812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,4,balanced,0.12469866871833801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,4,balanced,0.13865066568056741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,4,balanced,0.14966400464375815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,4,balanced,0.16272000471750894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,4,balanced,0.1750026742617289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,4,balanced,0.2800266742706299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,4,balanced,0.3100000023841858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,4,balanced,0.44620800018310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,4,balanced,0.5800586541493734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,8,balanced,0.041850666205088295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,8,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,8,balanced,0.04391466577847799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,8,balanced,0.05009066561857859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,8,balanced,0.05831466615200043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,8,balanced,0.07021866738796234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,8,balanced,0.0706879993279775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,8,balanced,0.07221866647402446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,8,balanced,0.07261333366235097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,8,balanced,0.0732426643371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,8,balanced,0.07363733152548473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,8,balanced,0.07433600227038066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,8,balanced,0.07869333525498708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,8,balanced,0.07906666894753774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,8,balanced,0.08688533306121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,8,balanced,0.08899733424186707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,8,balanced,0.09274133046468098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,8,balanced,0.10541333754857381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,8,balanced,0.11333333452542622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,8,balanced,0.14321066935857138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,8,balanced,0.15244799852371216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,8,balanced,0.2050293286641439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,8,balanced,0.24046399195988974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,8,balanced,0.3439519802729289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,4,power_law_1.2,0.05414400100708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,4,power_law_1.2,0.0607807993888855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,8,balanced,0.4161440134048462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,4,power_law_1.2,0.05281280279159546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,4,power_law_1.2,0.06622719764709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,8,balanced,0.5814453363418579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,4,power_law_1.2,0.06690559983253479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,4,power_law_1.2,0.06695680022239685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,8,balanced,0.7504533131917318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,4,power_law_1.2,0.06767359972000123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,4,power_law_1.2,0.06984959840774536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,4,power_law_1.2,0.07035520076751708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,4,power_law_1.2,0.07219840288162231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,4,power_law_1.2,0.07289599776268005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,4,power_law_1.2,0.07482879757881164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,4,power_law_1.2,0.07885439991950989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,4,power_law_1.2,0.08358399868011475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,4,power_law_1.2,0.09375360012054443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,4,power_law_1.2,0.09964159727096558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,4,power_law_1.2,0.1064255952835083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,4,power_law_1.2,0.12600959539413453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.2,0.14479999542236327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.2,0.20161280632019044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.2,0.23201279640197753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.2,0.29955201148986815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,2,power_law_1.2,0.03412480056285858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,2,power_law_1.2,0.03336319923400879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,2,power_law_1.2,0.03340159952640533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,2,power_law_1.2,0.03546879887580871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,2,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,2,power_law_1.2,0.036447998881340024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,2,power_law_1.2,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,2,power_law_1.2,0.03695360124111176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,2,power_law_1.2,0.03827199935913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,2,power_law_1.2,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,2,power_law_1.2,0.03896960020065308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,2,power_law_1.2,0.041043201088905336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,2,power_law_1.2,0.04108799993991852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,2,power_law_1.2,0.04142720103263855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.2,0.42863998413085935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,2,power_law_1.2,0.04848000109195709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,2,power_law_1.2,0.05023999810218811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,2,power_law_1.2,0.052960002422332765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,2,power_law_1.2,0.05982080101966858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,2,power_law_1.2,0.06204800009727478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,2,power_law_1.2,0.08136320114135742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,2,power_law_1.2,0.08824959993362427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,2,power_law_1.2,0.1206015944480896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,8,power_law_1.01,0.06687999963760376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,8,power_law_1.01,0.09268479943275451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,8,power_law_1.01,0.1423424005508423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,8,power_law_1.01,0.13338880538940429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,8,power_law_1.01,0.14759039878845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,8,power_law_1.01,0.1651968002319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,8,power_law_1.01,0.16019200086593627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,8,power_law_1.01,0.18136320114135743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,8,power_law_1.01,0.17632639408111572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,8,power_law_1.01,0.1721343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,8,power_law_1.01,0.16782079935073851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,2,power_law_1.2,0.13867520093917846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,8,power_law_1.01,0.1827839970588684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.2,0.5067776203155517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,8,power_law_1.01,0.17015680074691772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,8,power_law_1.01,0.1862912058830261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,8,power_law_1.01,0.19181439876556397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,8,power_law_1.01,0.2095103979110718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,8,power_law_1.01,0.20520319938659667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,8,power_law_1.01,0.223142409324646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,8,power_law_1.01,0.24395520687103273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,8,power_law_1.01,0.28183679580688475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,8,power_law_1.01,0.2969088077545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,8,power_law_1.01,0.3775552034378052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,8,power_law_1.01,0.41665282249450686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,8,power_law_1.01,0.5169151782989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,8,power_law_1.01,0.6534976005554199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,8,power_law_1.01,0.8761471748352051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,8,power_law_1.01,1.126425552368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,4,power_law_1.2,0.08281599879264831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,4,power_law_1.2,0.0983295977115631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,4,power_law_1.2,0.10430079698562622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,4,power_law_1.2,0.12502399682998658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,4,power_law_1.2,0.13982720375061036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,8,power_law_1.01,0.06785920262336731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,4,power_law_1.2,0.14143999814987182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,8,power_law_1.01,0.07715200185775757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,8,power_law_1.01,0.07322880029678344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,8,power_law_1.01,0.07812479734420777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,8,power_law_1.01,0.0838591992855072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,8,power_law_1.01,0.09190400242805481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,8,power_law_1.01,0.09401599764823913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,4,power_law_1.2,0.16205439567565919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,8,power_law_1.01,0.09477120041847228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,4,power_law_1.2,0.1684224009513855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,2,power_law_1.2,0.17715840339660643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,8,power_law_1.01,0.09477120041847228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,4,power_law_1.2,0.17657599449157715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,4,power_law_1.2,0.17360639572143555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,4,power_law_1.2,0.1820736050605774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,4,power_law_1.2,0.17166719436645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,4,power_law_1.2,0.1863103985786438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,8,power_law_1.01,0.09583359956741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,4,power_law_1.2,0.192576003074646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,8,power_law_1.01,0.09617919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,8,power_law_1.01,0.09917439818382263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,8,power_law_1.01,0.0992959976196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,8,power_law_1.01,0.10169600248336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,4,power_law_1.2,0.2008960008621216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,4,power_law_1.2,0.20341761112213136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,4,power_law_1.2,0.21024000644683838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,4,power_law_1.2,0.2283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,8,power_law_1.01,0.10748159885406494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,4,power_law_1.2,0.24873600006103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,8,power_law_1.01,0.10767359733581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,8,power_law_1.01,0.11433600187301636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,4,power_law_1.2,0.2934528112411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,8,power_law_1.01,0.12269439697265624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,4,power_law_1.2,0.34812800884246825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,8,power_law_1.01,0.14833920001983641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,4,power_law_1.2,0.4520127773284912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,8,power_law_1.01,0.18170880079269408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,8,power_law_1.01,0.1970752000808716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,4,power_law_1.2,0.5871359825134277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,8,power_law_1.01,0.2595455884933472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,4,power_law_1.2,0.8152383804321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,8,power_law_1.01,0.33822081089019773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,4,power_law_1.2,0.9926079750061035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,4,power_law_1.2,1.5308799743652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,4,power_law_1.2,1.810643196105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,8,power_law_1.01,0.48227200508117674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,8,power_law_1.01,0.6136256217956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.2,0.7688640117645263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,8,power_law_1.01,0.858732795715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,32,1,balanced,0.021802666286627453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,8,power_law_1.01,1.1385984420776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,32,1,balanced,0.023013333479563396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,32,1,balanced,0.023792001108328503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,32,1,balanced,0.02752533306678136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,32,1,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,32,1,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,32,1,balanced,0.058245331048965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,32,1,balanced,0.06011733412742615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,32,1,balanced,0.0637546678384145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,32,1,balanced,0.065610667069753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,32,1,balanced,0.07015466690063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,32,1,balanced,0.07460799813270569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,32,1,balanced,0.08036800225575765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,32,1,balanced,0.08885332942008972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,32,1,balanced,0.09129599730173747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,32,1,balanced,0.09670399626096089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,32,1,balanced,0.10932266712188721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,32,1,balanced,0.10866666833559673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,32,1,balanced,0.12165866295496623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,32,1,balanced,0.145989328622818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,32,1,balanced,0.16602133711179098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,32,1,balanced,0.2376586596171061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,2,power_law_1.2,0.24110078811645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,32,1,balanced,0.2691253423690796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,32,1,balanced,0.3152533372243245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,32,1,balanced,0.35711467266082764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,32,1,balanced,0.561621348063151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,32,1,balanced,0.6560480197270712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,64,balanced,0.05226666728655497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,64,balanced,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,64,balanced,0.05202133456865946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,64,balanced,0.055973331133524575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,64,balanced,0.056261335810025535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,64,balanced,0.06237333516279856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,64,balanced,0.06101333101590475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,64,balanced,0.06276266773541768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,64,balanced,0.06643733382225037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,64,balanced,0.06634133557478587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,64,balanced,0.06645333270231883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,64,balanced,0.06625600159168243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,64,balanced,0.06980266670385997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,64,balanced,0.0684746652841568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,64,balanced,0.07557866473992665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,64,balanced,0.07463466624418895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,64,balanced,0.07828266421953838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,64,balanced,0.08685866991678874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,64,balanced,0.09286399682362874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,64,balanced,0.11945066849390666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,64,balanced,0.13989866773287454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,64,balanced,0.18003199497858682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,64,balanced,0.21320533752441406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,64,balanced,0.2858826716740926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,64,balanced,0.35809600353240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,64,balanced,0.49872533480326336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,64,balanced,0.6387146711349487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.2,1.0531968116760253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,2,power_law_1.2,0.36090240478515623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.2,1.4158464431762696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,2,power_law_1.2,0.48842878341674806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,128,power_law_1.2,0.05390080213546753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,128,power_law_1.2,0.05235840082168579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,128,power_law_1.2,0.05055360198020935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,128,power_law_1.2,0.057599997520446776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,128,power_law_1.2,0.055801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,128,power_law_1.2,0.05672320127487183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,128,power_law_1.2,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,128,power_law_1.2,0.0582144021987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,128,power_law_1.2,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,128,power_law_1.2,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,128,power_law_1.2,0.06944000124931335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,128,power_law_1.2,0.07307519912719726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,128,power_law_1.2,0.07669119834899903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,128,power_law_1.2,0.07801600098609925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,128,power_law_1.2,0.08595200181007386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,128,power_law_1.2,0.09056640267372132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,128,power_law_1.2,0.09919999837875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,128,power_law_1.2,0.11735680103302001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,128,power_law_1.2,0.12237440347671509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,128,power_law_1.2,0.15884159803390502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,128,power_law_1.2,0.1765439987182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,128,power_law_1.2,0.2345344066619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,128,power_law_1.2,0.24140799045562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,128,power_law_1.2,0.329964804649353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,128,power_law_1.2,0.41403517723083494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,128,power_law_1.2,0.630022382736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,128,power_law_1.2,0.7833024024963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,2,power_law_1.2,0.025900799036026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,2,power_law_1.2,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,2,power_law_1.2,0.01844480037689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,2,power_law_1.2,0.05457280278205871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,2,power_law_1.2,0.02157440036535263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,2,power_law_1.2,0.0247296005487442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,2,power_law_1.2,0.07929599881172181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,2,power_law_1.2,0.02656640112400055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,2,power_law_1.2,0.10038399696350098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,2,power_law_1.2,0.03229439854621887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,2,power_law_1.2,0.1189695954322815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,2,power_law_1.2,0.04208639860153198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,2,power_law_1.2,0.1597759962081909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,2,power_law_1.2,0.17215360403060914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,2,power_law_1.2,0.04321280121803284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,2,power_law_1.2,0.18237440586090087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,2,power_law_1.2,0.04446719884872437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,2,power_law_1.2,0.04523519873619079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,2,power_law_1.2,0.0455487996339798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,2,power_law_1.2,0.04665600061416626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,2,power_law_1.2,0.1799232006072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,2,power_law_1.2,0.04793600142002106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,2,power_law_1.2,0.19552639722824097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,2,power_law_1.2,0.05061759948730469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,2,power_law_1.2,0.18904320001602173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,2,power_law_1.2,0.05103359818458557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,2,power_law_1.2,0.20129919052124023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,2,power_law_1.2,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,2,power_law_1.2,0.20865280628204347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,2,power_law_1.2,0.21172480583190917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,2,power_law_1.2,0.21930239200592042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,2,power_law_1.2,0.227891206741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,2,power_law_1.2,0.2534912109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,2,power_law_1.2,0.05569919943809509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,2,power_law_1.2,0.26280961036682127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,2,power_law_1.2,0.05931519865989685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,2,power_law_1.2,0.06311039924621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,2,power_law_1.2,0.5716159820556641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,2,power_law_1.2,0.09381120204925537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,2,power_law_1.2,0.11430399417877198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,2,power_law_1.2,0.631667184829712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,2,power_law_1.2,0.1041983962059021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,2,power_law_1.2,0.46435837745666503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,2,power_law_1.2,0.12316160202026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,2,power_law_1.2,0.49427199363708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,2,power_law_1.2,0.6013823986053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,2,power_law_1.2,0.67259521484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,2,power_law_1.2,0.9398591995239258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,2,power_law_1.2,1.1518400192260743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,2,power_law_1.2,0.14774399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,2,power_law_1.2,0.19258879423141478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,2,power_law_1.2,0.23668479919433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,16,power_law_1.2,0.12066559791564942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,16,power_law_1.2,0.13903360366821288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,16,power_law_1.2,0.14505599737167357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,16,power_law_1.2,0.1568511962890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,16,power_law_1.2,0.15368319749832154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,16,power_law_1.2,0.14812159538269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,16,power_law_1.2,0.14256000518798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,16,power_law_1.2,0.14479360580444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,16,power_law_1.2,0.14439680576324462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,16,power_law_1.2,0.15336320400238038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,16,power_law_1.2,0.1540735960006714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,16,power_law_1.2,0.1560320019721985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,16,power_law_1.2,0.1571776032447815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,16,power_law_1.2,0.1625856041908264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,16,power_law_1.2,0.17127039432525634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,16,power_law_1.2,0.1741055965423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,16,power_law_1.2,0.18554879426956178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,16,power_law_1.2,0.20624001026153566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,16,power_law_1.2,0.2334399938583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,1,power_law_1.2,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,16,power_law_1.2,0.2855999946594238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,1,power_law_1.2,0.038815999031066896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,1,power_law_1.2,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,16,power_law_1.2,0.3154047966003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,1,power_law_1.2,0.04565759897232056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,16,power_law_1.2,0.4333375930786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,1,power_law_1.2,0.046412798762321475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,1,power_law_1.2,0.049439999461174014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,16,power_law_1.2,0.4272448062896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,1,power_law_1.2,0.050169599056243894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,16,power_law_1.2,0.5236544132232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,16,power_law_1.2,0.688595199584961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,16,power_law_1.2,1.0169407844543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,1,power_law_1.2,0.04927999973297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,1,power_law_1.2,0.050419199466705325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,1,power_law_1.2,0.05081599950790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,1,power_law_1.2,0.05237119793891907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,1,power_law_1.2,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,1,power_law_1.2,0.06002560257911682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,8,balanced,0.03373866776625315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,1,power_law_1.2,0.06437119841575623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,8,balanced,0.03474666674931844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,8,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,1,power_law_1.2,0.07943040132522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,8,balanced,0.03812800099452337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,8,balanced,0.04423466821511587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,8,balanced,0.0446720023949941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,8,balanced,0.04576000074545542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,1,power_law_1.2,0.08391680121421814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,8,balanced,0.04667733112970988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,8,balanced,0.04632533093293508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,8,balanced,0.04594666759173075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,8,balanced,0.04755199948946635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,8,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,8,balanced,0.04965866605440775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,8,balanced,0.05027199784914652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,16,power_law_1.2,1.2357760429382325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,8,balanced,0.0568800022204717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,8,balanced,0.0572213331858317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,8,balanced,0.058133333921432495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,1,power_law_1.2,0.08533759713172913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,8,balanced,0.062090665102005005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,8,balanced,0.06742933392524719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,8,balanced,0.07654933134714763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,8,balanced,0.0824533353249232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,8,balanced,0.10700266559918721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,8,balanced,0.12071466445922852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,8,balanced,0.15744533141454062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,8,balanced,0.19368533293406168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,1,power_law_1.2,0.09878399968147278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,8,balanced,0.26869867245356244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,8,balanced,0.35627734661102295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.2,0.12385280132293701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,2,power_law_1.2,0.32257919311523436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.2,0.16398080587387084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.2,0.18496639728546144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.2,0.287718391418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.2,0.3643584012985229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,2,power_law_1.2,0.40665597915649415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.2,0.5317887783050537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,1,power_law_1.01,0.1605568051338196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,1,power_law_1.01,0.2480384111404419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,1,power_law_1.01,0.36903679370880127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,1,power_law_1.01,0.575494384765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,1,power_law_1.01,0.8000255584716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,1,power_law_1.01,1.0178943634033204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,1,power_law_1.01,1.3375871658325196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,1,power_law_1.01,1.4320384025573731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,1,power_law_1.01,1.4729599952697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,1,power_law_1.01,1.4624896049499512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,1,power_law_1.01,1.5805312156677247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,1,power_law_1.01,1.5409279823303224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,1,power_law_1.01,1.5880319595336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,1,power_law_1.01,1.643449592590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,1,power_law_1.01,1.626495933532715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,1,power_law_1.01,1.6530879974365233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,1,power_law_1.01,1.6323135375976563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,1,power_law_1.01,1.8160768508911134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.01,1.8676223754882812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.2,0.6775807857513427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.01,2.1083263397216796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.01,2.344371223449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.01,2.7691648483276365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.01,3.1941503524780273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.01,4.027974319458008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.01,4.890604782104492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.01,6.559763336181641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.01,8.142323303222657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,8,balanced,0.038549333810806274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,8,balanced,0.03790933390458425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.2,1.0148223876953124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,8,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,8,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,8,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,8,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,8,balanced,0.039818666875362396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,8,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,8,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,8,balanced,0.041477332512537636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,8,balanced,0.04170133173465729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,8,balanced,0.04177600145339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,8,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,8,balanced,0.044362664222717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,8,balanced,0.0470773329337438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,8,balanced,0.05002133548259735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,8,balanced,0.05053866902987162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,8,balanced,0.05574933191140493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,8,balanced,0.056474665800730385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,8,balanced,0.06771199901898702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,8,balanced,0.08111999928951263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,8,balanced,0.1106773316860199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,8,balanced,0.132314662138621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,8,balanced,0.17678399880727133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,8,balanced,0.21809067328770956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,8,balanced,0.3084320028622945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,8,balanced,0.38892801602681476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.2,1.332921600341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,1,balanced,0.021087999145189922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,1,balanced,0.02124800036350886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,1,balanced,0.021194666624069214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,1,balanced,0.023007998863856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,1,balanced,0.02499199906984965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,1,balanced,0.029472000896930695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,1,balanced,0.029157333076000214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,1,balanced,0.03014933317899704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,1,balanced,0.029120000700155895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,1,balanced,0.029690665503342945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,1,balanced,0.03259733319282532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,1,balanced,0.03137599925200144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,1,balanced,0.03357866654793421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,1,balanced,0.03585066646337509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,1,balanced,0.0743999977906545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,1,balanced,0.07659199833869934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,1,balanced,0.08070933322111766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,1,balanced,0.057328000664711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,1,balanced,0.06195199986298879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,1,balanced,0.08281066517035167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,1,balanced,0.10322667161623637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,1,balanced,0.1383093297481537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,1,balanced,0.16748799880345663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,1,balanced,0.2374239961306254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,1,balanced,0.3049013415972392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,1,balanced,0.4396853446960449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,1,balanced,0.5746666590372721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,2,power_law_1.2,0.021510399878025055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,2,power_law_1.2,0.02261119931936264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,2,power_law_1.2,0.02218240052461624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,2,power_law_1.2,0.023238399624824525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,2,power_law_1.2,0.029075199365615846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,2,power_law_1.2,0.0298880010843277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,2,power_law_1.2,0.028627198934555054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,2,power_law_1.2,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,2,power_law_1.2,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,2,power_law_1.2,0.031097599864006044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,2,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,2,power_law_1.2,0.073471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,2,power_law_1.2,0.07439360022544861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,2,power_law_1.2,0.0788927972316742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,2,power_law_1.2,0.05886719822883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,2,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,2,power_law_1.2,0.058982402086257935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,2,power_law_1.2,0.0728767991065979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.2,0.0904640018939972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.2,0.12588160037994384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.2,0.1505728006362915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,16,power_law_1.01,0.12488960027694702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,16,power_law_1.01,0.1313088059425354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,16,power_law_1.01,0.11178879737854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,16,power_law_1.01,0.1343999981880188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,16,power_law_1.01,0.13010560274124144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,16,power_law_1.01,0.13306879997253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,16,power_law_1.01,0.12912000417709352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.2,0.21107199192047119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,16,power_law_1.01,0.13842560052871705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,16,power_law_1.01,0.13605120182037353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,16,power_law_1.01,0.1360576033592224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,16,power_law_1.01,0.13612159490585327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,16,power_law_1.01,0.13896960020065308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,16,power_law_1.01,0.1392832040786743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,16,power_law_1.01,0.14506239891052247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,16,power_law_1.01,0.1499776005744934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,16,power_law_1.01,0.15271040201187133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,16,power_law_1.01,0.1664639949798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,16,power_law_1.01,0.1836351990699768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,16,power_law_1.01,0.194758403301239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,16,power_law_1.01,0.22728960514068602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,16,power_law_1.01,0.28273279666900636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,16,power_law_1.01,0.336025595664978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,16,power_law_1.01,0.3917952060699463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.2,0.26582400798797606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,16,power_law_1.01,0.5970176219940185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,16,power_law_1.01,0.7837247848510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,16,power_law_1.01,1.076249599456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,32,power_law_1.01,0.07847679853439331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,16,power_law_1.01,1.4352383613586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,32,power_law_1.01,0.07326080203056336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,32,power_law_1.01,0.07034239768981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,32,power_law_1.01,0.07200000286102295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,32,power_law_1.01,0.07094399929046631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,32,power_law_1.01,0.06843519806861878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,32,power_law_1.01,0.07281919717788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,32,power_law_1.01,0.07212160229682922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,32,power_law_1.01,0.07381119728088378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,32,power_law_1.01,0.07390720248222352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,32,power_law_1.01,0.07347840070724487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,32,power_law_1.01,0.0763264000415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,32,power_law_1.01,0.07440000176429748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,32,power_law_1.01,0.07475839853286743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,32,power_law_1.01,0.08269439935684204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,32,power_law_1.01,0.08350080251693726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,32,power_law_1.01,0.0856768012046814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,32,power_law_1.01,0.09836159944534302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.01,0.11076480150222778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.01,0.12821120023727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.01,0.14236799478530884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.01,0.18336000442504882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.01,0.21478400230407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.01,0.2959424018859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.01,0.375052809715271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.01,0.5346240043640137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.01,0.6673151969909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.2,0.3811072111129761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.2,0.49969282150268557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,16,power_law_1.2,0.15748480558395386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,16,power_law_1.2,0.15693440437316894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,16,power_law_1.2,0.16401920318603516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,16,power_law_1.2,0.15902719497680665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,16,power_law_1.2,0.12224639654159546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,16,power_law_1.2,0.13901439905166627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,16,power_law_1.2,0.1479871988296509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,16,power_law_1.2,0.14515199661254882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,16,power_law_1.2,0.14851200580596924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,16,power_law_1.2,0.148198401927948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,16,power_law_1.2,0.13996800184249877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,16,power_law_1.2,0.15664000511169435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,16,power_law_1.2,0.14561920166015624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,16,power_law_1.2,0.15354880094528198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,16,power_law_1.2,0.1513535976409912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,16,power_law_1.2,0.1545024037361145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,16,power_law_1.2,0.17531520128250122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,16,power_law_1.2,0.197542405128479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.2,0.2205440044403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.2,0.2631040096282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.2,0.2770751953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.2,0.3623807907104492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.2,0.424505615234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.2,0.6529664039611817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.2,0.8463935852050781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.2,0.7292799949645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.2,1.1299263954162597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,4,power_law_1.2,0.13648639917373656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,4,power_law_1.2,0.20781440734863282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,4,power_law_1.2,0.19328000545501708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,4,power_law_1.2,0.2802623987197876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,4,power_law_1.2,0.3507136106491089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,4,power_law_1.2,0.3377408027648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,4,power_law_1.2,0.48385281562805177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,4,power_law_1.2,0.48551039695739745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,4,power_law_1.2,0.45399041175842286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,4,power_law_1.2,0.5021759986877441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,4,power_law_1.2,0.48321919441223143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,4,power_law_1.2,0.44603519439697265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,4,power_law_1.2,0.5076863765716553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,4,power_law_1.2,0.4972479820251465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,4,power_law_1.2,0.5100736141204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,4,power_law_1.2,0.5291391849517822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,4,power_law_1.2,0.5531199932098388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,4,power_law_1.2,0.5683712005615235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.2,0.600710391998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.2,0.6960256099700928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.2,0.8060928344726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.2,1.592416000366211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.2,0.9242495536804199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,2,balanced,0.04576000074545542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,2,balanced,0.04789333542188009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,2,balanced,0.05187733471393585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,2,balanced,0.07422400017579396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,2,balanced,0.10493866602579753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,2,balanced,0.15651200215021768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,2,balanced,0.15986133615175882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,2,balanced,0.15969066818555197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,2,balanced,0.15987199544906616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,2,balanced,0.16332266728083292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,2,balanced,0.16270933548609415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,2,balanced,0.16404267152150473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,2,balanced,0.168122669061025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,2,balanced,0.17061332861582437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,2,balanced,0.1759679913520813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,2,balanced,0.1797013282775879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.2,1.099846363067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.2,1.396889591217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.2,1.6785856246948243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,2,balanced,0.1916853388150533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,2,balanced,0.2151306668917338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,2,balanced,0.23305066426595053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,2,balanced,0.26892799139022827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,2,balanced,0.29985066254933673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,2,balanced,0.40226133664449054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,2,balanced,0.453109343846639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,2,balanced,0.6479039986928304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,2,balanced,0.7822399934132894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,2,balanced,1.1207040150960286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.2,2.31921272277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,2,balanced,1.4208745956420898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.2,3.0544000625610352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.2,0.9597503662109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,4,power_law_1.2,0.04647679924964905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,4,power_law_1.2,0.05763840079307556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,4,power_law_1.2,0.059961599111557004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,4,power_law_1.2,0.0763584017753601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,4,power_law_1.2,0.08005120158195496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,4,power_law_1.2,0.08874880075454712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,4,power_law_1.2,0.10549759864807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,4,power_law_1.2,0.10718719959259033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,4,power_law_1.2,0.10254720449447632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,4,power_law_1.2,0.10177279710769653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,4,power_law_1.2,0.1090880036354065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,4,power_law_1.2,0.11202559471130372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,4,power_law_1.2,0.11196800470352172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,1,power_law_1.01,0.04111360013484955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,1,power_law_1.01,0.048198398947715757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,1,power_law_1.01,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,1,power_law_1.01,0.0775168001651764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,1,power_law_1.01,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,1,power_law_1.01,0.0847487986087799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,1,power_law_1.01,0.0859391987323761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,1,power_law_1.01,0.08441600203514099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,1,power_law_1.01,0.08482559919357299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,4,power_law_1.2,0.11276799440383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,1,power_law_1.01,0.08698239922523499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,4,power_law_1.2,0.12248320579528808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,1,power_law_1.01,0.08578559756278992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,4,power_law_1.2,0.12873599529266358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,1,power_law_1.01,0.09141759872436524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,4,power_law_1.2,0.13384319543838502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,1,power_law_1.01,0.09107199907302857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,4,power_law_1.2,0.15049599409103392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,4,power_law_1.2,0.15842560529708863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,1,power_law_1.01,0.09841920137405395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,4,power_law_1.2,0.20044798851013185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,1,power_law_1.01,0.10880000591278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,4,power_law_1.2,0.20056960582733155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,4,power_law_1.2,0.25004799365997316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,1,power_law_1.01,0.11532800197601319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,4,power_law_1.2,0.309497594833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,4,power_law_1.2,0.39509758949279783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,4,power_law_1.2,0.49288320541381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,4,power_law_1.2,0.6551616191864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,1,power_law_1.01,0.11426559686660767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,1,power_law_1.01,0.13123199939727784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,128,power_law_1.01,0.05151360034942627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,4,power_law_1.2,0.829593563079834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,128,power_law_1.01,0.052537602186203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,128,power_law_1.01,0.05451520085334778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,128,power_law_1.01,0.055129599571228025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,1,power_law_1.01,0.14273279905319214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,128,power_law_1.01,0.05356159806251526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,128,power_law_1.01,0.05356159806251526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,128,power_law_1.01,0.05432959794998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,128,power_law_1.01,0.055155199766159055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,128,power_law_1.01,0.05594879984855652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,128,power_law_1.01,0.05623679757118225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,128,power_law_1.01,0.058380800485610965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,1,power_law_1.01,0.18012160062789917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,128,power_law_1.01,0.06474239826202392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,128,power_law_1.01,0.06321920156478882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,128,power_law_1.01,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,128,power_law_1.01,0.07214720249176025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,128,power_law_1.01,0.07215359807014465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,128,power_law_1.01,0.07839360237121581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,128,power_law_1.01,0.09797760248184204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,128,power_law_1.01,0.10339839458465576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,128,power_law_1.01,0.1302783966064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,128,power_law_1.01,0.14974720478057862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,128,power_law_1.01,0.2026047945022583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,128,power_law_1.01,0.2158911943435669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,128,power_law_1.01,0.2766272068023682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,1,power_law_1.01,0.19998079538345337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,128,power_law_1.01,0.3299263954162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,128,power_law_1.01,0.4576064109802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,128,power_law_1.01,0.5598144054412841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,1,power_law_1.01,0.2647615909576416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,1,power_law_1.01,0.3265599966049194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,1,power_law_1.01,0.4362304210662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,1,power_law_1.01,0.5385216236114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,1,power_law_1.01,0.7493184089660645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,8,power_law_1.2,0.03054080009460449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,8,power_law_1.2,0.029139199852943422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,8,power_law_1.2,0.028454399108886717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,8,power_law_1.2,0.030752000212669373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,8,power_law_1.2,0.03059839904308319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,8,power_law_1.2,0.03167999982833862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,8,power_law_1.2,0.03256320059299469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,8,power_law_1.2,0.033817601203918454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,8,power_law_1.2,0.03594239950180054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,8,power_law_1.2,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,8,power_law_1.2,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,8,power_law_1.2,0.037561601400375365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,8,power_law_1.2,0.03984000086784363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,8,power_law_1.2,0.04087679982185364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,8,power_law_1.2,0.046463999152183535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,8,power_law_1.2,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,8,power_law_1.2,0.0532800018787384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,8,power_law_1.2,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,8,power_law_1.2,0.06989439725875854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,8,power_law_1.2,0.0487744003534317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,1,power_law_1.01,0.9771903991699219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,8,power_law_1.2,0.050355201959609984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,8,power_law_1.2,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,8,power_law_1.2,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,8,power_law_1.2,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,8,power_law_1.2,0.06247040033340454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,8,power_law_1.2,0.06412799954414368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,8,power_law_1.01,0.02033279985189438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,8,power_law_1.01,0.019897599518299103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,8,power_law_1.2,0.0871999979019165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,8,power_law_1.01,0.01984640061855316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,8,power_law_1.01,0.020153599977493285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,8,power_law_1.01,0.020665599405765532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,8,power_law_1.01,0.022252799570560457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,8,power_law_1.01,0.025593599677085875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,8,power_law_1.01,0.023321600258350374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,8,power_law_1.01,0.02503040134906769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,8,power_law_1.01,0.027801600098609925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,8,power_law_1.01,0.025862398743629455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,8,power_law_1.01,0.02800000011920929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,8,power_law_1.01,0.030028799176216127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,8,power_law_1.01,0.03043839931488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,8,power_law_1.01,0.0315775990486145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,8,power_law_1.2,0.10720000267028809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,8,power_law_1.01,0.0336896002292633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,8,power_law_1.01,0.05121279954910278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,8,power_law_1.01,0.06577919721603394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,8,power_law_1.2,0.06604160070419311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.01,0.07385600209236146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,8,power_law_1.2,0.06780800223350525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.01,0.08730880022048951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,8,power_law_1.2,0.06837760210037232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,8,power_law_1.2,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,8,power_law_1.2,0.07251840233802795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,8,power_law_1.2,0.07419520020484924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,8,power_law_1.2,0.07820799946784973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,8,power_law_1.2,0.08526719808578491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,8,power_law_1.2,0.08858879804611205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.01,0.07303680181503296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,8,power_law_1.2,0.09052159786224365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,8,power_law_1.2,0.1059648036956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,8,power_law_1.2,0.11527680158615113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,8,power_law_1.2,0.14406399726867675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,8,power_law_1.2,0.1656000018119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,8,power_law_1.2,0.22042880058288575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.01,0.08863999843597412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,8,power_law_1.2,0.28008959293365476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.01,0.10826239585876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,8,power_law_1.2,0.38197760581970214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.01,0.14657280445098878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,8,power_law_1.2,0.4975743770599365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.01,0.173363196849823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.01,0.2445823907852173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,8,power_law_1.2,0.7085440158843994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.01,0.3197887897491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,8,power_law_1.2,0.765715217590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,8,power_law_1.2,0.1435968041419983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,8,power_law_1.2,0.16969599723815917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,8,power_law_1.2,0.2329024076461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,8,power_law_1.2,0.28408958911895754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,8,power_law_1.2,0.41308159828186036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,2,power_law_1.01,0.07693439722061157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,2,power_law_1.01,0.09893119931221009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,2,power_law_1.01,0.14195200204849243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,2,power_law_1.01,0.17377279996871947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,2,power_law_1.01,0.2340480089187622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,2,power_law_1.01,0.3420480012893677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,2,power_law_1.01,0.3370687961578369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,2,power_law_1.01,0.34306559562683103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,2,power_law_1.01,0.33644800186157225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,2,power_law_1.01,0.3656512022018433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,2,power_law_1.01,0.3666815996170044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,2,power_law_1.01,0.3767296075820923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,2,power_law_1.01,0.37315199375152586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,2,power_law_1.01,0.3727168083190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,2,power_law_1.01,0.39127039909362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,2,power_law_1.01,0.4047935962677002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,2,power_law_1.01,0.42746877670288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,2,power_law_1.01,0.45667200088500975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,2,power_law_1.01,0.4772480010986328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,2,power_law_1.01,0.5218624114990235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,2,power_law_1.01,0.6034751892089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,2,power_law_1.01,0.6881663799285889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,2,power_law_1.01,0.8441920280456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,2,power_law_1.01,1.0564607620239257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,2,power_law_1.01,1.309875202178955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,2,power_law_1.01,1.6910272598266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,2,power_law_1.01,2.255264091491699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,1,balanced,0.019882666567961376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,1,balanced,0.019039999693632126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,1,balanced,0.019402666638294857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,1,balanced,0.021695998807748158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,1,balanced,0.027690666417280834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,1,balanced,0.02779199928045273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,1,balanced,0.027322667340437572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,1,balanced,0.02916266769170761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,1,balanced,0.029552000264326733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,1,balanced,0.02959466725587845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,1,balanced,0.0314026673634847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,1,balanced,0.03347733368476232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,1,balanced,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,1,balanced,0.03576533248027166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,1,balanced,0.03610666592915853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,1,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,1,balanced,0.03953066716591517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,1,balanced,0.050794666012128196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,1,balanced,0.05721066892147064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,1,balanced,0.0746666689713796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,1,balanced,0.08061866462230682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,1,balanced,0.10466133554776509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,1,balanced,0.10121066371599834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,1,balanced,0.15631999572118124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,1,balanced,0.16842132806777954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,1,balanced,0.23854400714238486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,1,balanced,0.3080853422482808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,8,power_law_1.2,0.570310401916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,1,balanced,0.09122666716575623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,1,balanced,0.09557867050170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,1,balanced,0.10922132929166158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,1,balanced,0.13687466581662497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,1,balanced,0.1914773384730021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,1,balanced,0.19248533248901367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,1,balanced,0.19590934117635092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,1,balanced,0.19780800739924112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,1,balanced,0.19924799601236978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,1,balanced,0.20186134179433188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,1,balanced,0.20483734210332236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,1,balanced,0.21118400494257608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,1,balanced,0.21423467000325522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,1,balanced,0.22033600012461343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,1,balanced,0.2206933299700419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,1,balanced,0.22738132874170938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,1,balanced,0.23632532358169556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,1,balanced,0.2680533329645793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,1,balanced,0.29999999205271405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,1,balanced,0.3664906819661458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,1,balanced,0.42083199818929035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,1,balanced,0.5978453159332275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,1,balanced,0.7168052991231283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,1,balanced,1.027136007944743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,1,balanced,1.334330717722575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,1,balanced,1.963088035583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,1,balanced,2.5964694023132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,8,power_law_1.01,0.033024001121521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,8,power_law_1.01,0.03255040049552917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,8,power_law_1.01,0.02901119887828827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,8,power_law_1.01,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,8,power_law_1.01,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,8,power_law_1.01,0.03320319950580597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,8,power_law_1.01,0.032179200649261476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,8,power_law_1.01,0.03397760093212128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,8,power_law_1.01,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,8,power_law_1.01,0.03808639943599701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,8,power_law_1.01,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,8,power_law_1.01,0.03891200125217438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,8,power_law_1.01,0.04160000085830688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,8,power_law_1.01,0.04010240137577057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,8,power_law_1.01,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,8,power_law_1.01,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,8,power_law_1.01,0.05744640231132507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,8,power_law_1.01,0.06959360241889953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,8,power_law_1.01,0.07836800217628478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,8,power_law_1.01,0.09741439819335937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,64,balanced,0.044010668992996216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,64,balanced,0.04368533194065094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,64,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,64,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,64,balanced,0.04590400060017904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,64,balanced,0.0498986691236496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,64,balanced,0.04933333396911621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,64,balanced,0.048026666045188904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,8,power_law_1.01,0.11267839670181275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,8,power_law_1.01,0.14701440334320068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,64,balanced,0.05232533315817515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,64,balanced,0.05142400165398916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,64,balanced,0.05179200073083242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,8,power_law_1.01,0.16547199487686157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,64,balanced,0.05394133428732554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,64,balanced,0.05413866539796194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,64,balanced,0.05792533357938131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,64,balanced,0.06406400104363759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,64,balanced,0.06005866825580597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,64,balanced,0.06437866886456807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,64,balanced,0.07232533395290375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,64,balanced,0.0743999977906545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,64,balanced,0.08448533217112224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,64,balanced,0.0950933297475179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,64,balanced,0.11771733562151591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,64,balanced,0.1381706694761912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,64,balanced,0.18476800123850504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,64,balanced,0.22844799359639487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,64,balanced,0.31619733572006226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,64,balanced,0.3986399968465169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,8,power_law_1.01,0.2245248079299927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,8,power_law_1.01,0.271616005897522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,8,power_law_1.01,0.4053760051727295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,64,power_law_1.01,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,64,power_law_1.01,0.0440447986125946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,64,power_law_1.01,0.04317440092563629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,64,power_law_1.01,0.04611839950084686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,64,power_law_1.01,0.04615679979324341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,64,power_law_1.01,0.048019200563430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,64,power_law_1.01,0.04864639937877655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,64,power_law_1.01,0.04938879907131195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,64,power_law_1.01,0.05114240050315857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,64,power_law_1.01,0.05103999972343445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,64,power_law_1.01,0.051769602298736575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,8,balanced,0.04874666531880697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,8,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,8,balanced,0.04621866842110952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,8,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,8,balanced,0.04958933095137278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,64,power_law_1.01,0.052223998308181765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,8,balanced,0.054005334774653115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,64,power_law_1.01,0.05211520195007324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,8,balanced,0.07549333572387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,8,balanced,0.06769600013891856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,8,balanced,0.08953600128491719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,8,balanced,0.08103999992211659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,8,balanced,0.07612266639868419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,8,balanced,0.13664000233014426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,8,balanced,0.1283199985822042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,8,balanced,0.10107733805974324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,8,balanced,0.14841066797574362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,8,balanced,0.12617599964141846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,8,balanced,0.16285333037376404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,8,balanced,0.21312000354131064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,8,balanced,0.2742026646931966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,8,balanced,0.3889919916788737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,8,balanced,0.5203839937845866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,8,balanced,0.7462186813354492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,8,balanced,0.9868533611297607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,64,power_law_1.01,0.05692160129547119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,8,balanced,1.4548800786336262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,64,power_law_1.01,0.06253439784049988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,64,power_law_1.01,0.059622400999069215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,8,balanced,1.9188693364461262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,64,power_law_1.01,0.06284160017967225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,64,power_law_1.01,0.07248640060424805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,8,balanced,2.87169615427653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,64,power_law_1.01,0.0760320007801056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,64,power_law_1.01,0.09438080191612244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,64,power_law_1.01,0.10583679676055908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,8,balanced,3.80401611328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,64,power_law_1.01,0.14183679819107056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,64,power_law_1.01,0.17074559926986693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,64,power_law_1.01,0.23725440502166747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,64,power_law_1.01,0.28767359256744385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,64,power_law_1.01,0.42797441482543946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,64,power_law_1.01,0.521446418762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,8,power_law_1.01,0.523744010925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,4,balanced,0.06842133402824402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,4,balanced,0.059994667768478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,4,balanced,0.08470933636029561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,4,balanced,0.08556266625722249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,4,balanced,0.08611733714739482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,4,balanced,0.08417066931724548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,4,balanced,0.0846666693687439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,4,balanced,0.08571199576059978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,4,balanced,0.08861333131790161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,4,balanced,0.09033600489298503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,4,balanced,0.0916426678498586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,4,balanced,0.09025599559148152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,4,balanced,0.09040000041325887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,4,balanced,0.09575999776522319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,4,balanced,0.10749333103497823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,4,balanced,0.10471466183662415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,4,balanced,0.10939199725786845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,4,balanced,0.12473066647847493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,4,balanced,0.1384053329626719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,4,balanced,0.16474666198094687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,4,balanced,0.1832853356997172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,4,balanced,0.2315466602643331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,4,balanced,0.2797706723213196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,4,balanced,0.3867786725362142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,4,balanced,0.4901493390401204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,4,balanced,0.6978027025858561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,4,balanced,0.8874026934305826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,2,power_law_1.01,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,2,power_law_1.01,0.09188479781150818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,2,power_law_1.01,0.10700800418853759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,2,power_law_1.01,0.14629119634628296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,2,power_law_1.01,0.17424000501632692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,2,power_law_1.01,0.2146752119064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,2,power_law_1.01,0.2673919916152954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,2,power_law_1.01,0.27292160987854003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,2,power_law_1.01,0.2731071949005127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,2,power_law_1.01,0.2788800001144409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,2,power_law_1.01,0.2879040002822876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,2,power_law_1.01,0.2962752103805542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,2,power_law_1.01,0.2983232021331787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,2,power_law_1.01,0.3209151983261108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,2,power_law_1.01,0.322707200050354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,2,power_law_1.01,0.3374783992767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,2,power_law_1.01,0.3606208086013794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,2,power_law_1.01,0.39574398994445803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.01,0.44437122344970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,1,power_law_1.01,0.0205375999212265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.01,0.530950403213501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,1,power_law_1.01,0.021484799683094025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.01,0.5469056129455566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,1,power_law_1.01,0.02595840096473694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,1,power_law_1.01,0.03170560002326965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,1,power_law_1.01,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,1,power_law_1.01,0.06558079719543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,1,power_law_1.01,0.07514879703521729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,1,power_law_1.01,0.08042240142822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.01,0.662278413772583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,1,power_law_1.01,0.08186240196228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,1,power_law_1.01,0.08325120210647582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,1,power_law_1.01,0.08579199910163879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,1,power_law_1.01,0.09300479888916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,1,power_law_1.01,0.0948736011981964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,1,power_law_1.01,0.09774720072746276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,1,power_law_1.01,0.10066560506820679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,1,power_law_1.01,0.10361599922180176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.01,0.7877823829650878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,1,power_law_1.01,0.11066880226135253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,1,power_law_1.01,0.13007359504699706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.01,1.0865216255187988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,1,power_law_1.01,0.14453760385513306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,1,power_law_1.01,0.1760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.01,1.311257553100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.01,1.9110719680786132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.01,2.4487743377685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,1,power_law_1.01,0.20578560829162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,1,power_law_1.01,0.2705024003982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,1,power_law_1.01,0.2940351963043213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,1,power_law_1.01,0.3947711944580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,1,power_law_1.01,0.5197696208953857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,1,power_law_1.01,0.780511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,1,power_law_1.01,0.9045439720153808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,16,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,16,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,16,balanced,0.04801600178082784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,16,balanced,0.07628266513347626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,16,balanced,0.10212799906730652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,16,balanced,0.11435199777285258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,16,balanced,0.11440533399581909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,16,balanced,0.11624000469843547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,16,balanced,0.11743999520937602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,16,balanced,0.11716799934705098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,16,balanced,0.11528000235557556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,16,balanced,0.11880532900492351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,16,balanced,0.11872532963752747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,16,balanced,0.11925333738327026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,16,balanced,0.12447466452916463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,16,balanced,0.12611200412114462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,16,balanced,0.12713600198427835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,16,balanced,0.1400266687075297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,16,balanced,0.1384053329626719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,16,balanced,0.16266666849454245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,16,balanced,0.15869866808255514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,16,balanced,0.2222879926363627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,16,balanced,0.20837867259979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,16,balanced,0.2783520023028056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,16,balanced,0.31491732597351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,16,balanced,0.46034133434295654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,16,balanced,0.5269866784413656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,2,power_law_1.01,0.03672960102558136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,2,power_law_1.01,0.043244799971580504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,2,power_law_1.01,0.04392960071563721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,2,power_law_1.01,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,2,power_law_1.01,0.05572479963302612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,2,power_law_1.01,0.05538560152053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,2,power_law_1.01,0.05886080265045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,2,power_law_1.01,0.060172802209854125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,2,power_law_1.01,0.06291199922561645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,2,power_law_1.01,0.06381440162658691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,2,power_law_1.01,0.06261759996414185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,2,power_law_1.01,0.06758400201797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,2,power_law_1.01,0.06751360297203064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,2,power_law_1.01,0.0731328010559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,2,power_law_1.01,0.03132160007953644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,2,power_law_1.01,0.0773248016834259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,2,power_law_1.01,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,2,power_law_1.01,0.08181759715080261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,2,power_law_1.01,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,4,power_law_1.2,0.02680320143699646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,2,power_law_1.01,0.049235200881958006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,2,power_law_1.01,0.08110719919204712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,2,power_law_1.01,0.05790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,4,power_law_1.2,0.025536000728607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,2,power_law_1.01,0.0695360004901886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,4,power_law_1.2,0.029919999837875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,2,power_law_1.01,0.09482880234718323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,4,power_law_1.2,0.03767040073871612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,4,power_law_1.2,0.04284160137176514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,4,power_law_1.2,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,4,power_law_1.2,0.04986239969730377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,4,power_law_1.2,0.05050879716873169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,4,power_law_1.2,0.052767997980117796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,4,power_law_1.2,0.0530239999294281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,4,power_law_1.2,0.055980801582336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,4,power_law_1.2,0.05906559824943543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,2,power_law_1.01,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,2,power_law_1.01,0.10742399692535401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,2,power_law_1.01,0.07027199864387512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,2,power_law_1.01,0.07466880083084107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,2,power_law_1.01,0.07457280158996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,2,power_law_1.01,0.1297279953956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,2,power_law_1.01,0.073471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,2,power_law_1.01,0.07535359859466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,2,power_law_1.01,0.07713279724121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,2,power_law_1.01,0.07995520234107971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,2,power_law_1.01,0.09223679900169372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,2,power_law_1.01,0.10022399425506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,2,power_law_1.01,0.1023103952407837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,2,power_law_1.01,0.11080960035324097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,2,power_law_1.01,0.12042239904403687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,2,power_law_1.01,0.14145920276641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,2,power_law_1.01,0.1569408059120178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,2,power_law_1.01,0.15237120389938355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,2,power_law_1.01,0.18450560569763183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,2,power_law_1.01,0.21008000373840333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,2,power_law_1.01,0.2790271997451782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,2,power_law_1.01,0.3332479953765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,2,power_law_1.01,0.47335038185119627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,2,power_law_1.01,0.5727424144744873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,4,power_law_1.2,0.06389759778976441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,4,power_law_1.2,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,4,power_law_1.2,0.06924800276756286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,4,power_law_1.2,0.07017599940299987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,4,power_law_1.2,0.07047680020332336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,4,power_law_1.2,0.07783679962158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,4,power_law_1.2,0.09217919707298279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,4,power_law_1.2,0.10523519515991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,2,power_law_1.01,0.1933632016181946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,4,power_law_1.2,0.11918079853057861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,4,power_law_1.2,0.1430400013923645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,4,power_law_1.2,0.16743040084838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,4,power_law_1.2,0.2239936113357544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,2,power_law_1.01,0.22629759311676026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,4,power_law_1.2,0.29244160652160645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,2,power_law_1.01,0.3176703929901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,2,power_law_1.01,0.3559168100357056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,2,balanced,0.06906133393446605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,2,balanced,0.0936959981918335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,2,balanced,0.14407466848691305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,2,balanced,0.2486720085144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,2,balanced,0.4440799951553345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,2,balanced,0.8457813262939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,2,balanced,0.8455519676208496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,2,balanced,0.8457919756571451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,2,balanced,0.851482629776001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,2,balanced,0.8548853397369385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,2,power_law_1.01,0.5963583946228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,2,balanced,0.8550506432851156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,2,balanced,0.8600320021311442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,2,balanced,0.8641173044840494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,2,balanced,0.8720906575520834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,2,balanced,0.8798080285390218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,2,balanced,0.887343962987264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,2,balanced,0.9007360140482584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,2,balanced,0.931114673614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,2,balanced,0.966480016708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,2,balanced,1.0149493217468262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,2,balanced,1.0892639954884846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,2,balanced,1.1819840272267659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,2,balanced,1.2837599913279216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,2,balanced,1.68067200978597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,2,balanced,1.7492106755574544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,2,balanced,2.774160067240397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,2,balanced,2.9487520853678384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,4,power_law_1.2,0.4060671806335449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,2,power_law_1.01,0.8108608245849609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,4,power_law_1.2,0.5325568199157715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,64,power_law_1.2,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,1,power_law_1.01,0.07749119997024537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,64,power_law_1.2,0.040454399585723874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,1,power_law_1.01,0.11912319660186768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,1,power_law_1.01,0.13619840145111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,64,power_law_1.2,0.04242559969425201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,1,power_law_1.01,0.21969280242919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,64,power_law_1.2,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,1,power_law_1.01,0.23336319923400878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,64,power_law_1.2,0.043424001336097716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,64,power_law_1.2,0.04776960015296936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,64,power_law_1.2,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,64,power_law_1.2,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,64,power_law_1.2,0.04813440144062042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,64,power_law_1.2,0.04867199957370758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,64,power_law_1.2,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,1,power_law_1.01,0.2265791893005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,64,power_law_1.2,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,1,power_law_1.01,0.24089601039886474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,64,power_law_1.2,0.05629439949989319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,1,power_law_1.01,0.2493567943572998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,64,power_law_1.2,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,1,power_law_1.01,0.24916479587554932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,64,power_law_1.2,0.0650816023349762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,1,power_law_1.01,0.2580352067947388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,64,power_law_1.2,0.06698240041732788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,1,power_law_1.01,0.2530816078186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,1,power_law_1.01,0.26813440322875975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,1,power_law_1.01,0.27111680507659913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,1,power_law_1.01,0.27596800327301024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,1,power_law_1.01,0.3010495901107788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,1,power_law_1.01,0.3073472023010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,1,power_law_1.01,0.32229759693145754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,64,power_law_1.2,0.0654528021812439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,64,power_law_1.2,0.07063680291175842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,64,power_law_1.2,0.07631999850273133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,64,power_law_1.2,0.093094402551651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,64,power_law_1.2,0.10722559690475464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,1,power_law_1.01,0.37392640113830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,64,power_law_1.2,0.1358016014099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,1,power_law_1.01,0.42020478248596194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,64,power_law_1.2,0.15859839916229249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,64,power_law_1.2,0.21733760833740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,64,power_law_1.2,0.2517695903778076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,64,power_law_1.2,0.3977663993835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,64,power_law_1.2,0.44963841438293456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,1,power_law_1.01,0.5138688087463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,1,power_law_1.01,0.6139455795288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,1,power_law_1.01,0.8171392440795898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,1,power_law_1.01,1.0340736389160157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,1,power_law_1.01,1.4604736328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,1,power_law_1.01,1.8598655700683593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,1,power_law_1.01,2.670911979675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,8,power_law_1.01,0.04643200039863586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,8,power_law_1.01,0.04830079972743988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,8,power_law_1.01,0.0491456001996994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,8,power_law_1.01,0.06256639957427979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,8,power_law_1.01,0.0605184018611908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,8,power_law_1.01,0.06439039707183838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,8,power_law_1.01,0.06622080206871032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,8,power_law_1.01,0.0638592004776001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,8,power_law_1.01,0.06635519862174988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,8,power_law_1.01,0.06643199920654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,8,power_law_1.01,0.06963840126991272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,8,power_law_1.01,0.07157760262489318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,8,power_law_1.01,0.07176960110664368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,8,power_law_1.01,0.074099200963974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,8,power_law_1.01,0.08243200182914734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,8,power_law_1.01,0.08697599768638611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,8,power_law_1.01,0.08785920143127442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,8,power_law_1.01,0.10318080186843873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.01,0.1155776023864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.01,0.1377087950706482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.01,0.14014079570770263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.01,0.19518719911575316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.01,0.2113408088684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.01,0.3349503993988037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.01,0.36110079288482666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.01,0.5454783916473389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.01,0.7365888118743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,1,power_law_1.01,3.437772750854492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,16,power_law_1.01,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,16,power_law_1.01,0.07020800113677979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,16,power_law_1.01,0.08343039751052857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,16,power_law_1.01,0.08720639944076539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,16,power_law_1.01,0.08805760145187377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,16,power_law_1.01,0.08483200073242188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,16,power_law_1.01,0.08626559972763062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,16,power_law_1.01,0.08479999899864196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,16,power_law_1.01,0.08755840063095092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,16,power_law_1.01,0.08396160006523132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,16,power_law_1.01,0.08351359963417053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,16,power_law_1.01,0.08937600255012512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,16,power_law_1.01,0.08844799995422363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,16,power_law_1.01,0.09356160163879394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,16,power_law_1.01,0.09811840057373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,16,power_law_1.01,0.10254080295562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,16,power_law_1.01,0.10773119926452637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,16,power_law_1.01,0.12023040056228637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,16,power_law_1.01,0.1324288010597229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,16,power_law_1.01,0.15226240158081056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,16,power_law_1.01,0.1739840030670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,16,power_law_1.01,0.21683199405670167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,16,power_law_1.01,0.2660991907119751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,16,power_law_1.01,0.3319551944732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,16,power_law_1.01,0.44021759033203123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,16,power_law_1.01,0.574073600769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,16,power_law_1.01,0.7428544044494629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,2,balanced,0.04367466767628988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,2,balanced,0.04618666569391886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,2,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,2,balanced,0.06946666538715363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,2,balanced,0.09329066673914592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,2,balanced,0.08932266632715861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,2,balanced,0.08669867118199666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,2,balanced,0.08674666285514832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,2,balanced,0.08520533641179402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,2,balanced,0.0851093331972758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,2,balanced,0.08715200424194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,2,balanced,0.09077866872151692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,2,balanced,0.09081600109736125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,2,balanced,0.09501333038012187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,2,balanced,0.09897067149480183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,2,balanced,0.09991466999053955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,2,balanced,0.10587732990582784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,2,balanced,0.11857600013415019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,2,balanced,0.12945600350697836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,2,balanced,0.15317866206169128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,2,balanced,0.17564266920089722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,2,balanced,0.2510240077972412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,2,balanced,0.293994665145874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,2,balanced,0.42721064885457355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,2,balanced,0.551637331644694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,2,balanced,0.8014986515045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,2,balanced,1.0351466337839763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,16,power_law_1.01,0.07427840232849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,16,power_law_1.01,0.07352960109710693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,16,power_law_1.01,0.06977919936180114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,16,power_law_1.01,0.07272319793701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,16,power_law_1.01,0.06554239988327026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,16,power_law_1.01,0.07559679746627808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,16,power_law_1.01,0.07264639735221863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,16,power_law_1.01,0.0770367980003357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,16,power_law_1.01,0.07652480006217957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,16,power_law_1.01,0.07991679906845092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,16,power_law_1.01,0.08208000063896179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,16,power_law_1.01,0.08329600095748901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,16,power_law_1.01,0.08350080251693726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,16,power_law_1.01,0.08659840226173401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,16,power_law_1.01,0.09707520008087159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,16,power_law_1.01,0.09681280255317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,16,power_law_1.01,0.09722239971160888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,16,power_law_1.01,0.1046720027923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,16,power_law_1.01,0.11887359619140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,16,power_law_1.01,0.13838720321655273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,16,power_law_1.01,0.14334720373153687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,16,power_law_1.01,0.18077440261840821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,16,power_law_1.01,0.21100800037384032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,16,power_law_1.01,0.2671744108200073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,16,power_law_1.01,0.37411201000213623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,16,power_law_1.01,0.47341442108154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,16,power_law_1.01,0.5395520210266114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,8,power_law_1.01,0.047251200675964354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,8,power_law_1.01,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,8,power_law_1.01,0.05015680193901062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,8,power_law_1.01,0.06103039979934692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,8,power_law_1.01,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,8,power_law_1.01,0.06364799737930298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,8,power_law_1.01,0.06544640064239501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,8,power_law_1.01,0.06736639738082886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,8,power_law_1.01,0.07066239714622498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,8,power_law_1.01,0.07123839855194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,8,power_law_1.01,0.07308160066604615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,8,power_law_1.01,0.07373440265655518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,8,power_law_1.01,0.07990400195121765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,8,power_law_1.01,0.08281599879264831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,8,power_law_1.01,0.09139840006828308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,8,power_law_1.01,0.09364479780197144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,8,power_law_1.01,0.09907839894294738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,8,power_law_1.01,0.11221760511398315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,8,power_law_1.01,0.1254591941833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,8,power_law_1.01,0.14958720207214354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,8,power_law_1.01,0.18636800050735475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,8,power_law_1.01,0.23348479270935057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,8,power_law_1.01,0.3068416118621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,8,power_law_1.01,0.4199552059173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,8,power_law_1.01,0.5344448089599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,8,power_law_1.01,0.7143936157226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,8,power_law_1.01,0.9679807662963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,64,power_law_1.01,0.028934401273727418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,64,power_law_1.01,0.028070399165153505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,64,power_law_1.01,0.027968001365661622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,64,power_law_1.01,0.030060800909996032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,64,power_law_1.01,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,64,power_law_1.01,0.045542401075363156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,64,power_law_1.01,0.06268159747123718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,64,power_law_1.01,0.06459519863128663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,64,power_law_1.01,0.07667840123176575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,64,power_law_1.01,0.07738239765167236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,64,power_law_1.01,0.0755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,64,power_law_1.01,0.07392640113830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,64,power_law_1.01,0.07197440266609192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,64,power_law_1.01,0.0701632022857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,64,power_law_1.01,0.06496000289916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,64,power_law_1.01,0.06725760102272034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,64,power_law_1.01,0.07841920256614685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,64,power_law_1.01,0.0770687997341156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.01,0.08575999736785889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.01,0.10073599815368653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.01,0.11395200490951538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.01,0.15280640125274658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.01,0.21166720390319824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.01,0.28818559646606445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.01,0.3454335927963257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.01,0.5349631786346436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.01,0.683519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,32,power_law_1.01,0.13115520477294923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,32,power_law_1.01,0.14734079837799072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,32,power_law_1.01,0.14744319915771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,32,power_law_1.01,0.14965120553970337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,32,power_law_1.01,0.14781440496444703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,32,power_law_1.01,0.13788800239562987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,32,power_law_1.01,0.1375872015953064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,32,power_law_1.01,0.1415552020072937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,32,power_law_1.01,0.14862719774246216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,32,power_law_1.01,0.14737919569015503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,32,power_law_1.01,0.14598400592803956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,32,power_law_1.01,0.148089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,32,power_law_1.01,0.15749119520187377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,32,power_law_1.01,0.1546880006790161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,32,power_law_1.01,0.1592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,32,power_law_1.01,0.16460800170898438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,32,power_law_1.01,0.17196160554885864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,32,power_law_1.01,0.18357759714126587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,32,power_law_1.01,0.19596799612045288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,32,power_law_1.01,0.21358079910278321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,32,power_law_1.01,0.25223679542541505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,32,power_law_1.01,0.30899200439453123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,32,power_law_1.01,0.3686271905899048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,32,power_law_1.01,0.4845759868621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,32,power_law_1.01,0.6431424140930175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,32,power_law_1.01,0.8782591819763184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,32,power_law_1.01,1.1132096290588378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,32,power_law_1.01,0.020185600221157073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,32,power_law_1.01,0.018598400056362152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,32,power_law_1.01,0.01855359971523285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,32,power_law_1.01,0.018572799861431122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,32,power_law_1.01,0.019020800292491914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,32,power_law_1.01,0.02069759964942932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,32,power_law_1.01,0.021971200406551362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,32,power_law_1.01,0.025177600979804992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,32,power_law_1.01,0.023955200612545014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,32,power_law_1.01,0.024108800292015075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,32,power_law_1.01,0.0378495991230011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,32,power_law_1.01,0.038387200236320494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,32,power_law_1.01,0.03880319893360138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,32,power_law_1.01,0.03772799968719483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,32,power_law_1.01,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,32,power_law_1.01,0.038387200236320494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,32,power_law_1.01,0.04141440093517303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,32,power_law_1.01,0.04152320027351379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.01,0.046009600162506104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.01,0.05907840132713318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.01,0.06756479740142822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.01,0.0822272002696991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.01,0.10057599544525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.01,0.13542399406433106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.01,0.16977920532226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.01,0.24625918865203858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.01,0.31344640254974365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,2,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,2,balanced,0.05598933498064677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,2,balanced,0.07468266785144806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,2,balanced,0.11448533336321513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,2,balanced,0.17778666814168295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,2,balanced,0.30609599749247235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,2,balanced,0.306442658106486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,2,balanced,0.30846933523813885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,2,balanced,0.31010667483011883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,2,balanced,0.31284799178441364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,2,balanced,0.31353066364924115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,2,balanced,0.31909332672754925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,2,balanced,0.32179733117421466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,2,balanced,0.3254879911740621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,2,balanced,0.33428800106048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,2,balanced,0.3375626802444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,2,balanced,0.3452320098876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,2,balanced,0.3659466505050659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,2,balanced,0.3830346663792928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,2,balanced,0.4209546645482381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,2,balanced,0.46114667256673175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,2,balanced,0.527023990948995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,2,balanced,0.6009440024693807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,2,balanced,0.793440024058024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,2,balanced,0.9093279838562012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,2,balanced,1.3222506841023762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,2,balanced,1.5871307055155437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,16,power_law_1.01,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,16,power_law_1.01,0.03136639893054962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,16,power_law_1.01,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,16,power_law_1.01,0.03331199884414673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,16,power_law_1.01,0.03434880077838898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,16,power_law_1.01,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,16,power_law_1.01,0.035743999481201175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,16,power_law_1.01,0.03596799969673157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,16,power_law_1.01,0.036339199542999266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,16,power_law_1.01,0.0364544004201889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,16,power_law_1.01,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,16,power_law_1.01,0.038899201154708865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,16,power_law_1.01,0.03984000086784363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,16,power_law_1.01,0.040992000699043275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,16,power_law_1.01,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,16,power_law_1.01,0.04391680061817169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,64,power_law_1.2,0.02421119958162308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,64,power_law_1.2,0.02423679977655411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,64,power_law_1.2,0.025190401077270507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,16,power_law_1.01,0.050521600246429446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,64,power_law_1.2,0.02561280131340027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,16,power_law_1.01,0.053964799642562865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,64,power_law_1.2,0.026284798979759216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,64,power_law_1.2,0.026790401339530943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,64,power_law_1.2,0.028332799673080444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,16,power_law_1.01,0.0633791983127594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,64,power_law_1.2,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,64,power_law_1.2,0.033471998572349546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,64,power_law_1.2,0.035488000512123107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,64,power_law_1.2,0.03487359881401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,64,power_law_1.2,0.03587839901447296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,64,power_law_1.2,0.0356799989938736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,64,power_law_1.2,0.033932799100875856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,64,power_law_1.2,0.037062400579452516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,16,power_law_1.01,0.07860479950904846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,64,power_law_1.2,0.03828479945659637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,16,power_law_1.01,0.0918079972267151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,16,power_law_1.01,0.11631360054016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,16,power_law_1.01,0.1407487988471985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,16,power_law_1.01,0.21391360759735106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,64,power_law_1.2,0.041196799278259276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,16,power_law_1.01,0.24934399127960205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,64,power_law_1.2,0.04351359903812409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,64,power_law_1.2,0.050496000051498416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,16,power_law_1.01,0.4344128131866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,64,power_law_1.2,0.05845119953155518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,64,power_law_1.2,0.0590399980545044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,16,power_law_1.01,0.516102409362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,64,power_law_1.2,0.0760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,64,power_law_1.2,0.09000319838523865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,64,power_law_1.2,0.1291327953338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,64,power_law_1.2,0.14899200201034546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,64,power_law_1.2,0.2169343948364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,64,power_law_1.01,0.08419200181961059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,64,power_law_1.01,0.08213760256767273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,64,power_law_1.01,0.07890560030937195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,64,power_law_1.01,0.08209279775619507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,64,power_law_1.01,0.07908480167388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,64,power_law_1.01,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,64,power_law_1.01,0.08243839740753174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,64,power_law_1.01,0.0824512004852295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,64,power_law_1.01,0.08522239923477173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,64,power_law_1.01,0.08145279884338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,64,power_law_1.01,0.08391680121421814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,64,power_law_1.01,0.0855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,64,power_law_1.01,0.08358399868011475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,64,power_law_1.01,0.08973439931869506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,64,power_law_1.01,0.09141759872436524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,64,power_law_1.01,0.09373440146446228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,64,power_law_1.01,0.10073599815368653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,64,power_law_1.01,0.10928000211715698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,64,power_law_1.01,0.11830400228500366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,64,power_law_1.01,0.14218879938125611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,64,power_law_1.01,0.16087679862976073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,64,power_law_1.01,0.20169599056243898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,64,power_law_1.01,0.24471681118011473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,64,power_law_1.01,0.3366976022720337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,64,power_law_1.01,0.4157120227813721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,64,power_law_1.01,0.573305606842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,64,power_law_1.01,0.7337984085083008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,64,power_law_1.2,0.2825727939605713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,8,balanced,0.02995733420054118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,8,balanced,0.027984000742435455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,8,balanced,0.029546665648619335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,8,balanced,0.031583999594052635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,8,balanced,0.03777066618204117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,8,balanced,0.039477333426475525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,8,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,8,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,8,balanced,0.03995733211437861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,8,balanced,0.039477333426475525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,8,balanced,0.04055999964475632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,8,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,8,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,8,balanced,0.041936000188191734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,8,balanced,0.048170665899912514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,8,balanced,0.04994133114814758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,8,balanced,0.05009600023428599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,8,balanced,0.05421866476535797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,8,balanced,0.06037333110968272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,8,balanced,0.07854400078455608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,8,balanced,0.08062933385372162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,8,balanced,0.10359467069307964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,8,balanced,0.11817600329717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,8,balanced,0.15477333466211954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,8,balanced,0.19553067286809286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,8,balanced,0.26743467648824054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,8,balanced,0.3323040008544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,2,power_law_1.2,0.045491200685501096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,2,power_law_1.2,0.04853120148181915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,2,power_law_1.2,0.053273600339889524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,2,power_law_1.2,0.0665727972984314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,2,power_law_1.2,0.07557759881019592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,2,power_law_1.2,0.09625599980354309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,2,power_law_1.2,0.09882240295410157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,2,power_law_1.2,0.10258560180664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,2,power_law_1.2,0.10677759647369385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,2,power_law_1.2,0.10849920511245728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,2,power_law_1.2,0.11238399744033814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,2,power_law_1.2,0.11857279539108276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,2,power_law_1.2,0.12496000528335571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,2,power_law_1.2,0.12634880542755128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,2,power_law_1.2,0.13564159870147705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,2,power_law_1.2,0.1435263991355896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,2,power_law_1.2,0.15529600381851197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,2,power_law_1.2,0.18547199964523314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,2,power_law_1.2,0.21711359024047852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,2,power_law_1.2,0.2635456085205078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,2,power_law_1.2,0.32140800952911375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,2,power_law_1.2,0.45310077667236326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,2,power_law_1.2,0.5561600208282471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,2,power_law_1.2,0.7597184181213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,2,power_law_1.2,1.050163173675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,2,power_law_1.2,1.396288013458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,2,power_law_1.2,1.8468799591064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,4,power_law_1.2,0.06881279945373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,4,power_law_1.2,0.06618239879608154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,4,power_law_1.2,0.07068799734115601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,4,power_law_1.2,0.08236160278320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,4,power_law_1.2,0.08419839739799499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,4,power_law_1.2,0.08696320056915283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,4,power_law_1.2,0.09524480104446412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,4,power_law_1.2,0.09919999837875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,4,power_law_1.2,0.11169279813766479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,4,power_law_1.2,0.11520639657974244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,4,power_law_1.2,0.09550719857215881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,4,power_law_1.2,0.10656640529632569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,4,power_law_1.2,0.10880000591278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,4,power_law_1.2,0.12559360265731812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,4,power_law_1.2,0.14781440496444703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,4,power_law_1.2,0.1529855966567993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,4,power_law_1.2,0.1545151948928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,4,power_law_1.2,0.173088002204895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,4,power_law_1.2,0.2327552080154419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,4,power_law_1.2,0.3254080057144165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,4,power_law_1.2,0.2991296052932739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,4,power_law_1.2,0.42597122192382814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,4,power_law_1.2,0.5210944175720215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,32,power_law_1.01,0.0849727988243103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,32,power_law_1.01,0.07680000066757202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,32,power_law_1.01,0.077292799949646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,32,power_law_1.01,0.07966079711914062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,32,power_law_1.01,0.07276160120964051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,32,power_law_1.01,0.08284800052642823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,32,power_law_1.01,0.0837823987007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,32,power_law_1.01,0.08525440096855164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,32,power_law_1.01,0.0858560025691986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,32,power_law_1.01,0.08422399759292602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,32,power_law_1.01,0.08643199801445008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,32,power_law_1.01,0.09033600091934205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,32,power_law_1.01,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,32,power_law_1.01,0.0927232027053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,32,power_law_1.01,0.10460799932479858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,32,power_law_1.01,0.11103359460830689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,32,power_law_1.01,0.11898239850997924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,32,power_law_1.01,0.13653759956359862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,1,balanced,0.02499733368555705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,1,balanced,0.02922133356332779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,32,power_law_1.01,0.14789119958877564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,1,balanced,0.04589866598447164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,1,balanced,0.06571199993292491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,1,balanced,0.10289067029953003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,32,power_law_1.01,0.17694079875946045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,1,balanced,0.13165866335233053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,1,balanced,0.1299626628557841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,32,power_law_1.01,0.20444159507751464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,4,power_law_1.2,0.7217663764953614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,1,balanced,0.13217600186665854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,1,balanced,0.13335999846458435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2688,1856,6,128,1,1,balanced,0.13667200009028116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2688,1856,6,128,1,1,balanced,0.14591466387112936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2688,1856,6,128,1,1,balanced,0.1537440021832784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2688,1856,6,128,1,1,balanced,0.14986133575439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2688,1856,6,128,1,1,balanced,0.1556000014146169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,32,power_law_1.01,0.26564478874206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2688,1856,6,128,1,1,balanced,0.1628159979979197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,32,power_law_1.01,0.31544320583343505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2688,1856,6,128,1,1,balanced,0.17896000544230142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2688,1856,6,128,1,1,balanced,0.1686506668726603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,4,power_law_1.2,1.187718391418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2688,1856,6,128,1,1,balanced,0.24149332443873087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2688,1856,6,128,1,1,balanced,0.21425066391626993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,32,power_law_1.01,0.4603263854980469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,32,power_law_1.01,0.5014527797698974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,32,power_law_1.01,0.7177792072296143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,32,power_law_1.01,1.0053248405456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,4,power_law_1.2,1.7259775161743165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,2,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,2,power_law_1.01,0.03549439907073974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,2,power_law_1.01,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,2,power_law_1.01,0.05301759839057922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,2,power_law_1.01,0.05198079943656921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,2,power_law_1.01,0.052960002422332765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,2,power_law_1.01,0.05792639851570129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,2,power_law_1.01,0.058457601070404056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,2,power_law_1.01,0.05921279788017273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,2,power_law_1.01,0.0595583975315094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,2,power_law_1.01,0.06106240153312683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,2,power_law_1.01,0.06273279786109924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,2,power_law_1.01,0.061887997388839724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,4,power_law_1.2,1.9686975479125977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,2,power_law_1.01,0.06838399767875672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,2,power_law_1.01,0.07492480278015137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,2,power_law_1.01,0.07832319736480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,2,power_law_1.01,0.0754368007183075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,2,power_law_1.01,0.08819839954376221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,2,power_law_1.01,0.10297600030899048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,2,power_law_1.01,0.1311743974685669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,2,power_law_1.01,0.15418239831924438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,2,power_law_1.01,0.2027967929840088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,2,power_law_1.01,0.25740799903869627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,2,power_law_1.01,0.31653759479522703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,2,power_law_1.01,0.3691648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,16,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,16,balanced,0.04363733530044556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,16,balanced,0.042538667718569435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,16,balanced,0.06001600126425425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,16,balanced,0.07630399862925212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,16,balanced,0.11466667056083679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,16,balanced,0.11760000387827556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,16,balanced,0.11320533355077107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,16,balanced,0.11437333623568217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,16,balanced,0.11382399996121724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,16,balanced,0.11521599690119426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,16,balanced,0.1144480009873708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,16,balanced,0.11586133639017741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,16,balanced,0.11755200227101643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,16,balanced,0.12294399738311768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,16,balanced,0.12178132931391399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,16,balanced,0.12487999598185222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,16,balanced,0.13025599718093872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,16,balanced,0.13383466998736063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,16,balanced,0.1456106702486674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,16,balanced,0.15727999806404114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,16,balanced,0.17583467562993368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,16,balanced,0.19165867567062378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,16,balanced,0.24900799989700317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,16,balanced,0.2837439974149068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,16,balanced,0.3739413420359294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,16,balanced,0.4633386532465617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,2,power_law_1.01,0.6451519966125489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,2,power_law_1.01,0.7763584136962891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,4,power_law_1.01,0.02327679991722107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,4,power_law_1.01,0.022815999388694764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,4,power_law_1.01,0.024691200256347655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,4,power_law_1.01,0.02930560111999512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,4,power_law_1.01,0.03240959942340851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,4,power_law_1.01,0.034643200039863584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,4,power_law_1.01,0.03895680010318756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,4,power_law_1.01,0.03973760008811951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,4,power_law_1.01,0.04073599874973297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,4,power_law_1.01,0.04264959990978241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,4,power_law_1.01,0.044249600172042845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,4,power_law_1.01,0.045587199926376346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,4,power_law_1.01,0.04731520116329193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,4,power_law_1.01,0.05048959851264954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,4,power_law_1.01,0.05868160128593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,4,power_law_1.01,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,4,power_law_1.01,0.06601600050926208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,4,power_law_1.01,0.09130880236625671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.01,0.10501760244369507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.01,0.12192000150680542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.01,0.1320896029472351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.01,0.15813119411468507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.01,0.18317439556121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.01,0.23101439476013183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.01,0.28929920196533204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.01,0.4003647804260254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.01,0.5074687957763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,8,power_law_1.2,0.06814079880714416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,8,power_law_1.2,0.07754240036010743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,8,power_law_1.2,0.08782079815864563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,8,power_law_1.2,0.09297279715538025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,8,power_law_1.2,0.1009984016418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,8,power_law_1.2,0.10574719905853272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,8,power_law_1.2,0.1149440050125122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,8,power_law_1.2,0.11530239582061767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,8,power_law_1.2,0.11310720443725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,8,power_law_1.2,0.1102336049079895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,8,power_law_1.2,0.10938880443572999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,8,power_law_1.2,0.1128383994102478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,8,power_law_1.2,0.11274880170822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,8,power_law_1.2,0.11519360542297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,8,power_law_1.2,0.127839994430542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,8,power_law_1.2,0.12951040267944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,8,power_law_1.2,0.13566720485687256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,8,power_law_1.2,0.14911359548568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,8,power_law_1.2,0.15953919887542725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,8,power_law_1.2,0.20451200008392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,8,power_law_1.2,0.2298815965652466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,8,power_law_1.2,0.31351680755615235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,8,power_law_1.2,0.39010560512542725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,4,balanced,0.03134933362404505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,4,balanced,0.030805334448814392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,4,balanced,0.03225066761175791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,4,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,4,balanced,0.03148799886306127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,4,balanced,0.03350933392842611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,4,balanced,0.033301333586374916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,4,balanced,0.03350933392842611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,4,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,4,balanced,0.033589333295822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,4,balanced,0.0339626669883728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,4,balanced,0.03573333223660787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,4,balanced,0.03533333291610082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,4,balanced,0.03621333340803782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,4,balanced,0.041050667564074196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,4,balanced,0.043552001317342125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,4,balanced,0.0436106671889623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,4,balanced,0.04799466828505198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,4,balanced,0.05171733101209005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,4,balanced,0.06233066817124685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,4,balanced,0.06642666459083557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,4,balanced,0.09708266456921895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,4,balanced,0.1253706713517507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,4,balanced,0.16471466422080994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,4,balanced,0.2021333376566569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,4,balanced,0.2788693308830261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,4,balanced,0.35341334342956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,8,power_law_1.2,0.5669760227203369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,8,power_law_1.2,0.7108096122741699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,8,power_law_1.2,1.045753574371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,8,power_law_1.2,1.327347183227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,1,power_law_1.2,0.02858879864215851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,1,power_law_1.2,0.02993920147418976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,1,power_law_1.2,0.03162240087985992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,1,power_law_1.2,0.0359935998916626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,1,power_law_1.2,0.036268800497055054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,1,power_law_1.2,0.03944959938526153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,1,power_law_1.2,0.04259200096130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,1,power_law_1.2,0.04222719967365265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,1,power_law_1.2,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,1,power_law_1.2,0.04709759950637817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,1,power_law_1.2,0.044659200310707095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,1,power_law_1.2,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,1,power_law_1.2,0.05235840082168579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,1,power_law_1.2,0.05301759839057922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,1,power_law_1.2,0.06472960114479065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,1,power_law_1.2,0.06722559928894042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,1,power_law_1.2,0.0736191987991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,1,power_law_1.2,0.09068160057067871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,1,power_law_1.2,0.09963520169258118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,1,power_law_1.2,0.12234879732131958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,1,power_law_1.2,0.13549439907073973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,1,power_law_1.2,0.17985919713974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,1,power_law_1.2,0.22346880435943603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,1,power_law_1.2,0.31508479118347166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,8,power_law_1.2,0.04652160108089447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,8,power_law_1.2,0.05066239833831787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,8,power_law_1.2,0.05839999914169312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,8,power_law_1.2,0.06033920049667359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,8,power_law_1.2,0.06736000180244446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,8,power_law_1.2,0.07594239711761475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,8,power_law_1.2,0.07617920041084289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,8,power_law_1.2,0.07879679799079894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,8,power_law_1.2,0.07884160280227662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,8,power_law_1.2,0.07917439937591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,8,power_law_1.2,0.0828607976436615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,8,power_law_1.2,0.0828927993774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,8,power_law_1.2,0.08212479948997498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,8,power_law_1.2,0.08609279990196228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,8,power_law_1.2,0.09068800210952759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,8,power_law_1.2,0.09810559749603272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,8,power_law_1.2,0.10490880012512208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,8,power_law_1.2,0.12289279699325562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,8,power_law_1.2,0.1287168025970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,8,power_law_1.2,0.15737600326538087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,8,power_law_1.2,0.18481919765472413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,8,power_law_1.2,0.2392319917678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,8,power_law_1.2,0.3100672006607056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,8,power_law_1.2,0.4180607795715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,8,power_law_1.2,0.5202239990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,8,power_law_1.2,0.7368768215179443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,8,power_law_1.2,1.0206208229064941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,1,power_law_1.2,0.4070079803466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,1,power_law_1.2,0.0574400007724762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,1,power_law_1.2,0.0769216001033783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,1,power_law_1.2,0.10416640043258667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,1,power_law_1.2,0.131167995929718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,1,power_law_1.2,0.1637887954711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,1,power_law_1.2,0.25766398906707766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,1,power_law_1.2,0.2811903953552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,1,power_law_1.2,0.2932159900665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,1,power_law_1.2,0.2966655969619751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,1,power_law_1.2,0.3023808002471924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,1,power_law_1.2,0.32051839828491213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,1,power_law_1.2,0.3254463911056519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,1,power_law_1.2,0.3357055902481079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,1,power_law_1.2,0.34072959423065186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,1,power_law_1.2,0.35820159912109373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,1,power_law_1.2,0.36401278972625734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,1,power_law_1.2,0.3895423889160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,1,power_law_1.2,0.4362304210662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,1,power_law_1.2,0.497376012802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,1,power_law_1.2,0.6082560062408447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,1,power_law_1.2,0.698419189453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,1,power_law_1.2,0.8725824356079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,1,power_law_1.2,1.1055423736572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,1,power_law_1.2,1.427558422088623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,1,power_law_1.2,1.8479679107666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,1,power_law_1.2,0.5812352180480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,1,power_law_1.2,2.4475263595581054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,1,power_law_1.2,2.9822080612182615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,32,power_law_1.01,0.10268800258636475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,32,power_law_1.01,0.0991104006767273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,32,power_law_1.01,0.09806079864501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,4,balanced,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,4,balanced,0.050767997900644936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,32,power_law_1.01,0.10581760406494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,4,balanced,0.069882666071256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,4,balanced,0.1016373336315155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,4,balanced,0.1486240029335022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,4,balanced,0.2513759930928548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,4,balanced,0.34969600041707355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,4,balanced,0.3516639868418376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,4,balanced,0.3551199833552043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,4,balanced,0.355461319287618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,4,balanced,0.35706134637196857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,4,balanced,0.3609706560770671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,4,balanced,0.3676533301671346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,4,balanced,0.36606399218241376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,4,balanced,0.3805919885635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,4,balanced,0.3803040186564128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,32,power_law_1.01,0.10071040391921997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,4,balanced,0.39698131879170734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,4,balanced,0.4168479839960734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,4,balanced,0.42876799901326496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,4,balanced,0.46720532576243085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,4,balanced,0.4962186813354492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,4,balanced,0.5816266536712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,4,balanced,0.6580853462219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,4,balanced,0.8057866891225179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,4,balanced,0.9921653270721436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,32,power_law_1.01,0.10426239967346192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,4,balanced,1.31223464012146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,32,power_law_1.01,0.1017024040222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,32,power_law_1.01,0.10243200063705445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,4,balanced,1.6850186983744304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,32,power_law_1.01,0.10546560287475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,32,power_law_1.01,0.1054527997970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,32,power_law_1.01,0.10627199411392212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,32,power_law_1.01,0.10746879577636718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,32,power_law_1.01,0.11294080018997192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,32,power_law_1.01,0.11185280084609986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,32,power_law_1.01,0.12244479656219483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,32,power_law_1.01,0.12666239738464355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,32,power_law_1.01,0.1369920015335083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,32,power_law_1.01,0.16150399446487426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,32,power_law_1.01,0.16522239446640014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,32,power_law_1.01,0.20291199684143066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,32,power_law_1.01,0.2357952117919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,32,power_law_1.01,0.30432639122009275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,32,power_law_1.01,0.30574080944061277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,32,power_law_1.01,0.38801279067993166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,32,power_law_1.01,0.5021503925323486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,32,power_law_1.01,0.6972544193267822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,32,power_law_1.01,0.8782336235046386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,2,power_law_1.2,0.07553279995918274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,2,power_law_1.2,0.09633920192718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,2,power_law_1.2,0.1246783971786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,2,power_law_1.2,0.18131200075149537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,2,power_law_1.2,0.20286719799041747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,2,power_law_1.2,0.32972159385681155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,2,power_law_1.2,0.35570559501647947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,2,power_law_1.2,0.33585920333862307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,2,power_law_1.2,0.3474751949310303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,2,power_law_1.2,0.3617791891098022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,2,power_law_1.2,0.36738560199737547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,2,power_law_1.2,0.37484800815582275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,2,power_law_1.2,0.37626240253448484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,2,power_law_1.2,0.37306239604949953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,2,power_law_1.2,0.406387186050415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,2,power_law_1.2,0.40195841789245607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,2,power_law_1.2,0.4348735809326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,2,power_law_1.2,0.4881472110748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,2,power_law_1.2,0.4872576236724854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,2,power_law_1.2,0.5516799926757813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,2,power_law_1.2,0.6231167793273926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,2,power_law_1.2,0.7231359958648682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,2,power_law_1.2,0.847481632232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,1,power_law_1.2,0.7561471939086915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,2,power_law_1.2,1.0804479598999024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,32,balanced,0.04587199787298838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,32,balanced,0.04598933458328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,32,balanced,0.046906664967536926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,32,balanced,0.05007466673851013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,32,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,32,balanced,0.057989334066708885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,32,balanced,0.057962665955225624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,32,balanced,0.05810666580994924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,32,balanced,0.05816000203291575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,32,balanced,0.059893334905306496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,32,balanced,0.058261334896087646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,32,balanced,0.06015466650327047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,32,balanced,0.061754668752352394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,32,balanced,0.06065600117047628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,32,balanced,0.06605866551399231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,32,balanced,0.06607466439406078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,32,balanced,0.06831466654936473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,32,balanced,0.07426666716734569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,32,balanced,0.07909866670767467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,32,balanced,0.0909546713034312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,32,balanced,0.11134399970372517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,32,balanced,0.13220799962679544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,32,balanced,0.14917866388956705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,32,balanced,0.1925813357035319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,32,balanced,0.22206399838129678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,32,balanced,0.2997973362604777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,32,balanced,0.38124267260233563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,2,power_law_1.2,1.3791104316711427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,2,power_law_1.2,1.9477504730224608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,32,power_law_1.01,0.044268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,32,power_law_1.01,0.03516800105571747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,32,power_law_1.01,0.03224320113658905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,32,power_law_1.01,0.032339200377464294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,2,power_law_1.2,2.318796730041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,32,power_law_1.01,0.033228799700737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,32,power_law_1.01,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,32,power_law_1.01,0.034969601035118106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,32,power_law_1.01,0.03484799861907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,32,power_law_1.01,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,32,power_law_1.01,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,32,power_law_1.01,0.038790398836135866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,32,power_law_1.01,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,32,power_law_1.01,0.04257920086383819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,32,power_law_1.01,0.04602240025997162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,32,power_law_1.01,0.05082240104675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,32,power_law_1.01,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,32,power_law_1.01,0.05366399884223938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,32,power_law_1.01,0.061510401964187625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,32,power_law_1.01,0.06357759833335877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,32,power_law_1.01,0.08001279830932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,32,power_law_1.01,0.08912640213966369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,32,power_law_1.01,0.11577600240707397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,32,power_law_1.01,0.12268160581588745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,32,power_law_1.01,0.16191359758377075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,32,power_law_1.01,0.19408639669418334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,32,power_law_1.01,0.31831040382385256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,32,power_law_1.01,0.3095936059951782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,128,balanced,0.04693333307902018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,128,balanced,0.04614399870236715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,128,balanced,0.04630400240421295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,128,balanced,0.05348266661167145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,128,balanced,0.05422399938106537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,128,balanced,0.05518933137257894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,128,balanced,0.05460800230503082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,128,balanced,0.056320001681645714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,128,balanced,0.05625600119431814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,128,balanced,0.05628266433874766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,128,balanced,0.05756799876689911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,128,balanced,0.0602453351020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,128,balanced,0.06001600126425425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,128,balanced,0.06025066475073496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,128,balanced,0.06620266536871593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,128,balanced,0.06820266445477803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,128,balanced,0.07028266787528992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,128,balanced,0.08522666494051616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,128,balanced,0.08668266733487447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,128,balanced,0.11734400192896526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,128,balanced,0.12800533572832742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,128,balanced,0.18284799655278525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,128,balanced,0.18074132998784384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,128,balanced,0.2361546754837036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,128,balanced,0.2706560095151265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,128,balanced,0.36156801382700604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,128,balanced,0.4431626796722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,8,power_law_1.01,0.04387840032577515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,8,power_law_1.01,0.04428800046443939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,8,power_law_1.01,0.04470399916172028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,8,power_law_1.01,0.04636160135269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,8,power_law_1.01,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,8,power_law_1.01,0.04636160135269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,8,power_law_1.01,0.045875200629234315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,8,power_law_1.01,0.04496000111103058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,8,power_law_1.01,0.044761601090431216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,8,power_law_1.01,0.04421760141849518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,8,power_law_1.01,0.04176000058650971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,8,power_law_1.01,0.05063679814338684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,8,power_law_1.01,0.05663359761238098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,8,power_law_1.01,0.04893440008163452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,8,power_law_1.01,0.04843519926071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,8,power_law_1.01,0.05736960172653198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,8,power_law_1.01,0.056576001644134524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,8,power_law_1.01,0.07503359913825988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,8,power_law_1.01,0.061433601379394534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,8,power_law_1.2,0.036601600050926206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,8,power_law_1.2,0.03672960102558136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,8,power_law_1.01,0.06973440051078797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,8,power_law_1.2,0.03614720106124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,8,power_law_1.2,0.038700801134109494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,8,power_law_1.2,0.03909119963645935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,8,power_law_1.2,0.039980798959732056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.01,0.08125439882278443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,8,power_law_1.2,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,8,power_law_1.2,0.04063999950885773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,8,power_law_1.01,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,8,power_law_1.2,0.04097279906272888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,8,power_law_1.01,0.08957440257072449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,8,power_law_1.2,0.04103679955005646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,8,power_law_1.01,0.09386879801750184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,8,power_law_1.2,0.04136320054531097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,8,power_law_1.2,0.042284798622131345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,8,power_law_1.2,0.04211199879646301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,8,power_law_1.2,0.04360319972038269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,8,power_law_1.2,0.048019200563430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.01,0.10606720447540283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,8,power_law_1.2,0.05012480020523071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,8,power_law_1.01,0.10580480098724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,8,power_law_1.2,0.05256320238113403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,8,power_law_1.2,0.05842559933662415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,8,power_law_1.01,0.10354559421539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,8,power_law_1.01,0.10169600248336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.01,0.125600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,8,power_law_1.01,0.09839360117912292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,8,power_law_1.01,0.10839040279388427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,8,power_law_1.01,0.10998400449752807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,8,power_law_1.01,0.11139839887619019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,8,power_law_1.01,0.11133439540863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,8,power_law_1.2,0.06621440052986145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,8,power_law_1.01,0.11649279594421387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,8,power_law_1.01,0.12383999824523925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,8,power_law_1.01,0.12715519666671754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,8,power_law_1.01,0.1331264019012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,8,power_law_1.01,0.14842239618301392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,8,power_law_1.01,0.1530751943588257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,8,power_law_1.01,0.18775039911270142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.01,0.1640895962715149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,8,power_law_1.01,0.20145280361175538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,8,power_law_1.01,0.25205121040344236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,8,power_law_1.01,0.2876287937164307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,8,power_law_1.01,0.3900736093521118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,8,power_law_1.01,0.49402880668640137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,8,power_law_1.01,0.7322239875793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,8,power_law_1.01,0.8306624412536621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,8,power_law_1.2,0.0887167990207672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,32,8,balanced,0.039850667119026184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,32,8,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,32,8,balanced,0.043568000197410583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,32,8,balanced,0.04786133269468943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,32,8,balanced,0.04974400003751119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,32,8,balanced,0.059290667374928795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,32,8,balanced,0.06115733087062836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,32,8,balanced,0.060975998640060425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,32,8,balanced,0.06192533175150553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,32,8,balanced,0.06206933160622915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,32,8,balanced,0.062352001667022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,32,8,balanced,0.06422933439413707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,32,8,balanced,0.06630933284759521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,32,8,balanced,0.06628266473611195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,32,8,balanced,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,32,8,balanced,0.07446933289368947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.01,0.19954559803009034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,32,8,balanced,0.07798933486143748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,8,power_law_1.2,0.1078976035118103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,32,8,balanced,0.0886346697807312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,32,8,balanced,0.0950986643632253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,32,8,balanced,0.1195199986298879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,32,8,balanced,0.1428053379058838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,32,8,balanced,0.19630932807922363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,32,8,balanced,0.2327679991722107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,32,8,balanced,0.3233120044072469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,32,8,balanced,0.4150826533635457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,32,8,balanced,0.5950186649958292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,32,8,balanced,0.7668373584747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.01,0.2726720094680786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,8,power_law_1.2,0.1471743941307068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.01,0.37163519859313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,8,power_law_1.2,0.17701760530471802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,8,power_law_1.2,0.24792320728302003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,2,balanced,0.023445333043734234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,2,balanced,0.023221333821614582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,2,balanced,0.027237333357334137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,2,balanced,0.03939733405907949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,2,balanced,0.05789866546789805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,2,balanced,0.06529599924882253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,2,balanced,0.06835199892520905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,2,balanced,0.07202133536338806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,2,balanced,0.07448000212510426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,2,balanced,0.07843199868996938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,2,balanced,0.08236800134181976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,2,balanced,0.0844533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,2,balanced,0.08988799651463826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,2,balanced,0.0888266662756602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,2,balanced,0.0946720043818156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,2,balanced,0.11338133613268535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,2,balanced,0.1165013313293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,2,balanced,0.1886026660601298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,2,balanced,0.19938133160273233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,2,balanced,0.22107199827829996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,2,balanced,0.22403732935587564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,2,balanced,0.25710399945576984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,2,balanced,0.26910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,2,balanced,0.4325600067774455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,2,balanced,0.4909493525822957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,2,balanced,0.7016159693400065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,2,balanced,0.9103786945343018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,8,power_law_1.2,0.3186367988586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.01,0.5611199855804443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,8,power_law_1.2,0.4749760150909424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,1,balanced,0.052015999952952065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,1,balanced,0.05141866703828176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,1,balanced,0.05788266658782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,1,balanced,0.06850133339564006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,1,balanced,0.10147733489672343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,1,balanced,0.10388267040252686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,1,balanced,0.10496000448862712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,1,balanced,0.10605866710344951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,1,balanced,0.10873599847157796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,1,balanced,0.10775466759999593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,1,balanced,0.1125440001487732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,1,balanced,0.11300266782442729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,1,balanced,0.11713600158691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,1,balanced,0.12091733018557231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,1,balanced,0.12880000472068787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,1,balanced,0.13371733824412027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,8,power_law_1.2,0.5836095809936523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,1,balanced,0.1458613375822703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.01,0.7177728176116943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,1,balanced,0.16875199476877847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,1,balanced,0.1918506622314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,1,balanced,0.23914666970570883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,1,balanced,0.2835200031598409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,1,balanced,0.41793068250020343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,1,balanced,0.5138346751530966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,1,balanced,0.7593920230865479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,1,balanced,0.9791946411132812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,4,power_law_1.2,0.04496000111103058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,1,balanced,1.433498700459798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,4,power_law_1.2,0.06322559714317322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,4,power_law_1.2,0.07680000066757202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,1,balanced,1.8862773577372234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,4,power_law_1.2,0.10919040441513062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,4,power_law_1.2,0.14345599412918092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,4,power_law_1.2,0.1572991967201233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,4,power_law_1.2,0.19286400079727173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,4,power_law_1.2,0.18101760149002075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,4,power_law_1.2,0.19434880018234252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,4,power_law_1.2,0.18831360340118408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,4,power_law_1.2,0.20101759433746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,4,power_law_1.2,0.21020801067352296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,4,power_law_1.2,0.22035839557647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,4,power_law_1.2,0.21234560012817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,4,power_law_1.2,0.22661759853363037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,4,power_law_1.2,0.22910718917846679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,4,power_law_1.2,0.2577343940734863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,4,power_law_1.2,0.26725120544433595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,4,power_law_1.2,0.28774399757385255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,4,power_law_1.2,0.33810560703277587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,4,power_law_1.2,0.43182082176208497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,4,power_law_1.2,0.4710527896881104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,4,power_law_1.2,0.5149695873260498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,4,power_law_1.2,0.6626815795898438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,4,power_law_1.2,0.8097920417785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,4,power_law_1.2,1.110041618347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,4,power_law_1.2,1.3964351654052733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,64,balanced,0.05221866567929586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,64,balanced,0.04997866849104563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,64,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,64,balanced,0.07230400045712788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,64,balanced,0.08039466540018718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,64,balanced,0.08205866813659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,64,balanced,0.08238400022188823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,64,balanced,0.08117333551247914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,64,balanced,0.08242133259773254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,64,balanced,0.08240533371766408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,64,balanced,0.08362133304278056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,64,balanced,0.08601066470146179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,64,balanced,0.0864533285299937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,64,balanced,0.0865066647529602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,64,balanced,0.09257599711418152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,64,balanced,0.09284266829490662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,64,balanced,0.09595200419425964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,64,balanced,0.11317333579063416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,64,balanced,0.1129866639773051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,64,balanced,0.1399626632531484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,64,balanced,0.14621866742769876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,64,balanced,0.19123733043670654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,64,balanced,0.18024533987045288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,64,balanced,0.242576003074646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,64,balanced,0.2490666707356771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,64,balanced,0.331167995929718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,64,balanced,0.38757868607838947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,2,power_law_1.2,0.021715199947357176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,2,power_law_1.2,0.02325119972229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,2,power_law_1.2,0.02744320034980774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,2,power_law_1.2,0.040556800365447995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,2,power_law_1.2,0.05320960283279419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,2,power_law_1.2,0.062438398599624634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,2,power_law_1.2,0.07167360186576843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,2,power_law_1.2,0.07948160171508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,2,power_law_1.2,0.07945600152015686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,2,power_law_1.2,0.08321279883384705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,2,power_law_1.2,0.089683198928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,2,power_law_1.2,0.08681600093841553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,2,power_law_1.2,0.09006080031394958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,2,power_law_1.2,0.09747200012207032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,2,power_law_1.2,0.0941760003566742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,2,power_law_1.2,0.10080000162124633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,2,power_law_1.2,0.10743039846420288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,2,power_law_1.2,0.13351680040359498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.2,0.13886719942092896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.2,0.24141440391540528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.2,0.285152006149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.2,0.2593087911605835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.2,0.286297607421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.2,0.3577023983001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.2,0.4466752052307129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.2,0.5943552017211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.2,0.7510335922241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,2,power_law_1.2,0.06603519916534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,2,power_law_1.2,0.07144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,2,power_law_1.2,0.07683839797973632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,2,power_law_1.2,0.08848000168800355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,2,power_law_1.2,0.10150400400161744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,2,power_law_1.2,0.1135103940963745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,2,power_law_1.2,0.1361407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,2,power_law_1.2,0.13964799642562867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,2,power_law_1.2,0.14180480241775512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,2,power_law_1.2,0.14806400537490844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,2,power_law_1.2,0.14878079891204835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,2,power_law_1.2,0.15657600164413452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,2,power_law_1.2,0.1623296022415161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,2,power_law_1.2,0.1665343999862671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,2,power_law_1.2,0.17548799514770508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,2,power_law_1.2,0.18078720569610596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,2,power_law_1.2,0.18940160274505616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,2,power_law_1.2,0.22110719680786134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,2,power_law_1.2,0.2462912082672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,2,power_law_1.2,0.29478399753570556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,2,power_law_1.2,0.35717120170593264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,2,power_law_1.2,0.4614528179168701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,2,power_law_1.2,0.5858496189117431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,2,power_law_1.2,0.7774720191955566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,2,power_law_1.2,1.0433728218078613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,2,power_law_1.2,1.4600383758544921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,2,power_law_1.2,2.0737791061401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,1,power_law_1.01,0.03600000143051148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,1,power_law_1.01,0.03761920034885406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,1,power_law_1.01,0.04455040097236633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,1,power_law_1.01,0.060787200927734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,1,power_law_1.01,0.0704576015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,1,power_law_1.01,0.09341440200805665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,1,power_law_1.01,0.09731199741363525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,1,power_law_1.01,0.0989184021949768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,1,power_law_1.01,0.10084480047225952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,1,power_law_1.01,0.10267519950866699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,1,power_law_1.01,0.10604159832000733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,1,power_law_1.01,0.10748159885406494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,1,power_law_1.01,0.11059199571609497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,1,power_law_1.01,0.12985600233078004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,1,power_law_1.01,0.14365439414978026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,1,power_law_1.01,0.14835840463638306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,1,power_law_1.01,0.16405119895935058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,1,power_law_1.01,0.18551039695739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,2,balanced,0.030250666042168934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,1,power_law_1.01,0.19671039581298827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,2,balanced,0.03173333406448364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,2,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,1,power_law_1.01,0.25201919078826907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,2,balanced,0.03588266670703888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,2,balanced,0.03540800015131632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,2,balanced,0.037290667494138084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,2,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,2,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,2,balanced,0.037802666425704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,2,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,2,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,2,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,2,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,2,balanced,0.0423573354880015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,2,balanced,0.05388266841570536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,2,balanced,0.054245332876841225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,2,balanced,0.05829333265622457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,2,balanced,0.06638933221499126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,2,balanced,0.06850133339564006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,2,balanced,0.08525866270065308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,2,balanced,0.09313066800435384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,2,balanced,0.11942399541536967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,1,power_law_1.01,0.2945152044296265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,2,balanced,0.14193600416183472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,2,balanced,0.18728532393773398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,2,balanced,0.22581332921981812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,2,balanced,0.31356799602508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,2,balanced,0.38861334323883057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,2,power_law_1.01,0.0605567991733551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,2,power_law_1.01,0.07613440155982971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,2,power_law_1.01,0.0914687991142273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,1,power_law_1.01,0.4016255855560303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,2,power_law_1.01,0.1250496029853821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,1,power_law_1.01,0.5147520065307617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,1,balanced,0.09187733133633931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,2,power_law_1.01,0.1606528043746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,1,balanced,0.09520000219345093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,1,balanced,0.10550933082898457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,2,power_law_1.01,0.1837183952331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,2,power_law_1.01,0.24707839488983155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,2,power_law_1.01,0.2610368013381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,2,power_law_1.01,0.2586368083953857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,2,power_law_1.01,0.2670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,2,power_law_1.01,0.2668800115585327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,1,power_law_1.01,0.6917568206787109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,1,balanced,0.12986666957537332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,2,power_law_1.01,0.27102079391479494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,1,balanced,0.17915733655293783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,1,balanced,0.18020800749460855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,2,power_law_1.01,0.2790591955184937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,1,balanced,0.18409067392349243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,1,balanced,0.18367999792099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,1,balanced,0.18754667043685913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,1,balanced,0.18851200739542642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,1,balanced,0.18930667638778687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,1,balanced,0.19615467389424643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,1,power_law_1.01,0.8837375640869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,2,power_law_1.01,0.2962048053741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,1,balanced,0.19672000408172607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,1,balanced,0.20127467314402261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,1,power_law_1.01,1.2361727714538575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,2,power_law_1.01,0.30293760299682615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,1,balanced,0.19734932978947958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,2,power_law_1.01,0.3168832063674927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,1,balanced,0.20223466555277506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,1,power_law_1.01,1.6424127578735352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,2,power_law_1.01,0.33509120941162107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,2,power_law_1.01,0.368287992477417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,2,power_law_1.01,0.36608641147613524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,2,power_law_1.01,0.4238912105560303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,2,power_law_1.01,0.45734400749206544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,1,balanced,0.2089866598447164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,1,balanced,0.22241065899531046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,1,balanced,0.23757867018381754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,1,balanced,0.27245332797368366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,2,power_law_1.01,0.550105619430542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,1,balanced,0.3044266700744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,1,balanced,0.43455998102823895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,2,power_law_1.01,0.655782413482666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,1,balanced,0.5083680152893066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,1,balanced,0.7145067056020101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,1,balanced,0.9350933233896891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,1,balanced,1.3470880190531414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,2,power_law_1.01,0.8701312065124511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,1,balanced,1.7657866477966309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,2,power_law_1.01,1.10066556930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,2,power_law_1.01,1.4326463699340821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,2,power_law_1.01,1.8039936065673827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,4,balanced,0.019215999792019527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,4,balanced,0.018960000326236088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,4,balanced,0.018789333601792652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,4,balanced,0.018895999838908512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,4,balanced,0.021327999730904896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,4,balanced,0.022272000710169475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,4,balanced,0.025360000630219776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,4,balanced,0.027034667630990345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,4,balanced,0.0271573339899381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,4,balanced,0.0272533322374026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,4,balanced,0.025333332518736523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,4,balanced,0.027082666754722595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,4,balanced,0.02923733244339625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,4,balanced,0.031311998764673867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,4,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,4,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,4,balanced,0.03822933385769526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,4,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,4,balanced,0.05201066533724467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,4,balanced,0.06610133250554402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,4,balanced,0.06645333270231883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,4,balanced,0.07607466479142506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,4,balanced,0.08513599634170532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,4,balanced,0.12410133083661397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,4,balanced,0.1442026694615682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,4,balanced,0.19867199659347534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,4,balanced,0.2560960054397583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,2,balanced,0.06899199883143108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,2,balanced,0.07133866846561432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,2,balanced,0.0809386670589447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,2,balanced,0.11188800136248271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,2,balanced,0.1546453336874644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,2,balanced,0.23850667476654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,2,balanced,0.3208799958229065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,2,balanced,0.32444266478220624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,2,balanced,0.32650667428970337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,2,balanced,0.33188267548878986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,2,balanced,0.3227146665255229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,2,balanced,0.3304746747016907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,2,balanced,0.3309333324432373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,2,balanced,0.333840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,2,balanced,0.3472586472829183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,2,balanced,0.34693864981333417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,2,balanced,0.3598453203837077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,2,balanced,0.38839467366536456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,2,balanced,0.4113493363062541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,2,balanced,0.46297065416971844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,2,balanced,0.5007413228352865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,2,balanced,0.6000159978866577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,2,balanced,0.7088373502095541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,2,balanced,0.9157866636912028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,16,power_law_1.2,0.06181120276451111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,2,balanced,1.218282699584961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,16,power_law_1.2,0.07709439992904663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,2,balanced,1.6284319559733074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,16,power_law_1.2,0.07185279726982116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,16,power_law_1.2,0.07788159847259521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,2,balanced,2.185093402862549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,16,power_law_1.2,0.07741439938545228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,16,power_law_1.2,0.07479680180549622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,16,power_law_1.2,0.0750976026058197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,16,power_law_1.2,0.07826560139656066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,16,power_law_1.2,0.0785152018070221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,16,power_law_1.2,0.07822080254554749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,16,power_law_1.2,0.07793279886245727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,16,power_law_1.2,0.08207359910011292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,16,power_law_1.2,0.08197759985923767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,16,power_law_1.2,0.08543999791145325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,16,power_law_1.2,0.09707520008087159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,16,power_law_1.2,0.09909759759902954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,16,power_law_1.2,0.10621440410614014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,16,power_law_1.2,0.12227200269699097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,16,power_law_1.2,0.12382080554962158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,16,power_law_1.2,0.15371520519256593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,16,power_law_1.2,0.17809280157089233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,16,power_law_1.2,0.2225600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,16,power_law_1.2,0.31756160259246824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,16,power_law_1.2,0.44496002197265627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,16,power_law_1.2,0.5570623874664307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,16,power_law_1.2,0.8642815589904785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,16,power_law_1.2,1.1353983879089355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,32,balanced,0.02926933268706004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,32,balanced,0.029301332930723827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,32,balanced,0.027621333797772724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,32,balanced,0.031258667508761086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,32,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,32,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,32,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,32,balanced,0.03544000039498011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,32,balanced,0.035360001027584076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,32,balanced,0.03534399966398875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,32,balanced,0.035546667873859406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,32,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,32,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,32,balanced,0.0377813329299291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,32,balanced,0.043568000197410583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,32,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,32,balanced,0.04351999859015147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,32,balanced,0.046037331223487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,32,balanced,0.04788800080617269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,32,balanced,0.053904001911481224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,32,balanced,0.05442133545875549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,32,balanced,0.0681279997030894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,32,balanced,0.0767146646976471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,32,balanced,0.09841066598892212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,32,balanced,0.12001599868138631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,32,balanced,0.16874667008717856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,32,balanced,0.21178666750590006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,128,power_law_1.2,0.10422400236129761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,128,power_law_1.2,0.10701440572738648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,128,power_law_1.2,0.10224000215530396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,128,power_law_1.2,0.10947840213775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,128,power_law_1.2,0.10533759593963624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,128,power_law_1.2,0.1059008002281189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,128,power_law_1.2,0.11079039573669433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,128,power_law_1.2,0.1168768048286438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,128,power_law_1.2,0.11106560230255128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,128,power_law_1.2,0.11559679508209228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,1,power_law_1.01,0.023904000222682954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,128,power_law_1.2,0.12003200054168701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,1,power_law_1.01,0.03297280073165894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,128,power_law_1.2,0.11279360055923462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,1,power_law_1.01,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,1,power_law_1.01,0.06851199865341187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,128,power_law_1.2,0.11729919910430908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,1,power_law_1.01,0.08944640159606934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,1,power_law_1.01,0.13386240005493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,1,power_law_1.01,0.13484159708023072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,128,power_law_1.2,0.11733119487762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,1,power_law_1.01,0.14376319646835328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,128,power_law_1.2,0.1194111943244934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,1,power_law_1.01,0.14617600440979003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2688,1856,6,128,1,1,power_law_1.01,0.14963200092315673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2688,1856,6,128,1,1,power_law_1.01,0.15588480234146118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,32,power_law_1.01,0.054745602607727054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2688,1856,6,128,1,1,power_law_1.01,0.1704576015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2688,1856,6,128,1,1,power_law_1.01,0.16072959899902345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,32,power_law_1.01,0.051123201847076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2688,1856,6,128,1,1,power_law_1.01,0.16813440322875978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2688,1856,6,128,1,1,power_law_1.01,0.18133759498596191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,32,power_law_1.01,0.049804800748825075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2688,1856,6,128,1,1,power_law_1.01,0.17463040351867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,128,power_law_1.2,0.11582080125808716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2688,1856,6,128,1,1,power_law_1.01,0.18792959451675414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,32,power_law_1.01,0.051609599590301515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2688,1856,6,128,1,1,power_law_1.01,0.22650880813598634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,128,power_law_1.2,0.11920640468597413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,32,power_law_1.01,0.051948797702789304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,128,power_law_1.2,0.13409279584884642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,128,power_law_1.2,0.16243840456008912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,128,power_law_1.2,0.19720959663391113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,128,power_law_1.2,0.20609281063079835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,128,power_law_1.2,0.25830399990081787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,128,power_law_1.2,0.3403327941894531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.01,0.2179327964782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,32,power_law_1.01,0.05438079833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,128,power_law_1.2,0.45493121147155763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,32,power_law_1.01,0.054681599140167236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,4,balanced,0.0417546679576238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,128,power_law_1.2,0.6234687805175781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,4,balanced,0.04195733368396759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,4,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,32,power_law_1.01,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,4,balanced,0.06005333364009857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,4,balanced,0.07638399799664815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,128,power_law_1.2,0.9340864181518554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,4,balanced,0.11863999565442403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,4,balanced,0.12144533793131511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,4,balanced,0.11760532855987549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,32,power_law_1.01,0.05679360032081604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,4,balanced,0.11678933103879292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,4,balanced,0.11866666873296101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,128,power_law_1.2,1.120793628692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,4,balanced,0.12005866567293803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,4,balanced,0.12052266796429952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,32,power_law_1.01,0.05857279896736145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,32,power_law_1.01,0.05800319910049438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,4,power_law_1.01,0.13354239463806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,4,balanced,0.12273066242535909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,4,balanced,0.12427199880282085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,4,balanced,0.12929067015647888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,4,balanced,0.13127467036247253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,4,balanced,0.13336533308029175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,4,balanced,0.14572266737620035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,4,balanced,0.14913066228230795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,32,power_law_1.01,0.05931519865989685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,4,balanced,0.17162134250005087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,4,balanced,0.18556267023086548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,32,power_law_1.01,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,4,balanced,0.22407466173171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,4,balanced,0.2600213289260864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,32,power_law_1.01,0.06180480122566223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,4,balanced,0.3441386620203654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,4,power_law_1.01,0.21211519241333007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,4,power_law_1.01,0.2663808107376099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,4,power_law_1.01,0.34420480728149416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,4,power_law_1.01,0.47385601997375487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,4,balanced,0.4031146764755249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,32,power_law_1.01,0.07185279726982116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,4,power_law_1.01,0.4994495868682861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,4,balanced,0.564026673634847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,32,power_law_1.01,0.0725055992603302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,4,balanced,0.7001279989878336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,4,power_law_1.01,0.6196095943450928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,32,power_law_1.01,0.0775168001651764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,4,power_law_1.01,0.5732351779937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,4,power_law_1.01,0.5780672073364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,4,power_law_1.01,0.6226816177368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,4,power_law_1.01,0.5943679809570312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,4,power_law_1.01,0.6454720020294189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,32,power_law_1.01,0.09294080138206481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,4,power_law_1.01,0.6497920036315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,32,power_law_1.01,0.0972927987575531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,4,power_law_1.01,0.6508863925933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,32,power_law_1.01,0.12104959487915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,4,power_law_1.01,0.6648575782775878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,32,power_law_1.01,0.1394047975540161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,4,power_law_1.01,0.6694528102874756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,32,power_law_1.01,0.17758079767227172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,4,power_law_1.01,0.7473663806915283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,32,power_law_1.01,0.21118080615997314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,4,power_law_1.01,0.7906432151794434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,32,power_law_1.01,0.2930943965911865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,4,power_law_1.01,0.8215231895446777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,32,power_law_1.01,0.390995192527771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,4,power_law_1.01,0.8885631561279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,32,power_law_1.01,0.5633344173431396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,4,power_law_1.01,0.9192000389099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,32,power_law_1.01,0.706118392944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,4,power_law_1.01,1.0760191917419433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,4,power_law_1.01,1.1964608192443849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,4,power_law_1.01,1.439948844909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,4,power_law_1.01,1.6452159881591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,4,power_law_1.01,2.2280576705932615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,4,power_law_1.01,2.6587007522583006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,32,power_law_1.2,0.06388480067253113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,32,power_law_1.2,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,32,power_law_1.2,0.06034560203552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,32,power_law_1.2,0.06446719765663148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,32,power_law_1.2,0.0645695984363556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,32,power_law_1.2,0.06467840075492859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,32,power_law_1.2,0.06633599996566772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,32,power_law_1.2,0.06646400094032287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,32,power_law_1.2,0.06820480227470398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,32,power_law_1.2,0.06878719925880432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,32,power_law_1.2,0.06883199810981751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,32,power_law_1.2,0.07095680236816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,2,power_law_1.01,0.043398401141166686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,2,power_law_1.01,0.04528000056743622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,2,power_law_1.01,0.04702720046043396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,2,power_law_1.01,0.054079997539520266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,2,power_law_1.01,0.06112639904022217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,2,power_law_1.01,0.06752640008926392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,2,power_law_1.01,0.07374079823493958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,2,power_law_1.01,0.07702400088310242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,2,power_law_1.01,0.0787775993347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,2,power_law_1.01,0.08010879755020142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,2,power_law_1.01,0.0814848005771637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,2,power_law_1.01,0.08831999897956848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,32,power_law_1.2,0.07256960272789001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,2,power_law_1.01,0.09000319838523865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,32,power_law_1.2,0.07349119782447815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,32,power_law_1.2,0.07882239818572997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,32,power_law_1.2,0.07964800000190735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,32,power_law_1.2,0.0825215995311737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,32,power_law_1.2,0.0942143976688385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,32,power_law_1.2,0.10945279598236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,2,power_law_1.01,0.09760640263557434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,2,power_law_1.01,0.10586880445480347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,2,power_law_1.01,0.11100800037384033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,2,power_law_1.01,0.11459200382232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,2,power_law_1.01,0.1328703999519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,2,power_law_1.01,0.14659839868545532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,2,power_law_1.01,0.18428159952163697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,32,power_law_1.2,0.12839679718017577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,2,power_law_1.01,0.21500799655914307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,32,power_law_1.2,0.14488960504531861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,2,power_law_1.01,0.2852799892425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,32,power_law_1.2,0.1966912031173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,2,power_law_1.01,0.35126399993896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,32,power_law_1.2,0.23848319053649902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,32,power_law_1.2,0.3409087896347046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,32,power_law_1.2,0.43708162307739257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,32,power_law_1.2,0.7045504093170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,2,power_law_1.01,0.4930880069732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,2,power_law_1.01,0.6314752101898193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,32,power_law_1.2,0.8998271942138671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,2,power_law_1.01,0.9132927894592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,2,power_law_1.01,1.2271103858947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,4,power_law_1.01,0.04170880019664765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,4,power_law_1.01,0.04917120039463043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,4,power_law_1.01,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,4,power_law_1.01,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,4,power_law_1.01,0.055846399068832396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,4,power_law_1.01,0.056979197263717654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,4,power_law_1.01,0.06105599999427795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,4,power_law_1.01,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,4,power_law_1.01,0.06110079884529114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,4,power_law_1.01,0.06202239990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,4,power_law_1.01,0.06372479796409607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,4,power_law_1.01,0.0666815996170044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,4,power_law_1.01,0.07099519968032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,4,power_law_1.01,0.07305600047111512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,4,power_law_1.01,0.08326399922370911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,4,power_law_1.01,0.08912640213966369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,4,power_law_1.01,0.09797120094299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,4,power_law_1.01,0.11768959760665894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,4,power_law_1.01,0.1272447943687439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,4,power_law_1.01,0.170579195022583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,4,power_law_1.01,0.19742079973220825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,4,power_law_1.01,0.31176960468292236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,4,power_law_1.01,0.33124480247497556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,4,power_law_1.01,0.46671361923217775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,4,power_law_1.01,0.6183743953704834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,64,power_law_1.2,0.028729599714279175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,64,power_law_1.2,0.02847360074520111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,64,power_law_1.2,0.02813439965248108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,64,power_law_1.2,0.03020159900188446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,64,power_law_1.2,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,64,power_law_1.2,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,64,power_law_1.2,0.06320639848709106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,64,power_law_1.2,0.06364160180091857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,64,power_law_1.2,0.07775999903678894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,64,power_law_1.2,0.07791360020637512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,64,power_law_1.2,0.07663999795913697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,64,power_law_1.2,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,64,power_law_1.2,0.0726207971572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,64,power_law_1.2,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,64,power_law_1.2,0.06583679914474487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,64,power_law_1.2,0.062431997060775755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,64,power_law_1.2,0.07294719815254211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,64,power_law_1.2,0.07349119782447815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.2,0.08125439882278443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,1,power_law_1.2,0.049107199907302855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.2,0.10417280197143555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,1,power_law_1.2,0.05235840082168579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.2,0.12177920341491699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,1,power_law_1.2,0.06338559985160827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.2,0.16624640226364135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,4,power_law_1.2,0.06997119784355163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,1,power_law_1.2,0.08186240196228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.2,0.21571199893951415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,4,power_law_1.2,0.07297919988632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,1,power_law_1.2,0.09921919703483581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,4,power_law_1.01,0.9256256103515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.2,0.27980799674987794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.2,0.4228479862213135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.2,0.5513663768768311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,4,power_law_1.2,0.08352640271186829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,4,power_law_1.2,0.09827200174331666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,4,power_law_1.2,0.11272959709167481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,4,power_law_1.2,0.12956160306930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,4,power_law_1.2,0.12777600288391114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,1,power_law_1.2,0.13319679498672485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,1,power_law_1.2,0.14312319755554198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,1,power_law_1.2,0.14709759950637818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,1,power_law_1.2,0.15133440494537354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,1,power_law_1.2,0.16325119733810425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,1,power_law_1.2,0.16860159635543823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.2,0.8146944046020508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,1,power_law_1.2,0.1774783968925476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,4,power_law_1.2,0.13560320138931276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,1,power_law_1.2,0.18658560514450073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,4,power_law_1.2,0.13745919466018677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,1,power_law_1.2,0.18708479404449463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,4,power_law_1.2,0.13935359716415405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,1,power_law_1.2,0.20166399478912353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,4,power_law_1.2,0.13911039829254152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,1,power_law_1.2,0.21096959114074706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,4,power_law_1.2,0.145305597782135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,1,power_law_1.2,0.22829439640045165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,4,power_law_1.2,0.15122560262680054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,1,power_law_1.2,0.27869439125061035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,4,power_law_1.2,0.15486079454421997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,4,power_law_1.2,0.1625920057296753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,1,power_law_1.2,0.3260159969329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,4,power_law_1.2,0.16837120056152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,1,power_law_1.2,0.4185344219207764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,4,power_law_1.2,0.18091520071029663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,4,power_law_1.2,0.2070080041885376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.2,0.23344640731811522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.2,0.2746687889099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.2,0.30307199954986574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,1,power_law_1.2,0.49364480972290037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.2,0.41039361953735354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,1,power_law_1.2,0.6609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.2,0.4945024013519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,1,power_law_1.2,0.8669952392578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.2,0.6426176071166992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,64,balanced,0.02739733209212621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.2,0.8234111785888671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.2,1.2694463729858398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,1,power_law_1.2,1.1689087867736816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,4,power_law_1.01,1.0706303596496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,1,power_law_1.2,1.6202943801879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,64,balanced,0.027237333357334137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,64,balanced,0.027471999327341717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,64,balanced,0.029301332930723827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,64,balanced,0.03159466634194056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,1,power_law_1.2,2.208038330078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,64,balanced,0.03166399896144867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,64,balanced,0.03193599979082743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,64,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,64,balanced,0.0317546675602595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,64,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.2,1.5300479888916017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,64,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,64,balanced,0.03685333331425985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,64,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,64,balanced,0.03704000016053518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,64,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,64,balanced,0.04176533222198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,64,balanced,0.04170133173465729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,64,balanced,0.04397333165009817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,64,balanced,0.04689066608746847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,64,balanced,0.053472002347310386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,64,balanced,0.05492266515890757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,64,balanced,0.06824533144632976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,64,balanced,0.07640533149242401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,64,balanced,0.09852266311645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,64,balanced,0.1176639993985494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,64,balanced,0.1588373382886251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,64,balanced,0.2015626629193624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,1,power_law_1.2,2.8350143432617188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,2,power_law_1.2,0.10277760028839111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,2,power_law_1.2,0.16037119626998902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,2,power_law_1.2,0.1879487991333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,2,power_law_1.2,0.2445375919342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,2,power_law_1.2,0.2321984052658081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,2,power_law_1.2,0.23491199016571046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,2,power_law_1.2,0.25827200412750245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,2,power_law_1.2,0.2582144021987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,2,power_law_1.2,0.2719487905502319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,2,power_law_1.2,0.27242240905761717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,2,power_law_1.2,0.2698879957199097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,2,power_law_1.2,0.2825344085693359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,2,power_law_1.2,0.285536003112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,2,power_law_1.2,0.34294400215148924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,2,power_law_1.2,0.3568000078201294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,2,power_law_1.2,0.37348480224609376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,2,power_law_1.2,0.39571199417114256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,2,power_law_1.2,0.4555520057678223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,2,power_law_1.2,0.4687295913696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,2,power_law_1.2,0.6407872200012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,2,power_law_1.2,0.731712007522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,2,power_law_1.2,1.057356834411621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,2,power_law_1.2,1.23057279586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,2,power_law_1.2,2.0099712371826173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,2,power_law_1.2,2.5754623413085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,16,power_law_1.01,0.07163519859313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,16,power_law_1.01,0.07620480060577392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,16,power_law_1.01,0.06944640278816223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,16,power_law_1.01,0.0773311972618103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,16,power_law_1.01,0.07550719976425171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,16,power_law_1.01,0.07793279886245727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,16,power_law_1.01,0.07912319898605347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,16,power_law_1.01,0.08035839796066284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,16,power_law_1.01,0.07759360074996949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,16,power_law_1.01,0.08067839741706848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,16,power_law_1.01,0.08302080035209655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,16,power_law_1.01,0.08462719917297364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,16,power_law_1.01,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,16,power_law_1.01,0.08754559755325317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,16,power_law_1.01,0.09312639832496643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,16,power_law_1.01,0.09371520280838012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,16,power_law_1.01,0.10351999998092651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,16,power_law_1.01,0.11535359621047973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,16,power_law_1.01,0.11739519834518433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,16,power_law_1.01,0.1388800024986267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,16,power_law_1.01,0.1581120014190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,16,power_law_1.01,0.18777600526809693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,16,power_law_1.01,0.22906239032745362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,16,power_law_1.01,0.33743360042572024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,16,power_law_1.01,0.39202558994293213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,16,power_law_1.01,0.5386176109313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,16,power_law_1.01,0.6701183795928956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,8,power_law_1.2,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,8,power_law_1.2,0.03431040048599243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,8,power_law_1.2,0.028255999088287354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,8,power_law_1.2,0.029734399914741517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,8,power_law_1.2,0.030067199468612672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,8,power_law_1.2,0.030393600463867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,8,power_law_1.2,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,8,power_law_1.2,0.03278720080852508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,8,power_law_1.2,0.03438720107078552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,8,power_law_1.2,0.03431679904460907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,8,power_law_1.2,0.03280639946460724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,8,power_law_1.2,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,8,power_law_1.2,0.03553279936313629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,8,power_law_1.2,0.03682560026645661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,8,power_law_1.2,0.04377599954605103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,8,power_law_1.2,0.04824320077896118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,8,power_law_1.2,0.04751999974250794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,8,power_law_1.2,0.05617920160293579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,8,power_law_1.2,0.05868800282478333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,8,power_law_1.2,0.06716799736022949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,16,power_law_1.01,0.04792959988117218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,16,power_law_1.01,0.06711040139198303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,16,power_law_1.01,0.06844800114631652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,16,power_law_1.01,0.07104640007019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,16,power_law_1.01,0.07170559763908387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,16,power_law_1.01,0.0697983980178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,16,power_law_1.01,0.0697983980178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,16,power_law_1.01,0.06894720196723939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,16,power_law_1.01,0.07247999906539918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,16,power_law_1.01,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,16,power_law_1.01,0.07234560251235962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,16,power_law_1.01,0.07724159955978394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,16,power_law_1.01,0.08087040185928344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,16,power_law_1.01,0.08228480219841003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,16,power_law_1.01,0.09073280096054077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,8,power_law_1.2,0.0900223970413208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,16,power_law_1.01,0.09376639723777772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,16,power_law_1.01,0.10584959983825684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,32,power_law_1.2,0.045747199654579164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,16,power_law_1.01,0.12357759475708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,32,power_law_1.2,0.0452672004699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,16,power_law_1.01,0.1270784020423889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,16,power_law_1.01,0.1574399948120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,16,power_law_1.01,0.1855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,16,power_law_1.01,0.23223679065704345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,16,power_law_1.01,0.29832959175109863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,32,power_law_1.2,0.04415999948978424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,16,power_law_1.01,0.40991997718811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,32,power_law_1.2,0.043833601474761966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,8,power_law_1.2,0.11455999612808228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,16,power_law_1.01,0.5066944122314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,32,power_law_1.2,0.044588801264762876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,32,power_law_1.2,0.050406402349472045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,16,power_law_1.01,0.7700607776641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,32,power_law_1.2,0.070387202501297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,32,power_law_1.2,0.06213120222091675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,16,power_law_1.01,0.9214912414550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,32,power_law_1.2,0.08607360124588012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,32,power_law_1.2,0.08695039749145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,32,power_law_1.2,0.08677759766578674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,32,power_law_1.2,0.09639679789543151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,32,power_law_1.2,0.09578239917755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,32,power_law_1.2,0.08949120044708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,32,power_law_1.2,0.089055997133255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,8,power_law_1.2,0.12692480087280272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,32,power_law_1.2,0.09836159944534302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,32,power_law_1.2,0.09431040287017822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,32,power_law_1.2,0.09229440093040467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,32,power_law_1.2,0.09808639883995056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,32,power_law_1.2,0.11772799491882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,32,power_law_1.2,0.1350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,32,power_law_1.2,0.19200639724731444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,32,power_law_1.2,0.21801600456237794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,32,power_law_1.2,0.29141759872436523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,32,power_law_1.2,0.3996608018875122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,32,power_law_1.2,0.5833856105804444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,32,power_law_1.2,0.7271552085876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,2,power_law_1.2,2.959519958496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,8,power_law_1.2,0.17344000339508056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,2,power_law_1.2,0.04567039906978607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,2,power_law_1.2,0.05162879824638367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,2,power_law_1.2,0.05428479909896851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,2,power_law_1.2,0.06394240260124207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,2,power_law_1.2,0.06321920156478882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,2,power_law_1.2,0.0651199996471405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,2,power_law_1.2,0.0656000018119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,2,power_law_1.2,0.06645119786262513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,2,power_law_1.2,0.06594560146331788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,2,power_law_1.2,0.06672000288963317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,2,power_law_1.2,0.06680960059165955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,2,power_law_1.2,0.07465599775314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,2,power_law_1.2,0.0865343987941742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,2,power_law_1.2,0.08799999952316284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,2,power_law_1.2,0.09338880181312562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,2,power_law_1.2,0.0946943998336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,2,power_law_1.2,0.09872000217437744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,2,power_law_1.2,0.11597440242767335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,2,power_law_1.2,4.59093132019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,8,power_law_1.2,0.2169408082962036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,2,power_law_1.2,0.1419327974319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,2,power_law_1.2,0.17153279781341552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,2,power_law_1.2,0.2109312057495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,2,power_law_1.2,0.35041279792785646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,8,power_law_1.2,0.2831167936325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,2,power_law_1.2,0.4292736053466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,2,power_law_1.2,0.6074175834655762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,2,power_law_1.2,0.6906047821044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,32,power_law_1.2,0.07825279831886292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,32,power_law_1.2,0.06591359972953796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,32,power_law_1.2,0.06453760266304016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,32,power_law_1.2,0.06771199703216553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,32,power_law_1.2,0.0661952018737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,32,power_law_1.2,0.06478080153465271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,32,power_law_1.2,0.06718720197677612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,32,power_law_1.2,0.06974719762802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,32,power_law_1.2,0.07041280269622803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,32,power_law_1.2,0.07106559872627258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,32,power_law_1.2,0.07310079932212829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,32,power_law_1.2,0.07207040190696716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,32,power_law_1.2,0.07257599830627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,32,power_law_1.2,0.08217599987983704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,32,power_law_1.2,0.08476799726486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,32,power_law_1.2,0.08999680280685425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,32,power_law_1.2,0.09313920140266418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,32,power_law_1.2,0.10550400018692016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,32,power_law_1.2,0.10754560232162476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,32,power_law_1.2,0.14142080545425414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,8,power_law_1.2,0.3845632076263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,32,power_law_1.2,0.15632640123367308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,32,power_law_1.2,0.20195200443267822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,32,power_law_1.2,0.2145024061203003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,32,power_law_1.2,0.32085120677948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,32,power_law_1.2,0.38543360233306884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,32,power_law_1.2,0.6398784160614014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,32,power_law_1.2,0.8370240211486817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,2,power_law_1.2,0.9208383560180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,32,balanced,0.04213866591453552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,32,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,32,balanced,0.043194666504859924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,32,balanced,0.04671466847260793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,32,balanced,0.04558933277924856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,32,balanced,0.049642667174339294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,32,balanced,0.05000533163547516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,32,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,32,balanced,0.05173333485921224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,32,balanced,0.051872000098228455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,32,balanced,0.05201066533724467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,32,balanced,0.052784000833829246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,32,balanced,0.05486933390299479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,32,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,32,balanced,0.060218666990598045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,32,balanced,0.06030400097370148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,32,balanced,0.0621013343334198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,32,balanced,0.06821866830190022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,32,balanced,0.07069333394368489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,32,balanced,0.08895466725031535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,32,balanced,0.10788800319035848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,32,balanced,0.1493280033270518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,32,balanced,0.17393600940704346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,32,balanced,0.23430399099985758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,32,balanced,0.2794293363889058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,32,balanced,0.3927146593729655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,32,balanced,0.4974986712137858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,2,power_law_1.2,1.5533056259155273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,4,balanced,0.03988266736268997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,4,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,4,balanced,0.04158399999141693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,4,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,4,balanced,0.047685335079828896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,4,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,4,balanced,0.0580213318268458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,4,balanced,0.058229332168896995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,4,balanced,0.05817066629727682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,4,balanced,0.0582826683918635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,4,balanced,0.05816000203291575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,4,balanced,0.05992533266544342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,4,balanced,0.0602453351020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,4,balanced,0.062447999914487205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,4,balanced,0.06611733138561249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,4,balanced,0.06817600131034851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,4,balanced,0.07047466437021892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,4,balanced,0.08066133161385854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,4,balanced,0.10089600086212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,4,balanced,0.12264532844225566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,4,balanced,0.12326400478680928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,4,balanced,0.15876266360282898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,4,balanced,0.19644800821940103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,4,balanced,0.2895413239796956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,4,balanced,0.3465546766916911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,2,power_law_1.2,0.0876416027545929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,4,balanced,0.4986826578776042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,2,power_law_1.2,0.0930624008178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,4,balanced,0.6410026550292969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,2,power_law_1.2,0.12099200487136841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,2,power_law_1.2,0.1499071955680847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,2,power_law_1.2,0.16435840129852294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,2,power_law_1.2,0.23230719566345215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,2,power_law_1.2,0.24623360633850097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,2,power_law_1.2,0.2534656047821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,2,power_law_1.2,0.25338239669799806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,2,power_law_1.2,0.2569983959197998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,2,power_law_1.2,0.2594624042510986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,2,power_law_1.2,0.26204159259796145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,2,power_law_1.2,0.25975039005279543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,2,power_law_1.2,0.2682687997817993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,2,power_law_1.2,0.27583999633789064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,2,power_law_1.2,0.2846400022506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,2,power_law_1.2,0.294323205947876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,2,power_law_1.2,0.3265984058380127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.2,0.35562241077423096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.2,0.4347392082214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.2,0.49475841522216796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.2,0.6368959903717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.2,0.7655231952667236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.2,1.0409024238586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.2,1.2263999938964845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.2,1.8421375274658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,1,balanced,0.031712000568707786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,1,balanced,0.035317334036032356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,1,balanced,0.05472533404827118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,1,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,1,balanced,0.05717866619427999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,1,balanced,0.059658666451772056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,1,balanced,0.05912533402442932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,1,balanced,0.0591839998960495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,1,balanced,0.060821334520975746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,1,balanced,0.05989866455396017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,1,balanced,0.06136533121267954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,1,balanced,0.06225066880385081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,1,balanced,0.061994666854540505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,1,balanced,0.0739573339621226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,1,balanced,0.07971199850241344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,1,balanced,0.07900266846021016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,1,balanced,0.08200533191363017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,1,balanced,0.0955573320388794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,1,balanced,0.09992532928784688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,1,balanced,0.12158399820327759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,1,balanced,0.14628266294797262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,1,balanced,0.19129600127538046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,1,balanced,0.22574933369954428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,1,balanced,0.29469867547353107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,1,balanced,0.3795413176218669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,1,balanced,0.5338079929351807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,1,balanced,0.6831093629201254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.2,2.370035171508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,8,balanced,0.05955733358860016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,8,balanced,0.06015466650327047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,8,balanced,0.06025066475073496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,8,balanced,0.05994666616121928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,8,balanced,0.05921066800753275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,8,balanced,0.05982933441797892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,8,balanced,0.06005333364009857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,8,balanced,0.061850666999816895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,8,balanced,0.0609493354956309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,8,balanced,0.06131733457247416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,8,balanced,0.06598933537801106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,8,balanced,0.06474666794141133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,8,balanced,0.06573866804440816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,8,balanced,0.07049066821734111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,8,balanced,0.07427200178305308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,8,balanced,0.07726933558781941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,8,balanced,0.07829333345095317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,8,balanced,0.09435199697812398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,8,balanced,0.09671466549237569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,8,balanced,0.11544000109036763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,8,balanced,0.14006400108337402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,8,balanced,0.17643733819325766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,8,balanced,0.21446933348973593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,8,balanced,0.290613333384196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,8,balanced,0.36154667536417645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,8,balanced,0.5131200154622396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,8,balanced,0.6676426728566488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,2,power_law_1.2,0.019865599274635316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,2,power_law_1.2,0.01935359984636307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,2,power_law_1.2,0.01995519995689392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,2,power_law_1.2,0.020294399559497835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,2,power_law_1.2,0.02136320024728775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,2,power_law_1.2,0.02102400064468384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,2,power_law_1.2,0.02064639925956726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,2,power_law_1.2,0.022380800545215608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,2,power_law_1.2,0.023481599986553192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,2,power_law_1.2,0.023737600445747374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,2,power_law_1.2,0.025811201333999632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,2,power_law_1.2,0.04865280091762543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,2,power_law_1.2,0.048902401328086854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,2,power_law_1.2,0.05095679759979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,2,power_law_1.2,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,2,power_law_1.2,0.037676799297332766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,2,power_law_1.2,0.041580799221992495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,2,power_law_1.2,0.0442111998796463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.2,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.2,0.06626560091972351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.2,0.07541760206222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.2,0.09971839785575867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,8,power_law_1.2,0.031539198756217954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,8,power_law_1.2,0.031539198756217954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,8,power_law_1.2,0.032307198643684386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,8,power_law_1.2,0.03404799997806549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,8,power_law_1.2,0.03506560027599335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,8,power_law_1.2,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,8,power_law_1.2,0.036313599348068236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,8,power_law_1.2,0.03726080060005188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.2,0.12458879947662353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,8,power_law_1.2,0.038073599338531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,8,power_law_1.2,0.03848319947719574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,8,power_law_1.2,0.0390720009803772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,8,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,8,balanced,0.03336533407370249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,8,balanced,0.03332799921433131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,8,balanced,0.035445332527160645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,8,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,8,power_law_1.2,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,8,balanced,0.03740799923737844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,8,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,8,balanced,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,8,power_law_1.2,0.04364160001277924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,8,power_law_1.2,0.04741120040416717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,8,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,8,balanced,0.03756266583998998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,8,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,8,power_law_1.2,0.05467519760131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,8,power_law_1.2,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.2,0.17171839475631714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,8,balanced,0.03987200061480204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,8,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,8,power_law_1.2,0.06665599942207337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,8,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,2,power_law_1.01,0.053446400165557864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,8,power_law_1.2,0.08359040021896362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,2,power_law_1.01,0.0745024025440216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,2,power_law_1.01,0.10254080295562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,2,power_law_1.01,0.12139519453048705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,8,balanced,0.04576000074545542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,2,power_law_1.01,0.1700543999671936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,8,balanced,0.04637866715590159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,2,power_law_1.01,0.20165760517120362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,2,power_law_1.01,0.22734720706939698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,2,power_law_1.01,0.29041280746459963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,8,power_law_1.2,0.09752960205078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,2,power_law_1.01,0.28217599391937254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,8,balanced,0.051818668842315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,2,power_law_1.01,0.2858367919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,2,power_law_1.01,0.30566399097442626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,2,power_law_1.01,0.31214079856872556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,2,power_law_1.01,0.32315518856048586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,8,power_law_1.2,0.13301119804382325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,2,power_law_1.01,0.3290303945541382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,2,power_law_1.01,0.351091194152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,2,power_law_1.01,0.36008319854736326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,2,power_law_1.01,0.3835135936737061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,8,balanced,0.06171200176080068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,8,balanced,0.0606879989306132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,8,balanced,0.07427733143170674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,2,power_law_1.01,0.4057727813720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,2,power_law_1.01,0.45951361656188966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,8,power_law_1.2,0.16680320501327514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,2,power_law_1.01,0.5407487869262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,2,power_law_1.01,0.5894015789031982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,2,power_law_1.01,0.7229887962341308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,2,power_law_1.01,0.9299455642700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,2,power_law_1.01,1.193843173980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.2,0.21010560989379884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,8,balanced,0.07736533383528392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,2,power_law_1.01,1.4355327606201171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,2,power_law_1.01,1.9175039291381837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,2,power_law_1.01,2.528268814086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,8,power_law_1.2,0.22739200592041015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,8,balanced,0.10506666700045268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,8,balanced,0.12202133735020955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,8,balanced,0.16513599952061972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,8,balanced,0.21176000436147055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,8,balanced,0.28356266021728516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,8,balanced,0.3714666763941447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,2,balanced,0.06555200119813283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,2,balanced,0.06865600248177846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,2,balanced,0.07850666840871175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,2,balanced,0.09970666964848836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,2,balanced,0.13598400354385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,2,balanced,0.1981173356374105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,2,balanced,0.1885813275973002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,8,power_law_1.2,0.2795711994171143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,2,balanced,0.18825600544611612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,2,balanced,0.18926932414372763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,2,balanced,0.18538665771484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,2,balanced,0.18590933084487915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,2,balanced,0.18660267194112143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,2,balanced,0.18997865915298462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,2,balanced,0.1922826568285624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.2,0.3065792083740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,2,balanced,0.19755733013153076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,2,balanced,0.19991467396418253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,2,balanced,0.20834134022394815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,2,balanced,0.22313066323598227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,2,balanced,0.2414026657740275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,2,balanced,0.2718026638031006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,2,balanced,0.3060266574223836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,2,balanced,0.3741653362909953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,2,balanced,0.4421173334121704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,2,balanced,0.6653600136439005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,2,balanced,0.8071093559265137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,2,balanced,1.1708906491597493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,2,balanced,1.5058026313781738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,8,power_law_1.01,0.05050879716873169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,8,power_law_1.01,0.056390398740768434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,8,power_law_1.01,0.06048640012741089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,8,power_law_1.01,0.06439039707183838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,8,power_law_1.01,0.07425280213356018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,8,power_law_1.01,0.06978560090065003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,8,power_law_1.01,0.07589120268821717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,8,power_law_1.01,0.07655680179595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,8,power_law_1.01,0.07601919770240784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,8,power_law_1.01,0.07701759934425353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,8,power_law_1.01,0.07825919985771179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,8,power_law_1.01,0.07986559867858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,8,power_law_1.01,0.08206719756126404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,8,power_law_1.01,0.08887680172920227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,8,power_law_1.01,0.09244800209999085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,8,power_law_1.01,0.10299520492553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,8,power_law_1.2,0.4569215774536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,8,power_law_1.01,0.10665600299835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,8,power_law_1.01,0.11605759859085082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,8,power_law_1.01,0.12752000093460084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,8,power_law_1.01,0.15175039768218995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,8,power_law_1.01,0.17537280321121215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,8,power_law_1.01,0.21590399742126465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,8,power_law_1.01,0.2675584077835083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,8,power_law_1.01,0.3572416067123413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,8,power_law_1.01,0.42062721252441404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,8,power_law_1.01,0.603116798400879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,8,power_law_1.01,0.7898303985595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,4,power_law_1.2,0.057920002937316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,4,power_law_1.2,0.06581760048866273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,4,power_law_1.2,0.0646336019039154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,4,power_law_1.2,0.07817599773406983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,4,power_law_1.2,0.07404800057411194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,4,power_law_1.2,0.07191680073738098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,4,power_law_1.2,0.07634559869766236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,4,power_law_1.2,0.072326397895813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,4,power_law_1.2,0.07489280104637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,4,power_law_1.2,0.07464960217475891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,4,power_law_1.2,0.07116159796714783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,4,power_law_1.2,0.07936639785766601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,4,power_law_1.2,0.07767680287361145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,4,power_law_1.2,0.07875840067863464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,4,power_law_1.2,0.09316480159759521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,4,power_law_1.2,0.09312000274658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,4,power_law_1.2,0.10852479934692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,4,power_law_1.2,0.12148480415344239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,4,power_law_1.2,0.14847359657287598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,4,power_law_1.2,0.165011203289032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,4,power_law_1.2,0.21989760398864747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,8,power_law_1.2,0.6099775791168213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,4,power_law_1.2,0.3043071985244751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,4,power_law_1.2,0.3443007946014404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,4,power_law_1.2,0.46769280433654786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,8,power_law_1.2,0.9733440399169921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.2,0.39390718936920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,8,balanced,0.042634665966033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,8,balanced,0.042026668787002563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,8,balanced,0.042037333051363625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,8,balanced,0.053472002347310386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,8,balanced,0.06877866884072621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,8,balanced,0.06826133529345195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,8,balanced,0.06963733335336049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,8,balanced,0.07150400181611379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,8,balanced,0.0703306645154953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,8,balanced,0.07047999898592631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,8,balanced,0.07008000214894612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,8,balanced,0.07310399909814198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,8,balanced,0.07283733288447063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,8,balanced,0.07284266750017802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,8,balanced,0.08065066734949748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,8,balanced,0.08163733283678691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,8,balanced,0.08200533191363017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,4,power_law_1.2,0.6100736141204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,8,balanced,0.08808533350626628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,8,balanced,0.09284266829490662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,8,balanced,0.10356266299883525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,8,balanced,0.11203733086585999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,8,balanced,0.14827199776967367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,8,balanced,0.16586666305859885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,8,balanced,0.2304640014966329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,8,balanced,0.29606399933497113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,8,balanced,0.417141318321228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,8,balanced,0.5429279804229736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,8,power_law_1.2,1.3178560256958007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,4,power_law_1.2,0.8926143646240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,4,power_law_1.2,1.16876163482666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,1,power_law_1.01,0.07336959838867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,1,power_law_1.01,0.10023679733276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,1,power_law_1.01,0.14752639532089235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,1,power_law_1.01,0.24944000244140624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,1,power_law_1.01,0.34796159267425536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,1,power_law_1.01,0.44977917671203616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,1,power_law_1.01,0.6279359817504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,1,power_law_1.01,0.6680831909179688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,1,power_law_1.01,0.6600128173828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,1,power_law_1.01,0.6897920131683349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,1,power_law_1.01,0.7127295970916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,1,power_law_1.01,0.7281343936920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,1,power_law_1.01,0.7582208156585694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,1,power_law_1.01,0.7695040225982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,1,power_law_1.01,0.7886144161224365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,1,power_law_1.01,0.7986176013946533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,1,power_law_1.01,0.8695103645324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,1,power_law_1.01,0.9615360260009765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,1,power_law_1.01,0.9677311897277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,16,power_law_1.01,0.05562880039215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,1,power_law_1.01,1.096895980834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,16,power_law_1.01,0.05925760269165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,1,power_law_1.01,1.190995216369629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,16,power_law_1.01,0.04999040067195892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,16,power_law_1.01,0.05304960012435913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,1,power_law_1.01,1.3816703796386718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,16,power_law_1.01,0.05375999808311462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,1,power_law_1.01,1.6102079391479491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,1,power_law_1.01,2.0204416275024415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,16,power_law_1.01,0.05541759729385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,16,power_law_1.01,0.05792639851570129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,16,power_law_1.01,0.058508801460266116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,1,power_law_1.01,2.4583744049072265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,16,power_law_1.01,0.05966079831123352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,16,power_law_1.01,0.05914239883422852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,16,power_law_1.01,0.06160640120506287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,16,power_law_1.01,0.06099839806556702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,64,power_law_1.2,0.040268799662590025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,16,power_law_1.01,0.06317440271377564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,64,power_law_1.2,0.040608000755310056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,64,power_law_1.2,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,64,power_law_1.2,0.04213759899139404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,64,power_law_1.2,0.04286080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,64,power_law_1.2,0.044460800290107724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,1,power_law_1.01,3.2791744232177735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,16,power_law_1.01,0.06753280162811279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,16,power_law_1.01,0.07182719707489013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,16,power_law_1.01,0.07514240145683289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,16,power_law_1.01,0.08392959833145142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,1,power_law_1.01,4.114483261108399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,16,power_law_1.01,0.09423360228538513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,64,power_law_1.2,0.04508799910545349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,16,power_law_1.01,0.09719039797782898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,64,power_law_1.2,0.04618239998817444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,16,power_law_1.01,0.11526399850845337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,16,power_law_1.01,0.13420159816741944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,16,power_law_1.01,0.16647039651870726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,16,power_law_1.01,0.20574719905853273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,16,power_law_1.01,0.2831360101699829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,16,power_law_1.01,0.32835841178894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,16,power_law_1.01,0.5314047813415528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,64,power_law_1.2,0.04789760112762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,16,power_law_1.01,0.6411712169647217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,64,power_law_1.2,0.04764800071716309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,64,power_law_1.2,0.04870400130748749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,64,balanced,0.05017599960168203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,64,power_law_1.2,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,64,power_law_1.2,0.049184000492095946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,64,balanced,0.05187733471393585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,64,balanced,0.05376000205675761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,64,balanced,0.06007466713587443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,64,balanced,0.06030400097370148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,64,balanced,0.0603413333495458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,64,balanced,0.06226666768391927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,64,balanced,0.06239999830722809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,64,balanced,0.06229333579540253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,64,power_law_1.2,0.05189759731292724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,64,balanced,0.06410666803518932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,64,balanced,0.06402133405208588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,64,balanced,0.066170667608579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,64,balanced,0.0663679987192154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,64,balanced,0.06644266843795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,64,balanced,0.07229866584142049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,64,balanced,0.07261333366235097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,64,balanced,0.07726400097211202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,64,balanced,0.0869706670443217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,64,balanced,0.09500267108281453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,64,balanced,0.1258026659488678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,64,balanced,0.13777066270510355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,64,balanced,0.1920213301976522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,64,balanced,0.18247467279434204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,64,power_law_1.2,0.05681279897689819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,64,balanced,0.23731732368469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,64,balanced,0.26239466667175293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,64,balanced,0.34946131706237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,64,balanced,0.42610132694244385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,64,power_law_1.2,0.056966400146484374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,64,power_law_1.2,0.06301440000534057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,64,power_law_1.2,0.06869120001792908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,64,power_law_1.2,0.07743359804153442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,64,power_law_1.2,0.09668480157852173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,64,power_law_1.2,0.1112064003944397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,64,power_law_1.2,0.14821759462356568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,64,power_law_1.2,0.18064639568328858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,64,power_law_1.2,0.2538559913635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,64,power_law_1.2,0.31064319610595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,64,power_law_1.2,0.5012288093566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,64,power_law_1.2,0.6817984104156494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,64,power_law_1.2,0.04943360090255737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,64,power_law_1.2,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,64,power_law_1.2,0.03080959916114807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,64,power_law_1.2,0.03377920091152191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,64,power_law_1.2,0.03115519881248474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,64,power_law_1.2,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,64,power_law_1.2,0.03272959887981415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,64,power_law_1.2,0.03264000117778778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,64,power_law_1.2,0.032864001393318173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,64,power_law_1.2,0.032627201080322264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,64,power_law_1.2,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,64,power_law_1.2,0.035897600650787356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,64,power_law_1.2,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,64,power_law_1.2,0.0400191992521286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,64,power_law_1.2,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,64,power_law_1.2,0.04537599980831146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,64,power_law_1.2,0.046163201332092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,64,power_law_1.2,0.05235199928283692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,64,power_law_1.2,0.05606399774551392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,64,power_law_1.2,0.06408960223197938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,64,power_law_1.2,0.06789119839668274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,64,power_law_1.2,0.08556159734725952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,64,power_law_1.2,0.11370240449905396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,64,power_law_1.2,0.13998080492019654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,64,power_law_1.2,0.18951679468154908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,8,balanced,0.0239680012067159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,8,balanced,0.02457600086927414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,8,balanced,0.027093333502610523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,8,balanced,0.03245333333810171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,8,balanced,0.04821866750717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,8,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,8,balanced,0.0624533345301946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,8,balanced,0.060218666990598045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,8,balanced,0.09004799524943034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,8,balanced,0.08506133159001668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,8,balanced,0.07750399907430013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,8,balanced,0.09822932879130046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,8,balanced,0.10038933157920837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,8,balanced,0.10014399886131287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,8,balanced,0.11432000001271565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,8,balanced,0.11318932970364888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,8,balanced,0.107424000898997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,8,balanced,0.11396267016728719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,8,balanced,0.13109333316485086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,8,balanced,0.14431466658910116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,8,balanced,0.16522133350372314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,8,balanced,0.19348265727361044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,8,balanced,0.21961599588394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,8,balanced,0.33987200260162354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,8,balanced,0.39825598398844403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,64,power_law_1.2,0.2632960081100464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,8,balanced,0.5772320032119751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,8,balanced,0.7590719858805338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,64,power_law_1.2,0.2916287899017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,32,power_law_1.01,0.06120960116386413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,32,power_law_1.01,0.0667136013507843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,32,power_law_1.01,0.06450560092926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,32,power_law_1.01,0.06842240095138549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,32,power_law_1.01,0.06843519806861878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,32,power_law_1.01,0.06888960003852844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,32,power_law_1.01,0.06835200190544129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,32,power_law_1.01,0.06945279836654664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,32,power_law_1.01,0.07188479900360108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,32,power_law_1.01,0.07085440158843995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,32,power_law_1.01,0.07041919827461243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,32,power_law_1.01,0.0754751980304718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,32,power_law_1.01,0.07336320281028748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,32,power_law_1.01,0.07996159791946411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,32,power_law_1.01,0.08680319786071777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,32,power_law_1.01,0.08908159732818603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,32,power_law_1.01,0.09683840274810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,32,power_law_1.01,0.1130560040473938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,32,power_law_1.01,0.11336959600448608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,32,power_law_1.01,0.14120320081710816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,32,power_law_1.01,0.16273280382156372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,32,power_law_1.01,0.20517759323120116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,32,power_law_1.01,0.25362560749053953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,32,power_law_1.01,0.33931519985198977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,32,power_law_1.01,0.4289087772369385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,32,power_law_1.01,0.644761610031128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,32,power_law_1.01,0.8643327713012695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,32,balanced,0.04578666885693868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,32,balanced,0.0481279989083608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,32,balanced,0.049786667029062905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,32,balanced,0.07150933146476746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,32,balanced,0.08208000163237254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,32,balanced,0.08392533659934998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,32,balanced,0.08476266264915466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,32,balanced,0.0825386643409729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,32,balanced,0.08482666810353597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,32,balanced,0.08342933654785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,32,balanced,0.084389328956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,32,balanced,0.08784533540407817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,32,balanced,0.08694932858149211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,32,balanced,0.08793600400288899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,32,balanced,0.09305066863695781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,32,balanced,0.09409600496292114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,32,balanced,0.09918399651845296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,32,balanced,0.11133333047231038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,32,balanced,0.11732266346613567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,32,balanced,0.14267733693122864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,32,balanced,0.16089066863059998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,32,balanced,0.21554134289423624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,32,balanced,0.191103994846344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,32,balanced,0.2504853407541911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,32,balanced,0.2795253396034241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,32,balanced,0.37942934036254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,32,balanced,0.44208534558614093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,32,4,power_law_1.2,0.017555199563503265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,32,4,power_law_1.2,0.017497600615024568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,32,4,power_law_1.2,0.01774719953536987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,32,4,power_law_1.2,0.017849600315093993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,32,4,power_law_1.2,0.018572799861431122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,32,4,power_law_1.2,0.018668800592422485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,32,4,power_law_1.2,0.02077440023422241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,32,4,power_law_1.2,0.020147199928760528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,32,4,power_law_1.2,0.03824000060558319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,32,4,power_law_1.2,0.0377023994922638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,32,4,power_law_1.2,0.03785600066184998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,32,4,power_law_1.2,0.030054399371147157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,32,4,power_law_1.2,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,32,4,power_law_1.2,0.030022400617599487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,2,power_law_1.01,0.09099519848823548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,32,4,power_law_1.2,0.030675199627876282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,2,power_law_1.01,0.11845120191574096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,2,power_law_1.01,0.13441280126571656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,32,4,power_law_1.2,0.03147520124912262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,2,power_law_1.01,0.17750400304794312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,2,power_law_1.01,0.2197632074356079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,32,4,power_law_1.2,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,2,power_law_1.01,0.2461632013320923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,16,power_law_1.01,0.0204927995800972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,16,power_law_1.01,0.019833600521087645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,2,power_law_1.01,0.3040319919586182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,16,power_law_1.01,0.018035200238227845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,16,power_law_1.01,0.018329599499702455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,16,power_law_1.01,0.018956799805164338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,16,power_law_1.01,0.019443200528621675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,16,power_law_1.01,0.02067199945449829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,16,power_law_1.01,0.020550400018692017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,16,power_law_1.01,0.02067199945449829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,16,power_law_1.01,0.021222400665283202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,16,power_law_1.01,0.023852799832820893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,16,power_law_1.01,0.050783997774124144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,16,power_law_1.01,0.024639999866485594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,2,power_law_1.01,0.30451838970184325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,16,power_law_1.01,0.04170880019664765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,16,power_law_1.01,0.03740800023078918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,2,power_law_1.01,0.30441598892211913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,16,power_law_1.01,0.03993600010871887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,2,power_law_1.01,0.3109888076782227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,16,power_law_1.01,0.03738879859447479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,16,power_law_1.01,0.04069119989871979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,2,power_law_1.01,0.33679358959197997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,16,power_law_1.01,0.03989759981632233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,2,power_law_1.01,0.341919994354248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,32,4,power_law_1.2,0.03738879859447479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,16,power_law_1.01,0.02513279914855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,16,power_law_1.01,0.03766399919986725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,16,power_law_1.01,0.03743360042572021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,16,power_law_1.01,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,16,power_law_1.01,0.038873600959777835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,32,4,power_law_1.2,0.04099839925765991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,16,power_law_1.01,0.041043201088905336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,16,power_law_1.01,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,2,power_law_1.01,0.3319360017776489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,16,power_law_1.01,0.05123839974403381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,2,power_law_1.01,0.34488959312438966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,2,power_law_1.01,0.3375808000564575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,2,power_law_1.01,0.3573184013366699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,2,power_law_1.01,0.3668544054031372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,16,power_law_1.01,0.04101119935512543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,2,power_law_1.01,0.38401920795440675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,16,power_law_1.01,0.04132480025291443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.01,0.42517762184143065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,16,power_law_1.01,0.0403328001499176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,16,power_law_1.01,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,16,power_law_1.01,0.04126720130443573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.01,0.48131837844848635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,16,power_law_1.01,0.07230079770088196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,16,power_law_1.01,0.0426367998123169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,16,power_law_1.01,0.08598399758338929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,16,power_law_1.01,0.04323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,16,power_law_1.01,0.11532800197601319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,16,power_law_1.01,0.04261760115623474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,16,power_law_1.01,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,16,power_law_1.01,0.05064319968223572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,16,power_law_1.01,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,32,4,power_law_1.2,0.04686079919338226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,16,power_law_1.01,0.05727360248565674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,16,power_law_1.01,0.06640639901161194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,16,power_law_1.01,0.07772799730300903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.01,0.5619711875915527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,16,power_law_1.01,0.08079360127449035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,16,power_law_1.01,0.10209920406341552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.01,0.6816895961761474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,16,power_law_1.01,0.1175104022026062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.01,0.8138367652893066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.01,1.077894401550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,16,power_law_1.01,0.14046080112457277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.01,1.3291263580322266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,16,power_law_1.01,0.1999168038368225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,16,power_law_1.01,0.15529600381851197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,16,power_law_1.01,0.24953598976135255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.01,1.8205055236816405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,32,4,power_law_1.2,0.05007359981536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.01,2.4148544311523437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,16,power_law_1.01,0.18792959451675414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,16,power_law_1.01,0.28780159950256345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,16,power_law_1.01,0.3246016025543213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,32,4,power_law_1.2,0.06977919936180114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,32,4,power_law_1.2,0.07948160171508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,32,4,power_law_1.2,0.10718079805374145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,32,4,power_law_1.2,0.13386240005493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,32,power_law_1.2,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,32,power_law_1.2,0.05807999968528747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,32,power_law_1.2,0.05800319910049438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,32,power_law_1.2,0.060819202661514284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,32,power_law_1.2,0.06074879765510559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,32,power_law_1.2,0.061689597368240354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,32,power_law_1.2,0.06154239773750305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,32,power_law_1.2,0.064300799369812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,32,power_law_1.2,0.0633408010005951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,32,power_law_1.2,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,32,power_law_1.2,0.06514559984207154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,32,power_law_1.2,0.06816639900207519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,32,power_law_1.2,0.06761599779129028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,32,power_law_1.2,0.0692799985408783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,32,power_law_1.2,0.07637119889259339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,32,power_law_1.2,0.07838079929351807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,32,power_law_1.2,0.08749439716339111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,32,power_law_1.2,0.09976959824562073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,32,power_law_1.2,0.10828800201416015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,32,power_law_1.2,0.1352895975112915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,32,power_law_1.2,0.15889919996261598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,32,power_law_1.2,0.20274560451507567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,32,power_law_1.2,0.23700480461120604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,32,power_law_1.2,0.3491647958755493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,32,power_law_1.2,0.4127232074737549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,32,power_law_1.2,0.5811840057373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,32,power_law_1.2,0.7677440166473388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,8,power_law_1.2,0.07518079876899719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,8,power_law_1.2,0.0960640013217926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,8,power_law_1.2,0.10289920568466186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,8,power_law_1.2,0.10426239967346192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,8,power_law_1.2,0.12826240062713623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,8,power_law_1.2,0.11932159662246704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,8,power_law_1.2,0.13570560216903688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,8,power_law_1.2,0.13649280071258546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,8,power_law_1.2,0.13456000089645387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,8,power_law_1.2,0.13151999711990356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,8,power_law_1.2,0.13363840579986572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,8,power_law_1.2,0.13867520093917846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,8,power_law_1.2,0.13950719833374023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,8,power_law_1.2,0.14165760278701783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,8,power_law_1.2,0.14746240377426148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,32,4,power_law_1.2,0.18952959775924683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,8,power_law_1.2,0.15130239725112915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,8,power_law_1.2,0.15556479692459108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,8,power_law_1.2,0.16328959465026854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,8,power_law_1.2,0.18316800594329835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,16,power_law_1.01,0.06550400257110596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,8,power_law_1.2,0.20613760948181153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,16,power_law_1.01,0.09537919759750366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,8,power_law_1.2,0.23939199447631837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,16,power_law_1.01,0.11902719736099243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,8,power_law_1.2,0.30266880989074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,16,power_law_1.01,0.11852799654006958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,8,power_law_1.2,0.3382719993591309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,8,power_law_1.2,0.4843135833740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,8,power_law_1.2,0.632044792175293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,8,power_law_1.2,1.000704002380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,16,power_law_1.01,0.12040319442749023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,16,power_law_1.01,0.1261952042579651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,16,power_law_1.01,0.09957119822502136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,16,power_law_1.01,0.11630079746246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,16,power_law_1.01,0.11572480201721191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,16,power_law_1.01,0.10603519678115844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,8,power_law_1.2,1.220639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,16,power_law_1.01,0.11985280513763427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,16,power_law_1.01,0.1118783950805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,16,power_law_1.01,0.12373759746551513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,16,power_law_1.01,0.1256832003593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,16,power_law_1.01,0.13077759742736816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,16,power_law_1.01,0.13514879941940308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,16,power_law_1.01,0.1356927990913391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,16,power_law_1.01,0.1541759967803955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,16,power_law_1.01,0.16737279891967774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,16,power_law_1.01,0.19271680116653442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,16,power_law_1.01,0.20031359195709228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,16,power_law_1.01,0.27280640602111816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,16,power_law_1.01,0.29919359683990476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,16,power_law_1.01,0.4017792224884033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,16,power_law_1.01,0.4887551784515381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,16,power_law_1.01,0.67674241065979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,16,power_law_1.01,0.8085760116577149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,32,4,power_law_1.2,0.23582720756530762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,64,power_law_1.2,0.14513920545578002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,64,power_law_1.2,0.048640000820159915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,64,power_law_1.2,0.044121599197387694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,64,power_law_1.2,0.04643200039863586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,64,power_law_1.2,0.04655359983444214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,64,power_law_1.2,0.047737601399421695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,64,power_law_1.2,0.04892799854278564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,64,power_law_1.2,0.04947839975357056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,64,power_law_1.2,0.05086719989776611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,64,power_law_1.2,0.05017600059509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,64,power_law_1.2,0.05172479748725891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,64,power_law_1.2,0.05382400155067444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,64,power_law_1.2,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,64,power_law_1.2,0.05598719716072083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,64,power_law_1.2,0.06320000290870667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,64,power_law_1.2,0.06367999911308289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,64,power_law_1.2,0.06828160285949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,64,power_law_1.2,0.08013439774513245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,64,power_law_1.2,0.08996480107307434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,64,power_law_1.2,0.11463040113449097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,64,power_law_1.2,0.14447360038757323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,64,power_law_1.2,0.18008320331573485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,64,power_law_1.2,0.2267008066177368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,64,power_law_1.2,0.3177023887634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,64,power_law_1.2,0.38645119667053224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,64,power_law_1.2,0.5559679985046386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,64,power_law_1.2,0.7520832061767578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,1,balanced,0.05429866909980774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,1,balanced,0.07525866727034251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,1,balanced,0.11814399560292561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,1,balanced,0.12008000413576762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,1,balanced,0.12210667133331299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,1,balanced,0.12268267075220744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,1,balanced,0.12384532888730367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,1,balanced,0.12550933162371317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,1,balanced,0.1271519958972931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,1,balanced,0.12769599755605063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,1,balanced,0.13020267089207968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,1,balanced,0.1313866674900055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,1,balanced,0.13387200236320496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,1,balanced,0.13988799850145975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,1,balanced,0.1577173372109731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,1,balanced,0.16014400124549866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,1,balanced,0.15873066584269205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,1,balanced,0.19262933731079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,1,balanced,0.20122667153676352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,1,balanced,0.262501339117686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,1,balanced,0.29291200637817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,32,power_law_1.01,0.0599232017993927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,32,power_law_1.01,0.0554751992225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,1,balanced,0.4094773530960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,32,power_law_1.01,0.05497599840164184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,32,power_law_1.01,0.05782399773597717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,1,balanced,0.5113866726557413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,1,balanced,0.7309973239898682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,32,power_law_1.01,0.06034560203552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,1,balanced,0.9394559860229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,32,power_law_1.01,0.06287999749183655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,1,balanced,1.3675467173258464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,32,power_law_1.01,0.06349440217018128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,1,balanced,1.7968319257100422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,32,power_law_1.01,0.06399359703063964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,32,power_law_1.01,0.06467199921607972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,32,power_law_1.01,0.0636031985282898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,32,power_law_1.01,0.06375679969787598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,32,power_law_1.01,0.0723583996295929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,32,power_law_1.01,0.07184640169143677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,32,power_law_1.01,0.07393280267715455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,32,power_law_1.01,0.08596479892730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,32,power_law_1.01,0.08811519742012024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,32,power_law_1.01,0.0867904007434845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,32,power_law_1.01,0.10257279872894287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,32,power_law_1.01,0.11040639877319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,32,power_law_1.01,0.13036160469055175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,32,power_law_1.01,0.14540159702301025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,32,power_law_1.01,0.197817599773407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,32,power_law_1.01,0.22206079959869385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,32,power_law_1.01,0.33983359336853025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,32,power_law_1.01,0.4411647796630859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,32,power_law_1.01,0.6372159957885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,32,power_law_1.01,0.7750912189483643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,64,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,64,power_law_1.01,0.03452160060405731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,64,power_law_1.01,0.03390080034732819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,64,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,64,power_law_1.01,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,64,power_law_1.01,0.03685120046138764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,64,power_law_1.01,0.03664000034332275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,64,power_law_1.01,0.036883199214935304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,64,power_law_1.01,0.037241598963737486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,64,power_law_1.01,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,64,power_law_1.01,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,64,power_law_1.01,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,64,power_law_1.01,0.03827199935913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,64,power_law_1.01,0.03856000006198883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,64,power_law_1.01,0.041433599591255185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,64,power_law_1.01,0.04223999977111816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,64,power_law_1.01,0.0434688001871109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,64,power_law_1.01,0.046828800439834596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,64,power_law_1.01,0.051481598615646364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,64,power_law_1.01,0.05678719878196716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,64,power_law_1.01,0.06109439730644226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,64,power_law_1.01,0.07265920042991639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,64,power_law_1.01,0.0847487986087799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,64,power_law_1.01,0.11096320152282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,64,power_law_1.01,0.1350335955619812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,64,power_law_1.01,0.17163519859313964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,64,power_law_1.01,0.23418879508972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,16,power_law_1.2,0.07333120107650756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,16,power_law_1.2,0.05240960121154785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,16,power_law_1.2,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,16,power_law_1.2,0.0547327995300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,16,power_law_1.2,0.04386560022830963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,16,power_law_1.2,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,16,power_law_1.2,0.05447040200233459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,16,power_law_1.2,0.05578879714012146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,16,power_law_1.2,0.053286397457122804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,16,power_law_1.2,0.055308800935745236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,16,power_law_1.2,0.05489919781684875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,16,power_law_1.2,0.057171201705932616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,16,power_law_1.2,0.05767679810523987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,16,power_law_1.2,0.06020479798316956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,16,power_law_1.2,0.06601600050926208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,16,power_law_1.2,0.06593279838562012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,16,power_law_1.2,0.06779519915580749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,16,power_law_1.2,0.07491199970245362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,16,power_law_1.2,0.08218880295753479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,16,power_law_1.2,0.08681600093841553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,16,power_law_1.2,0.09512959718704224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,16,power_law_1.2,0.12272640466690063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,16,power_law_1.2,0.13930879831314086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,16,power_law_1.2,0.17244800329208373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,16,power_law_1.2,0.2479423999786377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,8,power_law_1.2,0.04012160003185272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,8,power_law_1.2,0.03770880103111267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,8,power_law_1.2,0.03903999924659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,8,power_law_1.2,0.042316800355911253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,8,power_law_1.2,0.04137600064277649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,8,power_law_1.2,0.04447360038757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,16,power_law_1.2,0.30522880554199217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,8,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,8,power_law_1.2,0.045510399341583255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,8,power_law_1.2,0.045484799146652224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,8,power_law_1.2,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,8,power_law_1.2,0.04744960069656372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,8,power_law_1.2,0.048076799511909483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,8,power_law_1.2,0.04828799962997436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,8,power_law_1.2,0.04926080107688904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,8,power_law_1.2,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,16,power_law_1.2,0.3570240020751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,8,power_law_1.2,0.056006401777267456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,8,power_law_1.2,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,8,power_law_1.2,0.06544640064239501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,8,power_law_1.2,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,8,power_law_1.2,0.07837439775466919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,8,power_law_1.2,0.09572479724884034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,8,power_law_1.2,0.12459520101547242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,8,power_law_1.2,0.15504640340805054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,8,power_law_1.2,0.19704960584640502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,8,power_law_1.2,0.2634687900543213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,1,balanced,0.07037333150704701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,1,balanced,0.09816533327102661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,1,balanced,0.15158933401107788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,1,balanced,0.2593440016110738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,1,balanced,0.47315200169881183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,1,balanced,0.5799253384272257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,1,balanced,0.5831413269042969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,1,balanced,0.5889226595560709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,1,balanced,0.5899626811345419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,1,balanced,0.5941119988759359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,1,balanced,0.5988373359044393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,1,balanced,0.6039359966913859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,1,balanced,0.6047999858856201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,1,balanced,0.6175573269526163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,1,balanced,0.625706672668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,1,balanced,0.6365493138631185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,1,balanced,0.6509493192036947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,1,balanced,0.7416266600290934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,1,balanced,0.7164106369018555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,1,balanced,0.8767840067545573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,1,balanced,0.856175978978475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,1,balanced,1.1013387044270833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,1,balanced,1.1523040135701497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,1,balanced,1.6955466270446777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,1,balanced,1.9170400301615398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,1,balanced,3.156794548034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,1,balanced,3.57692813873291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,8,power_law_1.2,0.054764801263809205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,8,power_law_1.2,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,8,power_law_1.2,0.054662400484085084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,8,power_law_1.2,0.05912320017814636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,8,power_law_1.2,0.058905601501464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,8,power_law_1.2,0.0597055971622467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,8,power_law_1.2,0.06577919721603394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,8,power_law_1.2,0.06598399877548218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,8,power_law_1.2,0.06683520078659058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,8,power_law_1.2,0.06922879815101624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,8,power_law_1.2,0.06924160122871399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,8,power_law_1.2,0.07544959783554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,8,power_law_1.2,0.0749888002872467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,8,power_law_1.2,0.0804032027721405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,8,power_law_1.2,0.09535359740257263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,8,power_law_1.2,0.09533439874649048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,8,power_law_1.2,0.10843520164489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,8,power_law_1.2,0.1438912034034729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,8,power_law_1.2,0.16615040302276612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,8,power_law_1.2,0.22347519397735596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,8,power_law_1.2,0.2643455982208252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,8,power_law_1.2,0.39606399536132814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,8,power_law_1.2,0.36636159420013426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,8,power_law_1.2,0.49907841682434084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,8,power_law_1.2,0.5003967761993409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,8,power_law_1.2,0.754585599899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,8,power_law_1.2,0.8840448379516601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,2,balanced,0.031189332405726116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,2,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,2,balanced,0.05821333328882853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,2,balanced,0.08431999882062276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,2,balanced,0.1418773333231608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,2,balanced,0.24822400013605753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,2,balanced,0.2528533339500427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,2,balanced,0.26017600297927856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,2,balanced,0.26765867074330646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,2,balanced,0.2752373417218526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,2,balanced,0.29021867116292316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,2,balanced,0.2731893261273702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,2,balanced,0.2807040015856425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,2,balanced,0.2962613304456075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,2,balanced,0.30007465680440265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,2,balanced,0.3025386730829875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,2,balanced,0.31519466638565063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,2,balanced,0.4005599816640218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,2,balanced,0.39002132415771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,2,balanced,0.7314826647440592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,2,balanced,0.8024213314056396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,2,balanced,0.642735997835795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,2,balanced,0.6784959634145101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,2,balanced,0.7471520105997721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,2,balanced,0.8318400382995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,2,balanced,1.3972585995992024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,2,balanced,1.530687967936198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,1,balanced,0.03622400015592575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,1,balanced,0.038015998899936676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,1,balanced,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,1,balanced,0.05649066468079885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,1,balanced,0.084906667470932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,1,balanced,0.08515200018882751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,1,balanced,0.08705066641171773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,1,balanced,0.08902399738629659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,1,balanced,0.09035199880599976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,1,balanced,0.0928000013033549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,1,balanced,0.09515200058619182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,1,balanced,0.0979253351688385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,1,balanced,0.10244266192118327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,1,balanced,0.10839466253916423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,1,balanced,0.11500266194343567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,1,balanced,0.1222879985968272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,1,balanced,0.13030933340390524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,1,balanced,0.15453333655993143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,1,balanced,0.18609599272410074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,1,balanced,0.23562133312225342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,1,balanced,0.2786506613095601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,1,balanced,0.4135040044784546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,1,balanced,0.5104373296101888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,1,balanced,0.7497279644012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,1,balanced,0.977733294169108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,1,balanced,1.4473120371500652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,1,balanced,1.8753172556559246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,8,power_law_1.2,1.2711039543151856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,16,power_law_1.2,0.021036800742149354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,16,power_law_1.2,0.022431999444961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,16,power_law_1.2,0.022144000232219695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,16,power_law_1.2,0.022278399765491487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,16,power_law_1.2,0.022784000635147093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,16,power_law_1.2,0.02308479994535446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,16,power_law_1.2,0.0255295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,16,power_law_1.2,0.025868800282478333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,16,power_law_1.2,0.03164800107479095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,16,power_law_1.2,0.031200000643730165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,16,power_law_1.2,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,16,power_law_1.2,0.052716797590255736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,16,power_law_1.2,0.05363199710845947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,16,power_law_1.2,0.05334399938583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,16,power_law_1.2,0.05244799852371216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,16,power_law_1.2,0.05571200251579285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,16,power_law_1.2,0.05761280059814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,16,power_law_1.2,0.05872640013694763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,16,power_law_1.2,0.06356480121612548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,16,power_law_1.2,0.07340160012245178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,16,power_law_1.2,0.08409600257873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,16,power_law_1.2,0.1029695987701416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,16,power_law_1.2,0.1284351944923401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,16,power_law_1.2,0.16151679754257203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,16,power_law_1.2,0.20341119766235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,16,power_law_1.2,0.30927999019622804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,16,power_law_1.2,0.3933248043060303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,8,power_law_1.2,1.736524772644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,1,power_law_1.01,0.05111039876937866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,1,power_law_1.01,0.058873599767684935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,1,power_law_1.01,0.0775168001651764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,1,power_law_1.01,0.10030080080032348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,1,power_law_1.01,0.12535040378570556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,1,power_law_1.01,0.1779263973236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,1,power_law_1.01,0.18372479677200318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,1,power_law_1.01,0.19256319999694824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,1,power_law_1.01,0.19820799827575683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,1,power_law_1.01,0.2015615940093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,1,power_law_1.01,0.21172480583190917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,1,power_law_1.01,0.2244800090789795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,1,power_law_1.01,0.22961280345916749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,1,power_law_1.01,0.2359231948852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,1,power_law_1.01,0.24810240268707276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,1,power_law_1.01,0.2554559946060181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,1,power_law_1.01,0.2779455900192261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,1,power_law_1.01,0.319648003578186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,1,power_law_1.01,0.36212480068206787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,1,power_law_1.01,0.4485184192657471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,1,power_law_1.01,0.5184639930725098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,1,power_law_1.01,0.6776127815246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,1,power_law_1.01,0.881107234954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,1,power_law_1.01,1.1786368370056153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,1,power_law_1.01,1.6134208679199218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,1,power_law_1.01,2.142207908630371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,1,power_law_1.01,2.629855918884277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,16,power_law_1.01,0.09137920141220093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,16,power_law_1.01,0.1352959990501404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,16,power_law_1.01,0.12510080337524415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,16,power_law_1.01,0.13109760284423827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,16,power_law_1.01,0.1330623984336853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,16,power_law_1.01,0.12462079524993896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,16,power_law_1.01,0.12597119808197021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,16,power_law_1.01,0.11813119649887086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,16,power_law_1.01,0.125273597240448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,16,power_law_1.01,0.13345279693603515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,16,power_law_1.01,0.12915199995040894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,16,power_law_1.01,0.1278656005859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,16,power_law_1.01,0.1323583960533142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,16,power_law_1.01,0.13344000577926635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,16,power_law_1.01,0.14971519708633424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,16,power_law_1.01,0.15249919891357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,16,power_law_1.01,0.1606528043746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,16,power_law_1.01,0.17548799514770508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,16,power_law_1.01,0.1883263945579529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,16,power_law_1.01,0.22620799541473388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,16,power_law_1.01,0.24757759571075438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,16,power_law_1.01,0.3148672103881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,16,power_law_1.01,0.35852799415588377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,16,power_law_1.01,0.5174143791198731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,16,power_law_1.01,0.6181248188018799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,16,power_law_1.01,0.8122367858886719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,16,power_law_1.01,1.118835163116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,8,balanced,0.057946667075157166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,8,balanced,0.06854933500289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,8,balanced,0.08536000053087871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,8,balanced,0.12469866871833801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,8,balanced,0.19081066052118936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,8,balanced,0.3008693257967631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,8,balanced,0.2964479923248291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,8,balanced,0.2865813374519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,8,balanced,0.27292799949645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,8,balanced,0.2719786763191223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,8,balanced,0.266213337580363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,8,balanced,0.27190933624903363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,8,balanced,0.2670186758041382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,8,balanced,0.26552534103393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,8,balanced,0.2690933346748352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,8,balanced,0.260368009408315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,8,balanced,0.2807040015856425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,8,balanced,0.2786506613095601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,8,balanced,0.2754613359769185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,8,balanced,0.3019040028254191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,8,balanced,0.3043733239173889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,8,balanced,0.3264906605084737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,8,balanced,0.3587626616160075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,8,balanced,0.590282678604126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,8,balanced,0.6089706818262736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,8,balanced,0.8420000076293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,8,balanced,1.0831519762674968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,32,balanced,0.056794668237368263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,32,balanced,0.04734399914741516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,32,balanced,0.04675200084845225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,32,balanced,0.05031999945640564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,32,balanced,0.07141866783301036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,32,balanced,0.09870933492978413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,32,balanced,0.09990933537483215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,32,balanced,0.09882666667302449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,32,balanced,0.09891200065612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,32,balanced,0.10133866469065349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,32,balanced,0.1009386678536733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,32,balanced,0.10223467151323955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,32,balanced,0.1032533347606659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,32,balanced,0.10469333330790202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,32,balanced,0.10739733775456746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,32,balanced,0.1102133293946584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,32,balanced,0.11189333597819011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,32,balanced,0.11773332953453064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,32,balanced,0.12130666772524516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,32,balanced,0.13408000270525613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,32,balanced,0.14422399799029031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,32,balanced,0.16540799538294473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,32,balanced,0.1878933310508728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,32,balanced,0.22931732734044394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,32,balanced,0.2715573310852051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,32,balanced,0.367029349009196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,32,balanced,0.45772798856099445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,16,power_law_1.2,0.02336000055074692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,16,power_law_1.2,0.024172799289226533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,16,power_law_1.2,0.025945600867271424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,16,power_law_1.2,0.02346239984035492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,16,power_law_1.2,0.024358400702476503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,16,power_law_1.2,0.03573119938373566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,16,power_law_1.2,0.046188798546791074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,16,power_law_1.2,0.04343039989471435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,16,power_law_1.2,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,4,balanced,0.020389333367347717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,4,balanced,0.019285333653291065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,4,balanced,0.019071999937295914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,4,balanced,0.020949333906173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,4,balanced,0.02475733309984207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,4,balanced,0.025125332176685333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,4,balanced,0.025770666698614757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,4,balanced,0.02677333354949951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,4,balanced,0.027429332335789997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,4,balanced,0.02714666724205017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,4,balanced,0.028773332635561626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,4,balanced,0.029546665648619335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,4,balanced,0.03128000100453695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,4,balanced,0.03143466760714849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,4,balanced,0.03196266790231069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,4,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,4,balanced,0.039477333426475525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,4,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,4,balanced,0.062261333068211876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,4,balanced,0.07875733574231465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,4,balanced,0.0783679982026418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,4,balanced,0.0844693382581075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,4,balanced,0.09364799658457439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,4,balanced,0.1437066694100698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,4,balanced,0.16077333688735962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,4,balanced,0.22714134057362875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,4,balanced,0.2929973403612773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,4,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,4,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,4,balanced,0.04805333415667216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,4,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,4,balanced,0.06222933530807495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,4,balanced,0.09437867005666097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,4,balanced,0.09485333164532979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,4,balanced,0.09782933195432027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,4,balanced,0.09914132952690125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,4,balanced,0.10052266716957092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,4,balanced,0.10126399993896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,4,balanced,0.10630399982134502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,4,balanced,0.10874666770299275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,4,balanced,0.10569066802660625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,4,balanced,0.11553600430488586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,4,balanced,0.11734400192896526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,4,balanced,0.12982933719952902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,4,balanced,0.14663466811180115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,4,balanced,0.1586079994837443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,4,balanced,0.18726933002471924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,4,balanced,0.21439999341964722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,4,balanced,0.2929653326670329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,4,balanced,0.34699734052022296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,4,balanced,0.4970133304595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,4,balanced,0.6200106541315714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,4,balanced,0.905407985051473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,4,balanced,1.1709386507670085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,32,power_law_1.01,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,32,power_law_1.01,0.056704002618789676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,32,power_law_1.01,0.05553920269012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,32,power_law_1.01,0.060652798414230345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,32,power_law_1.01,0.0582144021987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,32,power_law_1.01,0.05073919892311096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,32,power_law_1.01,0.05495679974555969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,32,power_law_1.01,0.05978239774703979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,32,power_law_1.01,0.059059202671051025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,32,power_law_1.01,0.06082559823989868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,32,power_law_1.01,0.0624064028263092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,32,power_law_1.01,0.06483839750289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,32,power_law_1.01,0.06647679805755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,32,power_law_1.01,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,32,power_law_1.01,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,32,power_law_1.01,0.07669119834899903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,32,power_law_1.01,0.07512320280075073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,32,power_law_1.01,0.08424320220947265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,32,power_law_1.01,0.08910719752311706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,32,power_law_1.01,0.10144000053405762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,32,power_law_1.01,0.11112320423126221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,32,power_law_1.01,0.13587199449539183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,32,power_law_1.01,0.15083520412445067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,32,power_law_1.01,0.1958847999572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,32,power_law_1.01,0.2656768083572388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,32,power_law_1.01,0.35288960933685304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,32,power_law_1.01,0.4615808010101318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,1,power_law_1.01,0.05724800229072571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,1,power_law_1.01,0.06197119951248169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,1,power_law_1.01,0.06778879761695862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,1,power_law_1.01,0.08673920035362244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,1,power_law_1.2,0.0983295977115631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,1,power_law_1.01,0.12113920450210572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,1,power_law_1.2,0.12300159931182861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,1,power_law_1.01,0.1455296039581299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,1,power_law_1.2,0.18725119829177855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,1,power_law_1.01,0.15928319692611695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,1,power_law_1.2,0.2841408014297485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,1,power_law_1.01,0.19102079868316652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,1,power_law_1.01,0.20027520656585693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,1,power_law_1.2,0.3695296049118042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,1,power_law_1.01,0.20689918994903564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,1,power_law_1.2,0.6451327800750732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,1,power_law_1.01,0.2131648063659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,1,power_law_1.2,0.70065279006958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,1,power_law_1.01,0.2282111883163452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,1,power_law_1.01,0.23601279258728028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,1,power_law_1.2,0.7265408039093018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,1,power_law_1.01,0.25786240100860597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,1,power_law_1.2,0.7538623809814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,1,power_law_1.01,0.27583999633789064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,1,power_law_1.01,0.2928639888763428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,1,power_law_1.2,0.7698624134063721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,1,power_law_1.01,0.33011839389801023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,1,power_law_1.2,0.7908800125122071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,1,power_law_1.01,0.40844159126281737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,1,power_law_1.2,0.826848030090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,1,power_law_1.2,0.8381440162658691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,1,power_law_1.2,0.8678784370422363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,1,power_law_1.2,0.8850815773010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,1,power_law_1.01,0.3947200059890747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,1,power_law_1.2,0.911616039276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,1,power_law_1.01,0.4982016086578369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,1,power_law_1.2,0.9856639862060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,1,power_law_1.2,1.1217087745666503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,1,power_law_1.2,1.1378368377685546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,1,power_law_1.01,0.5832640171051026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,1,power_law_1.01,0.7718336105346679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,1,power_law_1.01,0.9749759674072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,1,power_law_1.2,1.3667519569396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,1,power_law_1.01,1.3279423713684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,1,power_law_1.2,1.44967041015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,1,power_law_1.01,1.8217599868774415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,1,power_law_1.2,1.814124870300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,1,power_law_1.01,2.5098623275756835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,1,power_law_1.2,2.0183103561401365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,1,power_law_1.01,3.1069440841674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,1,power_law_1.2,2.605228805541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,1,power_law_1.2,3.4997886657714843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,1,power_law_1.2,4.721996688842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,1,power_law_1.2,5.678291320800781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,8,balanced,0.04562666515509287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,8,balanced,0.04796266555786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,8,balanced,0.07158400118350983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,8,balanced,0.10345600048700969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,8,balanced,0.1771999994913737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,8,balanced,0.18032532930374146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,8,balanced,0.18523732821146646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,8,balanced,0.186191995938619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,8,balanced,0.18849599361419678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,8,balanced,0.1858560045560201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,8,balanced,0.18595200777053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,8,balanced,0.18779200315475464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,8,balanced,0.1891040007273356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,8,balanced,0.1922880013783773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,8,balanced,0.19538132349650064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,8,balanced,0.1970613400141398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,8,balanced,0.19917333126068115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,8,balanced,0.21677333116531372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,8,balanced,0.21504533290863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,8,balanced,0.2495786746342977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,8,balanced,0.24374399582544962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,8,balanced,0.31360532840092975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,8,balanced,0.3161333401997884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,8,balanced,0.44096533457438153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,8,balanced,0.4802986780802409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,8,balanced,0.7410453160603842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,8,balanced,0.8255519866943359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,8,power_law_1.01,0.06719359755516052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,8,power_law_1.01,0.08862720131874084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,8,power_law_1.01,0.0821120023727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,8,power_law_1.01,0.0957759976387024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,8,power_law_1.01,0.10146559476852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,8,power_law_1.01,0.11033600568771362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,8,power_law_1.01,0.10801279544830322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,8,power_law_1.01,0.1078976035118103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,8,power_law_1.01,0.10784640312194824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,8,power_law_1.01,0.10925439596176148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,8,power_law_1.01,0.11233279705047608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,8,power_law_1.01,0.1131327986717224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,8,power_law_1.01,0.114956796169281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,8,power_law_1.01,0.11811200380325318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,8,power_law_1.01,0.12867200374603271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,8,power_law_1.01,0.13125760555267335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,8,power_law_1.01,0.13992960453033448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,8,power_law_1.01,0.16236799955368042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,8,power_law_1.01,0.17571840286254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,8,power_law_1.01,0.21878399848937988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,8,power_law_1.01,0.2652672052383423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,8,power_law_1.01,0.33133440017700194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,8,power_law_1.01,0.38093440532684325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,8,power_law_1.01,0.49864959716796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,8,power_law_1.01,0.5809343814849853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,8,power_law_1.01,0.8216192245483398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,8,power_law_1.01,1.0576640129089356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,64,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,64,balanced,0.051882664362589516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,64,balanced,0.053354665637016296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,64,balanced,0.05588266750176748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,64,balanced,0.05762666463851929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,64,balanced,0.06615466872851054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,64,balanced,0.06551466882228851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,64,balanced,0.06887466708819072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,64,balanced,0.06446933249632518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,64,balanced,0.06617600222428639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,64,balanced,0.07054933408896129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,64,balanced,0.06832000116507213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,64,balanced,0.07050133248170216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,64,balanced,0.06898666421572368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,64,balanced,0.07451733450094859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,64,balanced,0.07673066854476929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,64,balanced,0.08188266555468242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,64,balanced,0.09276266892751057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,64,balanced,0.1035146713256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,64,balanced,0.12130133310953777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,64,balanced,0.13583466410636902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,64,balanced,0.17286400000254312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,64,balanced,0.20459733406702676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,64,balanced,0.2722666660944621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,64,balanced,0.3364746570587158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,64,balanced,0.47489599386850995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,64,balanced,0.6033120155334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,16,balanced,0.04780800143877665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,16,balanced,0.04789866507053375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,16,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,16,balanced,0.05324266850948334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,16,balanced,0.0741066684325536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,16,balanced,0.10039466619491577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,16,balanced,0.10309867064158122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,16,balanced,0.10347732901573181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,16,balanced,0.10452799995740254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,16,balanced,0.10523200035095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,16,balanced,0.10669333736101787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,16,balanced,0.10788266857465108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,16,balanced,0.11019200086593628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,16,balanced,0.11195733149846394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,16,balanced,0.11549333731333415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,16,balanced,0.11949867010116577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,16,balanced,0.12450133760770161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,16,balanced,0.1337493360042572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,16,balanced,0.1444000005722046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,16,balanced,0.1641386648019155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,16,balanced,0.18904000520706177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,16,balanced,0.22849599520365396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,16,balanced,0.2638026674588521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,16,balanced,0.3559733231862386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,16,balanced,0.43459733327229816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,16,balanced,0.6029013395309448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,16,balanced,0.7601280212402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,2,power_law_1.01,0.04821119904518127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,2,power_law_1.01,0.05381119847297668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,2,power_law_1.01,0.06972799897193908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,2,power_law_1.01,0.08739839792251587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,2,power_law_1.01,0.10487040281295776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,2,power_law_1.01,0.12403839826583862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,2,power_law_1.01,0.14991999864578248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,2,power_law_1.01,0.15555200576782227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,2,power_law_1.01,0.15701760053634645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,2,power_law_1.01,0.16568319797515868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,2,power_law_1.01,0.16471680402755737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,2,power_law_1.01,0.1729472041130066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,2,power_law_1.01,0.17909760475158693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,2,power_law_1.01,0.19484800100326538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,2,power_law_1.01,0.20769920349121093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,2,power_law_1.01,0.21835520267486572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,2,power_law_1.01,0.22367360591888427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,2,power_law_1.01,0.25788800716400145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,2,power_law_1.01,0.2942528009414673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,2,power_law_1.01,0.34791040420532227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,2,power_law_1.01,0.3998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,2,power_law_1.01,0.5092864036560059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,2,power_law_1.01,0.6603519916534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,2,power_law_1.01,0.8606528282165528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,2,power_law_1.01,1.0888511657714843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,2,power_law_1.01,1.53253116607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,2,power_law_1.01,1.987321662902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,16,power_law_1.01,0.07071359753608704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,16,power_law_1.01,0.09926400184631348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,16,power_law_1.01,0.09359359741210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,16,power_law_1.01,0.09875199794769288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,16,power_law_1.01,0.09999359846115112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,16,power_law_1.01,0.08643839955329895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,16,power_law_1.01,0.09195520281791687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,16,power_law_1.01,0.09596160054206848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,16,power_law_1.01,0.10208640098571778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,16,power_law_1.01,0.09879680275917054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,16,power_law_1.01,0.10268160104751586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,1,power_law_1.2,0.0945472002029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,16,power_law_1.01,0.10205440521240235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,1,power_law_1.2,0.136518394947052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,16,power_law_1.01,0.10560640096664428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,16,power_law_1.01,0.10746239423751831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,16,power_law_1.01,0.11434240341186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,16,balanced,0.04607999821503957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,16,balanced,0.04734933376312256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,16,balanced,0.051498666405677795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,16,balanced,0.07247466842333476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,16,balanced,0.09664533535639445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,16,balanced,0.149536003669103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,16,balanced,0.1504906713962555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,16,balanced,0.15095466375350952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,16,balanced,0.15152532855669656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,16,balanced,0.15238933761914572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,16,balanced,0.15385066469510397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,1,power_law_1.2,0.21577599048614501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,16,balanced,0.15366933743158975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,16,balanced,0.15617600083351135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,16,balanced,0.1564906636873881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,1,power_law_1.2,0.34685440063476564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,16,power_law_1.01,0.11939200162887573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,16,power_law_1.01,0.12440320253372192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,16,power_law_1.01,0.14449280500411987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,16,power_law_1.01,0.14967679977416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,16,power_law_1.01,0.18245760202407837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,16,balanced,0.16014933586120605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,16,balanced,0.16103999813397726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,16,power_law_1.01,0.2049407958984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,1,power_law_1.2,0.5086016178131103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,16,power_law_1.01,0.26652801036834717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,1,power_law_1.2,0.6760704040527343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,16,power_law_1.01,0.3286976099014282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,1,power_law_1.2,1.0049728393554687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,16,balanced,0.16594133774439493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,16,balanced,0.1730133295059204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,16,balanced,0.17729065815607706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,16,balanced,0.1930346687634786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,16,balanced,0.2072640061378479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,16,balanced,0.23310933510462442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,16,balanced,0.25417067607243854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,16,balanced,0.324015994866689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,16,balanced,0.35491732756296795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,16,balanced,0.5143946806589762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,1,power_law_1.2,1.0757696151733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,16,power_law_1.01,0.4463359832763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,16,balanced,0.5951840082804362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,16,power_law_1.01,0.5400576114654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,1,power_law_1.2,1.1234111785888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,16,power_law_1.01,0.7640384197235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,1,power_law_1.2,1.1610495567321777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,16,power_law_1.01,1.0199808120727538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,1,power_law_1.2,1.1886655807495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,1,power_law_1.2,1.2193856239318848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,1,power_law_1.2,1.2562944412231445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,1,power_law_1.2,1.2931200027465821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,1,power_law_1.2,1.314848041534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,1,power_law_1.2,1.3691136360168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,1,power_law_1.2,1.463987159729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,1,power_law_1.2,1.6198848724365233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,1,power_law_1.2,1.6405824661254882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,1,power_law_1.2,1.8444864273071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,1,power_law_1.2,1.9018688201904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,1,power_law_1.2,2.2241920471191405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,1,power_law_1.2,2.4587263107299804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,1,power_law_1.2,3.0688127517700194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,1,power_law_1.2,3.573990249633789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,1,power_law_1.2,4.673318481445312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,1,power_law_1.2,5.718278503417968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,1,power_law_1.2,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,1,power_law_1.2,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,1,power_law_1.2,0.043756800889968875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,1,power_law_1.2,0.05672320127487183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,1,power_law_1.2,0.06925439834594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,1,power_law_1.2,0.0927616000175476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,1,power_law_1.2,0.09678080081939697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,1,power_law_1.2,0.10083839893341065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,1,power_law_1.2,0.1044927954673767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,1,power_law_1.2,0.10860799551010132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,1,power_law_1.2,0.11433600187301636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,1,power_law_1.2,0.11972479820251465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,1,power_law_1.2,0.125382399559021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,1,power_law_1.2,0.1255679965019226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,1,power_law_1.2,0.13601919412612914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,1,power_law_1.2,0.14414080381393432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,1,power_law_1.2,0.1566655993461609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,1,power_law_1.2,0.1934656023979187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,1,power_law_1.2,0.22019200325012206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,1,power_law_1.2,0.28437759876251223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,1,power_law_1.2,0.3426431894302368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,1,power_law_1.2,0.46286721229553224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,1,power_law_1.2,0.5886144161224365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,1,power_law_1.2,0.8233407974243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,1,power_law_1.2,1.0639360427856446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,1,power_law_1.2,1.5253503799438477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,16,balanced,0.04814933240413666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,16,balanced,0.04967466493447622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,16,balanced,0.049685334165891014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,16,balanced,0.054661333560943604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,16,balanced,0.05519466598828634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,16,balanced,0.07085866729418437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,16,balanced,0.07295999924341838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,16,balanced,0.07422400017579396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,16,balanced,0.07400000095367432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,16,balanced,0.07471466561158498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,16,balanced,0.07487466434637706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,16,balanced,0.07631466786066692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,16,balanced,0.07913599908351898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,16,balanced,0.07867200175921123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,16,balanced,0.0848479966322581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,16,balanced,0.08682133754094441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,16,balanced,0.09086933732032776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,16,balanced,0.0995253324508667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,16,balanced,0.1072266697883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,16,balanced,0.12164800365765889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,16,balanced,0.14638933539390564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,16,balanced,0.17293866475423178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,16,balanced,0.19441600640614828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,16,balanced,0.2589120070139567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,16,balanced,0.3163359959920247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,16,balanced,0.43429867426554364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,16,balanced,0.5540266831715902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,1,power_law_1.2,1.982700729370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,2,power_law_1.2,0.019283199310302736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,2,power_law_1.2,0.019046400487422944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,2,power_law_1.2,0.018643200397491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,2,power_law_1.2,0.018572799861431122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,2,power_law_1.2,0.01940480023622513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,2,power_law_1.2,0.020211200416088104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,2,power_law_1.2,0.025862398743629455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,2,power_law_1.2,0.024249599874019624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,2,power_law_1.2,0.02815999984741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,2,power_law_1.2,0.02675839960575104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,2,power_law_1.2,0.02807680070400238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,2,power_law_1.2,0.02956799864768982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,2,power_law_1.2,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,2,power_law_1.2,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,2,power_law_1.2,0.044921600818634035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,2,power_law_1.2,0.04232960045337677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,2,power_law_1.2,0.043884798884391785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,2,power_law_1.2,0.05740159749984741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.2,0.06506879925727845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.2,0.08570240139961242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.2,0.10700160264968872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.2,0.14433280229568482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,2,power_law_1.2,0.04997760057449341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,2,power_law_1.2,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,2,power_law_1.2,0.06026880145072937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,2,power_law_1.2,0.07313920259475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,2,power_law_1.2,0.09303680062294006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,2,power_law_1.2,0.11192959547042847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,2,power_law_1.2,0.12124799489974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,2,power_law_1.2,0.14294400215148925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,2,power_law_1.2,0.14320640563964843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,2,power_law_1.2,0.14632320404052734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,2,power_law_1.2,0.15173120498657228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,2,power_law_1.2,0.1548095941543579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,2,power_law_1.2,0.1687551975250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.2,0.17891839742660523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,2,power_law_1.2,0.16874239444732667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,2,power_law_1.2,0.1833024024963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,2,power_law_1.2,0.18528640270233154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,2,power_law_1.2,0.20247039794921876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,2,power_law_1.2,0.23064959049224854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,2,power_law_1.2,0.25960960388183596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,2,power_law_1.2,0.3187328100204468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,2,power_law_1.2,0.3638848066329956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,2,power_law_1.2,0.4682303905487061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,2,power_law_1.2,0.5914432048797608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,2,power_law_1.2,0.8271103858947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.2,0.2536384105682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,2,power_law_1.2,0.9884799957275391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,2,power_law_1.2,1.445849609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,2,power_law_1.2,1.858892822265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,8,power_law_1.01,0.03511680066585541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,8,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,8,power_law_1.01,0.03557760119438171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,8,power_law_1.01,0.03930880129337311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,8,power_law_1.01,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,8,power_law_1.01,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,8,power_law_1.01,0.04046719968318939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,8,power_law_1.01,0.041145598888397215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,8,power_law_1.01,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,8,power_law_1.01,0.04519039988517761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,8,power_law_1.01,0.04609920084476471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,8,power_law_1.01,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,8,power_law_1.01,0.05111680030822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,8,power_law_1.01,0.05016319751739502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,8,power_law_1.01,0.060063999891281125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,8,power_law_1.01,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,8,power_law_1.01,0.06238080263137817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,8,power_law_1.01,0.07663360238075256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,8,power_law_1.01,0.09273599982261657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,8,power_law_1.01,0.11336319446563721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,8,power_law_1.01,0.1255552053451538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.2,0.32451839447021485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,8,power_law_1.01,0.15760639905929566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,8,power_law_1.01,0.20992639064788818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,8,power_law_1.01,0.26559360027313234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,8,power_law_1.01,0.3619647979736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,2,balanced,0.044031997521718345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,2,balanced,0.048341333866119385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,2,balanced,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,2,balanced,0.08229333162307739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,2,balanced,0.11729066570599873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,2,balanced,0.13217600186665854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,2,balanced,0.1350879967212677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,8,power_law_1.01,0.4982016086578369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,8,power_law_1.01,0.6403071880340576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,2,balanced,0.13369599978129068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,2,balanced,0.13327999909718832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,8,balanced,0.04364799956480662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,2,balanced,0.13502933581670126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,8,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,8,balanced,0.04381333291530609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,8,balanced,0.050026665131251015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,8,balanced,0.06026133398214976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,8,balanced,0.06819200019041698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,8,balanced,0.0688213308652242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,8,balanced,0.07021866738796234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,8,balanced,0.06946666538715363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,8,balanced,0.07054933408896129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,8,balanced,0.07063999772071838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,8,balanced,0.07152000069618225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,8,balanced,0.07338666419188182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,8,balanced,0.07625600198904674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,8,balanced,0.07828799883524577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,8,balanced,0.08099199831485748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,8,balanced,0.08404800295829773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,8,balanced,0.09259733557701111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,2,balanced,0.1353653371334076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,2,balanced,0.1386666695276896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,2,balanced,0.1418453355630239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,2,balanced,0.1469066639741262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,2,balanced,0.15171200037002563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,2,balanced,0.15862933794657388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,2,balanced,0.16698133945465088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,2,balanced,0.18823466698328653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,2,balanced,0.20065067211786905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,2,balanced,0.26500799258550006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,2,balanced,0.2873013416926066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,2,balanced,0.4163680076599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,2,balanced,0.44463467597961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,8,balanced,0.09507733583450317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,8,balanced,0.12204266587893169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,8,balanced,0.13156267007191977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,8,balanced,0.17436800400416055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,8,balanced,0.18759999672571817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,8,balanced,0.26502933104832965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,8,balanced,0.3081173300743103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,8,balanced,0.43935998280843097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.2,0.47266559600830077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,8,balanced,0.5370239814122518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,2,balanced,0.6796693007151285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,2,balanced,0.7805226643880209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,2,balanced,1.2230453491210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,2,balanced,1.4481493631998699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.2,0.6195775985717773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,2,power_law_1.2,0.07284479737281799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,64,power_law_1.2,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,2,power_law_1.2,0.0788096010684967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,64,power_law_1.2,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,2,power_law_1.2,0.09004160165786743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,64,power_law_1.2,0.052108800411224364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,2,power_law_1.2,0.11361919641494751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,64,power_law_1.2,0.05425919890403748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,64,power_law_1.2,0.05057920217514038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,64,power_law_1.2,0.05319679975509643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,2,power_law_1.2,0.13575040102005004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,64,power_law_1.2,0.055731201171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,2,power_law_1.2,0.16071679592132568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,2,power_law_1.2,0.1795199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,64,power_law_1.2,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,64,power_law_1.2,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,2,power_law_1.2,0.20931200981140136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,64,power_law_1.2,0.05553920269012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,64,power_law_1.2,0.055878400802612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,64,power_law_1.2,0.05793920159339905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,64,power_law_1.2,0.05886719822883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,64,power_law_1.2,0.05792639851570129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,64,power_law_1.2,0.06272000074386597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,64,power_law_1.2,0.06369280219078063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,64,power_law_1.2,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,2,power_law_1.2,0.2056191921234131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,64,power_law_1.2,0.07584000229835511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,64,power_law_1.2,0.08085119724273682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,2,power_law_1.2,0.21504640579223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,64,power_law_1.2,0.10001280307769775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,2,power_law_1.2,0.21879680156707765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,2,power_law_1.2,0.22019200325012206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,2,power_law_1.2,0.2228991985321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,2,power_law_1.2,0.24216320514678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,64,power_law_1.2,0.10909440517425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,2,power_law_1.2,0.2517568111419678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,2,power_law_1.2,0.260479998588562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,2,power_law_1.2,0.2711040019989014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,2,power_law_1.2,0.29063680171966555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,2,power_law_1.2,0.32872960567474363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,64,power_law_1.2,0.1376960039138794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,64,power_law_1.2,0.16279040575027465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,64,power_law_1.2,0.2085952043533325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,64,power_law_1.2,0.2792639970779419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,64,power_law_1.2,0.41095681190490724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,64,power_law_1.2,0.5114880084991456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,2,power_law_1.2,0.3882816076278687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,2,power_law_1.2,0.4467328071594238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,2,power_law_1.2,0.5686528205871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,2,power_law_1.2,0.6915071964263916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,2,power_law_1.2,0.9242560386657714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,2,power_law_1.2,1.2146368026733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,2,power_law_1.2,1.7471744537353515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,2,power_law_1.2,2.1751424789428713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,2,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,2,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,2,balanced,0.04164800047874451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,2,balanced,0.04560000201066335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,2,balanced,0.047797332207361855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,2,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,2,balanced,0.04353600243727366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,2,balanced,0.04586666822433472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,2,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,2,balanced,0.043840001026789345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,2,balanced,0.04558399816354116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,2,balanced,0.045519997676213585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,2,balanced,0.04651199777921041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,2,balanced,0.047653332352638245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,2,balanced,0.052042668064435325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,2,balanced,0.05399466554323832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,2,balanced,0.05613866448402405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,2,balanced,0.06805333495140076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,2,balanced,0.07854400078455608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,2,balanced,0.10086400310198466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,2,balanced,0.11652800440788269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,2,balanced,0.154448002576828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,2,balanced,0.18440532684326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,2,balanced,0.2599253257115682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,2,balanced,0.3329226573308309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,2,balanced,0.4809066851933797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,2,balanced,0.6185866594314575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,8,balanced,0.03382399926582972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,8,power_law_1.01,0.01791359931230545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,8,balanced,0.035445332527160645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,8,balanced,0.03934400031963984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,8,power_law_1.01,0.01879040002822876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,8,balanced,0.044693330923716225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,8,balanced,0.060991997520128884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,8,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,8,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,8,balanced,0.05693866809209188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,8,balanced,0.053898667295773826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,8,balanced,0.05486933390299479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,8,balanced,0.05416533350944519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,8,balanced,0.05407999952634176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,8,balanced,0.05608533322811127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,8,balanced,0.05622933308283488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,8,balanced,0.05871466795603434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,8,balanced,0.06018133461475372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,8,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,8,balanced,0.06614399949709575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,8,balanced,0.06631466746330261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,8,power_law_1.01,0.018092800676822663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,8,balanced,0.07301866511503856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,8,balanced,0.07896000146865845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,8,power_law_1.01,0.01907840073108673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,8,power_law_1.01,0.01880960017442703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,8,power_law_1.01,0.0191551998257637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,8,power_law_1.01,0.02173440009355545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,8,power_law_1.01,0.020371200144290925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,8,power_law_1.01,0.022259199619293214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,8,power_law_1.01,0.023769600689411162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,8,balanced,0.10549333691596985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,8,power_law_1.01,0.022457599639892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,8,power_law_1.01,0.02504960000514984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,8,power_law_1.01,0.026617598533630372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,8,power_law_1.01,0.024940800666809083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,4,power_law_1.01,0.13038079738616942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,4,power_law_1.01,0.16812800168991088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,4,power_law_1.01,0.15433599948883056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,4,power_law_1.01,0.18927359580993652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,8,balanced,0.1164959967136383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,8,balanced,0.15581867098808289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,8,balanced,0.19106133778889975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,8,balanced,0.26333866516749066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,8,balanced,0.32973333199818927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,8,power_law_1.01,0.026374399662017822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,8,power_law_1.01,0.027871999144554137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,8,power_law_1.01,0.03811840116977692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,4,power_law_1.01,0.18602240085601807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,4,power_law_1.01,0.17374080419540405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,4,power_law_1.01,0.17656320333480835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,4,power_law_1.01,0.17445759773254393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,4,power_law_1.01,0.17681920528411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,4,power_law_1.01,0.17046400308609008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,4,power_law_1.01,0.16309759616851807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,4,power_law_1.01,0.17723519802093507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,8,power_law_1.01,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,4,power_law_1.01,0.20821120738983154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,8,power_law_1.01,0.049728000164031984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,4,power_law_1.01,0.19815679788589477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,4,power_law_1.01,0.21919360160827636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,4,power_law_1.01,0.22631680965423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,8,power_law_1.01,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,4,power_law_1.01,0.26878719329833983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,8,power_law_1.01,0.052243202924728394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,8,power_law_1.01,0.06316800117492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,8,power_law_1.01,0.07331839799880982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,8,power_law_1.01,0.09556480050086975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,8,power_law_1.01,0.12353919744491577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,8,power_law_1.01,0.16234240531921387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,8,power_law_1.01,0.2092672109603882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,4,power_law_1.01,0.3617856025695801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.01,0.3725759983062744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,32,power_law_1.01,0.1437183976173401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,32,power_law_1.01,0.13646719455718995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,32,power_law_1.01,0.13664000034332274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,32,power_law_1.01,0.13553279638290405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.01,0.46369280815124514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,32,power_law_1.01,0.13018239736557008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,32,power_law_1.01,0.12602239847183228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,32,power_law_1.01,0.12747520208358765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,32,power_law_1.01,0.1152959942817688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,32,power_law_1.01,0.12833280563354493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,32,power_law_1.01,0.13071999549865723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,32,power_law_1.01,0.12121599912643433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,32,power_law_1.01,0.1283776044845581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,32,power_law_1.01,0.1275712013244629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,32,power_law_1.01,0.12376960515975952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,32,power_law_1.01,0.13038719892501832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,32,power_law_1.01,0.13312000036239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,32,power_law_1.01,0.1324031949043274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,32,power_law_1.01,0.14401919841766359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.01,0.16669440269470215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.01,0.1879040002822876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.01,0.2212991952896118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.01,0.2692543983459473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.01,0.32354559898376467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.01,0.4327104091644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.01,0.6351935863494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.01,0.5441408157348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.01,0.7442560195922852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.01,0.9833344459533692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.01,0.9643839836120606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.01,1.265824031829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.01,1.9250495910644532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.01,2.039423942565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.01,3.1359167098999023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.01,4.08990707397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,4,balanced,0.07845333218574524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,4,balanced,0.08051733175913493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,4,balanced,0.08890666564305623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,4,balanced,0.11178132891654968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,4,balanced,0.14755200346310934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,4,balanced,0.2221013307571411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,4,balanced,0.2219946583112081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,4,balanced,0.2219840089480082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,4,balanced,0.2254400054613749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,4,balanced,0.2222399910291036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,4,balanced,0.22566399971644083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,4,balanced,0.23073599735895792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,4,balanced,0.22879467407862344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,4,balanced,0.23253866036732992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,4,balanced,0.23501867055892944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,2,power_law_1.2,0.14209920167922974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,2,power_law_1.2,0.1672063946723938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,2,power_law_1.2,0.2661439895629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,2,power_law_1.2,0.2757951974868774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,2,power_law_1.2,0.29723520278930665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,2,power_law_1.2,0.29395840167999265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,2,power_law_1.2,0.29421439170837405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,2,power_law_1.2,0.2883584022521973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,2,power_law_1.2,0.28292479515075686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,2,power_law_1.2,0.27878398895263673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,2,power_law_1.2,0.26635520458221434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,2,power_law_1.2,0.2738687992095947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,2,power_law_1.2,0.3121344089508057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,2,power_law_1.2,0.30186240673065184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,2,power_law_1.2,0.351475191116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,2,power_law_1.2,0.3504575967788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,2,power_law_1.2,0.399833607673645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,2,power_law_1.2,0.46202239990234373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.2,0.5080448150634765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.2,0.798195219039917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.2,0.8611583709716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.2,1.349395179748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.2,1.6905279159545898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.2,2.136710357666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,64,power_law_1.01,0.03288959860801697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,64,power_law_1.01,0.03175680041313171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,64,power_law_1.01,0.026655998826026917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,64,power_law_1.01,0.02917119860649109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,64,power_law_1.01,0.029139199852943422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,64,power_law_1.01,0.030777600407600404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,64,power_law_1.01,0.030713599920272828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,64,power_law_1.01,0.030041599273681642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,64,power_law_1.01,0.030720001459121703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,4,power_law_1.2,0.04119040071964264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,4,power_law_1.2,0.04156799912452698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,64,power_law_1.01,0.03048959970474243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,64,power_law_1.01,0.030707201361656188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,4,power_law_1.2,0.04239999949932098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,4,power_law_1.2,0.046163201332092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,4,power_law_1.2,0.04874880015850067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,4,power_law_1.2,0.05578240156173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,4,power_law_1.2,0.05620480179786682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,4,power_law_1.2,0.05720319747924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,4,power_law_1.2,0.0565887987613678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,8,power_law_1.01,0.056531202793121335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,4,power_law_1.2,0.05742719769477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,8,power_law_1.01,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,64,power_law_1.01,0.030707201361656188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,8,power_law_1.01,0.05921279788017273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.2,2.847251129150391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,8,power_law_1.01,0.06783360242843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,8,power_law_1.01,0.0699072003364563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,8,power_law_1.01,0.07624959945678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,8,power_law_1.01,0.07816960215568543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,4,power_law_1.2,0.05772799849510193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,8,power_law_1.01,0.08083840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,4,power_law_1.2,0.05925760269165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,64,power_law_1.01,0.031244799494743347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,64,power_law_1.01,0.03258239924907684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,64,power_law_1.01,0.036134400963783266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,64,power_law_1.01,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,64,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,64,power_law_1.01,0.04127359986305237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,64,power_law_1.01,0.04119040071964264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,8,power_law_1.01,0.07921280264854431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,64,power_law_1.01,0.04640640020370483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,8,power_law_1.01,0.08201599717140198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,8,power_law_1.01,0.0822272002696991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,8,power_law_1.01,0.08574720025062561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,8,power_law_1.01,0.08642560243606567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,8,power_law_1.01,0.08843520283699036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,8,power_law_1.01,0.0971455991268158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,4,power_law_1.2,0.06113280057907104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,4,power_law_1.2,0.06169599890708923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,64,power_law_1.01,0.04570879936218262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,4,power_law_1.2,0.06676480174064636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,64,power_law_1.01,0.055846399068832396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,4,power_law_1.2,0.06918399930000305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,64,power_law_1.01,0.05854719877243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,4,power_law_1.2,0.07293440103530884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,64,power_law_1.01,0.09102079868316651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,4,power_law_1.2,0.07916160225868225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,64,power_law_1.01,0.09599360227584838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,4,power_law_1.2,0.08972160220146179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,64,power_law_1.01,0.12709759473800658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,8,power_law_1.01,0.10146559476852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,8,power_law_1.01,0.10518399477005005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,8,power_law_1.01,0.12259839773178101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.01,0.13583359718322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.01,0.16548479795455934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,4,power_law_1.2,0.11256320476531982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.01,0.17461119890213012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,64,power_law_1.01,0.15322879552841187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.01,0.237280011177063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.01,0.2942975997924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.01,0.37837440967559816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,4,power_law_1.2,0.12798080444335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.01,0.48504319190979006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,4,power_law_1.2,0.166156804561615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.01,0.775596809387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.01,0.9271743774414063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,4,power_law_1.2,0.21610240936279296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,4,power_law_1.2,0.28065919876098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,4,power_law_1.2,0.3650880098342896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,4,power_law_1.2,0.5471680164337158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.2,3.9464126586914063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,4,power_law_1.2,0.6957568168640137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,4,power_law_1.2,0.054016000032424925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,4,power_law_1.2,0.05667200088500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,4,power_law_1.2,0.05795199871063232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,4,power_law_1.2,0.06647679805755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,4,power_law_1.2,0.07169280052185059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,4,power_law_1.2,0.07845119833946228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,4,power_law_1.2,0.08978559970855712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,4,power_law_1.2,0.08330240249633789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,4,power_law_1.2,0.08862720131874084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,4,power_law_1.2,0.08986240029335021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,4,power_law_1.2,0.092467200756073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,4,power_law_1.2,0.0932416021823883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,4,power_law_1.2,0.09775999784469605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,4,power_law_1.2,0.09854080080986023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,4,power_law_1.2,0.10752639770507813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,4,power_law_1.2,0.11282559633255004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,4,power_law_1.2,0.11482880115509034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,4,power_law_1.2,0.1268288016319275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,64,power_law_1.01,0.04392960071563721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,4,power_law_1.2,0.14152319431304933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,64,power_law_1.01,0.041971200704574586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,4,power_law_1.2,0.17233279943466187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,4,power_law_1.2,0.2055039882659912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,4,power_law_1.2,0.24284799098968507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,4,power_law_1.2,0.31191039085388184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,4,power_law_1.2,0.4333183765411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,4,power_law_1.2,0.5850048065185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,64,power_law_1.01,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,64,power_law_1.01,0.044435200095176694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,4,power_law_1.2,0.8296768188476562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,4,power_law_1.2,1.0820992469787598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,64,power_law_1.01,0.043161600828170776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,64,power_law_1.01,0.044838398694992065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,64,power_law_1.01,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,64,power_law_1.01,0.044563201069831845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.2,4.940339279174805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,64,power_law_1.01,0.04475519955158234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,64,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,64,power_law_1.01,0.04550400078296661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,64,power_law_1.01,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,64,power_law_1.01,0.046035200357437134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,64,power_law_1.01,0.045638400316238406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,64,power_law_1.01,0.0497408002614975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,64,power_law_1.01,0.051500797271728516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,64,power_law_1.01,0.05602560043334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,64,power_law_1.01,0.06061440110206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.01,0.06441599726676941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.01,0.07432960271835327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.01,0.08851199746131896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.01,0.11048320531845093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.01,0.1313920021057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.01,0.18466559648513795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.01,0.22469758987426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.01,0.3438080072402954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.01,0.4033535957336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,16,power_law_1.2,0.08799999952316284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,16,power_law_1.2,0.08981119990348815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,16,power_law_1.2,0.08654720187187195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,16,power_law_1.2,0.09312639832496643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,16,power_law_1.2,0.09283199906349182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,16,power_law_1.2,0.09472000002861022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,16,power_law_1.2,0.09336320161819459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,16,power_law_1.2,0.0948032021522522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,16,power_law_1.2,0.09619200229644775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,16,power_law_1.2,0.09631999731063842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,16,power_law_1.2,0.09761919975280761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,16,power_law_1.2,0.10032639503479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,16,power_law_1.2,0.1007040023803711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,16,power_law_1.2,0.10295039415359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,16,power_law_1.2,0.11086080074310303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,16,power_law_1.2,0.12527999877929688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,16,power_law_1.2,0.130950403213501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,16,power_law_1.2,0.14488960504531861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,16,power_law_1.2,0.1637503981590271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,16,power_law_1.2,0.20330240726470947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,16,power_law_1.2,0.22696321010589598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,16,power_law_1.2,0.3307391881942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,16,power_law_1.2,0.4100992202758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,16,power_law_1.2,0.5597375869750977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,16,power_law_1.2,0.7345088005065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,16,power_law_1.2,1.2088512420654296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,16,power_law_1.2,1.5337984085083007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,8,power_law_1.2,0.10651520490646363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,8,power_law_1.2,0.12344319820404052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,8,power_law_1.2,0.13538559675216674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,2,power_law_1.2,0.0722815990447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,8,power_law_1.2,0.16736639738082887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,2,power_law_1.2,0.08220800161361694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,8,power_law_1.2,0.17669119834899902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,2,power_law_1.2,0.099891197681427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,8,power_law_1.2,0.18343039751052856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,2,power_law_1.2,0.12879999876022338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,8,power_law_1.2,0.18131840229034424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,8,power_law_1.2,0.18629759550094604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,8,power_law_1.2,0.18124799728393554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,8,power_law_1.2,0.18525439500808716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,8,power_law_1.2,0.19296640157699585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,8,power_law_1.2,0.2028736114501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,8,power_law_1.2,0.192249596118927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,2,power_law_1.2,0.1529855966567993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,8,power_law_1.2,0.20647039413452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,2,power_law_1.2,0.2203200101852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,2,power_law_1.2,0.231443190574646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,2,power_law_1.2,0.23559041023254396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,2,power_law_1.2,0.23761279582977296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,2,power_law_1.2,0.2441472053527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,2,power_law_1.2,0.25382399559020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,8,power_law_1.2,0.22009599208831787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,2,power_law_1.2,0.2593791961669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,8,power_law_1.2,0.22661120891571046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,2,power_law_1.2,0.2682431936264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,8,power_law_1.2,0.2442624092102051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,2,power_law_1.2,0.2823040008544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,8,power_law_1.2,0.27130880355834963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,8,power_law_1.2,0.3035072088241577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,8,power_law_1.2,0.3924288034439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,8,power_law_1.2,0.4411456108093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,8,power_law_1.2,0.537004804611206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,2,power_law_1.2,0.2925184011459351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,8,power_law_1.2,0.6517888069152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,2,power_law_1.2,0.30697600841522216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,2,power_law_1.2,0.3349760055541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,2,power_law_1.2,0.38414080142974855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,2,power_law_1.2,0.42465920448303224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,2,power_law_1.2,0.5231040000915528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,2,power_law_1.2,0.640064001083374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,8,power_law_1.2,0.8454272270202636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,2,power_law_1.2,0.8216128349304199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,8,power_law_1.2,1.1572992324829101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,2,power_law_1.2,0.9897919654846191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,2,power_law_1.2,1.3372223854064942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,2,power_law_1.2,1.5911423683166503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,8,power_law_1.2,1.616499137878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,2,power_law_1.2,2.183795166015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,2,power_law_1.2,2.9720767974853515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,2,balanced,0.08083199958006541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,2,balanced,0.09913600484530131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,2,balanced,0.1431839962800344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,2,balanced,0.2125493288040161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,2,balanced,0.3493493398030599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,2,balanced,0.3863999843597412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,2,balanced,0.37838931878407794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,2,balanced,0.3920266628265381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,2,balanced,0.3881066640218099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,2,balanced,0.37809598445892334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,2,balanced,0.3805653254191081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,2,balanced,0.40752001603444415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,2,balanced,0.38973867893218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,2,balanced,0.4073226849238078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,2,balanced,0.4083626667658488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,2,balanced,0.3860693375269572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,2,balanced,0.41410664717356366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,2,balanced,0.4021386702855428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,2,balanced,0.4128640095392863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,2,balanced,0.4461919864018758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,2,balanced,0.47381333510080975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,2,balanced,0.7948106924692789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,8,power_law_1.2,1.964543914794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,2,balanced,0.8208213647206625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,2,balanced,1.170192003250122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,2,balanced,1.5363945960998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,2,balanced,1.9002933502197266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,2,balanced,2.6386987368265786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,1,power_law_1.01,0.02736639976501465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,1,power_law_1.01,0.026995199918746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,1,power_law_1.01,0.026982399821281432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,1,power_law_1.01,0.031999999284744264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,1,power_law_1.01,0.044633600115776065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,1,power_law_1.01,0.06108800172805786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,1,power_law_1.01,0.052076798677444455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,1,power_law_1.01,0.06104959845542908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,1,power_law_1.01,0.05134720206260681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,1,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,1,power_law_1.01,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,1,power_law_1.01,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,1,power_law_1.01,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,1,power_law_1.01,0.07631360292434693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,1,power_law_1.01,0.16008960008621215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,1,power_law_1.01,0.18511999845504762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,1,power_law_1.01,0.21697919368743895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,1,power_law_1.01,0.11831680536270142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.01,0.1439039945602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.01,0.20083200931549072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.01,0.2571199893951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.01,0.3590143918991089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.01,0.46782717704772947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,8,balanced,0.061610668897628784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,8,balanced,0.06409599880377452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,8,balanced,0.0747573326031367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,8,balanced,0.09272000193595886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,8,balanced,0.13124799728393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,8,balanced,0.14055466651916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,8,balanced,0.14131733775138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,8,balanced,0.1415786643822988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,8,balanced,0.14421866337458292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,8,balanced,0.14220266540845236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,8,balanced,0.14326933026313782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,8,balanced,0.14678933223088583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,8,balanced,0.14567466576894125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,8,balanced,0.14917866388956705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,8,balanced,0.1572106679280599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,8,balanced,0.1585973302523295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,8,balanced,0.16102400422096252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,8,balanced,0.17299733559290567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,8,balanced,0.17882666985193887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,8,balanced,0.19313599665959677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,8,balanced,0.21288534005482992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,8,balanced,0.2876266638437907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,8,balanced,0.32518933216730755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,8,balanced,0.43672001361846924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,8,balanced,0.568122665087382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,8,balanced,0.7703359921773275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,8,balanced,1.0138133366902669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.01,0.6771903991699219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.01,0.8897919654846191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.01,1.3123647689819335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,32,power_law_1.2,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,32,power_law_1.2,0.0840831995010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,32,power_law_1.2,0.08426880240440368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,32,power_law_1.2,0.08812159895896912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,32,power_law_1.2,0.08824959993362427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,32,power_law_1.2,0.07779200077056884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,32,power_law_1.2,0.0716159999370575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,32,power_law_1.2,0.07947520017623902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,32,power_law_1.2,0.0864192008972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,32,power_law_1.2,0.08312320113182067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,32,power_law_1.2,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,32,power_law_1.2,0.08897280097007751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,32,power_law_1.2,0.08613759875297547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,32,power_law_1.2,0.09029759764671326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,32,power_law_1.2,0.09818239808082581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,32,power_law_1.2,0.09805439710617066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,32,power_law_1.2,0.10392960309982299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,32,power_law_1.2,0.12297600507736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,32,power_law_1.2,0.12992639541625978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,32,power_law_1.2,0.15427839756011963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,32,power_law_1.2,0.17075200080871583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,32,power_law_1.2,0.22931199073791503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,32,power_law_1.2,0.2565887928009033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,32,power_law_1.2,0.3468096017837524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,32,power_law_1.2,0.4290624141693115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,32,power_law_1.2,0.5801983833312988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,32,power_law_1.2,0.7932223796844482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.01,1.7330368041992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,8,power_law_1.01,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,8,power_law_1.01,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,8,power_law_1.01,0.06364799737930298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,8,power_law_1.01,0.06812160015106201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,8,power_law_1.01,0.07265920042991639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,8,power_law_1.01,0.07455360293388366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,8,power_law_1.01,0.07652480006217957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,8,power_law_1.01,0.07432960271835327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,8,power_law_1.01,0.0729856014251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,8,power_law_1.01,0.07269120216369629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,8,power_law_1.01,0.07832959890365601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,8,power_law_1.01,0.08174719810485839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,8,power_law_1.01,0.08421760201454162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,8,power_law_1.01,0.08578559756278992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,8,power_law_1.01,0.0911296010017395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,8,power_law_1.01,0.0932864010334015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,8,power_law_1.01,0.09600639939308167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,8,power_law_1.01,0.10659199953079224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,32,power_law_1.2,0.07226240038871765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,8,power_law_1.01,0.1098431944847107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,32,power_law_1.2,0.0748799979686737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,8,power_law_1.01,0.13014400005340576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,32,power_law_1.2,0.07186560034751892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,8,power_law_1.01,0.14270080327987672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,32,power_law_1.2,0.07170559763908387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,8,power_law_1.01,0.17418240308761596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,8,power_law_1.01,0.20039041042327882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,8,power_law_1.01,0.2667072057723999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,8,power_law_1.01,0.33512320518493655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,8,power_law_1.01,0.45696001052856444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,8,power_law_1.01,0.591206407546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,32,power_law_1.2,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,32,power_law_1.2,0.06894720196723939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,32,power_law_1.2,0.06811519861221313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,32,power_law_1.2,0.07386239767074584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,32,power_law_1.2,0.07160959839820862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,32,power_law_1.2,0.07472000122070313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,32,power_law_1.2,0.07477759718894958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,32,power_law_1.2,0.07408000230789184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,32,power_law_1.2,0.07793279886245727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,32,power_law_1.2,0.0780672013759613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,32,power_law_1.2,0.08369280099868774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,32,power_law_1.2,0.08446720242500305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,32,power_law_1.2,0.09148160219192505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,32,power_law_1.2,0.10406399965286255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,32,power_law_1.2,0.11187200546264649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,32,power_law_1.2,0.14024319648742675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,32,power_law_1.2,0.15822720527648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,32,power_law_1.2,0.20428800582885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,32,power_law_1.2,0.25883519649505615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,32,power_law_1.2,0.3377216100692749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,32,power_law_1.2,0.4587520122528076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,32,power_law_1.2,0.6392064094543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,32,power_law_1.2,0.819865608215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,8,balanced,0.049839998284975685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,8,balanced,0.05594133337338766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,8,balanced,0.07034666836261749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,8,balanced,0.09850666920344035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,8,balanced,0.13294399778048197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,8,balanced,0.15851199626922607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,8,balanced,0.16135467092196146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,8,balanced,0.16263467073440552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,8,balanced,0.16336533427238464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,8,balanced,0.16366400321324667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,8,balanced,0.16383467117945352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,8,balanced,0.16645333170890808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,8,balanced,0.16775999466578165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,8,balanced,0.17095466454823813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,8,balanced,0.1780959963798523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,8,balanced,0.18109333515167236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,8,balanced,0.18797866503397623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,8,balanced,0.21768534183502197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,8,balanced,0.21658132473627725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,8,balanced,0.2735946575800578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,8,balanced,0.2874240080515544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,8,balanced,0.4092959960301717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,8,balanced,0.38438932100931805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,8,balanced,0.5470186471939087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,8,balanced,0.6127733389536539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,8,balanced,0.8767093022664388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,8,balanced,1.041642665863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,4,power_law_1.01,0.07049599885940552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,4,power_law_1.01,0.09245439767837524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,4,power_law_1.01,0.11256959438323974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,4,power_law_1.01,0.13201919794082642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,4,power_law_1.01,0.14742399454116822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,4,power_law_1.01,0.15011199712753295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,4,power_law_1.01,0.20775680541992186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,4,power_law_1.01,0.21389439105987548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,4,power_law_1.01,0.2322688102722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,4,power_law_1.01,0.22175359725952148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,4,power_law_1.01,0.22449920177459717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,4,power_law_1.01,0.21845760345458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,4,power_law_1.01,0.2228032112121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,4,power_law_1.01,0.24225280284881592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,4,power_law_1.01,0.25352959632873534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,4,power_law_1.01,0.2561280012130737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,4,power_law_1.01,0.2762752056121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,4,power_law_1.01,0.2913088083267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,4,power_law_1.01,0.30475521087646484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,4,power_law_1.01,0.35149440765380857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,4,power_law_1.01,0.38794240951538084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,4,power_law_1.01,0.4879295825958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,4,power_law_1.01,0.5431488037109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,4,power_law_1.01,0.6828864097595215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,4,power_law_1.01,0.8660544395446778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,4,power_law_1.01,1.1120256423950194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,4,power_law_1.01,1.4206015586853027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,64,power_law_1.2,0.01674239933490753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,64,power_law_1.2,0.015897600352764128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,64,power_law_1.2,0.016371199488639833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,64,power_law_1.2,0.017056000232696534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,64,power_law_1.2,0.018905599415302277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,64,power_law_1.2,0.031302401423454286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,64,power_law_1.2,0.02629759907722473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,64,power_law_1.2,0.02593280076980591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,64,power_law_1.2,0.026579201221466064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,64,power_law_1.2,0.026815998554229736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,64,power_law_1.2,0.026067200303077697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,64,power_law_1.2,0.026521599292755126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,64,power_law_1.2,0.02733440101146698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,64,power_law_1.2,0.02757120132446289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,64,power_law_1.2,0.028435200452804565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,64,power_law_1.2,0.028780800104141236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,64,power_law_1.2,0.02958720028400421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,64,power_law_1.2,0.03218559920787811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,64,power_law_1.2,0.03390080034732819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,64,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,64,power_law_1.2,0.04259200096130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,64,power_law_1.2,0.04865280091762543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,64,power_law_1.2,0.060115200281143186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,64,power_law_1.2,0.07713919878005981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,64,power_law_1.2,0.09710720181465149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,64,power_law_1.2,0.1307968020439148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,64,power_law_1.2,0.168614399433136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,32,balanced,0.053082664807637535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,32,balanced,0.05213333169619242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,32,balanced,0.05388799806435903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,32,balanced,0.05996799965699514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,32,balanced,0.062181333700815834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,32,balanced,0.07009066641330719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,32,balanced,0.07369066774845123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,32,balanced,0.07441066702206929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,32,balanced,0.07494933406511943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,32,balanced,0.07634666562080383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,32,balanced,0.07435733576615651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,32,balanced,0.080335999528567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,32,balanced,0.08001066744327545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,32,balanced,0.07864533364772797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,32,balanced,0.08749866485595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,32,balanced,0.08477333188056946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,32,balanced,0.09302933017412822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,32,balanced,0.0995199978351593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,32,balanced,0.10698133707046509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,32,balanced,0.128330667813619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,32,balanced,0.14410666624704996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,32,balanced,0.18120000759760538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,32,balanced,0.21317332983016968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,32,balanced,0.2789919972419739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,32,balanced,0.3614613215128581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,32,balanced,0.491157333056132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,32,balanced,0.6317013502120972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,1,power_law_1.01,0.05287039875984192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,1,power_law_1.01,0.07194240093231201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,1,power_law_1.01,0.0981760025024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,1,power_law_1.01,0.1356927990913391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,1,power_law_1.01,0.17543040513992308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,1,power_law_1.01,0.2688767910003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,1,power_law_1.01,0.28008959293365476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,1,power_law_1.01,0.2829056024551392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,1,power_law_1.01,0.2974080085754395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,1,power_law_1.01,0.3049855947494507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,1,power_law_1.01,0.30979840755462645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,1,power_law_1.01,0.32245759963989257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,1,power_law_1.01,0.3244992017745972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,1,power_law_1.01,0.3335488080978394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,1,power_law_1.01,0.34703359603881834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,1,power_law_1.01,0.3621567964553833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,1,power_law_1.01,0.38445439338684084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,1,power_law_1.01,0.4268159866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,1,power_law_1.01,0.4790463924407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,1,power_law_1.01,0.5841663837432861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,1,power_law_1.01,0.6799744129180908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,1,power_law_1.01,0.8583423614501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,1,power_law_1.01,1.0858367919921874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,1,power_law_1.01,1.4191935539245606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,1,power_law_1.01,1.8192832946777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,1,power_law_1.01,2.4165376663208007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,1,power_law_1.01,2.960825538635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,16,power_law_1.01,0.03345920145511627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,16,power_law_1.01,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,16,power_law_1.01,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,16,power_law_1.01,0.03325439989566803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,16,power_law_1.01,0.03290880024433136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,16,power_law_1.01,0.03468799889087677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,16,power_law_1.01,0.0348800003528595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,16,power_law_1.01,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,16,power_law_1.01,0.036447998881340024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,16,power_law_1.01,0.037248000502586365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,16,power_law_1.01,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,16,power_law_1.01,0.039129599928855896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,16,power_law_1.01,0.03984000086784363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,16,power_law_1.01,0.040320000052452086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,16,power_law_1.01,0.04387840032577515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,16,power_law_1.01,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,16,power_law_1.01,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,16,power_law_1.01,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,16,power_law_1.01,0.06259840130805969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,16,power_law_1.01,0.0738752007484436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,16,power_law_1.01,0.08133119940757752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,16,power_law_1.01,0.11946879625320435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,16,power_law_1.01,0.14982399940490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,16,power_law_1.01,0.2044287919998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,16,power_law_1.01,0.25382399559020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,16,power_law_1.01,0.3627327919006348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,16,power_law_1.01,0.5121024131774903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,8,power_law_1.2,0.08774399757385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,8,power_law_1.2,0.0887935996055603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,8,power_law_1.2,0.09742720127105713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,8,power_law_1.2,0.11381759643554687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,8,power_law_1.2,0.12588800191879274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,8,power_law_1.2,0.12617599964141846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,8,power_law_1.2,0.1276479959487915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,8,power_law_1.2,0.125600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,8,power_law_1.2,0.12876800298690796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,8,power_law_1.2,0.1397055983543396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,8,power_law_1.2,0.13691519498825072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,8,power_law_1.2,0.14156160354614258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,8,power_law_1.2,0.13943040370941162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,8,power_law_1.2,0.14169600009918212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,8,power_law_1.2,0.1491520047187805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,8,power_law_1.2,0.15901440382003784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,8,power_law_1.2,0.1624511957168579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,8,power_law_1.2,0.191539204120636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.2,0.2141632080078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.2,0.2627903938293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.2,0.26844799518585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.2,0.3455104112625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.2,0.42714881896972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.2,0.6042496204376221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.2,0.8501248359680176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,8,power_law_1.01,0.030272001028060914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.2,1.1310208320617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,8,power_law_1.01,0.03221119940280914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,8,power_law_1.01,0.03201279938220978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,8,power_law_1.01,0.034668800234794614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,8,power_law_1.01,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,8,power_law_1.01,0.03691520094871521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,8,power_law_1.01,0.03713279962539673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,8,power_law_1.01,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,8,power_law_1.01,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,8,power_law_1.01,0.03898240029811859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,8,power_law_1.01,0.0409280002117157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,8,power_law_1.01,0.042368000745773314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,8,power_law_1.01,0.04357120096683502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,8,power_law_1.01,0.04417920112609863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,8,power_law_1.01,0.04812160134315491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,8,power_law_1.01,0.05084159970283508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,8,power_law_1.01,0.05381119847297668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,8,power_law_1.01,0.06441599726676941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,8,power_law_1.01,0.06743680238723755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,8,power_law_1.01,0.08506240248680115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,8,power_law_1.01,0.10183680057525635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,8,power_law_1.01,0.13925119638442993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,8,power_law_1.01,0.18145279884338378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,8,power_law_1.01,0.2569472074508667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,8,power_law_1.01,0.3189055919647217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,8,power_law_1.01,0.47150721549987795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,8,power_law_1.01,0.5914688110351562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.2,1.5233216285705566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,32,balanced,0.045994664231936135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,32,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,32,balanced,0.04961599906285604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,32,balanced,0.05811200042565664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,32,balanced,0.06091733276844025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,32,balanced,0.06201600035031637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,32,balanced,0.06202666461467743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,32,balanced,0.060864001512527466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,32,balanced,0.06409066418806712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,32,balanced,0.06225066880385081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,32,balanced,0.06372266511122386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,32,balanced,0.06681600213050842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,32,balanced,0.06635733445485432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,32,balanced,0.06860800087451935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,32,balanced,0.07240533332029979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,32,balanced,0.0738453318675359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,32,balanced,0.07964266836643219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,32,balanced,0.09879466891288757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,32,balanced,0.09893332918485005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,1,power_law_1.01,0.04922240078449249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,32,balanced,0.1276693344116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,32,balanced,0.14323733250300089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,1,power_law_1.01,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,32,balanced,0.19359999895095825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,1,power_law_1.01,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,1,balanced,0.07030933101971944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,1,balanced,0.0941546658674876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,1,balanced,0.13772799571355185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,1,balanced,0.22080000241597494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,1,balanced,0.3871413469314575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,1,balanced,0.7268266677856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,1,balanced,0.7252106666564941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,32,balanced,0.17631999651590982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,1,power_law_1.01,0.09123200178146362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,1,balanced,0.7320319811503092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,1,power_law_1.01,0.10800000429153442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,32,balanced,0.22749867041905722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,1,balanced,0.7358933289845785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,1,power_law_1.01,0.15002880096435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,1,balanced,0.7406773567199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,1,power_law_1.01,0.1570623993873596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,32,balanced,0.2630400061607361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,1,power_law_1.01,0.16222720146179198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,32,balanced,0.35072000821431476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,1,power_law_1.01,0.16741759777069093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,32,balanced,0.4220000108083089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,1,power_law_1.01,0.17152639627456664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,1,balanced,0.7427519957224528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,1,power_law_1.01,0.173689603805542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,1,balanced,0.7488373120625814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,1,power_law_1.01,0.18049919605255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,1,balanced,0.7565386295318604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,1,power_law_1.01,0.1834496021270752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,1,power_law_1.01,0.19271680116653442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,1,balanced,0.7654879887898763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,32,power_law_1.2,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,1,balanced,0.7758826414744059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,32,power_law_1.2,0.033843201398849485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,1,balanced,0.7878613471984863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,32,power_law_1.2,0.03170560002326965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,1,balanced,0.8020479679107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,1,balanced,0.8416586716969808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,32,power_law_1.2,0.033452799916267394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,1,power_law_1.01,0.19855999946594238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,1,balanced,0.8814293543497721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,1,power_law_1.01,0.20734078884124757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,1,balanced,0.9599413077036539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,1,balanced,1.01638396581014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,1,balanced,1.1524746417999268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,1,balanced,1.289685328801473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,32,power_law_1.2,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,1,power_law_1.01,0.21721599102020264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,1,power_law_1.01,0.24949119091033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.01,0.2799743890762329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.01,0.34766719341278074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.01,0.38387839794158934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.01,0.49480319023132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,32,power_law_1.2,0.034995201230049136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,1,balanced,1.6417120297749836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,32,power_law_1.2,0.0350847989320755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.01,0.604863977432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,32,power_law_1.2,0.035385599732398985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,1,balanced,1.933450698852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.01,0.8188223838806152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.01,1.0292736053466798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,1,balanced,2.742394765218099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.01,1.4550527572631835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,1,balanced,3.366565386454264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,32,power_law_1.2,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,32,power_law_1.2,0.03738240003585815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,32,power_law_1.2,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,32,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.01,1.8857599258422852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,32,power_law_1.2,0.04220159947872162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,32,power_law_1.2,0.047654399275779726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,32,power_law_1.2,0.055174398422241214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,32,power_law_1.2,0.05321599841117859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,32,power_law_1.2,0.05858560204505921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,32,power_law_1.2,0.0648256003856659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,32,power_law_1.2,0.07002879977226258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,32,power_law_1.2,0.08789119720458985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,32,power_law_1.2,0.1029695987701416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,32,power_law_1.2,0.14049279689788818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,32,power_law_1.2,0.1395840048789978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,32,power_law_1.2,0.22158079147338866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,32,power_law_1.2,0.2573888063430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,8,power_law_1.01,0.09520639777183533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,8,power_law_1.01,0.14353280067443847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,8,power_law_1.01,0.17022080421447755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,8,power_law_1.01,0.17512320280075072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,8,power_law_1.01,0.22617599964141846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,8,power_law_1.01,0.1833024024963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,8,power_law_1.01,0.2221695899963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,8,power_law_1.01,0.21018879413604735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,8,power_law_1.01,0.2250432014465332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,8,power_law_1.01,0.23498239517211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,8,power_law_1.01,0.21381120681762694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,8,power_law_1.01,0.22386560440063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,8,power_law_1.01,0.2369152069091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,8,power_law_1.01,0.23522560596466063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,8,power_law_1.01,0.23875839710235597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,8,power_law_1.01,0.24248321056365968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,8,power_law_1.01,0.255295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,8,power_law_1.01,0.29461119174957273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,8,power_law_1.01,0.2937279939651489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,8,power_law_1.01,0.34796159267425536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,8,power_law_1.01,0.36620800495147704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,8,power_law_1.01,0.45512962341308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,8,power_law_1.01,0.47648000717163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,8,power_law_1.01,0.6180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,8,power_law_1.01,0.7515967845916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,8,power_law_1.01,1.0004351615905762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,8,power_law_1.01,1.2122367858886718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,32,power_law_1.2,0.40189437866210936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,32,power_law_1.2,0.44525442123413084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,4,power_law_1.2,0.0627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,4,power_law_1.2,0.06696320176124573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,4,power_law_1.2,0.06518399715423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,4,power_law_1.2,0.07811840176582337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,4,power_law_1.2,0.08399360179901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,4,power_law_1.2,0.08293759822845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,4,power_law_1.2,0.09784960150718688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,4,power_law_1.2,0.10252159833908081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,4,power_law_1.2,0.09945600032806397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,4,power_law_1.2,0.10336639881134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,4,power_law_1.2,0.10264960527420045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,4,power_law_1.2,0.10288000106811523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,4,power_law_1.2,0.1090880036354065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,4,power_law_1.2,0.10982400178909302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,4,power_law_1.2,0.11646720170974731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,4,power_law_1.2,0.12136319875717164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,4,power_law_1.2,0.12255359888076782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,4,power_law_1.2,0.1384703993797302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,4,power_law_1.2,0.14991359710693358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,4,balanced,0.03989866624275843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,4,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,4,balanced,0.04364799956480662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,4,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,4,balanced,0.062405332922935486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,4,balanced,0.08871466914812724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,4,balanced,0.08877333005269368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,4,balanced,0.09077333410580952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,4,balanced,0.09143466750780742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,4,balanced,0.09181867043177287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,4,balanced,0.09248532851537068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,4,balanced,0.09459199508031209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,4,balanced,0.097680002450943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,4,balanced,0.09698133667310078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,4,balanced,0.10647466778755188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,4,balanced,0.10710400342941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,4,balanced,0.11165866255760193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,4,balanced,0.1254026691118876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,4,balanced,0.13205867012341818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,4,balanced,0.03375466664632162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,4,balanced,0.15380799770355225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,4,balanced,0.17057599623998007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,4,balanced,0.22416534026463827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,4,balanced,0.2575146754582723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,4,balanced,0.3591359853744507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,4,balanced,0.44582398732503253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,4,power_law_1.2,0.17950719594955444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,4,power_law_1.2,0.20373759269714356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,4,balanced,0.6432640155156454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,4,balanced,0.03199466566244761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,4,power_law_1.2,0.2528831958770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,4,balanced,0.8181013266245524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,4,power_law_1.2,0.3107392072677612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,4,power_law_1.2,0.4504767894744873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,4,power_law_1.2,0.5302783966064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,4,power_law_1.2,0.7617472171783447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,4,balanced,0.03292799989382426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,4,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,4,balanced,0.03578133384386698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,4,power_law_1.2,0.9768832206726075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,4,balanced,0.036389333506425224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,4,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,4,balanced,0.037589333951473236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,4,balanced,0.03739733248949051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,4,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,4,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,4,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,4,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,4,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,4,balanced,0.04180799921353658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,4,balanced,0.04355733096599579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,4,balanced,0.04548266530036926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,4,balanced,0.047914668917655945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,4,balanced,0.04785066843032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,4,balanced,0.057962665955225624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,4,balanced,0.06007466713587443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,4,balanced,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,4,balanced,0.08683199683825175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,4,balanced,0.1206719974676768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,4,balanced,0.1425279974937439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,4,balanced,0.19476799170176187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,4,balanced,0.24330667654673258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,8,power_law_1.2,0.03676159977912903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,8,power_law_1.2,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,8,power_law_1.2,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,8,power_law_1.2,0.037567999958992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,8,power_law_1.2,0.0377344012260437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,8,power_law_1.2,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,8,power_law_1.2,0.03852800130844116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,8,power_law_1.2,0.03880319893360138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,8,power_law_1.2,0.039980798959732056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,8,power_law_1.2,0.04142720103263855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,256,power_law_1.01,0.055846399068832396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,8,power_law_1.2,0.04136320054531097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,256,power_law_1.01,0.056518399715423585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,256,power_law_1.01,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,256,power_law_1.01,0.059033602476119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,256,power_law_1.01,0.060153597593307497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,256,power_law_1.01,0.061350399255752565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,256,power_law_1.01,0.06282879710197449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,256,power_law_1.01,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,256,power_law_1.01,0.06168320178985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,256,power_law_1.01,0.061343997716903687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,256,power_law_1.01,0.06227200031280518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,8,power_law_1.2,0.04594559967517853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,8,power_law_1.2,0.047712001204490664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,256,power_law_1.01,0.06336640119552613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,256,power_law_1.01,0.06389120221138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,256,power_law_1.01,0.06419839859008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,256,power_law_1.01,0.06782079935073852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,256,power_law_1.01,0.06804479956626892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,256,power_law_1.01,0.06947839856147767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,256,power_law_1.01,0.0830784022808075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,256,power_law_1.01,0.08835840225219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,256,power_law_1.01,0.10533759593963624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,8,power_law_1.2,0.05413119792938233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,256,power_law_1.01,0.12120959758758545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,256,power_law_1.01,0.16013439893722534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,256,power_law_1.01,0.19286400079727173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,256,power_law_1.01,0.24887681007385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,256,power_law_1.01,0.32565760612487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,8,power_law_1.2,0.059628802537918094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,256,power_law_1.01,0.44676480293273924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,256,power_law_1.01,0.5766079902648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,8,power_law_1.2,0.06796159744262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,8,power_law_1.2,0.07402880191802978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,8,power_law_1.2,0.095769602060318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,8,power_law_1.2,0.11889280080795288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,8,power_law_1.2,0.1610111951828003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,8,power_law_1.2,0.20156800746917725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,8,power_law_1.2,0.2756351947784424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,8,power_law_1.2,0.36452479362487794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,1,power_law_1.2,0.07263360023498536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,1,power_law_1.2,0.09354879856109619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,1,power_law_1.2,0.13229440450668334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,1,power_law_1.2,0.2036288022994995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,1,power_law_1.2,0.2818687915802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,1,power_law_1.2,0.3499392032623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,1,power_law_1.2,0.5223999977111816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,1,power_law_1.2,0.5586175918579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,1,power_law_1.2,0.5712639808654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,1,power_law_1.2,0.599891185760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,1,power_law_1.2,0.6124735832214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,1,power_law_1.2,0.6457344055175781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,1,power_law_1.2,0.643993616104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,1,power_law_1.2,0.6855487823486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,1,power_law_1.2,0.7023104190826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,1,power_law_1.2,0.7440256118774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,1,power_law_1.2,0.7597760200500489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,1,power_law_1.2,0.8434111595153808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,1,power_law_1.2,0.8872575759887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,1,power_law_1.2,1.006112003326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,1,power_law_1.2,1.0683903694152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,1,power_law_1.2,1.2539967536926269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,1,power_law_1.2,1.468569564819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,1,power_law_1.2,1.8298303604125976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,1,power_law_1.2,0.05231999754905701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,1,power_law_1.2,2.2101055145263673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,1,power_law_1.2,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,1,power_law_1.2,0.0789247989654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,1,power_law_1.2,0.11230720281600952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,1,power_law_1.2,2.961664009094238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,1,power_law_1.2,0.13935359716415405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,1,power_law_1.2,0.17984000444412232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,1,power_law_1.2,0.23795840740203858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,1,power_law_1.2,3.6732479095458985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,1,power_law_1.2,0.25435519218444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,1,power_law_1.2,0.2619904041290283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,1,power_law_1.2,0.26754560470581057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,1,power_law_1.2,0.28095359802246095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,1,power_law_1.2,0.2881983995437622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,1,power_law_1.2,0.2965375900268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,1,power_law_1.2,0.3090559959411621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,1,power_law_1.2,0.3236736059188843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,1,power_law_1.2,0.33946239948272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,1,power_law_1.2,0.35443840026855467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,1,power_law_1.2,0.40019841194152833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,1,power_law_1.2,0.42620158195495605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,1,power_law_1.2,0.5031040191650391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,1,power_law_1.2,0.5616896152496338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,1,power_law_1.2,0.7105472087860107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,1,power_law_1.2,0.8519871711730957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,1,power_law_1.2,1.1327743530273438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,8,power_law_1.2,0.5950079917907715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,1,power_law_1.2,1.4361087799072265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,1,power_law_1.2,1.9990911483764648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,1,power_law_1.2,2.538982391357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,8,power_law_1.2,0.7880256175994873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,32,power_law_1.2,0.05486080050468445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,32,power_law_1.2,0.03585279881954193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,32,power_law_1.2,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,32,power_law_1.2,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,8,power_law_1.2,1.1345919609069823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,32,power_law_1.2,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,32,power_law_1.2,0.033913600444793704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,32,power_law_1.2,0.03439359962940216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,32,power_law_1.2,0.03475840091705322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,32,power_law_1.2,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,32,power_law_1.2,0.03560320138931274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,32,power_law_1.2,0.037503999471664426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,32,power_law_1.2,0.041228801012039185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,32,power_law_1.2,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,32,power_law_1.2,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,32,power_law_1.2,0.052832001447677614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,32,power_law_1.2,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,32,power_law_1.2,0.061222398281097413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,32,power_law_1.2,0.06869120001792908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,32,power_law_1.2,0.06878079771995545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,32,power_law_1.2,0.07661439776420594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,32,power_law_1.2,0.08064000010490417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,32,power_law_1.2,0.09300479888916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,32,power_law_1.2,0.12184319496154786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,32,power_law_1.2,0.16026240587234497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,2,power_law_1.2,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,2,power_law_1.2,0.07456640005111695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,2,power_law_1.2,0.09468160271644592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,32,power_law_1.2,0.2232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,2,power_law_1.2,0.07422720193862915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,2,power_law_1.2,0.08144000172615051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,2,power_law_1.2,0.10659840106964111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,2,power_law_1.2,0.12125439643859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,2,power_law_1.2,0.14409600496292113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,2,power_law_1.2,0.15447039604187013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,2,power_law_1.2,0.17431039810180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,2,power_law_1.2,0.19207040071487427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,32,power_law_1.2,0.3302272081375122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,2,power_law_1.2,0.27167999744415283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,2,power_law_1.2,0.26725120544433595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,2,power_law_1.2,0.21066880226135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,2,power_law_1.2,0.27584640979766845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,2,power_law_1.2,0.2715775966644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,2,power_law_1.2,0.2847232103347778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,2,power_law_1.2,0.2871167898178101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,2,power_law_1.2,0.29703679084777834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,2,power_law_1.2,0.30316159725189207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,2,power_law_1.2,0.2903295993804932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,2,power_law_1.2,0.32083840370178224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,2,power_law_1.2,0.2895103931427002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,2,power_law_1.2,0.30084478855133057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,2,power_law_1.2,0.3159359931945801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,2,power_law_1.2,0.3220992088317871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,2,power_law_1.2,0.3287168025970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,2,power_law_1.2,0.32714879512786865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,2,power_law_1.2,0.35576961040496824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,2,power_law_1.2,0.33296639919281007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,2,power_law_1.2,0.34967041015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,2,power_law_1.2,0.3616192102432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,2,power_law_1.2,0.38555519580841063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,2,power_law_1.2,0.4155968189239502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,2,power_law_1.2,0.46648321151733396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,2,power_law_1.2,0.5644415855407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,2,power_law_1.2,0.6123007774353028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,2,power_law_1.2,0.7687359809875488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,2,power_law_1.2,0.36066560745239257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,2,power_law_1.2,0.39692161083221433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,2,power_law_1.2,0.4347519874572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,2,power_law_1.2,0.5346879959106445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,2,power_law_1.2,0.9700608253479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,2,power_law_1.2,1.256588840484619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,2,power_law_1.2,0.5856895923614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,2,power_law_1.2,0.7414527893066406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,2,power_law_1.2,1.4275839805603028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,2,power_law_1.2,0.7776000022888183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,2,power_law_1.2,1.9978303909301758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,2,power_law_1.2,1.0164735794067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,2,power_law_1.2,1.3405759811401368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,2,power_law_1.2,2.692620849609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,2,power_law_1.2,1.7061952590942382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,32,power_law_1.2,0.3792959928512573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,2,power_law_1.2,2.3817216873168947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,8,power_law_1.2,1.5674304008483886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,1,balanced,0.050255998969078064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,1,balanced,0.05193066596984863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,1,balanced,0.06052800019582113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,1,balanced,0.0853706697622935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,1,balanced,0.12308800220489502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,1,balanced,0.19523199399312338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,1,balanced,0.19643733898798624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,64,power_law_1.01,0.029382398724555968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,1,balanced,0.1968160072962443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,1,balanced,0.1991680065790812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,64,power_law_1.01,0.02969599962234497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,64,power_law_1.01,0.027968001365661622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,64,power_law_1.01,0.029209598898887634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,64,power_law_1.01,0.030316799879074097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,1,balanced,0.2011573314666748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,64,power_law_1.01,0.030854400992393494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,64,power_law_1.01,0.031244799494743347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,64,power_law_1.01,0.03157120048999786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,64,power_law_1.01,0.0313728004693985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,64,power_law_1.01,0.03237119913101196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,64,power_law_1.01,0.03272959887981415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,64,power_law_1.01,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,64,power_law_1.01,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,1,balanced,0.2060906688372294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,1,balanced,0.21309866507848105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,1,balanced,0.2098133365313212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,1,balanced,0.21789334217707315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,1,balanced,0.2261120080947876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,1,balanced,0.23050665855407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,1,balanced,0.23842666546503702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,1,balanced,0.26206399997075397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,1,balanced,0.2799999912579854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,1,balanced,0.3306879997253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,1,balanced,0.3786986668904622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,1,balanced,0.4606666564941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,64,power_law_1.01,0.035750401020050046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,1,balanced,0.5473759969075521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,64,power_law_1.01,0.038790398836135866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,1,balanced,0.7729547023773193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,64,power_law_1.01,0.03927040100097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,1,balanced,0.9323039849599203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,1,balanced,1.3312266667683919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,1,balanced,1.7147893905639648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,64,power_law_1.01,0.041843199729919435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,64,power_law_1.01,0.04480000138282776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,64,power_law_1.01,0.04674560129642487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,64,power_law_1.01,0.05745279788970947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,64,power_law_1.01,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,64,power_law_1.01,0.09139840006828308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,64,power_law_1.01,0.09738879799842834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,64,power_law_1.01,0.1349184036254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,64,power_law_1.01,0.17004799842834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,64,power_law_1.01,0.18339840173721314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,64,power_law_1.01,0.24931840896606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,2,power_law_1.01,0.04785279929637909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,2,power_law_1.01,0.05550079941749573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,2,power_law_1.01,0.06369919776916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,1,power_law_1.2,0.10002559423446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,2,power_law_1.01,0.07751039862632751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,1,power_law_1.2,0.14855680465698243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,2,power_law_1.01,0.09241600036621093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,1,power_law_1.2,0.24555521011352538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,1,power_law_1.2,0.36625919342041013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,1,power_law_1.2,0.4866496086120605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,1,power_law_1.2,0.8554368019104004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,1,power_law_1.2,0.9152447700500488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,2,power_law_1.01,0.11866879463195801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,2,power_law_1.01,0.12165759801864624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,2,power_law_1.01,0.12603520154953002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,2,power_law_1.01,0.13043199777603148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,2,power_law_1.01,0.13128960132598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,2,power_law_1.01,0.14021120071411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,2,power_law_1.01,0.1484992027282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,2,power_law_1.01,0.15484800338745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,1,power_law_1.2,0.974886417388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,2,power_law_1.01,0.15303679704666137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,2,power_law_1.01,0.16284799575805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,2,power_law_1.01,0.16698240041732787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,2,power_law_1.01,0.18052480220794678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,2,power_law_1.01,0.21130239963531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,2,power_law_1.01,0.23491840362548827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,2,power_law_1.01,0.2926271915435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,1,power_law_1.2,0.9727744102478028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,2,power_law_1.01,0.36263039112091067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,2,power_law_1.01,0.476364803314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,2,power_law_1.01,0.5974976062774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,2,power_law_1.01,0.7747200012207032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,2,power_law_1.01,1.0272576332092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,1,power_law_1.2,0.993228816986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,2,power_law_1.01,1.4425536155700684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,1,power_law_1.2,1.0430912017822265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,1,power_law_1.2,1.076204776763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,2,power_law_1.01,1.858470344543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,1,power_law_1.2,1.0859647750854493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,1,power_law_1.2,1.1198016166687013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,1,power_law_1.2,1.129043197631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,1,power_law_1.2,1.1880000114440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,1,power_law_1.2,1.2348671913146974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,1,power_law_1.2,1.3644224166870118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,1,power_law_1.2,1.4118656158447265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,1,power_law_1.2,1.6241792678833007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,1,power_law_1.2,1.5971263885498046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,1,power_law_1.2,1.8899200439453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,1,power_law_1.2,2.2173824310302734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,1,power_law_1.2,2.6886463165283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,1,power_law_1.2,3.4544639587402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,1,power_law_1.2,4.422208023071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,1,power_law_1.2,5.620281600952149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,64,balanced,0.023738667368888855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,64,balanced,0.02319466571013133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,64,balanced,0.023077333966890972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,64,balanced,0.023530667026837666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,64,balanced,0.025439999997615814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,64,balanced,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,64,balanced,0.055200000603993736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,64,balanced,0.05583466589450836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,64,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,64,balanced,0.053157334526379905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,64,balanced,0.051685333251953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,64,balanced,0.04996266464392344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,64,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,64,balanced,0.04565866788228353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,64,balanced,0.06189866860707601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,64,balanced,0.0625439981619517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,64,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,64,balanced,0.07632533212502797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,64,balanced,0.08272000153859456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,64,balanced,0.10936533411343892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,64,balanced,0.13428266843159994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,64,balanced,0.18828266859054565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,64,balanced,0.23099732398986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,64,balanced,0.33800001939137775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,64,balanced,0.4338719844818115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,64,balanced,0.6585226853688558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,64,balanced,0.8753493626912435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,4,power_law_1.01,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,4,power_law_1.01,0.046412798762321475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,4,power_law_1.01,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,4,power_law_1.01,0.05137919783592224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,4,power_law_1.01,0.057631999254226685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,4,power_law_1.01,0.05768960118293762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,4,power_law_1.01,0.06305279731750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,4,power_law_1.01,0.06846079826354981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,4,power_law_1.01,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,4,power_law_1.01,0.06929919719696045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,4,power_law_1.01,0.06886399984359741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,4,power_law_1.01,0.07004799842834472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,4,power_law_1.01,0.07085440158843995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,4,power_law_1.01,0.07676799893379212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,4,power_law_1.01,0.0815168023109436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,4,power_law_1.01,0.0840831995010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,4,power_law_1.01,0.09182720184326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,4,power_law_1.01,0.10566400289535523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,4,power_law_1.01,0.11077120304107665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,4,power_law_1.01,0.12997119426727294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,4,power_law_1.01,0.15041919946670532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,4,power_law_1.01,0.1857599973678589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,8,power_law_1.01,0.05940480232238769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,8,power_law_1.01,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,8,power_law_1.01,0.051795202493667605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,8,power_law_1.01,0.060710400342941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,8,power_law_1.01,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,8,power_law_1.01,0.06279039978981019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,32,power_law_1.2,0.13781119585037233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,4,power_law_1.01,0.2404927968978882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,4,power_law_1.01,0.3114880084991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,4,power_law_1.01,0.4365695953369141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,4,power_law_1.01,0.6077568054199218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,8,power_law_1.01,0.06403840184211732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,8,power_law_1.01,0.0651968002319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,32,power_law_1.2,0.16686079502105713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,8,power_law_1.01,0.0675711989402771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,4,power_law_1.01,0.6983232021331787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,8,power_law_1.01,0.06755840182304382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,8,power_law_1.01,0.07031679749488831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,8,power_law_1.01,0.07217919826507568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,8,power_law_1.01,0.072953599691391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,8,power_law_1.01,0.07749119997024537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,8,power_law_1.01,0.08309760093688964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,8,power_law_1.01,0.0912447988986969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,32,power_law_1.2,0.1652735948562622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,8,power_law_1.01,0.09095680117607116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,8,power_law_1.01,0.09887359738349914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,32,power_law_1.2,0.16789120435714722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,8,power_law_1.01,0.11336959600448608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,8,power_law_1.01,0.1302783966064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,8,power_law_1.01,0.16310399770736694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,8,power_law_1.01,0.19803520441055297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,8,power_law_1.01,0.2432960033416748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,8,power_law_1.01,0.31261439323425294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,8,power_law_1.01,0.39054720401763915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,32,power_law_1.2,0.166758394241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,8,power_law_1.01,0.5853824138641357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,32,power_law_1.2,0.14018559455871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,8,power_law_1.01,0.7787519931793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,32,power_law_1.2,0.13757439851760864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,32,power_law_1.2,0.15413119792938232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,32,power_law_1.2,0.139302396774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,32,power_law_1.2,0.15176960229873657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,32,power_law_1.2,0.1463871955871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,32,power_law_1.2,0.15589120388031005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,32,power_law_1.2,0.1446079969406128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,32,power_law_1.2,0.1611456036567688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,32,power_law_1.2,0.1699072003364563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,32,power_law_1.2,0.17075200080871583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,32,power_law_1.2,0.1610111951828003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,32,power_law_1.2,0.17950719594955444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,32,power_law_1.2,0.2088831901550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,32,power_law_1.2,0.23921918869018555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,32,power_law_1.2,0.2703999996185303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,32,power_law_1.2,0.3251712083816528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,32,power_law_1.2,0.381331205368042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,32,power_law_1.2,0.5381824016571045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,32,power_law_1.2,0.63854079246521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,32,power_law_1.2,0.9260479927062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,32,power_law_1.2,1.1864255905151366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,2,power_law_1.01,0.021216000616550445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,2,power_law_1.01,0.02184319943189621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,2,power_law_1.01,0.02465279996395111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,2,power_law_1.01,0.030124801397323608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,2,power_law_1.01,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,2,power_law_1.01,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,2,power_law_1.01,0.05329279899597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,2,power_law_1.01,0.05598719716072083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,2,power_law_1.01,0.05847679972648621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,2,power_law_1.01,0.06255999803543091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,2,power_law_1.01,0.06606079936027527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,2,power_law_1.01,0.06112639904022217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,2,power_law_1.01,0.066048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,2,power_law_1.01,0.0709824025630951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,2,power_law_1.01,0.07055360078811646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,2,power_law_1.01,0.07167360186576843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,2,power_law_1.01,0.07907840013504028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,2,power_law_1.01,0.09560959935188293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,2,power_law_1.01,0.10484479665756226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,1,power_law_1.01,0.0520576000213623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,2,power_law_1.01,0.16134400367736818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,1,power_law_1.01,0.05611519813537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,2,power_law_1.01,0.17467520236968995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,1,power_law_1.01,0.06670079827308655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,2,power_law_1.01,0.19921280145645143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,2,power_law_1.01,0.22231040000915528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,2,power_law_1.01,0.28243839740753174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,2,power_law_1.01,0.33482239246368406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,2,power_law_1.01,0.44837121963500975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,1,power_law_1.01,0.08448640108108521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,1,power_law_1.01,0.09699199795722961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,2,power_law_1.01,0.5667136192321778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,1,power_law_1.01,0.13018879890441895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,1,power_law_1.01,0.1358464002609253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,1,power_law_1.01,0.14042240381240845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,1,power_law_1.01,0.14282239675521852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,1,power_law_1.01,0.14423680305480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,1,power_law_1.01,0.15170559883117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,1,power_law_1.01,0.15352959632873536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,1,power_law_1.01,0.1584831953048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,1,power_law_1.01,0.1641343951225281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,1,power_law_1.01,0.17358720302581787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,1,power_law_1.01,0.17897599935531616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,1,power_law_1.01,0.1998271942138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,1,power_law_1.01,0.23486080169677734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.01,0.2719360113143921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.01,0.34637439250946045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.01,0.4054719924926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.01,0.5583871841430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.01,0.702566385269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.01,0.9871423721313477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.01,1.2664064407348632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.01,1.8226303100585937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.01,2.3838272094726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,16,power_law_1.2,0.03343360126018524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,16,power_law_1.2,0.03283840119838714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,16,power_law_1.2,0.03170560002326965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,16,power_law_1.2,0.03289600014686585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,16,power_law_1.2,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,16,power_law_1.2,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,16,power_law_1.2,0.0349375993013382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,16,power_law_1.2,0.03611519932746887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,16,power_law_1.2,0.036499199271202085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,16,power_law_1.2,0.037484800815582274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,16,power_law_1.2,0.038073599338531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,16,power_law_1.2,0.039340800046920775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,16,power_law_1.2,0.039827200770378116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,16,power_law_1.2,0.04089600145816803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,16,power_law_1.2,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,16,power_law_1.2,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,16,power_law_1.2,0.04952960014343262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,16,power_law_1.2,0.05910400152206421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,16,power_law_1.2,0.06437119841575623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,16,power_law_1.2,0.07665280103683472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,16,power_law_1.2,0.0874239981174469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,16,power_law_1.2,0.13776639699935914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,16,power_law_1.2,0.16437120437622071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,16,power_law_1.2,0.2395008087158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,16,power_law_1.2,0.3043200016021729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,16,power_law_1.2,0.4721856117248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,128,power_law_1.2,0.0775871992111206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,128,power_law_1.2,0.07383679747581481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,128,power_law_1.2,0.07292799949645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,128,power_law_1.2,0.06926720142364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,128,power_law_1.2,0.07674239873886109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,128,power_law_1.2,0.0803264021873474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,128,power_law_1.2,0.07901440262794494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,128,power_law_1.2,0.07994239926338195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,128,power_law_1.2,0.08231679797172546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,128,power_law_1.2,0.08012160062789916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,128,power_law_1.2,0.08179839849472045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,128,power_law_1.2,0.08803840279579163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,128,power_law_1.2,0.0889024019241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,128,power_law_1.2,0.09386240243911743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,128,power_law_1.2,0.10032639503479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,128,power_law_1.2,0.10275839567184449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,128,power_law_1.2,0.10933760404586793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,128,power_law_1.2,0.1325376033782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,128,power_law_1.2,0.14542720317840577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,128,power_law_1.2,0.17119359970092773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,128,power_law_1.2,0.19303679466247559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,128,power_law_1.2,0.26099839210510256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,128,power_law_1.2,0.2612031936645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,128,power_law_1.2,0.36175999641418455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,128,power_law_1.2,0.46541438102722166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,16,power_law_1.2,0.5944767951965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,128,power_law_1.2,0.6511295795440674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,128,power_law_1.2,0.8059328079223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,32,power_law_1.2,0.0723583996295929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,32,power_law_1.2,0.08299520015716552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,32,power_law_1.2,0.07145599722862243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,32,power_law_1.2,0.07783039808273315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,32,power_law_1.2,0.07160959839820862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,32,power_law_1.2,0.06819199919700622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,32,power_law_1.2,0.0711679995059967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,32,power_law_1.2,0.07290239930152893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,32,power_law_1.2,0.0726207971572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,32,power_law_1.2,0.07656959891319275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,32,power_law_1.2,0.0807807981967926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,32,power_law_1.2,0.07989119887351989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,32,power_law_1.2,0.08306559920310974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,32,power_law_1.2,0.08386560082435608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,32,power_law_1.2,0.09773439764976502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,32,power_law_1.2,0.09878399968147278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,32,power_law_1.2,0.0970687985420227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,32,power_law_1.2,0.11046400070190429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,32,power_law_1.2,0.11100159883499146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,32,power_law_1.2,0.1381183981895447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,32,power_law_1.2,0.1456447958946228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,32,power_law_1.2,0.1884160041809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,32,power_law_1.2,0.21310079097747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,32,power_law_1.2,0.27937281131744385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,32,power_law_1.2,0.3327296018600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,32,power_law_1.2,0.5097472190856933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,32,power_law_1.2,0.5979008197784423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,8,balanced,0.03284800052642822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,8,balanced,0.03162133445342382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,8,balanced,0.03148266673088074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,8,balanced,0.033610666791598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,8,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,8,balanced,0.03521066655715307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,8,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,8,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,8,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,8,balanced,0.0355679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,8,balanced,0.037178667883078255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,8,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,8,balanced,0.03782933453718821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,8,balanced,0.03765333443880081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,8,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,8,balanced,0.045968001087506614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,8,balanced,0.04378133515516917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,8,balanced,0.05604266623655955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,8,balanced,0.052330667773882546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,8,balanced,0.07167999943097432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,8,balanced,0.08272533118724823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,8,balanced,0.1058026651541392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,8,balanced,0.12352533141771953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,8,balanced,0.16246400276819864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,8,balanced,0.2039733330408732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,8,balanced,0.2767733335494995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,8,balanced,0.34039465586344403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,8,power_law_1.01,0.04779520034790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,8,power_law_1.01,0.05028480291366577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,8,power_law_1.01,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,8,power_law_1.01,0.0580672025680542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,8,power_law_1.01,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,8,power_law_1.01,0.06366080045700073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,8,power_law_1.01,0.06424959897994995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,8,power_law_1.01,0.06367999911308289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,8,power_law_1.01,0.0685696005821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,8,power_law_1.01,0.06644480228424073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,8,power_law_1.01,0.06910719871520996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,8,power_law_1.01,0.0711296021938324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,8,power_law_1.01,0.0717248022556305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,8,power_law_1.01,0.07499520182609558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,8,power_law_1.01,0.08167679905891419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,8,power_law_1.01,0.08575999736785889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,8,power_law_1.01,0.0913536012172699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,8,power_law_1.01,0.10440959930419921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,8,power_law_1.01,0.11091840267181396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,8,power_law_1.01,0.13994879722595216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,8,power_law_1.01,0.15500160455703735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,8,power_law_1.01,0.21592319011688232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,8,power_law_1.01,0.25213439464569093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,8,power_law_1.01,0.33431038856506345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,8,power_law_1.01,0.4361279964447021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,8,power_law_1.01,0.6470399856567383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,8,power_law_1.01,0.8010623931884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,4,power_law_1.2,0.07045120000839233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,4,power_law_1.2,0.07878400087356567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,4,power_law_1.2,0.09169279932975768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,4,power_law_1.2,0.11103999614715576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,4,power_law_1.2,0.10559359788894654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,4,power_law_1.2,0.14414080381393432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,4,power_law_1.2,0.1515071988105774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,4,power_law_1.2,0.14750080108642577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,4,power_law_1.2,0.15443840026855468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,4,power_law_1.2,0.15356160402297975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,4,power_law_1.2,0.15751680135726928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,4,power_law_1.2,0.15809279680252075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,4,power_law_1.2,0.15767040252685546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,4,power_law_1.2,0.1664896011352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,4,power_law_1.2,0.1717695951461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,4,power_law_1.2,0.17256319522857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,4,power_law_1.2,0.17564799785614013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,4,power_law_1.2,0.2032831907272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,4,power_law_1.2,0.21880319118499755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,4,power_law_1.2,0.26746881008148193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,4,power_law_1.2,0.3252608060836792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,4,power_law_1.2,0.40176639556884763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,4,power_law_1.2,0.5559872150421142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,4,power_law_1.2,0.7316736221313477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,4,power_law_1.2,1.1105279922485352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,16,power_law_1.01,0.03697279989719391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,4,power_law_1.2,1.447059154510498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,16,power_law_1.01,0.03664000034332275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,16,power_law_1.01,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,16,power_law_1.01,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,16,power_law_1.01,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,16,power_law_1.01,0.03635199964046478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,16,power_law_1.01,0.03589119911193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,16,power_law_1.01,0.03591679930686951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,16,power_law_1.01,0.03672960102558136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,16,power_law_1.01,0.037222400307655334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,16,power_law_1.01,0.03940480053424835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,16,power_law_1.01,0.04121600091457367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,16,power_law_1.01,0.041920000314712526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,16,power_law_1.01,0.048511999845504764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,16,power_law_1.01,0.05583999752998352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,16,power_law_1.01,0.05632640123367309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,16,power_law_1.01,0.0542464017868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,16,power_law_1.01,0.06251519918441772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,16,power_law_1.01,0.06958720088005066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,16,power_law_1.01,0.07566720247268677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,16,power_law_1.01,0.08860160112380981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,16,power_law_1.01,0.1254271984100342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,16,power_law_1.01,0.12532479763031007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,16,power_law_1.01,0.15851520299911498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,16,power_law_1.01,0.19112319946289064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,16,power_law_1.01,0.3245824098587036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,16,power_law_1.01,0.37470719814300535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,64,power_law_1.01,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,64,power_law_1.01,0.03194240033626557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,64,power_law_1.01,0.03194240033626557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,64,power_law_1.01,0.03407360017299652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,64,power_law_1.01,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,64,power_law_1.01,0.03661440014839172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,64,power_law_1.01,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,64,power_law_1.01,0.0378495991230011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,64,power_law_1.01,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,64,power_law_1.01,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,64,power_law_1.01,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,64,power_law_1.01,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,64,power_law_1.01,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,64,power_law_1.01,0.046214398741722104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,64,power_law_1.01,0.05466880202293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,64,power_law_1.01,0.052908802032470705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,64,power_law_1.01,0.05694720149040222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,64,power_law_1.01,0.06750079989433289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,64,power_law_1.01,0.08012160062789916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,64,power_law_1.01,0.09233919978141784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,64,power_law_1.01,0.11866879463195801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,64,power_law_1.01,0.16323839426040648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,64,power_law_1.01,0.21150081157684325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,64,power_law_1.01,0.2995327949523926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,64,power_law_1.01,0.3463360071182251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,64,power_law_1.01,0.5257791996002197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,64,power_law_1.01,0.8364735603332519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,4,power_law_1.2,1.9213247299194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,32,power_law_1.01,0.03269760012626648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,32,power_law_1.01,0.03052160143852234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,32,power_law_1.01,0.031251201033592226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,32,power_law_1.01,0.03128960132598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,32,power_law_1.01,0.03153280019760132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,32,power_law_1.01,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,32,power_law_1.01,0.033036801218986514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,32,power_law_1.01,0.034067198634147644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,32,power_law_1.01,0.034745600819587705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,32,power_law_1.01,0.03470720052719116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,32,power_law_1.01,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,32,power_law_1.01,0.037036800384521486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,32,power_law_1.01,0.037785598635673524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,32,power_law_1.01,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,32,power_law_1.01,0.04325760006904602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,32,power_law_1.01,0.04352000057697296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,32,power_law_1.01,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,32,power_law_1.01,0.05006719827651977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,32,power_law_1.01,0.05799040198326111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,32,power_law_1.01,0.06344320178031922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,32,power_law_1.01,0.08259199857711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,32,power_law_1.01,0.1091264009475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,32,power_law_1.01,0.12872320413589478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,32,power_law_1.01,0.1873088002204895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,32,power_law_1.01,0.23728640079498292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,32,power_law_1.01,0.3463360071182251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,32,power_law_1.01,0.43739519119262693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,8,balanced,0.0562720000743866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,8,balanced,0.058005332946777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,8,balanced,0.06150400141874949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,8,balanced,0.07229333122571309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,8,balanced,0.08854933579762776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,8,balanced,0.11829866965611775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,8,balanced,0.11843199531237285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,8,balanced,0.11819199721018474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,8,balanced,0.12113066514333089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,8,balanced,0.11872000495592754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,8,balanced,0.12082133690516154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,8,balanced,0.12276800473531087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,8,balanced,0.1221440037091573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,8,balanced,0.12593066692352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,8,balanced,0.13260799646377563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,8,balanced,0.13436266779899597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,8,balanced,0.1388800044854482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,8,balanced,0.15850667158762613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,8,balanced,0.1649066706498464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,8,balanced,0.17907732725143433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,8,balanced,0.2044693430264791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,8,balanced,0.24027733008066812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,8,balanced,0.28057066599527997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,8,balanced,0.3926933209101359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,8,balanced,0.47629865010579425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,8,balanced,0.6781546274820963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,8,balanced,0.8782613277435303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,16,power_law_1.2,0.045158401131629944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,16,power_law_1.2,0.03240320086479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,16,power_law_1.2,0.03123840093612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,16,power_law_1.2,0.033164799213409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,16,power_law_1.2,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,16,power_law_1.2,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,16,power_law_1.2,0.033036801218986514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,16,power_law_1.2,0.0332863986492157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,16,power_law_1.2,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,16,power_law_1.2,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,16,power_law_1.2,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,16,power_law_1.2,0.03704319894313812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,16,power_law_1.2,0.037478399276733396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,16,power_law_1.2,0.03985919952392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,16,power_law_1.2,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,16,power_law_1.2,0.04679679870605469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,16,power_law_1.2,0.048902401328086854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,16,power_law_1.2,0.05095679759979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,16,power_law_1.2,0.05304319858551025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,16,power_law_1.2,0.06618880033493042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,16,power_law_1.2,0.0755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,16,power_law_1.2,0.09565439820289612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,16,power_law_1.2,0.11756800413131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,16,power_law_1.2,0.14792959690093993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,16,power_law_1.2,0.1878592014312744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,16,power_law_1.2,0.2639807939529419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,16,power_law_1.2,0.34707839488983155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,16,power_law_1.2,0.038790398836135866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,16,power_law_1.2,0.03589119911193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,16,power_law_1.2,0.0347135990858078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,16,power_law_1.2,0.03708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,16,power_law_1.2,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,16,power_law_1.2,0.03918719887733459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,16,power_law_1.2,0.03902080059051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,16,power_law_1.2,0.03978880047798157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,16,power_law_1.2,0.03923200070858002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,16,power_law_1.2,0.04002560079097748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,16,power_law_1.2,0.040249601006507874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,16,power_law_1.2,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,16,power_law_1.2,0.04392960071563721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,16,power_law_1.2,0.045664000511169436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,16,power_law_1.2,0.051744002103805545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,16,power_law_1.2,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,16,power_law_1.2,0.061260801553726194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,16,power_law_1.2,0.0720575988292694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,16,power_law_1.2,0.08484479784965515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,16,power_law_1.2,0.10497280359268188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,1,power_law_1.01,0.02739199995994568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,16,power_law_1.2,0.12681599855422973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,1,power_law_1.01,0.046726399660110475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,1,power_law_1.01,0.06530560255050659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,16,power_law_1.2,0.16892160177230836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,1,power_law_1.01,0.08503040075302123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,16,power_law_1.2,0.21489279270172118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,1,power_law_1.01,0.11697280406951904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,1,power_law_1.01,0.16651519536972045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,1,power_law_1.01,0.17523839473724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,1,power_law_1.01,0.1808832049369812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,1,power_law_1.01,0.18807040452957152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,1,power_law_1.01,0.19064320325851442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,1,power_law_1.01,0.20407040119171144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,1,power_law_1.01,0.19888639450073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,1,power_law_1.01,0.206496000289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,16,power_law_1.2,0.33123838901519775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,1,power_law_1.01,0.215667200088501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,1,power_law_1.01,0.21215999126434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,1,power_law_1.01,0.21408638954162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,1,power_law_1.01,0.23324799537658691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,1,power_law_1.01,0.24427518844604493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.01,0.26508800983428954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.01,0.6220672130584717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.01,0.7083583831787109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.01,0.42266240119934084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.01,0.49215359687805177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,16,power_law_1.2,0.42768001556396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.01,0.6006847858428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.01,0.7391551971435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.01,1.0046015739440919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.01,1.2709055900573731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,8,power_law_1.01,0.04622719883918762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,8,power_law_1.01,0.051046401262283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,8,power_law_1.01,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,8,power_law_1.01,0.060288000106811526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,8,power_law_1.01,0.06631680130958557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,8,power_law_1.01,0.07512959837913513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,8,power_law_1.01,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,8,power_law_1.01,0.07575039863586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,8,power_law_1.01,0.07729920148849487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,8,power_law_1.01,0.08049280047416688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,8,power_law_1.01,0.08036479949951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,8,power_law_1.01,0.0813759982585907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,8,power_law_1.01,0.07968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,8,power_law_1.01,0.08669440150260925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,8,power_law_1.01,0.09150720238685608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,8,power_law_1.01,0.09356160163879394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,8,power_law_1.01,0.1054144024848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,8,power_law_1.01,0.11564160585403442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,8,power_law_1.01,0.124236798286438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,8,power_law_1.01,0.15434880256652833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,8,power_law_1.01,0.18072960376739503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,16,power_law_1.2,0.09283840060234069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,16,power_law_1.2,0.12066559791564942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,16,power_law_1.2,0.11281280517578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,16,power_law_1.2,0.12047359943389893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,16,power_law_1.2,0.12186239957809449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,16,power_law_1.2,0.08986240029335021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,16,power_law_1.2,0.1229248046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,16,power_law_1.2,0.11552000045776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,16,power_law_1.2,0.11557120084762573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,16,power_law_1.2,0.1150015950202942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,16,power_law_1.2,0.11498240232467652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,8,power_law_1.01,0.22969601154327393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,16,power_law_1.2,0.11134719848632812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,8,power_law_1.01,0.2913151979446411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,16,power_law_1.2,0.1190335988998413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,16,power_law_1.2,0.6728064060211182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,16,power_law_1.2,0.11807359457015991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,16,power_law_1.2,0.12480640411376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,16,power_law_1.2,0.1196671962738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,16,power_law_1.2,0.13084160089492797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,16,power_law_1.2,0.14351999759674072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,16,power_law_1.2,0.16087039709091186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,16,power_law_1.2,0.17982720136642455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,8,power_law_1.01,0.38536319732666013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,16,power_law_1.2,0.20142080783843994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,8,power_law_1.01,0.4967807769775391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,8,power_law_1.01,0.7141823768615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,8,power_law_1.01,0.8673151969909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,16,power_law_1.2,0.2500672101974487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,16,power_law_1.2,0.3324287891387939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,16,power_law_1.2,0.43418240547180176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,16,power_law_1.2,0.5242879867553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,16,power_law_1.2,0.8049856185913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,16,power_law_1.2,0.9064512252807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,16,power_law_1.2,0.9056320190429688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,32,power_law_1.01,0.15038080215454103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,32,power_law_1.01,0.14958720207214354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,32,power_law_1.01,0.1488319993019104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,32,power_law_1.01,0.15245440006256103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,32,power_law_1.01,0.14431999921798705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,32,power_law_1.01,0.14177279472351073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,8,balanced,0.0624533345301946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,32,power_law_1.01,0.15073920488357545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,8,balanced,0.06225066880385081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,32,power_law_1.01,0.15024640560150146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,8,balanced,0.06243200103441874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,8,balanced,0.06534933547178905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,32,power_law_1.01,0.1496448040008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,8,balanced,0.068271999557813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,8,balanced,0.0674720009167989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,32,power_law_1.01,0.15466879606246947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,32,power_law_1.01,0.15381759405136108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,32,power_law_1.01,0.15699199438095093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,32,power_law_1.01,0.1575999975204468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,8,balanced,0.06733866532643636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,8,balanced,0.06809600194295247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,8,balanced,0.07044800122578938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,8,balanced,0.0702453354994456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,8,balanced,0.07049066821734111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,8,balanced,0.07022933165232341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,8,balanced,0.07205333312352498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,8,balanced,0.07218666871388753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,8,balanced,0.08066666622956593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,8,balanced,0.08273600041866302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,8,balanced,0.08668266733487447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,8,balanced,0.09778133034706116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,8,balanced,0.09886933366457622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,8,balanced,0.12779733538627625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,8,balanced,0.1434879998366038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,32,power_law_1.01,0.1637760043144226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,8,balanced,0.19342400630315146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,8,balanced,0.22557334105173746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,8,balanced,0.30728532870610553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,8,balanced,0.3798826535542806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,8,balanced,0.5426293214162191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,8,balanced,0.7006879647572836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,32,power_law_1.01,0.1704256057739258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,32,power_law_1.01,0.17698559761047364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,32,power_law_1.01,0.19175039529800414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,32,power_law_1.01,0.22287359237670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,32,power_law_1.01,0.22199039459228515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,32,power_law_1.01,0.2799936056137085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,32,power_law_1.01,0.31324799060821534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,32,power_law_1.01,0.4024640083312988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,32,power_law_1.01,0.3997632026672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,32,power_law_1.01,0.49799041748046874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,32,power_law_1.01,0.6496895790100098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,32,power_law_1.01,0.8430720329284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,32,power_law_1.01,1.0066495895385743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,1,power_law_1.2,0.07556480169296265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,1,power_law_1.2,0.09779199957847595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,1,power_law_1.2,0.14979840517044068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,1,power_law_1.2,0.20575358867645263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,1,power_law_1.2,0.2684031963348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,1,power_law_1.2,0.46552319526672364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,1,power_law_1.2,0.49580159187316897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,1,power_law_1.2,0.5153791904449463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,1,power_law_1.2,0.522380781173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,1,power_law_1.2,0.5441408157348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,1,power_law_1.2,0.5567615985870361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,1,power_law_1.2,0.5821375846862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,1,power_law_1.2,0.5947968006134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,1,power_law_1.2,0.5972224235534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,1,power_law_1.2,0.6244863986968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,1,power_law_1.2,0.6534336090087891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,1,power_law_1.2,0.6919616222381592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,1,power_law_1.2,0.7819583892822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,1,power_law_1.2,0.7963456153869629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,1,power_law_1.2,0.9500351905822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,1,power_law_1.2,1.0774592399597167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,1,power_law_1.2,1.3403776168823243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,1,power_law_1.2,1.4821632385253907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,1,power_law_1.2,1.9090431213378907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,1,power_law_1.2,2.6764095306396483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,1,power_law_1.2,3.531148910522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,1,power_law_1.2,4.376729583740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,32,power_law_1.01,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,32,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,32,power_law_1.01,0.03269760012626648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,32,power_law_1.01,0.03578880131244659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,32,power_law_1.01,0.03505280017852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,32,power_law_1.01,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,32,power_law_1.01,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,32,power_law_1.01,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,32,power_law_1.01,0.036748799681663516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,32,power_law_1.01,0.037241598963737486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,32,power_law_1.01,0.03856000006198883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,32,power_law_1.01,0.03935360014438629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,32,power_law_1.01,0.03954559862613678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,32,power_law_1.01,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,32,power_law_1.01,0.045875200629234315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,32,power_law_1.01,0.046988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,32,power_law_1.01,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,32,power_law_1.01,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,32,power_law_1.01,0.054092800617218016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,32,power_law_1.01,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,32,power_law_1.01,0.07198079824447631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,32,power_law_1.01,0.09018239974975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,32,power_law_1.01,0.10849920511245728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,32,power_law_1.01,0.1385472059249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,32,power_law_1.01,0.20055680274963378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,32,power_law_1.01,0.2571007966995239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,32,power_law_1.01,0.33343360424041746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,4,power_law_1.2,0.02683520019054413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,4,power_law_1.2,0.025196799635887147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,4,power_law_1.2,0.02481919974088669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,4,power_law_1.2,0.02964479923248291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,4,power_law_1.2,0.03086079955101013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,4,power_law_1.2,0.03276160061359405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,4,power_law_1.2,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,4,power_law_1.2,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,4,power_law_1.2,0.08120319843292237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,4,power_law_1.2,0.08050559759140015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,4,power_law_1.2,0.08001279830932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,4,power_law_1.2,0.08449919819831848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,4,power_law_1.2,0.08001919984817504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,4,power_law_1.2,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,4,power_law_1.2,0.0880128026008606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,4,power_law_1.2,0.07340160012245178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,4,power_law_1.2,0.09409279823303222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,4,power_law_1.2,0.11743359565734864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.2,0.14214400053024293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.2,0.19984639883041383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.2,0.2549760103225708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.2,0.35817599296569824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.2,0.4684927940368652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,16,balanced,0.033333333830038704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.2,0.6745728015899658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,16,balanced,0.02956266701221466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,16,balanced,0.032255999743938446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,16,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,16,balanced,0.0553653339544932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,16,balanced,0.05774400134881338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,16,balanced,0.05713599920272827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,16,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,16,balanced,0.05736533304055532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,16,balanced,0.05810666580994924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,16,balanced,0.05773333211739858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,16,balanced,0.05895466605822245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,16,balanced,0.05789333085219065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,16,balanced,0.060677334666252136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,16,balanced,0.06487999856472015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,16,balanced,0.06460799773534139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,16,balanced,0.06596800188223521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,16,balanced,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,16,balanced,0.07041066884994507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,16,balanced,0.07462933162848155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,16,balanced,0.07817066709200542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,16,balanced,0.09703466296195984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,16,balanced,0.10585600137710571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,16,balanced,0.12713066736857095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,16,balanced,0.15679466724395752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,16,balanced,0.1982133388519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,16,balanced,0.2313493291536967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,8,power_law_1.01,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,8,power_law_1.01,0.048537600040435794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,8,power_law_1.01,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,8,power_law_1.01,0.05367680191993714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,8,power_law_1.01,0.057183998823165896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.2,0.8879167556762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,8,power_law_1.01,0.058246397972106935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,8,power_law_1.01,0.05845119953155518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,8,power_law_1.01,0.06111999750137329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,8,power_law_1.01,0.06172159910202026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,8,power_law_1.01,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,8,power_law_1.01,0.06549760103225707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,8,power_law_1.01,0.06940799951553345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,8,power_law_1.01,0.07152640223503112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,8,power_law_1.01,0.07700480222702026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,8,power_law_1.01,0.0843455970287323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,8,power_law_1.01,0.08661119937896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,8,power_law_1.01,0.08869760036468506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,8,power_law_1.01,0.10902400016784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,8,power_law_1.01,0.11850880384445191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,8,power_law_1.01,0.14477440118789672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,8,power_law_1.01,0.17697919607162477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,8,power_law_1.01,0.2302016019821167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,8,power_law_1.01,0.269324803352356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,8,power_law_1.01,0.38111999034881594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,8,power_law_1.01,0.4876863956451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,8,power_law_1.01,0.740499210357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,8,power_law_1.01,0.8907584190368653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,8,balanced,0.05677333474159241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,8,balanced,0.056234667698542275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,8,balanced,0.05815466741720835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,8,balanced,0.06631466746330261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,8,balanced,0.07701866825421651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,8,balanced,0.09790399670600891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,8,balanced,0.09708799918492635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,8,balanced,0.09630399942398071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,8,balanced,0.10002133250236511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,8,balanced,0.09698133667310078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,8,balanced,0.09860266248385112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,8,balanced,0.10285866260528564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,8,balanced,0.10264000296592712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,8,balanced,0.10526399811108907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,8,balanced,0.11248000462849934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,8,balanced,0.1143839955329895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,8,balanced,0.11784000198046367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,8,balanced,0.12947733203570047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,8,balanced,0.13378133376439413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,8,balanced,0.15471999843915304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.2,1.3060544013977051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,8,balanced,0.17471466461817423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,8,balanced,0.2100693384806315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,8,balanced,0.2508053382237752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,8,balanced,0.3648800055185954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,8,balanced,0.44784533977508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,8,balanced,0.6331679821014404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,8,balanced,0.8262773354848226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,1,power_law_1.2,0.019852800667285918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,1,power_law_1.2,0.02094080001115799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,1,power_law_1.2,0.02513279914855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,1,power_law_1.2,0.031200000643730165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,1,power_law_1.01,0.1631872057914734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,1,power_law_1.2,0.04588159918785095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,1,power_law_1.01,0.20619521141052247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,1,power_law_1.2,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,1,power_law_1.01,0.2798336029052734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,1,power_law_1.2,0.07332479953765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,1,power_law_1.2,0.07952640056610108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,1,power_law_1.01,0.4193920135498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,1,power_law_1.2,0.08217599987983704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,1,power_law_1.2,0.08151040077209473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,1,power_law_1.2,0.08516479730606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,1,power_law_1.01,0.5627520084381104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,1,power_law_1.2,0.08973439931869506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,1,power_law_1.2,0.09537919759750366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,1,power_law_1.01,0.9067008018493652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,1,power_law_1.2,0.09850239753723145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,1,power_law_1.2,0.10044159889221191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,1,power_law_1.2,0.10447360277175903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,1,power_law_1.2,0.1124351978302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,1,power_law_1.2,0.13066879510879517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,1,power_law_1.2,0.144595205783844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,1,power_law_1.2,0.17785600423812867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,1,power_law_1.2,0.20618879795074463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,1,power_law_1.01,0.9597760200500488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,1,power_law_1.2,0.27255680561065676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,1,power_law_1.01,0.9800831794738769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,1,power_law_1.01,1.020479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,1,power_law_1.01,1.0500864028930663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,1,power_law_1.01,1.0663743972778321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,1,power_law_1.2,0.2995199918746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,1,power_law_1.2,0.39818239212036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,1,power_law_1.2,0.5299327850341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,1,power_law_1.2,0.7859392166137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,1,power_law_1.01,1.0920703887939454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,1,power_law_1.2,0.9034303665161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,1,power_law_1.01,1.128211212158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,1,power_law_1.01,1.1493120193481445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,1,power_law_1.01,1.1921600341796874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,1,power_law_1.01,1.2220864295959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,1,power_law_1.01,1.2998016357421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,1,power_law_1.01,1.4364864349365234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.01,1.585747241973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.01,1.930361557006836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.01,2.0812864303588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.01,2.626367950439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.01,3.204172897338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.01,4.275788879394531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.01,5.3341632843017575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,1,power_law_1.01,7.551443481445313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,1,power_law_1.01,9.681708526611327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.2,1.7218240737915038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,8,power_law_1.01,0.026368001103401185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,8,power_law_1.01,0.023257599771022798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,8,power_law_1.01,0.02258560061454773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,8,power_law_1.01,0.024108800292015075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,8,power_law_1.01,0.026073598861694337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,8,power_law_1.01,0.027603200078010558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,8,power_law_1.01,0.02768000066280365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,8,power_law_1.01,0.027673599123954774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,8,power_law_1.01,0.031948798894882204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,8,power_law_1.01,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,8,power_law_1.01,0.03244799971580505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,8,power_law_1.01,0.03485440015792847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,8,power_law_1.01,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,8,power_law_1.01,0.03596160113811493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,8,power_law_1.01,0.03993600010871887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,8,power_law_1.01,0.04042240083217621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,8,power_law_1.01,0.044582399725914004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,8,power_law_1.01,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,8,power_law_1.01,0.05550720095634461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,8,power_law_1.01,0.0679423987865448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,8,power_law_1.01,0.07145599722862243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,8,power_law_1.01,0.07823359966278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,8,power_law_1.01,0.10448640584945679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,8,power_law_1.01,0.1251456022262573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,8,power_law_1.01,0.16888320446014404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,8,power_law_1.01,0.23214080333709716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,8,power_law_1.01,0.2914112091064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,2,power_law_1.2,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,2,power_law_1.2,0.06152960062026978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,2,power_law_1.2,0.07988479733467102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,2,power_law_1.2,0.10456960201263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,2,power_law_1.2,0.1224511981010437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,2,power_law_1.2,0.17790080308914186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,2,power_law_1.2,0.19258879423141478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,2,power_law_1.2,0.1960319995880127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,2,power_law_1.2,0.20624639987945556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,2,power_law_1.2,0.20628480911254882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,2,power_law_1.2,0.21186559200286864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,2,power_law_1.2,0.21790080070495604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,2,power_law_1.2,0.2155071973800659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,2,power_law_1.2,0.23043839931488036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,2,power_law_1.2,0.23676159381866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,2,power_law_1.2,0.2510335922241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,2,power_law_1.2,0.2576256036758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,2,power_law_1.2,0.28652799129486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,2,power_law_1.2,0.316595196723938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,2,power_law_1.2,0.35871999263763427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,2,power_law_1.2,0.40159358978271487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,2,power_law_1.2,0.5668799877166748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,2,power_law_1.2,0.6370175838470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,2,power_law_1.2,0.8658368110656738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,2,power_law_1.2,1.0422847747802735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,2,power_law_1.2,1.5524800300598145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,8,power_law_1.01,0.043609601259231565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,8,power_law_1.01,0.04424319863319397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,8,power_law_1.01,0.04213759899139404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,8,power_law_1.01,0.04528000056743622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,8,power_law_1.01,0.04505600035190582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,8,power_law_1.01,0.04827519953250885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,8,power_law_1.01,0.049619200825691226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,8,power_law_1.01,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,8,power_law_1.01,0.048742398619651794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,8,power_law_1.01,0.05009920001029968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,8,power_law_1.01,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,8,power_law_1.01,0.05093119740486145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,8,power_law_1.01,0.050521600246429446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,8,power_law_1.01,0.0514240026473999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,8,power_law_1.01,0.05594239830970764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,8,power_law_1.01,0.05797119736671448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,8,power_law_1.01,0.060166400671005246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,8,power_law_1.01,0.0640064001083374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,8,power_law_1.01,0.06933760046958923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,8,power_law_1.01,0.08209919929504395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,8,power_law_1.01,0.09253119826316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,8,power_law_1.01,0.1218176007270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,8,power_law_1.01,0.1509055972099304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,8,power_law_1.01,0.20016639232635497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,8,power_law_1.01,0.23621759414672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,8,power_law_1.01,0.37572479248046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,8,power_law_1.01,0.4505152225494385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,8,power_law_1.01,0.020326399803161622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,8,power_law_1.01,0.02447360008955002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,8,power_law_1.01,0.02691200077533722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,8,power_law_1.01,0.02879360020160675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,8,power_law_1.01,0.03086079955101013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,8,power_law_1.01,0.03216640055179596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,8,power_law_1.01,0.03619840145111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,8,power_law_1.01,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,8,power_law_1.01,0.04403199851512909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,8,power_law_1.01,0.045561599731445315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,8,power_law_1.01,0.043699198961257936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,8,power_law_1.01,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,8,power_law_1.01,0.05748479962348938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,8,power_law_1.01,0.05592319965362549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,8,power_law_1.01,0.07208319902420043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,8,power_law_1.01,0.07133439779281617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,8,power_law_1.01,0.07299839854240417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,8,power_law_1.01,0.08807039856910706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,8,power_law_1.01,0.08247680068016053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,8,power_law_1.01,0.08710399866104127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,8,power_law_1.01,0.08309760093688964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,8,power_law_1.01,0.10109440088272095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,8,power_law_1.01,0.1118399977684021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,8,power_law_1.01,0.14469759464263915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,8,power_law_1.01,0.1739583969116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,8,power_law_1.01,0.24289278984069823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,8,power_law_1.01,0.30154240131378174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,2,power_law_1.2,1.915443229675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,64,power_law_1.01,0.04769279956817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,64,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,64,power_law_1.01,0.043961599469184875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,64,power_law_1.01,0.04642559885978699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,64,power_law_1.01,0.04649600088596344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,64,power_law_1.01,0.04689280092716217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,64,power_law_1.01,0.047865599393844604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,64,power_law_1.01,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,64,power_law_1.01,0.04903680086135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,64,power_law_1.01,0.04965760111808777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,64,power_law_1.01,0.05004159808158874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,64,power_law_1.01,0.05125759840011597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,64,power_law_1.01,0.05243520140647888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,64,power_law_1.01,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,64,power_law_1.01,0.05727360248565674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,64,power_law_1.01,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,64,power_law_1.01,0.060524797439575194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,64,power_law_1.01,0.06743680238723755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,64,power_law_1.01,0.07228800058364868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,64,power_law_1.01,0.0843392014503479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,64,power_law_1.01,0.09070720076560974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,64,power_law_1.01,0.11406079530715943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,64,power_law_1.01,0.1336192011833191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,64,power_law_1.01,0.185862398147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,64,power_law_1.01,0.22257280349731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,64,power_law_1.01,0.3425600051879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,64,power_law_1.01,0.41108479499816897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,2,power_law_1.01,0.0692031979560852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,2,power_law_1.01,0.07556480169296265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,2,power_law_1.01,0.0837440013885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,2,power_law_1.01,0.09585919976234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,2,power_law_1.01,0.1108672022819519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,2,power_law_1.01,0.11687040328979492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,2,power_law_1.01,0.1367616057395935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,2,power_law_1.01,0.14044159650802612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,2,power_law_1.01,0.14679039716720582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,2,power_law_1.01,0.15089279413223267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,2,power_law_1.01,0.15162880420684816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,2,power_law_1.01,0.153331196308136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,2,power_law_1.01,0.15319039821624755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,2,power_law_1.01,0.1620736002922058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,2,power_law_1.01,0.1664576053619385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,2,power_law_1.01,0.16844160556793214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,2,power_law_1.01,0.17704960107803344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,2,power_law_1.01,0.19435520172119142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,2,power_law_1.01,0.2187135934829712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,2,power_law_1.01,0.26098558902740476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,2,power_law_1.01,0.2960959911346436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,2,power_law_1.01,0.3875904083251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,2,power_law_1.01,0.47312002182006835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,2,power_law_1.01,0.6454592227935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,2,power_law_1.01,0.8325759887695312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,2,power_law_1.01,1.1483776092529296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,2,power_law_1.01,1.436070442199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,1,power_law_1.2,0.02083200067281723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,1,power_law_1.2,0.022886399924755097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,1,power_law_1.2,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,1,power_law_1.2,0.04656000137329101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,1,power_law_1.2,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,1,power_law_1.2,0.08800640106201171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,1,power_law_1.2,0.10709760189056397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,1,power_law_1.2,0.11519360542297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,1,power_law_1.2,0.11882879734039306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,1,power_law_1.2,0.12089600563049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,1,power_law_1.2,0.12737280130386353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,1,power_law_1.2,0.13304959535598754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,1,power_law_1.2,0.13648639917373656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,1,power_law_1.2,0.1373695969581604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,1,power_law_1.2,0.14789119958877564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,1,power_law_1.2,0.14752639532089235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,1,power_law_1.2,0.15728000402450562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,1,power_law_1.2,0.1880959987640381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.2,0.19911680221557618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.2,0.2442239999771118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.2,0.29134719371795653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.2,0.3741055965423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.2,0.3884160041809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.2,0.5366015911102295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.2,0.6513728141784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.2,0.9435647964477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.2,1.213926410675049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,32,power_law_1.2,0.049267199635505673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,32,power_law_1.2,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,32,power_law_1.2,0.04533120095729828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,32,power_law_1.2,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,32,power_law_1.2,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,32,power_law_1.2,0.04801279902458191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,32,power_law_1.2,0.04959999918937683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,32,power_law_1.2,0.05113599896430969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,32,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,32,power_law_1.2,0.05305600166320801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,32,power_law_1.2,0.05440639853477478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,32,power_law_1.2,0.05516800284385681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,32,power_law_1.2,0.057196801900863646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,32,power_law_1.2,0.05926399827003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,32,power_law_1.2,0.06394240260124207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,32,power_law_1.2,0.06556159853935242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,32,power_law_1.2,0.06883199810981751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,32,power_law_1.2,0.07811840176582337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,32,power_law_1.2,0.093094402551651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,32,power_law_1.2,0.11768319606781005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,32,power_law_1.2,0.14492160081863403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,32,power_law_1.2,0.18787839412689208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,32,power_law_1.2,0.22534399032592772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,32,2,power_law_1.01,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,32,2,power_law_1.01,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,32,2,power_law_1.01,0.034815999865531924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,32,2,power_law_1.01,0.03790720105171204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,32,power_law_1.2,0.3190079927444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,32,2,power_law_1.01,0.03871360123157501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,32,2,power_law_1.01,0.03911679983139038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,32,power_law_1.2,0.4119679927825928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,32,power_law_1.2,0.5679359912872315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,32,2,power_law_1.01,0.03982079923152924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,32,2,power_law_1.01,0.039699199795722964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,32,2,power_law_1.01,0.0403328001499176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,32,2,power_law_1.01,0.04023680090904236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,32,2,power_law_1.01,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,32,2,power_law_1.01,0.04568960070610047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,32,2,power_law_1.01,0.05028480291366577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,32,2,power_law_1.01,0.05368319749832153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,32,2,power_law_1.01,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,32,2,power_law_1.01,0.06087039709091187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,32,2,power_law_1.01,0.06666880249977111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,32,2,power_law_1.01,0.07887359857559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,32,2,power_law_1.01,0.08869119882583618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,32,power_law_1.2,0.8762432098388672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,32,2,power_law_1.01,0.12229119539260865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,32,2,power_law_1.01,0.1432255983352661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,32,2,power_law_1.01,0.19913599491119385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,32,2,power_law_1.01,0.2856640100479126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,32,2,power_law_1.01,0.3576895952224731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,32,2,power_law_1.01,0.4857664108276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,32,2,balanced,0.019248000035683315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,32,2,balanced,0.019050666441520054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,32,2,balanced,0.019071999937295914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,32,2,balanced,0.019002666076024372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,32,2,balanced,0.019082666685183842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,32,2,balanced,0.019189332922299702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,32,2,balanced,0.01995733380317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,32,2,balanced,0.022341333329677582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,32,2,balanced,0.0233599990606308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,32,2,balanced,0.02310933421055476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,32,2,balanced,0.023034666975339253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,32,2,balanced,0.039664000272750854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,32,2,balanced,0.041120000183582306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,32,2,balanced,0.04161066561937332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,32,2,balanced,0.033146666983763375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,32,2,balanced,0.032698666055997215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,32,2,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,32,2,balanced,0.039162665605545044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,32,2,balanced,0.03965866565704346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,32,2,balanced,0.047925333182017006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,32,2,balanced,0.05176533261934916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,32,2,balanced,0.07064533233642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,32,2,balanced,0.08143466711044312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,32,2,balanced,0.10930132865905762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,32,2,balanced,0.13555733362833658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,32,2,balanced,0.19049600760142008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,32,2,power_law_1.01,0.889465618133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,32,2,balanced,0.23778132597605386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,4,power_law_1.2,0.027987200021743774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,4,power_law_1.2,0.027744001150131224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,4,power_law_1.2,0.038624000549316403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,4,power_law_1.2,0.055027198791503903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,4,power_law_1.2,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,4,power_law_1.2,0.057843202352523805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,4,power_law_1.2,0.06791039705276489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,4,power_law_1.2,0.07882879972457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,4,power_law_1.2,0.08288000226020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,4,power_law_1.2,0.09712640047073365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,4,power_law_1.2,0.10261119604110717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,4,power_law_1.2,0.04469760060310364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,4,power_law_1.2,0.10056320428848267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,4,power_law_1.2,0.05624319911003113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,4,power_law_1.2,0.05687040090560913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,4,power_law_1.2,0.10224000215530396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,4,power_law_1.2,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,4,power_law_1.2,0.1029312014579773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,4,power_law_1.2,0.061587202548980716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,4,power_law_1.2,0.10325759649276733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,4,power_law_1.2,0.10832639932632446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,4,power_law_1.2,0.10802559852600098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,4,power_law_1.2,0.11228159666061402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,4,power_law_1.2,0.11690239906311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,4,power_law_1.2,0.12113920450210572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,4,power_law_1.2,0.06475520133972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,4,power_law_1.2,0.12718080282211303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,4,power_law_1.2,0.06405119895935059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,4,power_law_1.2,0.14883840084075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,4,power_law_1.2,0.06646400094032287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,4,power_law_1.2,0.06775040030479432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,4,power_law_1.2,0.16517759561538697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,4,power_law_1.2,0.07093120217323304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,4,power_law_1.2,0.1952191948890686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,4,power_law_1.2,0.07710719704627991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,4,power_law_1.2,0.08297600150108338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,4,power_law_1.2,0.2424384117126465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,4,power_law_1.2,0.33563520908355715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,2,power_law_1.2,0.04503040015697479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,2,power_law_1.2,0.06516479849815368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,4,power_law_1.2,0.08632320165634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,2,power_law_1.2,0.07079039812088013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,4,power_law_1.2,0.09630079865455628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,2,power_law_1.2,0.0833407998085022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.2,0.18038400411605834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,2,power_law_1.2,0.08528640270233154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.2,0.22274560928344728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,2,power_law_1.2,0.0871616005897522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.2,0.15674879550933837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,4,power_law_1.2,0.4264319896697998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.2,0.1865280032157898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.2,0.21591041088104249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,2,power_law_1.2,0.0899839997291565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,2,power_law_1.2,0.0911616027355194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.2,0.2973311901092529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,2,power_law_1.2,0.09059200286865235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,2,power_law_1.2,0.09619200229644775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,2,power_law_1.2,0.0967743992805481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.2,0.3533375978469849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,4,power_law_1.2,0.6139455795288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,2,power_law_1.2,0.1021888017654419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,2,power_law_1.2,0.1037824034690857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,2,power_law_1.2,0.1093824028968811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,2,power_law_1.2,0.11826560497283936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,32,2,power_law_1.01,1.1196479797363281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,2,power_law_1.2,0.13658879995346068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,2,power_law_1.2,0.13757439851760864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.2,0.49822077751159666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,4,power_law_1.2,0.872435188293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,2,power_law_1.2,0.16298240423202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,2,power_law_1.2,0.19548799991607665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,2,power_law_1.2,0.22772479057312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,8,power_law_1.2,0.04340479969978332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,8,power_law_1.2,0.042131200432777405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,8,power_law_1.2,0.04387840032577515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,8,power_law_1.2,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,8,power_law_1.2,0.043699198961257936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,2,power_law_1.2,0.27854719161987307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,8,power_law_1.2,0.049107199907302855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,8,power_law_1.2,0.04968959987163544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,8,power_law_1.2,0.048870399594306946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,8,power_law_1.2,0.04983679950237274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,8,power_law_1.2,0.05075200200080872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,8,power_law_1.2,0.05055999755859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,8,power_law_1.2,0.051596802473068235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,8,power_law_1.2,0.05208960175514221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,8,power_law_1.2,0.05314559936523437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,8,power_law_1.2,0.05790079832077026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,8,power_law_1.2,0.05809919834136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,8,power_law_1.2,0.06116480231285095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,8,power_law_1.2,0.06858239769935608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,8,power_law_1.2,0.07066239714622498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,4,power_law_1.2,1.0750911712646485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,8,power_law_1.2,0.088019198179245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,8,power_law_1.2,0.10103039741516114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,8,power_law_1.2,0.12853120565414428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,2,power_law_1.2,0.39017601013183595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,8,power_law_1.2,0.1536128044128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,8,power_law_1.01,0.03699199855327606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,8,power_law_1.2,0.21699841022491456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,8,power_law_1.01,0.034176000952720643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,8,power_law_1.01,0.03196159899234772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,8,power_law_1.01,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,8,power_law_1.01,0.03497599959373474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,8,power_law_1.01,0.03666560053825378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,8,power_law_1.01,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,2,power_law_1.2,0.43702402114868166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,8,power_law_1.01,0.03856000006198883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,8,power_law_1.01,0.039417600631713866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,8,power_law_1.01,0.039929598569869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,8,power_law_1.01,0.041536000370979306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,8,power_law_1.01,0.04264320135116577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,8,power_law_1.01,0.04377599954605103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,8,power_law_1.01,0.04874880015850067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,8,power_law_1.01,0.05448319911956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,8,power_law_1.01,0.0576960027217865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,8,power_law_1.01,0.05761920213699341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,8,power_law_1.2,0.25201919078826907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,8,power_law_1.01,0.06442239880561829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,8,power_law_1.01,0.08422399759292602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,8,power_law_1.01,0.09919999837875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,8,power_law_1.2,0.4109248161315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,8,power_law_1.01,0.1000704050064087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,8,power_law_1.01,0.1303231954574585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,8,power_law_1.01,0.15826560258865358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,8,power_law_1.01,0.20327680110931395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,8,power_law_1.01,0.26065280437469485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,8,power_law_1.01,0.3833024024963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,8,power_law_1.01,0.4554751873016357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.2,0.7036287784576416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,2,power_law_1.2,0.7790207862854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,8,power_law_1.2,0.4957888126373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,4,power_law_1.2,1.5038528442382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,2,power_law_1.2,0.9232895851135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,2,power_law_1.2,1.4444607734680175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,32,power_law_1.2,0.07605760097503662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,32,power_law_1.2,0.07511039972305297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,32,power_law_1.2,0.07537919878959656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,32,power_law_1.2,0.08140799999237061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,32,power_law_1.2,0.07819520235061646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,32,power_law_1.2,0.07923200130462646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,32,power_law_1.2,0.07957760095596314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,32,power_law_1.2,0.07975040078163147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,32,power_law_1.2,0.08013439774513245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,32,power_law_1.2,0.07960320115089417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,32,power_law_1.2,0.08193280100822449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,32,power_law_1.2,0.08421120047569275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,32,power_law_1.2,0.08536319732666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,32,power_law_1.2,0.0868607997894287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,32,power_law_1.2,0.10067199468612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,32,power_law_1.2,0.10201599597930908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,32,power_law_1.2,0.10648959875106812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,32,power_law_1.2,0.12651519775390624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,32,power_law_1.2,0.13064320087432862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,32,power_law_1.2,0.16147199869155884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,32,power_law_1.2,0.19004160165786743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,32,power_law_1.2,0.2518143892288208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,32,power_law_1.2,0.2610687971115112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,32,power_law_1.2,0.3556096076965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,32,power_law_1.2,0.4607552051544189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,32,power_law_1.2,0.6500864028930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,32,power_law_1.2,0.6707520008087158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,2,power_law_1.2,1.455718421936035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,64,power_law_1.01,0.06592640280723572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,64,power_law_1.01,0.06689919829368592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,64,power_law_1.01,0.06811519861221313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,64,power_law_1.01,0.07071359753608704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,64,power_law_1.01,0.06859520077705383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,64,power_law_1.01,0.07187839746475219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,64,power_law_1.01,0.07144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,64,power_law_1.01,0.0712768018245697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,64,power_law_1.01,0.0717631995677948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,64,power_law_1.01,0.07283200025558471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,64,power_law_1.01,0.07482240200042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,64,power_law_1.01,0.0755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,64,power_law_1.01,0.07807999849319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,64,power_law_1.01,0.07935360074043274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,64,power_law_1.01,0.08721920251846313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,64,power_law_1.01,0.09029759764671326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,64,power_law_1.01,0.09604480266571044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,64,power_law_1.01,0.11047680377960205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,64,power_law_1.01,0.12691839933395385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,64,power_law_1.01,0.1530176043510437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,64,power_law_1.01,0.18588160276412963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,64,power_law_1.01,0.24479360580444337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,64,power_law_1.01,0.25617280006408694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,64,power_law_1.01,0.3122175931930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,64,power_law_1.01,0.3782399892807007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,64,power_law_1.01,0.5158143997192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,64,power_law_1.01,0.6804096221923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,32,balanced,0.017903999735911686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,32,balanced,0.01764800027012825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,32,balanced,0.01740266631046931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,32,balanced,0.019424000134070713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,32,balanced,0.019551999866962433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,32,balanced,0.021589333812395733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,32,balanced,0.03194133440653483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,32,balanced,0.03203733265399933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,32,balanced,0.029648000995318096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,32,balanced,0.029722665747006733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,32,balanced,0.0295413335164388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,32,balanced,0.02945599953333537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,32,balanced,0.02972800036271413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,32,balanced,0.02985599885384242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,32,balanced,0.03193599979082743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,32,balanced,0.03180799881617228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,32,balanced,0.03178666780392329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,32,balanced,0.038015998899936676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,32,balanced,0.04005866746107737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,32,balanced,0.04624533156553904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,32,balanced,0.053317333261171974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,32,balanced,0.06378133098284404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,32,balanced,0.07860266665617625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,32,balanced,0.10326400399208069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,32,balanced,0.1318186620871226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,32,balanced,0.18079467614491782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,32,balanced,0.2325119972229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,2,power_law_1.2,0.032953599095344545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,2,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,2,power_law_1.2,0.043782401084899905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,2,power_law_1.2,0.05740799903869629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,2,power_law_1.2,0.05745279788970947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,2,power_law_1.2,0.08119040131568908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,2,power_law_1.2,0.08303999900817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,2,power_law_1.2,0.08442879915237426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,2,power_law_1.2,0.088646399974823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,2,power_law_1.2,0.08894720077514648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,2,power_law_1.2,0.09665279984474182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,2,power_law_1.2,0.1
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,2,power_law_1.2,0.1053887963294983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,2,power_law_1.2,0.10483839511871337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,2,power_law_1.2,0.11342079639434814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,2,power_law_1.2,0.11804800033569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,2,power_law_1.2,0.12918399572372435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,2,power_law_1.2,0.15114879608154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,2,power_law_1.2,0.1657472014427185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,2,power_law_1.2,0.2176448106765747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,2,power_law_1.2,0.24544000625610352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,2,power_law_1.2,0.32028799057006835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,2,power_law_1.2,0.3832639932632446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,16,power_law_1.01,0.03185279965400696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,16,power_law_1.01,0.02906239926815033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,16,power_law_1.01,0.029356798529624938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,16,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,2,power_law_1.2,0.5548416137695312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,16,power_law_1.01,0.03226880133152008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,16,power_law_1.01,0.03372159898281098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,16,power_law_1.01,0.033606401085853575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,16,power_law_1.01,0.03372800052165985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,16,power_law_1.01,0.03377920091152191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,16,power_law_1.01,0.03444480001926422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,16,power_law_1.01,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,2,power_law_1.2,0.6857088088989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,16,power_law_1.01,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,16,power_law_1.01,0.03578880131244659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,16,power_law_1.01,0.038073599338531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,8,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,8,balanced,0.054117331902186074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,8,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,8,balanced,0.06066666543483734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,8,balanced,0.0658079981803894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,16,power_law_1.01,0.04078719913959503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,16,power_law_1.01,0.041459199786186215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,8,balanced,0.08524800340334575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,8,balanced,0.10651733477910359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,8,balanced,0.10818666219711304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,8,balanced,0.10830932855606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,8,balanced,0.10993599891662598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,8,balanced,0.11009066303571065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,16,power_law_1.01,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,16,power_law_1.01,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,2,balanced,0.03341866781314214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,2,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,2,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,8,balanced,0.1132586697737376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,2,balanced,0.03978666663169861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,8,balanced,0.11559999982515971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,2,balanced,0.062128002444903054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,2,balanced,0.06346666812896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,2,balanced,0.062261333068211876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,2,balanced,0.06332799792289734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,2,balanced,0.06587199866771698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,2,balanced,0.06586133440335591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,2,balanced,0.0664106657107671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,2,balanced,0.06833066542943318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,2,balanced,0.07146666447321574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,2,balanced,0.07453866799672444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,2,balanced,0.07937600215276082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,2,balanced,0.08342933654785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,16,power_law_1.01,0.05562880039215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,16,power_law_1.01,0.06696959733963012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,16,power_law_1.01,0.08291199803352356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,16,power_law_1.01,0.10179840326309204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,16,power_law_1.01,0.1287168025970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,2,balanced,0.08597333232561748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,16,power_law_1.01,0.17801599502563475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,2,power_law_1.2,1.0046015739440919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,16,power_law_1.01,0.22366719245910643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,8,balanced,0.11522666613260905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,8,balanced,0.12248532970746358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,8,balanced,0.12472533186276753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,16,power_law_1.01,0.3425600051879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,8,balanced,0.13223466277122498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,8,balanced,0.14335999886194864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,8,balanced,0.15533333023389181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,8,balanced,0.17847466468811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,8,balanced,0.1923733353614807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,8,balanced,0.2353066603342692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,8,balanced,0.2794026732444763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,8,balanced,0.36696000893910724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,8,balanced,0.494106650352478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,8,balanced,0.6628533204396566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,8,balanced,0.8577226797739664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,16,power_law_1.01,0.43396477699279784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,2,balanced,0.09734933574994405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,2,balanced,0.10655466715494792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,2,balanced,0.13326399525006613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,2,balanced,0.14913066228230795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,2,balanced,0.19761067628860474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,2,balanced,0.23270400365193686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,2,balanced,0.344490647315979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,2,balanced,0.44895466168721515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,2,balanced,0.6560159921646118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,2,balanced,0.8665386835734049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,2,power_law_1.2,1.3092608451843262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,128,power_law_1.2,0.05144960284233093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,128,power_law_1.2,0.05016319751739502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,128,power_law_1.2,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,128,power_law_1.2,0.0543936014175415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,128,power_law_1.2,0.053491199016571046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,128,power_law_1.2,0.0548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,128,power_law_1.2,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,128,power_law_1.2,0.05559679865837097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,128,power_law_1.2,0.05631999969482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,128,power_law_1.2,0.05658239722251892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,128,power_law_1.2,0.058374398946762086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,128,power_law_1.2,0.06170240044593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,128,power_law_1.2,0.06361600160598754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,128,power_law_1.2,0.06490240097045899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,128,power_law_1.2,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,128,power_law_1.2,0.07468799948692321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,128,power_law_1.2,0.08254719972610473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,128,power_law_1.2,0.09578239917755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,128,power_law_1.2,0.10587519407272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,128,power_law_1.2,0.1415168046951294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,128,power_law_1.2,0.15966720581054689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,128,power_law_1.2,0.21684479713439941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,128,power_law_1.2,0.23054080009460448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,128,power_law_1.2,0.30083839893341063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,128,power_law_1.2,0.3827455997467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,128,power_law_1.2,0.5283520221710205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,128,power_law_1.2,0.677568006515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,2,balanced,0.019386666516462963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,2,balanced,0.0216799999276797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,2,balanced,0.024656000236670177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,2,balanced,0.029370665550231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,2,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,2,balanced,0.047541335225105286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,2,balanced,0.04951466619968414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,2,balanced,0.05125333368778229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,2,balanced,0.05222400029500326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,2,balanced,0.053354665637016296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,2,balanced,0.053269331653912864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,2,balanced,0.055717334151268005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,2,balanced,0.05835733314355215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,2,balanced,0.05816533168156942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,2,balanced,0.05977066854635874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,2,balanced,0.06348800162474315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,2,balanced,0.07519466678301494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,2,balanced,0.07149866720040639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,2,balanced,0.09132267038027446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,2,balanced,0.1339946687221527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,2,balanced,0.14377599954605103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,2,balanced,0.14216533303260803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,2,balanced,0.14564800262451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,2,balanced,0.2558453281720479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,4,power_law_1.01,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,2,balanced,0.2530933419863383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,2,balanced,0.3636213143666585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,4,power_law_1.01,0.07671040296554565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,2,balanced,0.47299734751383465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,4,power_law_1.01,0.0954367995262146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,4,power_law_1.01,0.1328511953353882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,4,power_law_1.01,0.1610751986503601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,4,power_law_1.01,0.18220800161361694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,4,power_law_1.01,0.2076864004135132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,4,power_law_1.01,0.1978432059288025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,4,power_law_1.01,0.20927999019622803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,4,power_law_1.01,0.21422719955444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,4,power_law_1.01,0.21670401096343994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,4,power_law_1.01,0.24143359661102295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,4,power_law_1.01,0.24228479862213134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,4,power_law_1.01,0.24084479808807374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,4,power_law_1.01,0.2627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,4,power_law_1.01,0.27319679260253904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,4,power_law_1.01,0.27551999092102053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,4,power_law_1.01,0.28305280208587646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.01,0.3124608039855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.01,0.36805119514465334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.01,0.45459837913513185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.01,0.5122943878173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.01,0.5822463989257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.01,0.7274816036224365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.01,0.8586239814758301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.01,1.1710847854614257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.01,1.4654208183288575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,32,power_law_1.01,0.07553920149803162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,32,power_law_1.01,0.07406719923019409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,32,power_law_1.01,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,32,power_law_1.01,0.08131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,32,power_law_1.01,0.07690879702568054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,32,power_law_1.01,0.07787520289421082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,32,power_law_1.01,0.08028799891471863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,32,power_law_1.01,0.08106880187988282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,32,power_law_1.01,0.0809216022491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,32,power_law_1.01,0.0839743971824646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,32,power_law_1.01,0.08339840173721313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,32,power_law_1.01,0.08669440150260925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,32,power_law_1.01,0.08558719754219055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,32,power_law_1.01,0.09118720293045043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,32,power_law_1.01,0.09678720235824585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,32,power_law_1.01,0.1008255958557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,32,power_law_1.01,0.10479999780654907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,32,power_law_1.01,0.12346880435943604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,32,power_law_1.01,0.1323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,32,power_law_1.01,0.16197760105133058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,32,power_law_1.01,0.19511040449142455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,32,power_law_1.01,0.252620792388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,32,power_law_1.01,0.27305600643157957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,32,power_law_1.01,0.3536319971084595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,32,power_law_1.01,0.4416895866394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,32,power_law_1.01,0.6512000083923339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,32,power_law_1.01,0.7636544227600097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,128,power_law_1.2,0.03158400058746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,128,power_law_1.2,0.031219199299812317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,128,power_law_1.2,0.03131519854068756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,128,power_law_1.2,0.03358719944953918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,128,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,128,power_law_1.2,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,128,power_law_1.2,0.03498879969120026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,128,power_law_1.2,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,128,power_law_1.2,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,128,power_law_1.2,0.03535360097885132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,128,power_law_1.2,0.036652800440788266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,128,power_law_1.2,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,128,power_law_1.2,0.036985599994659425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,128,power_law_1.2,0.03857919871807099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,128,power_law_1.2,0.040275201201438904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,128,power_law_1.2,0.041536000370979306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,128,power_law_1.2,0.04157440066337585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,128,power_law_1.2,0.04567039906978607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,128,balanced,0.01939733326435089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,128,balanced,0.01951466624935468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,128,power_law_1.2,0.048403200507164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,128,balanced,0.021205333371957142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,128,balanced,0.023423999547958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,128,power_law_1.2,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,128,power_law_1.2,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,128,power_law_1.2,0.0781823992729187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,128,power_law_1.2,0.0998848021030426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,128,power_law_1.2,0.14665600061416625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,128,power_law_1.2,0.18176640272140504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,1,balanced,0.021365332106749218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,1,balanced,0.021157334248224895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,1,balanced,0.021087999145189922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,1,balanced,0.021007999777793884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,1,balanced,0.02103466788927714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,1,balanced,0.02207999924818675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,1,balanced,0.021312000850836437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,1,balanced,0.023311999936898548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,1,balanced,0.023029332359631855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,1,balanced,0.02306666721900304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,1,balanced,0.024538666009902954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,1,balanced,0.02550933261712392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,1,balanced,0.027248000105222065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,1,balanced,0.028832000990708668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,1,balanced,0.05197866757710775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,1,balanced,0.0535093347231547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,1,balanced,0.05514666438102722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,1,balanced,0.043552001317342125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,1,balanced,0.0459146648645401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,1,balanced,0.05611733098824819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,1,balanced,0.06436266501744588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,1,balanced,0.08895466725031535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,1,balanced,0.10285866260528564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,1,balanced,0.13779733578364053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,1,balanced,0.17492266496022543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,1,balanced,0.24682132403055826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,1,balanced,0.30807467301686603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,128,power_law_1.2,0.26112639904022217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,128,power_law_1.2,0.3336639881134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,2,power_law_1.2,0.18781440258026122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,2,power_law_1.2,0.2705535888671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,2,power_law_1.2,0.4024384021759033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,2,power_law_1.2,0.595686388015747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,2,power_law_1.2,0.7824319839477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,2,power_law_1.2,1.3209471702575684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,32,power_law_1.2,0.044819200038909913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,2,power_law_1.2,1.3516544342041015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,32,power_law_1.2,0.04705919921398163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,2,power_law_1.2,1.342636775970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,32,power_law_1.2,0.04162560105323791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,32,power_law_1.2,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,2,power_law_1.2,1.4337535858154298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,2,power_law_1.2,1.4264703750610352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,2,power_law_1.2,1.4555264472961427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,32,power_law_1.2,0.04351359903812409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,32,power_law_1.2,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,2,power_law_1.2,1.5200960159301757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,32,power_law_1.2,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,32,power_law_1.2,0.04692479968070984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,32,power_law_1.2,0.0469184011220932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,32,power_law_1.2,0.04983679950237274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,32,power_law_1.2,0.05139200091361999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,32,power_law_1.2,0.053260797262191774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,32,power_law_1.2,0.054016000032424925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,32,power_law_1.2,0.05988479852676391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,32,power_law_1.2,0.06442880034446716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,32,power_law_1.2,0.06442880034446716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,32,power_law_1.2,0.06483200192451477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,2,power_law_1.2,1.5501567840576171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,32,power_law_1.2,0.07054719924926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,32,power_law_1.2,0.076665598154068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,32,power_law_1.2,0.09319040179252625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,32,power_law_1.2,0.10654720067977905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,32,power_law_1.2,0.1302783966064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,32,power_law_1.2,0.15329920053482055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,32,power_law_1.2,0.20742399692535402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,32,power_law_1.2,0.23905279636383056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,2,power_law_1.2,1.5939583778381348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,32,power_law_1.2,0.40312957763671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,32,power_law_1.2,0.4382847785949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,2,power_law_1.2,1.6541120529174804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,2,power_law_1.2,1.7000896453857421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,2,power_law_1.2,1.784204864501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,2,power_law_1.2,1.9557376861572267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,2,power_law_1.2,2.0884544372558596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,2,power_law_1.2,2.3938880920410157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,2,power_law_1.2,2.466502380371094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,2,power_law_1.2,2.925958442687988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,2,power_law_1.2,3.128646469116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,2,power_law_1.2,3.8381248474121095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,2,power_law_1.2,5.084524917602539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,2,power_law_1.2,6.542144012451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,2,power_law_1.2,7.6489410400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,2,balanced,0.06860800087451935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,2,balanced,0.07500266532103221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,2,balanced,0.09682666261990865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,2,balanced,0.13753066460291544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,2,balanced,0.19904534022013345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,2,balanced,0.3272320032119751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,2,balanced,0.4527039925257365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,2,balanced,0.455130656560262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,2,balanced,0.45738665262858075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,2,balanced,0.4622613191604614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,2,balanced,0.4636799891789754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,2,balanced,0.47466667493184406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,2,balanced,0.48316800594329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,2,balanced,0.4806986649831136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,2,balanced,0.49026668071746826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,2,balanced,0.5001866817474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,2,balanced,0.5145493348439535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,2,balanced,0.5397706826527914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,2,balanced,0.5727573235829672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,2,balanced,0.6301013231277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,2,balanced,0.669818639755249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,2,balanced,0.7743199666341146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,2,balanced,0.8958186308542887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,2,balanced,1.0965973536173503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,2,balanced,1.4358879725138347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,2,balanced,1.8497227032979329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,2,balanced,2.4933066368103027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,32,balanced,0.03994133323431015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,32,balanced,0.0390133336186409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,32,balanced,0.04014399896065394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,32,balanced,0.043280000487963356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,32,balanced,0.04322133461634318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,32,balanced,0.047653332352638245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,32,balanced,0.048351998130480446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,32,balanced,0.04970133304595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,32,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,32,balanced,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,32,balanced,0.0499893327554067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,32,balanced,0.051813334226608276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,32,balanced,0.05249066650867462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,32,balanced,0.053354665637016296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,32,balanced,0.058330665032068886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,32,balanced,0.05858133236567179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,4,power_law_1.01,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,32,balanced,0.060453335444132485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,32,balanced,0.06659199794133504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,32,balanced,0.07258133093516032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,4,power_law_1.01,0.04734080135822296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,32,balanced,0.08429867029190063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,32,balanced,0.09805333614349365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,32,balanced,0.12723199526468912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,4,power_law_1.01,0.04966399967670441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,32,balanced,0.15056533614794412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,32,balanced,0.20153599977493286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,4,power_law_1.01,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,32,balanced,0.24610666433970133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,32,balanced,0.335258682568868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,4,power_law_1.01,0.05685120224952698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,32,balanced,0.4314613342285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,4,power_law_1.01,0.060012799501419065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,4,power_law_1.01,0.0692736029624939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,4,power_law_1.01,0.06686720252037048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,4,power_law_1.01,0.06858239769935608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,4,power_law_1.01,0.06917120218276977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,4,power_law_1.01,0.07293440103530884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,4,power_law_1.01,0.07525119781494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,4,power_law_1.01,0.07691519856452941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,32,balanced,0.025248001019159954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,4,power_law_1.01,0.0833407998085022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,32,balanced,0.023455999791622162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,32,balanced,0.023071999351183575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,32,balanced,0.025248001019159954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,32,balanced,0.0271519993742307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,32,balanced,0.029461334149042766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,32,balanced,0.02957333376010259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,32,balanced,0.029509333272775013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,32,balanced,0.029152000943819683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,32,balanced,0.029311999678611755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,32,balanced,0.02977599948644638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,32,balanced,0.031221332649389904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,32,balanced,0.031285333136717476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,32,balanced,0.03368533402681351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,32,balanced,0.03588266670703888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,32,balanced,0.03733866661787033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,32,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,32,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,32,balanced,0.039850667119026184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,32,balanced,0.042810668547948204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,32,balanced,0.04424533247947693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,32,balanced,0.04794133206208547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,32,balanced,0.05138133466243744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,32,balanced,0.06675200164318085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,32,balanced,0.07428800066312154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,32,balanced,0.09059733152389526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,32,balanced,0.11012799541155498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,4,power_law_1.01,0.08910719752311706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,4,power_law_1.01,0.09264640212059021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,4,power_law_1.01,0.09941759705543518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,4,power_law_1.01,0.11363840103149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,4,power_law_1.01,0.12231680154800414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,4,power_law_1.01,0.1468224048614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,4,power_law_1.01,0.16755199432373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,4,power_law_1.01,0.22057600021362306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,4,power_law_1.01,0.2706496000289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,4,power_law_1.01,0.35863680839538575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,4,power_law_1.01,0.4719679832458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,4,power_law_1.01,0.6755519866943359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,4,power_law_1.01,0.8703295707702636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,8,power_law_1.01,0.07966079711914062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,8,power_law_1.01,0.08872320055961609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,8,power_law_1.01,0.07786239981651306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,8,power_law_1.01,0.09607040286064147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,8,power_law_1.01,0.09730560183525086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,8,power_law_1.01,0.1033471941947937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,8,power_law_1.01,0.10291199684143067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,8,power_law_1.01,0.10178560018539429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,8,power_law_1.01,0.10432640314102173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,8,power_law_1.01,0.10408320426940917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,8,power_law_1.01,0.10687359571456909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,4,16,power_law_1.2,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,8,power_law_1.01,0.10954879522323609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,4,16,power_law_1.2,0.07234560251235962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,4,16,power_law_1.2,0.08058239817619324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,4,16,power_law_1.2,0.08980479836463928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,4,16,power_law_1.2,0.08880000114440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,8,power_law_1.01,0.10977280139923096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,4,16,power_law_1.2,0.08600320219993592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,8,power_law_1.01,0.110316801071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,4,16,power_law_1.2,0.08563200235366822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,8,power_law_1.01,0.12310400009155273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,4,16,power_law_1.2,0.0886847972869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,4,16,power_law_1.2,0.08578559756278992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,4,16,power_law_1.2,0.08728320002555848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,4,16,power_law_1.2,0.08947200179100037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,4,16,power_law_1.2,0.09238399863243103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,8,power_law_1.01,0.12104320526123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,4,16,power_law_1.2,0.09100160002708435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,8,power_law_1.01,0.13010560274124144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,4,16,power_law_1.2,0.09752320051193238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,8,power_law_1.01,0.1475648045539856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,4,16,power_law_1.2,0.10408960580825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,8,power_law_1.01,0.16117759943008422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,4,16,power_law_1.2,0.11108479499816895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,8,power_law_1.01,0.18681600093841552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,4,16,power_law_1.2,0.12092159986495972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,8,power_law_1.01,0.20965759754180907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,4,16,power_law_1.2,0.13763200044631957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,4,16,power_law_1.2,0.14227839708328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,8,power_law_1.01,0.28687999248504636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,4,16,power_law_1.2,0.1801599979400635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,8,power_law_1.01,0.37336320877075196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,4,16,power_law_1.2,0.2088383913040161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,8,power_law_1.01,0.5144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,4,16,power_law_1.2,0.27884159088134763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,4,16,power_law_1.2,0.34751999378204346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,8,power_law_1.01,0.6438784122467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,4,16,power_law_1.2,0.4678336143493652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,8,power_law_1.01,0.9399680137634278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,4,16,power_law_1.2,0.5718463897705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,8,power_law_1.01,1.2523072242736817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,4,16,power_law_1.2,0.8725055694580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,4,16,power_law_1.2,1.0571328163146974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,32,power_law_1.01,0.09589759707450866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,32,power_law_1.01,0.1131775975227356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,32,power_law_1.01,0.10931199789047241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,32,power_law_1.01,0.11429760456085206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,32,power_law_1.01,0.1143231987953186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,32,power_law_1.01,0.10987520217895508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,32,power_law_1.01,0.10113279819488526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,32,power_law_1.01,0.10728319883346557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,32,power_law_1.01,0.10618879795074462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,32,power_law_1.01,0.10831359624862671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,32,power_law_1.01,0.10680960416793824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,32,power_law_1.01,0.11070079803466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,32,power_law_1.01,0.10986239910125732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,32,power_law_1.01,0.11267839670181275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,32,power_law_1.01,0.1221824049949646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,32,power_law_1.01,0.12303359508514404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,32,power_law_1.01,0.1317952036857605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,32,power_law_1.01,0.13143680095672608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,32,power_law_1.01,0.14410879611968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,32,power_law_1.01,0.16635520458221437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,32,power_law_1.01,0.18566399812698364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,32,power_law_1.01,0.23665280342102052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,32,power_law_1.01,0.29115519523620603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,32,power_law_1.01,0.3806207895278931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,32,power_law_1.01,0.4686783790588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,32,power_law_1.01,0.6658815860748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,32,power_law_1.01,0.8308863639831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,128,power_law_1.01,0.017049600183963776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,128,power_law_1.01,0.016326400637626647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,128,power_law_1.01,0.017958399653434754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,128,power_law_1.01,0.017350399494171144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,128,power_law_1.01,0.019366399943828584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,128,power_law_1.01,0.024985599517822265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,128,power_law_1.01,0.02499839961528778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,128,power_law_1.01,0.025388801097869874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,128,power_law_1.01,0.025433599948883057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,128,power_law_1.01,0.025203201174736022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,128,power_law_1.01,0.025331199169158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,128,power_law_1.01,0.02563839852809906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,128,power_law_1.01,0.026015999913215637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,128,power_law_1.01,0.026918399333953857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,128,power_law_1.01,0.028479999303817748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,128,power_law_1.01,0.02849920094013214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,128,power_law_1.01,0.02933120131492615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,128,power_law_1.01,0.03255040049552917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,128,power_law_1.01,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,128,power_law_1.01,0.037503999471664426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,128,power_law_1.01,0.03922559916973114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,128,power_law_1.01,0.04577920138835907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,128,power_law_1.01,0.050835198163986205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,128,power_law_1.01,0.06355199813842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,128,power_law_1.01,0.08140799999237061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,128,power_law_1.01,0.10861439704895019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,128,power_law_1.01,0.1366528034210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,4,power_law_1.01,0.06567040085792542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,4,power_law_1.01,0.07780479788780212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,4,power_law_1.01,0.0706496000289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,4,power_law_1.01,0.08286719918251037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,4,power_law_1.01,0.0782975971698761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,4,power_law_1.01,0.08499839901924133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,4,power_law_1.01,0.09049599766731262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,4,power_law_1.01,0.10138239860534667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,4,power_law_1.01,0.10554879903793335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,4,power_law_1.01,0.1133247971534729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,4,power_law_1.01,0.10062719583511352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,4,power_law_1.01,0.10675200223922729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,4,power_law_1.01,0.10880000591278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,4,power_law_1.01,0.1192512035369873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,4,power_law_1.01,0.1423616051673889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,4,power_law_1.01,0.15456639528274535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,4,power_law_1.01,0.14377599954605103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,4,power_law_1.01,0.1701375961303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,4,power_law_1.01,0.21196160316467286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,4,power_law_1.01,0.2593791961669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,4,power_law_1.01,0.3096191883087158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,4,power_law_1.01,0.3953984022140503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,4,power_law_1.01,0.5214911937713623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,4,power_law_1.01,0.7315392017364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,4,power_law_1.01,1.0749504089355468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,4,power_law_1.2,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,4,power_law_1.2,0.03709439933300018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,4,power_law_1.2,0.03790079951286316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,4,power_law_1.2,0.039603200554847715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,4,power_law_1.2,0.03909760117530823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,4,power_law_1.2,0.039673599600791934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,4,power_law_1.2,0.040012800693511964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,4,power_law_1.2,0.040780800580978396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,4,power_law_1.2,0.041145598888397215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,4,power_law_1.2,0.04074240028858185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,4,power_law_1.2,0.040268799662590025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,4,power_law_1.2,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,4,power_law_1.2,0.04348799884319306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,4,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,4,power_law_1.01,1.5360511779785155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,2,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,2,balanced,0.04938133557637533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,2,balanced,0.05819733440876007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,2,balanced,0.07902400195598602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,2,balanced,0.114138662815094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,2,balanced,0.11030399799346924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,2,balanced,0.10822932918866475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,2,balanced,0.10682132840156555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,2,balanced,0.10498133301734924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,2,balanced,0.10527466734250386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,2,balanced,0.11069333553314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,2,balanced,0.10807466506958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,2,balanced,0.11129599809646606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,2,balanced,0.11543466647466023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,2,balanced,0.11852266391118367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,2,balanced,0.12423466642697652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,4,power_law_1.2,0.049446401000022885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,4,power_law_1.2,0.05058559775352478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,4,power_law_1.2,0.055129599571228025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,4,power_law_1.2,0.06076800227165222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,2,balanced,0.12060800194740295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,2,balanced,0.1339306632677714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,2,balanced,0.14201600352923074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,2,balanced,0.16593600312868753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,2,balanced,0.18659200270970663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,2,balanced,0.27053866783777875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,2,balanced,0.3120853304862976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,2,balanced,0.4506133397420247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.2,0.06688640117645264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,2,balanced,0.5839093526204427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,2,balanced,0.8358986377716064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,2,balanced,1.0682133038838704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.2,0.0796992003917694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,2,balanced,0.04242133100827535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,2,balanced,0.04394133388996124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,2,balanced,0.049882665276527405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,2,balanced,0.05801066756248474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,2,balanced,0.07844266792138417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,2,balanced,0.08046933511892955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,2,balanced,0.07948266466458638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,2,balanced,0.08159466584523518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,2,balanced,0.08283733328183492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,2,balanced,0.08268266419569652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,2,balanced,0.08404800295829773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,2,balanced,0.08684800068537395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,2,balanced,0.08944533268610637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,2,balanced,0.0918346643447876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,2,balanced,0.09583466251691182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,2,balanced,0.10172800223032634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,2,balanced,0.10865066448847453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,2,balanced,0.12410133083661397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,2,balanced,0.1369706690311432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,2,balanced,0.1641866664091746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,2,balanced,0.1886133352915446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,2,balanced,0.26739199956258136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,2,balanced,0.3219093283017476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.2,0.09066240191459655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,2,balanced,0.46342933177948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,2,balanced,0.605130672454834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,2,balanced,0.8838559786478678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,2,balanced,1.1533493200937908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.2,0.12157440185546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,4,power_law_1.01,1.900556755065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.2,0.13498239517211913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.2,0.1916416049003601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,64,power_law_1.01,0.10759040117263793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,64,power_law_1.01,0.09169920086860657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,64,power_law_1.01,0.09145600199699402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,64,power_law_1.01,0.09928960204124451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,64,power_law_1.01,0.09535999894142151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,64,power_law_1.01,0.09555839896202087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,64,power_law_1.01,0.10065920352935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,64,power_law_1.01,0.09770879745483399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,64,power_law_1.01,0.09905920028686524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,64,power_law_1.01,0.09652479887008666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,64,power_law_1.01,0.09724799990653991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,64,power_law_1.01,0.10184320211410522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,64,power_law_1.01,0.10513919591903687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,64,power_law_1.01,0.10212479829788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.2,0.24972159862518312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,64,power_law_1.01,0.10713599920272827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,64,power_law_1.01,0.11022720336914063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,64,power_law_1.01,0.11955840587615967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,64,power_law_1.01,0.13765759468078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,64,power_law_1.01,0.1429759979248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,64,power_law_1.01,0.1823807954788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,64,power_law_1.01,0.18842240571975707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,64,power_law_1.01,0.24982399940490724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,64,power_law_1.01,0.2488640069961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,64,power_law_1.01,0.34554879665374755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,64,power_law_1.01,0.45888638496398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,64,power_law_1.01,0.651852798461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,64,power_law_1.01,0.767142391204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,2,power_law_1.2,0.10611200332641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,2,power_law_1.2,0.11327359676361085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,2,power_law_1.2,0.12878079414367677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,2,power_law_1.2,0.1610303997993469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,2,power_law_1.2,0.17405439615249635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,2,power_law_1.2,0.20167040824890137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,2,power_law_1.2,0.2544640064239502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,2,power_law_1.2,0.2571327924728394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,2,power_law_1.2,0.27375359535217286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,2,power_law_1.2,0.27278718948364256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,2,power_law_1.2,0.2848063945770264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,2,power_law_1.2,0.30556800365448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,2,power_law_1.2,0.2892735958099365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,2,power_law_1.2,0.311411190032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,2,power_law_1.2,0.31875200271606446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,2,power_law_1.2,0.33237760066986083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,2,power_law_1.2,0.33889920711517335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,2,power_law_1.2,0.38234241008758546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,2,power_law_1.2,0.4052095890045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,2,power_law_1.2,0.4993728160858154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,2,power_law_1.2,0.5743040084838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,2,power_law_1.2,0.7316351890563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,2,power_law_1.2,0.8537216186523438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,2,power_law_1.2,1.210649585723877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,32,power_law_1.01,0.06963199973106385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,32,power_law_1.01,0.06792320013046264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,32,power_law_1.01,0.06786559820175171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,32,power_law_1.01,0.0726527988910675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,32,power_law_1.01,0.06680319905281067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.2,0.3928832054138184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,32,power_law_1.01,0.06390399932861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,32,power_law_1.01,0.07246720194816589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,32,power_law_1.01,0.07395840287208558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,32,power_law_1.01,0.07598080039024353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,32,power_law_1.01,0.07697280049324036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,2,power_law_1.2,1.5259391784667968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,32,power_law_1.01,0.07593600153923034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,32,power_law_1.01,0.08090879917144775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,32,power_law_1.01,0.08203520178794861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,32,power_law_1.01,0.08385279774665833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,32,power_law_1.01,0.09128320217132568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,32,power_law_1.01,0.09345920085906982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,32,power_law_1.01,0.09957759976387023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,32,power_law_1.01,0.11388159990310669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,32,power_law_1.01,0.12095999717712402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,32,power_law_1.01,0.149452805519104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,32,power_law_1.01,0.1649791955947876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,32,power_law_1.01,0.2157439947128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,32,power_law_1.01,0.26421120166778567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,32,power_law_1.01,0.37287039756774903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,32,power_law_1.01,0.47092480659484864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,32,power_law_1.01,0.6464640140533447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,32,power_law_1.01,0.8499520301818848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,8,balanced,0.019440000255902607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,8,balanced,0.01893866683046023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,8,balanced,0.01893866683046023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,8,balanced,0.02107733239730199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,8,balanced,0.025242666403452556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,8,balanced,0.025392000873883564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,8,balanced,0.027610667049884796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,8,balanced,0.027119999130566914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,8,balanced,0.02771199991305669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,8,balanced,0.02874133239189784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,8,balanced,0.029904000461101532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,8,balanced,0.031184000273545582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,8,balanced,0.03181866556406021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,8,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,8,balanced,0.03978666663169861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,8,balanced,0.04004266609748205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,8,balanced,0.05449066559473673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,8,balanced,0.06025599936644236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,8,balanced,0.07144533097743988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,8,balanced,0.09364799658457439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,8,balanced,0.07684266567230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,8,balanced,0.09221866726875305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,8,balanced,0.1069599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,8,balanced,0.1548746625582377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,8,balanced,0.1830186645189921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,8,balanced,0.26021866003672284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,8,balanced,0.3372480074564616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.2,0.439635181427002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,4,balanced,0.03194133440653483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,4,balanced,0.02962133288383484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,4,balanced,0.03160533308982849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,4,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,4,balanced,0.04154133299986521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,4,balanced,0.04180799921353658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,4,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,4,balanced,0.042133331298828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,4,balanced,0.04252799848715464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,4,balanced,0.042362665136655174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,4,balanced,0.04213866591453552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,4,balanced,0.04418133199214935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,4,balanced,0.04582933088143667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,4,balanced,0.046629334489504494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,4,balanced,0.052042668064435325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,4,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,4,balanced,0.056176001826922096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,4,balanced,0.06366399923960368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,4,balanced,0.07249600191911061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,4,balanced,0.09526399771372478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,4,balanced,0.10178666313489278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,4,balanced,0.1320319970448812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,4,balanced,0.15385599931081137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,4,balanced,0.21177067359288534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,4,balanced,0.26021333535512287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,4,balanced,0.3766080141067505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,4,balanced,0.4905279874801636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,2,power_law_1.01,0.05023360252380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,2,power_law_1.01,0.06961920261383056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,2,power_law_1.01,0.08065279722213745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,2,power_law_1.01,0.1022528052330017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,2,power_law_1.01,0.11886080503463745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,2,power_law_1.01,0.1683583974838257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,2,power_law_1.01,0.16940159797668458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,2,power_law_1.01,0.16974079608917236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,2,power_law_1.01,0.17342079877853395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,8,power_law_1.2,0.06290559768676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,8,power_law_1.2,0.06265599727630615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,8,power_law_1.2,0.06172159910202026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,8,power_law_1.2,0.06625279784202576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,8,power_law_1.2,0.06670079827308655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,8,power_law_1.2,0.06782720088958741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,8,power_law_1.2,0.06789759993553161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,2,power_law_1.01,0.18033920526504515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,2,power_law_1.01,0.184934401512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,2,power_law_1.01,0.18306560516357423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,2,power_law_1.01,0.19324159622192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,2,power_law_1.01,0.19785599708557128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,2,power_law_1.01,0.20778241157531738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,2,power_law_1.01,0.21235198974609376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,2,power_law_1.01,0.22704639434814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,8,power_law_1.2,0.06898559927940369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,2,power_law_1.01,0.25360000133514404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,2,power_law_1.01,0.27778561115264894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,2,power_law_1.01,0.3255359888076782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,2,power_law_1.01,0.39165439605712893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,2,power_law_1.01,0.48969597816467286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,8,power_law_1.2,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,2,power_law_1.01,0.6393152236938476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,8,power_law_1.2,0.07485439777374267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,8,power_law_1.2,0.07643520236015319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,2,power_law_1.01,0.8289279937744141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,2,power_law_1.01,1.0462911605834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,2,power_law_1.01,1.432857608795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,8,power_law_1.2,0.08235520124435425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,2,power_law_1.01,1.7958976745605468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,8,power_law_1.2,0.08291839957237243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,8,power_law_1.2,0.09950079917907714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,8,power_law_1.2,0.12057600021362305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,8,power_law_1.2,0.1317055940628052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,8,power_law_1.2,0.13839999437332154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,8,power_law_1.2,0.17879040241241456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,8,power_law_1.2,0.2297215938568115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,8,power_law_1.2,0.34270079135894777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,8,power_law_1.2,0.4020671844482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,8,power_law_1.2,0.605951976776123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,8,power_law_1.2,0.7440832138061524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,16,balanced,0.029616000751654308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,16,balanced,0.029637334247430164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,16,balanced,0.02941333254178365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,16,balanced,0.03148266673088074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,16,balanced,0.034245334565639496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,16,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,16,balanced,0.03443733354409536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,16,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,16,balanced,0.0356480007370313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,16,balanced,0.035599999129772186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,16,balanced,0.035573333501815796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,16,balanced,0.037903999288876854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,16,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,16,balanced,0.03749333322048187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,16,balanced,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,16,balanced,0.0433599998553594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,16,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,16,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,16,balanced,0.04985066751639048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,16,balanced,0.05412800113360087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,16,balanced,0.056159997979799904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,16,balanced,0.07296533385912578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,16,balanced,0.0849173367023468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,16,balanced,0.11125333110491435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,16,balanced,0.13582932949066162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,16,balanced,0.18525866667429605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,16,balanced,0.23198399941126505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,8,power_law_1.2,1.0861696243286132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,2,power_law_1.01,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,2,power_law_1.01,0.03343999981880188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,2,power_law_1.01,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,2,power_law_1.01,0.035385599732398985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,2,power_law_1.01,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,2,power_law_1.01,0.03658879995346069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,2,power_law_1.01,0.037376001477241516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,8,power_law_1.2,1.509382438659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,2,power_law_1.01,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,2,power_law_1.01,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,2,power_law_1.01,0.03780480027198792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,2,power_law_1.01,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,2,power_law_1.01,0.04026240110397339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,2,power_law_1.01,0.041459199786186215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,2,power_law_1.01,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,2,power_law_1.01,0.047635200619697574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,2,power_law_1.01,0.04899199903011322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,2,power_law_1.01,0.052985602617263795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,2,power_law_1.01,0.06197119951248169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,2,power_law_1.01,0.06389120221138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,2,power_law_1.01,0.0778432011604309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,4,power_law_1.01,0.01940480023622513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,2,power_law_1.01,0.09274240136146546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,4,power_law_1.01,0.01847040057182312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,4,power_law_1.01,0.01889919936656952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,4,power_law_1.01,0.018828800320625304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,4,power_law_1.01,0.01913599967956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,4,power_law_1.01,0.01966720074415207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,4,power_law_1.01,0.022342400252819063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,4,power_law_1.01,0.022137600183486938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,4,power_law_1.01,0.046163201332092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,4,power_law_1.01,0.045824000239372255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,4,power_law_1.01,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,4,power_law_1.01,0.03546879887580871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,4,power_law_1.01,0.035180801153182985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,4,power_law_1.01,0.03537920117378235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,4,power_law_1.01,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,4,power_law_1.01,0.0363072007894516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,4,power_law_1.01,0.03983359932899475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,4,power_law_1.01,0.04227199852466583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,8,power_law_1.2,2.299590492248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,2,power_law_1.01,0.12649600505828856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.01,0.04581120014190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.01,0.052671998739242554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,128,power_law_1.01,0.027321600914001466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,128,power_law_1.01,0.026022401452064515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,128,power_law_1.01,0.025183999538421632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,128,power_law_1.01,0.027270400524139406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,128,power_law_1.01,0.027929601073265076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,128,power_law_1.01,0.02943359911441803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,128,power_law_1.01,0.02956160008907318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,128,power_law_1.01,0.028940799832344054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,128,power_law_1.01,0.029631999135017396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,128,power_law_1.01,0.029875200986862183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,128,power_law_1.01,0.03118079900741577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,128,power_law_1.01,0.03215999901294708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,2,power_law_1.01,0.1423616051673889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,128,power_law_1.01,0.032307198643684386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,128,power_law_1.01,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,128,power_law_1.01,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,128,power_law_1.01,0.036556801199913024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,128,power_law_1.01,0.03735679984092712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,128,power_law_1.01,0.040550398826599124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.01,0.06169599890708923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,128,power_law_1.01,0.04368639886379242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,128,power_law_1.01,0.04989440143108368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,128,power_law_1.01,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,128,power_law_1.01,0.06069120168685913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,128,power_law_1.01,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,128,power_law_1.01,0.09466879963874816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,128,power_law_1.01,0.12501120567321777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,128,power_law_1.01,0.18935680389404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,128,power_law_1.01,0.2282560110092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.01,0.08524799942970276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,2,power_law_1.01,0.19307520389556884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.01,0.0991424024105072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.01,0.1346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.01,0.17224960327148436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,2,power_law_1.01,0.23987839221954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,1,power_law_1.2,0.13660160303115845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,1,power_law_1.2,0.19990400075912476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,1,power_law_1.2,0.24929919242858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,1,power_law_1.2,0.40615038871765136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,1,power_law_1.2,0.4281280040740967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,1,power_law_1.2,0.43546237945556643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,1,power_law_1.2,0.4475711822509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,1,power_law_1.2,0.45958399772644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,1,power_law_1.2,0.4757631778717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,1,power_law_1.2,0.4652095794677734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,1,power_law_1.2,0.48372478485107423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.01,0.24522879123687744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,1,power_law_1.2,0.49538559913635255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,1,power_law_1.2,0.5042175769805908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,1,power_law_1.2,0.510099220275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,1,power_law_1.2,0.560748815536499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,1,power_law_1.2,0.5665791988372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,2,power_law_1.01,0.33992319107055663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,1,power_law_1.2,0.6240896224975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,1,power_law_1.2,0.7007359981536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,1,power_law_1.2,0.8124992370605468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,1,power_law_1.2,0.9614912033081054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,1,power_law_1.2,1.1914175987243651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,8,power_law_1.2,3.0613183975219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,2,power_law_1.01,0.4230463981628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,1,power_law_1.2,1.5998527526855468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.01,0.3053760051727295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,1,power_law_1.2,1.9457792282104491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,2,power_law_1.01,0.067603200674057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,2,power_law_1.01,0.08914560079574585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,2,power_law_1.01,0.10400639772415161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,2,power_law_1.01,0.13787519931793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,2,power_law_1.01,0.1627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,2,power_law_1.01,0.18075519800186157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,2,power_law_1.01,0.23722879886627196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,1,power_law_1.2,2.6954559326171874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,2,power_law_1.01,0.2547264099121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,2,power_law_1.01,0.2513024091720581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,2,power_law_1.01,0.24789121150970458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,2,power_law_1.01,0.25885438919067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,2,power_law_1.01,0.2723583936691284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,2,power_law_1.01,0.28072960376739503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,2,power_law_1.01,0.2804543972015381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,2,power_law_1.01,0.29228160381317136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,2,power_law_1.01,0.3028480052947998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,2,power_law_1.01,0.32746880054473876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,2,power_law_1.01,0.3619136095046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,2,power_law_1.01,0.3950079917907715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,2,power_law_1.01,0.47399678230285647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,2,power_law_1.01,0.48840961456298826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,2,power_law_1.01,0.6109248161315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,2,power_law_1.01,0.737388801574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,2,power_law_1.01,0.9644800186157226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,2,power_law_1.01,1.2426495552062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,2,power_law_1.01,1.664076805114746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,2,power_law_1.01,2.132633590698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,1,power_law_1.2,3.718751907348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,1,balanced,0.017685333887736004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,1,balanced,0.019029332945744198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,1,balanced,0.020992000897725422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,1,balanced,0.023402666052182514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,1,balanced,0.030975999931494396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,1,balanced,0.03158933420976003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,1,balanced,0.03286933402220408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,1,balanced,0.03395200024048487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,1,balanced,0.03489600121974945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,1,balanced,0.03495999922355016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,1,balanced,0.037274666130542755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,1,balanced,0.03938133269548416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,1,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,1,balanced,0.04351999859015147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,1,balanced,0.04464533428351084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,1,balanced,0.04517866671085358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,1,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,1,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,1,balanced,0.058789332707722984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,1,balanced,0.09697600205739339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,1,balanced,0.1076586643854777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,1,balanced,0.10736533006032307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,1,balanced,0.11528000235557556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,1,balanced,0.186463991800944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,1,balanced,0.19640533129374185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,1,balanced,0.2763200004895528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,1,balanced,0.3574879964192708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,1,power_law_1.2,5.293964767456055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,16,balanced,0.025360000630219776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,16,balanced,0.025472000241279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,16,balanced,0.02521066615978877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,16,balanced,0.027221334477265675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,16,balanced,0.04012266546487808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,16,balanced,0.05517866710821787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,16,balanced,0.061893333991368614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,16,balanced,0.061162665486335754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,16,balanced,0.08066133161385854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,16,balanced,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,16,balanced,0.07719466586907704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,16,balanced,0.11965866883595784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,16,balanced,0.12269866466522217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,16,balanced,0.12318933010101318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,16,balanced,0.11312533418337505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,16,balanced,0.11346133550008138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,16,balanced,0.12618666887283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,16,balanced,0.12692266702651978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,16,balanced,0.14314132928848267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,16,balanced,0.1665279964605967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,16,balanced,0.19276267290115356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,16,balanced,0.24155199527740479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,16,balanced,0.2949440081914266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,16,balanced,0.38494932651519775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,16,balanced,0.48339200019836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,16,balanced,0.7501546541849772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,16,balanced,0.9528586864471436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,1,power_law_1.2,6.890835571289062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,4,power_law_1.01,0.05169919729232788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,4,power_law_1.01,0.06098560094833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,4,power_law_1.01,0.07018880248069763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,4,power_law_1.01,0.07118719816207886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,4,power_law_1.01,0.09440640211105347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,4,power_law_1.01,0.10796799659729003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,4,power_law_1.01,0.11655679941177369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,4,power_law_1.01,0.11769599914550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,4,power_law_1.01,0.11688319444656373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,4,power_law_1.01,0.11849600076675415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,4,power_law_1.01,0.12231680154800414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,4,power_law_1.01,0.12677760124206544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,4,power_law_1.01,0.13147519826889037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,4,power_law_1.01,0.1356927990913391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,4,power_law_1.01,0.1418944001197815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,4,power_law_1.01,0.14965759515762328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,4,power_law_1.01,0.1525439977645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,4,power_law_1.01,0.1721984028816223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,4,power_law_1.01,0.18536959886550902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,4,power_law_1.01,0.2287168025970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,4,power_law_1.01,0.24189438819885253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,4,power_law_1.01,0.32335360050201417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,4,power_law_1.01,0.3909440040588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,4,power_law_1.01,0.538969612121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,4,power_law_1.01,0.6344831943511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,4,power_law_1.01,0.9452608108520508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,4,power_law_1.01,1.2093567848205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,128,power_law_1.2,0.031097599864006044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,128,power_law_1.2,0.030931198596954347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,128,power_law_1.2,0.03126400113105774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,128,power_law_1.2,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,128,power_law_1.2,0.03377920091152191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,128,power_law_1.2,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,128,power_law_1.2,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,128,power_law_1.2,0.03524479866027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,128,power_law_1.2,0.035129600763320924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,128,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,128,power_law_1.2,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,128,power_law_1.2,0.036595198512077334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,128,power_law_1.2,0.03688960075378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,128,power_law_1.2,0.03754880130290985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,128,power_law_1.2,0.039705601334571836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,128,power_law_1.2,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,128,power_law_1.2,0.041196799278259276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,128,power_law_1.2,0.044326400756835936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,128,power_law_1.2,0.049798399209976196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,128,power_law_1.2,0.054995197057724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,128,power_law_1.2,0.06032639741897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,128,power_law_1.2,0.07182080149650574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,128,power_law_1.2,0.08516479730606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,128,power_law_1.2,0.11355520486831665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,2,power_law_1.2,0.02707200050354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,2,power_law_1.2,0.03852800130844116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,2,power_law_1.2,0.050432002544403075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,2,power_law_1.2,0.061990398168563846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,2,power_law_1.2,0.07441279888153077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,128,power_law_1.2,0.14135040044784547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,2,power_law_1.2,0.10408320426940917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,2,power_law_1.2,0.10978560447692871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,2,power_law_1.2,0.11028480529785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,2,power_law_1.2,0.11656320095062256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,2,power_law_1.2,0.11608320474624634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,2,power_law_1.2,0.11976319551467896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,2,power_law_1.2,0.1229632019996643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,2,power_law_1.2,0.11935360431671142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,2,power_law_1.2,0.12688000202178956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,2,power_law_1.2,0.1509376049041748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,2,power_law_1.2,0.13882880210876464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,2,power_law_1.2,0.14499839544296264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,2,power_law_1.2,0.08906239867210389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,2,power_law_1.2,0.30344960689544676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,2,power_law_1.2,0.13327360153198242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.2,0.34771840572357177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,2,power_law_1.2,0.18777600526809693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.2,0.2257983922958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,2,power_law_1.2,0.24296319484710693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.2,0.26295039653778074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,2,power_law_1.2,0.29261438846588134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,2,power_law_1.2,0.42615041732788084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,2,power_law_1.2,0.524563217163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,2,power_law_1.2,0.5623104095458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,2,power_law_1.2,0.5820672035217285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,2,power_law_1.2,0.6061376094818115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,2,power_law_1.2,0.6408448219299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,2,power_law_1.2,0.6417727947235108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,2,power_law_1.2,0.6321407794952393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,2,power_law_1.2,0.6750207901000976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,2,power_law_1.2,0.6759039878845214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,2,power_law_1.2,0.7048704147338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,2,power_law_1.2,0.7548672199249268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.2,0.2963007926940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,2,power_law_1.2,0.8338175773620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.2,0.32787840366363524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,2,power_law_1.2,0.8612735748291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,2,power_law_1.2,0.9149248123168945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,2,power_law_1.2,0.9474559783935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,2,power_law_1.2,1.0855039596557616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,2,power_law_1.2,1.2898816108703612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.2,0.4369152069091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,2,power_law_1.2,1.5695103645324706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,2,power_law_1.2,1.866815948486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,2,power_law_1.2,2.361529541015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,2,power_law_1.2,3.0229055404663088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.2,0.5477759838104248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,128,power_law_1.2,0.19026559591293335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,4,balanced,0.05449600021044413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,4,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,4,balanced,0.062047998110453285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,4,balanced,0.07049599786599477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,4,balanced,0.08250133196512859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,4,balanced,0.1085599958896637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,4,balanced,0.10672533512115479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,4,balanced,0.10530666510264079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,4,balanced,0.10466667016347249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,4,balanced,0.10593066612879436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,4,balanced,0.10713600118954976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,4,balanced,0.10635733604431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,4,balanced,0.11248532931009929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,4,balanced,0.111135999361674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,4,balanced,0.11558933059374492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,4,balanced,0.1169599990049998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,4,balanced,0.12111467123031616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,4,balanced,0.13315733273824057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,4,balanced,0.14179199934005737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,4,balanced,0.1609173317750295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,4,balanced,0.180293341477712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,4,balanced,0.21967999140421549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,4,balanced,0.2574026584625244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,4,balanced,0.37381335099538165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.2,0.7418560028076172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,4,balanced,0.4567253192265828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,4,balanced,0.6472320159276327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,4,balanced,0.8403093020121256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,32,power_law_1.01,0.021721599996089934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,32,power_law_1.01,0.0204352006316185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,32,power_law_1.01,0.020467199385166168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,32,power_law_1.01,0.023731200397014617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,32,power_law_1.01,0.02038400024175644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,32,power_law_1.01,0.022316800057888032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,32,power_law_1.01,0.02186879962682724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,32,power_law_1.01,0.025702399015426636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,32,power_law_1.01,0.02563839852809906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,32,power_law_1.01,0.025516799092292784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,32,power_law_1.01,0.04394879937171936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,32,power_law_1.01,0.043315199017524716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,32,power_law_1.01,0.041945600509643556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,32,power_law_1.01,0.042847999930381776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,32,power_law_1.01,0.041068801283836366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,32,power_law_1.01,0.0445248007774353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,32,power_law_1.01,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,32,power_law_1.01,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,32,power_law_1.01,0.050297600030899045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,32,power_law_1.01,0.061766397953033444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,32,power_law_1.01,0.07018240094184876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,32,power_law_1.01,0.09120000004768372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,32,power_law_1.01,0.11291520595550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,32,power_law_1.01,0.15267839431762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,32,power_law_1.01,0.1866047978401184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.2,0.94650239944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,32,power_law_1.01,0.27068800926208497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,128,power_law_1.2,0.2537472009658813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,32,power_law_1.01,0.3409024000167847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,1,balanced,0.07851733267307281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,1,balanced,0.08250666658083598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,1,balanced,0.08991466959317525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,1,balanced,0.11386133233706157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,1,balanced,0.1500746707121531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,1,balanced,0.21574399868647257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,1,balanced,0.22147732973098755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,1,balanced,0.2222506602605184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,1,balanced,0.22310932477315268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,1,balanced,0.2248106598854065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,1,balanced,0.22652800877888998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,1,balanced,0.2311413288116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,1,balanced,0.23842666546503702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,1,balanced,0.2418986757596334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,1,balanced,0.2526773413022359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,1,balanced,0.25814932584762573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,1,balanced,0.2725119988123576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,1,balanced,0.30740267038345337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,1,balanced,0.33822401364644367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,1,balanced,0.40488000710805255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,1,balanced,0.46669332186381024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,1,balanced,0.5953119993209839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,1,balanced,0.7228426933288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,1,balanced,1.0635680357615154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,1,balanced,1.3299626509348552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,1,balanced,1.9324533144632976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,1,balanced,2.511903921763102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,8,power_law_1.01,0.024051199853420257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,8,power_law_1.01,0.03054719865322113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,8,power_law_1.01,0.030329599976539612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,8,power_law_1.01,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,8,power_law_1.01,0.03711360096931458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,8,power_law_1.01,0.04008319973945618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,8,power_law_1.01,0.046054399013519286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,8,power_law_1.01,0.0434688001871109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,8,power_law_1.01,0.04615679979324341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,8,power_law_1.01,0.057606399059295654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,8,power_law_1.01,0.05578240156173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,8,power_law_1.01,0.05677440166473389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,8,power_law_1.01,0.08735359907150268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,8,power_law_1.01,0.08687360286712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,8,power_law_1.01,0.08692479729652405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,8,power_law_1.01,0.09297279715538025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,8,power_law_1.01,0.0945792019367218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,8,power_law_1.01,0.08687360286712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.01,0.08886399865150452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.01,0.09976320266723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.01,0.11660159826278686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.01,0.14401919841766359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.01,0.16659200191497803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.01,0.22037758827209472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.01,0.2689536094665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.01,0.36976640224456786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.01,0.5168767929077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,2,power_law_1.01,0.06000639796257019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,2,power_law_1.01,0.07697920203208923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,2,power_law_1.01,0.0945855975151062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,2,power_law_1.01,0.12131199836730958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,2,power_law_1.01,0.1539199948310852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,2,power_law_1.01,0.191046404838562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,2,power_law_1.01,0.23416318893432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,2,power_law_1.01,0.24353280067443847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,2,power_law_1.01,0.2513535976409912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,2,power_law_1.01,0.26401920318603517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,2,power_law_1.01,0.2599423885345459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,2,power_law_1.01,0.26871678829193113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,2,power_law_1.01,0.2766144037246704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,2,power_law_1.01,0.292140793800354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,2,power_law_1.01,0.2850303888320923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,2,power_law_1.01,0.2931519985198975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,2,power_law_1.01,0.3104192018508911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,2,power_law_1.01,0.34993278980255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,2,power_law_1.01,0.35080320835113527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,2,power_law_1.01,0.3992640018463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,2,power_law_1.01,0.4142655849456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,2,power_law_1.01,0.4897600173950195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,2,power_law_1.01,0.5726592063903808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,2,power_law_1.01,0.719155216217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,2,power_law_1.01,0.8696063995361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,2,power_law_1.01,1.1506239891052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,2,power_law_1.01,1.4915455818176269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,16,balanced,0.021674667795499165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,16,balanced,0.021226666867733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,16,balanced,0.019088000059127808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,16,balanced,0.02123733361562093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,2,balanced,0.02651199946800868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,16,balanced,0.021231998999913532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,2,balanced,0.02916266769170761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,16,balanced,0.02279466638962428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,16,balanced,0.02342933416366577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,16,balanced,0.023237332701683044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,16,balanced,0.02531733363866806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,16,balanced,0.02332266668478648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,16,balanced,0.027530667682488758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,16,balanced,0.030069333811601002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,16,balanced,0.029466666281223297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,16,balanced,0.04075733323891958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,16,balanced,0.04720533390839895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,16,balanced,0.04655466477076212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,2,balanced,0.044826666514078774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,2,balanced,0.06205866734186808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,2,balanced,0.0976746678352356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,2,balanced,0.16635732849438986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,2,balanced,0.17063466707865396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,2,balanced,0.174453337987264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,2,balanced,0.17854400475819907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,2,balanced,0.18413867553075156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,2,balanced,0.19217065970102945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,2,balanced,0.18429332971572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,2,balanced,0.1881493330001831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,2,balanced,0.19665066401163736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,2,balanced,0.23360532522201538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,2,balanced,0.23356799284617105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,2,balanced,0.2238453427950541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,16,balanced,0.04517333209514618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,16,balanced,0.04892266790072123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,16,balanced,0.05635733405749003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,16,balanced,0.07085866729418437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,16,balanced,0.08071466783682506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,16,balanced,0.10268800457318623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,16,balanced,0.11759466926256816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,16,balanced,0.1684053341547648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,16,balanced,0.20637333393096924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,16,balanced,0.29452266295750934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,16,balanced,0.3824319839477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,2,balanced,0.34071465333302814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,2,balanced,0.3163466652234395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,2,balanced,0.49293335278828937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,2,balanced,0.5388960043589274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,2,balanced,0.5597173372904459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,2,balanced,0.566927989323934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,2,balanced,0.5161546468734741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,2,balanced,0.5132533311843872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,4,power_law_1.2,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,4,power_law_1.2,0.048492801189422605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,2,balanced,0.9892426331837972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,4,power_law_1.2,0.05529599785804749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,2,balanced,0.9350346724192301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,4,power_law_1.2,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,4,power_law_1.2,0.07121279835700989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,4,power_law_1.2,0.08436480164527893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,4,power_law_1.2,0.0949184000492096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,8,power_law_1.2,0.03802880048751831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,8,power_law_1.2,0.033907198905944826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,8,power_law_1.2,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,8,power_law_1.2,0.044563201069831845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,8,power_law_1.2,0.04246399998664856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,8,power_law_1.2,0.05467519760131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,8,power_law_1.2,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,8,power_law_1.2,0.05694720149040222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,8,power_law_1.2,0.05616000294685364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,8,power_law_1.2,0.05619840025901794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,8,power_law_1.2,0.05788159966468811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,8,power_law_1.2,0.05942400097846985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,8,power_law_1.2,0.06103039979934692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,1,balanced,0.11964799960454305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,1,balanced,0.15762133399645487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,1,balanced,0.2217866579691569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,1,balanced,0.3662293354670207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,1,balanced,0.6380586624145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,4,power_law_1.2,0.10014079809188843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,1,balanced,1.0562506516774495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,4,power_law_1.2,0.09936000108718872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,1,balanced,1.0306453704833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,4,power_law_1.2,0.09786880016326904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,1,balanced,1.0026506582895915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,4,power_law_1.2,0.10038399696350098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,1,balanced,1.0156373182932537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,8,power_law_1.2,0.06161919832229614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,4,power_law_1.2,0.10522880554199218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,1,balanced,0.9917706648508707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,8,power_law_1.2,0.07191039919853211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,4,power_law_1.2,0.10682879686355591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,8,power_law_1.2,0.07514240145683289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,1,balanced,0.977242628733317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,4,power_law_1.2,0.1106943964958191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,8,power_law_1.2,0.07253119945526124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,1,balanced,0.973370631535848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,4,power_law_1.2,0.1185215950012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,8,power_law_1.2,0.0806656002998352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,1,balanced,0.9845279852549235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,8,power_law_1.2,0.09960319995880126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,1,balanced,0.9976480007171631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,8,power_law_1.2,0.11330560445785523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,1,balanced,0.9757973353068033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,1,balanced,1.00873597462972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,1,balanced,1.0256053606669109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,1,balanced,1.0437226295471191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,8,power_law_1.2,0.12554240226745605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,4,power_law_1.2,0.12235519886016846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,4,power_law_1.2,0.134060800075531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,4,power_law_1.2,0.15389440059661866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,4,power_law_1.2,0.1670912027359009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,4,power_law_1.2,0.19705599546432495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,4,power_law_1.2,0.22622079849243165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,4,power_law_1.2,0.28282239437103274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,4,power_law_1.2,0.35808639526367186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,4,power_law_1.2,0.4902912139892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,8,power_law_1.2,0.15383679866790773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,4,power_law_1.2,0.6528512001037597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,4,power_law_1.2,0.8574912071228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,1,balanced,1.0742506980895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,8,power_law_1.2,0.20051200389862062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,1,balanced,1.1351306438446045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,4,power_law_1.2,1.0978495597839355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,8,power_law_1.2,0.25167999267578123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,1,balanced,1.192080020904541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,1,balanced,1.329749345779419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,1,balanced,1.5199306805928547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,1,balanced,2.3868799209594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,1,balanced,2.615685304005941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,1,balanced,3.7020692825317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,1,balanced,4.8501386642456055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,8,power_law_1.2,0.32673919200897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,8,power_law_1.2,0.4368000030517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,32,balanced,0.06549333532651265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,32,balanced,0.05878399809201559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,32,balanced,0.05964266757170359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,32,balanced,0.06644266843795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,32,balanced,0.08028799792130788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,16,power_law_1.01,0.1021888017654419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,16,power_law_1.01,0.10914560556411743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,16,power_law_1.01,0.10528000593185424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,16,power_law_1.01,0.11136640310287475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,32,balanced,0.11414933204650879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,16,power_law_1.01,0.10603519678115844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,32,balanced,0.11569600303967793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,32,balanced,0.11820800105730693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,16,power_law_1.01,0.10867199897766114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,16,power_law_1.01,0.10887680053710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,16,power_law_1.01,0.11273599863052368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,16,power_law_1.01,0.11173759698867798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,16,power_law_1.01,0.10863360166549682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,16,power_law_1.01,0.11096320152282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,16,power_law_1.01,0.11198079586029053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,32,balanced,0.11421866218249004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,16,power_law_1.01,0.11630719900131226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,32,balanced,0.11388799548149109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,32,balanced,0.12240533034006755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,32,balanced,0.11726933717727661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,16,power_law_1.01,0.12049920558929443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,32,balanced,0.11451733112335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,32,balanced,0.1197760005791982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,32,balanced,0.1237493356068929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,32,balanced,0.12587199608484903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,32,balanced,0.1269599994023641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,8,power_law_1.2,0.673529577255249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,32,balanced,0.1378613313039144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,32,balanced,0.14410666624704996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,32,balanced,0.16119999686876932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,32,balanced,0.17669866482416788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,32,balanced,0.21358400583267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,32,balanced,0.2479520042737325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,32,balanced,0.3409279982248942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,32,balanced,0.4060533444086711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,32,balanced,0.5652159849802653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,16,power_law_1.01,0.13395839929580688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,32,balanced,0.7154666582743326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,16,power_law_1.01,0.1356287956237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,16,power_law_1.01,0.14295040369033812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,16,power_law_1.01,0.16068480014801026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,16,power_law_1.01,0.17011840343475343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,16,power_law_1.01,0.2237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,16,power_law_1.01,0.22394239902496338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,16,power_law_1.01,0.2875072002410889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,16,power_law_1.01,0.293887996673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,16,power_law_1.01,0.41211519241333006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,16,power_law_1.01,0.496127986907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,16,power_law_1.01,0.650816011428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,16,power_law_1.01,0.9210240364074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,16,power_law_1.2,0.0580672025680542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,16,power_law_1.2,0.06513280272483826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,16,power_law_1.2,0.059628802537918094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,16,power_law_1.2,0.06665599942207337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,16,power_law_1.2,0.0657920002937317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,16,power_law_1.2,0.065830397605896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,16,power_law_1.2,0.06704000234603882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,16,power_law_1.2,0.06738560199737549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,16,power_law_1.2,0.07047039866447449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,16,power_law_1.2,0.07066879868507385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,16,power_law_1.2,0.06977279782295227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,16,power_law_1.2,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,16,power_law_1.2,0.07370880246162415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,16,power_law_1.2,0.07642239928245545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,16,power_law_1.2,0.08627840280532836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,16,power_law_1.2,0.0899839997291565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,16,power_law_1.2,0.09537919759750366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,16,power_law_1.2,0.10547840595245361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,16,power_law_1.2,0.11912959814071655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,16,power_law_1.2,0.14074239730834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,16,power_law_1.2,0.16343040466308595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,16,power_law_1.2,0.2181312084197998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,16,power_law_1.2,0.2857151985168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,16,power_law_1.2,0.4213568210601807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,16,power_law_1.2,0.5207104206085205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,16,power_law_1.2,0.821504020690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,16,power_law_1.2,1.1469183921813966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,1,balanced,0.053226664662361145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,1,balanced,0.05601066847642263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,1,balanced,0.06232533355553945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,1,balanced,0.09276266892751057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,1,balanced,0.12787733475367227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,1,balanced,0.19569599628448486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,1,balanced,0.19788267215092978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,1,balanced,0.20119466384251913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,1,balanced,0.20428800582885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,1,balanced,0.20707732439041138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,1,balanced,0.20779200394948324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,1,balanced,0.2142560084660848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,1,balanced,0.22006932894388834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,1,balanced,0.22361600399017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,1,balanced,0.23587199052174887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,1,balanced,0.24267733097076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,1,balanced,0.2583893338839213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,1,balanced,0.3010186751683553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,1,balanced,0.3310239911079407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,1,balanced,0.4044853448867798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,1,balanced,0.464352011680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,1,balanced,0.6622293392817179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,1,balanced,0.7647679646809896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,1,balanced,1.140496015548706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,1,balanced,1.4146614074707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,1,balanced,2.059765338897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,1,balanced,2.707770665486654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,32,1,power_law_1.01,0.020454399287700653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,32,1,power_law_1.01,0.021158400177955627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,32,1,power_law_1.01,0.02398719936609268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,32,1,power_law_1.01,0.028832000494003297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,32,1,power_law_1.01,0.0381056010723114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,32,1,power_law_1.01,0.05324800014495849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,32,1,power_law_1.01,0.06469759941101075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,32,1,power_law_1.01,0.06762880086898804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,32,1,power_law_1.01,0.07059839963912964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,2,power_law_1.2,0.01681919991970062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,32,1,power_law_1.01,0.0725823998451233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,2,power_law_1.2,0.016748799383640288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,32,1,power_law_1.01,0.07830399870872498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,2,power_law_1.2,0.017612800002098083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,32,1,power_law_1.01,0.0832319974899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,2,power_law_1.2,0.0204352006316185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,32,1,power_law_1.01,0.09134079813957215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,2,power_law_1.2,0.02136960029602051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,32,1,power_law_1.01,0.10058239698410035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,2,power_law_1.2,0.023788799345493317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,32,1,power_law_1.01,0.10324480533599853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,32,1,power_law_1.01,0.109990394115448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,32,1,power_law_1.01,0.12676479816436767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,32,1,power_law_1.01,0.12318079471588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,32,1,power_law_1.01,0.14142719507217408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,32,1,power_law_1.01,0.18203519582748412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,16,balanced,0.03152533372243246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,16,balanced,0.02939733366171519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,32,1,power_law_1.01,0.19971840381622313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,2,power_law_1.2,0.024665600061416625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,2,power_law_1.2,0.0247871994972229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,2,power_law_1.2,0.026764801144599913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,2,power_law_1.2,0.026681599020957947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,2,power_law_1.2,0.028095999360084535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,2,power_law_1.2,0.02781440019607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,2,power_law_1.2,0.027897599339485168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,16,balanced,0.03152533372243246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,16,balanced,0.03562666724125544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,32,1,power_law_1.01,0.2830976009368896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,16,balanced,0.05776533484458923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,16,balanced,0.05794133245944977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,16,balanced,0.05740800003210703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,16,balanced,0.059903999169667564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,16,balanced,0.06009600063165029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,16,balanced,0.059450666109720864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,16,balanced,0.061253334085146584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,16,balanced,0.06162666777769724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,16,balanced,0.06364266574382782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,16,balanced,0.06322133541107178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,16,balanced,0.06769066552321117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,16,balanced,0.07045866549015045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,2,power_law_1.2,0.03047040104866028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,16,balanced,0.07284266750017802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,16,balanced,0.07825600107510884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,16,balanced,0.07952000200748444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,16,balanced,0.09067199627558391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,16,balanced,0.10089600086212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,16,balanced,0.12346133589744568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,16,balanced,0.14246400197347006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,16,balanced,0.18954133987426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,16,balanced,0.2347573240598043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,16,balanced,0.3194719950358073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,16,balanced,0.3898826837539673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,2,power_law_1.2,0.035641598701477054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,2,power_law_1.2,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,2,power_law_1.2,0.03880319893360138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,2,power_law_1.2,0.05976319909095764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,2,power_law_1.2,0.0637440025806427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,2,power_law_1.2,0.07144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,2,power_law_1.2,0.07185279726982116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,2,power_law_1.2,0.08506240248680115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,32,1,power_law_1.01,0.3376703977584839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,32,1,power_law_1.01,0.38705921173095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,2,power_law_1.2,0.09813119769096375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,32,1,power_law_1.01,0.45267200469970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,32,1,power_law_1.01,0.6016448020935059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,32,1,power_law_1.01,0.7515711784362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,2,power_law_1.2,0.12904959917068481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,2,power_law_1.2,0.1627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,4,4,balanced,0.04173333446184794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,4,4,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,4,4,balanced,0.05171733101209005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,4,4,balanced,0.06637866795063019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,4,4,balanced,0.09191999832789104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,4,4,balanced,0.08662399649620056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,4,4,balanced,0.08703999718030293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,4,4,balanced,0.0827893316745758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,4,4,balanced,0.08578133583068848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,4,4,balanced,0.08102933565775554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,4,4,balanced,0.0823466678460439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,4,4,balanced,0.08423466483751933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,4,4,balanced,0.08993599812189738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,4,4,balanced,0.0855466624101003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,4,4,balanced,0.09499200185139973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,4,4,balanced,0.09300800164540608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,4,4,balanced,0.09671466549237569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,4,4,balanced,0.10365333159764607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,4,4,balanced,0.1097920040289561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,4,4,balanced,0.12711466352144876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,32,balanced,0.0306986669699351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,32,balanced,0.028197333216667175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,32,balanced,0.028192001084486645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,32,balanced,0.03207999964555105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,32,balanced,0.033610666791598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,32,balanced,0.033770665526390076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,32,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,32,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,32,balanced,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,32,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,32,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,32,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,32,balanced,0.03781333317359289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,32,balanced,0.03774933268626531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,32,balanced,0.039877332746982574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,32,balanced,0.041690667470296226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,32,balanced,0.04372799893220266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,4,4,balanced,0.1418186624844869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,32,balanced,0.04457599918047587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,4,4,balanced,0.19836799303690592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,32,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,4,4,balanced,0.2286240061124166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,4,4,balanced,0.312063992023468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,4,4,balanced,0.39498666922251385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,4,4,balanced,0.5580319960912069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,4,4,balanced,0.730410655339559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,32,balanced,0.05208533505598704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,32,balanced,0.056277334690093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,32,balanced,0.07188799977302551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,32,balanced,0.09278933207194011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,32,balanced,0.12556800246238708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,32,balanced,0.1586026648680369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,32,balanced,0.2146079937616984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,32,balanced,0.2712799906730652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,4,balanced,0.031957333286603294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,4,balanced,0.029658667743206024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,4,balanced,0.03143999973932902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,4,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,4,balanced,0.03496533383925756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,4,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,4,balanced,0.03374933451414108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,4,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,4,balanced,0.03589866558710734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,4,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,4,balanced,0.037690666814645134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,4,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,4,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,4,balanced,0.03990933299064636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,4,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,4,balanced,0.046911999583244324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,4,balanced,0.048010667165120445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,4,balanced,0.054117331902186074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,4,balanced,0.05611200133959452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,4,balanced,0.07833066582679749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,4,balanced,0.08762133121490479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,4,balanced,0.11408533652623494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,4,balanced,0.1364479959011078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,4,balanced,0.1816906730333964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,4,balanced,0.21918932596842447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,4,balanced,0.3036479949951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,4,balanced,0.39163732528686523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,2,power_law_1.2,0.2103872060775757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,16,power_law_1.01,0.06079360246658325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,16,power_law_1.01,0.05923839807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,16,power_law_1.01,0.05537279844284058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,16,power_law_1.01,0.06359040141105651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,16,power_law_1.01,0.05355520248413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,16,power_law_1.01,0.06327679753303528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,16,power_law_1.01,0.06335359811782837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,16,power_law_1.01,0.06590080261230469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,16,power_law_1.01,0.06471679806709289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,16,power_law_1.01,0.06548479795455933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,16,power_law_1.01,0.06961280107498169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,16,power_law_1.01,0.07343999743461609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,16,power_law_1.01,0.07500799894332885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,16,power_law_1.01,0.07445759773254394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,16,power_law_1.01,0.0833728015422821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,16,power_law_1.01,0.08542720079421998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,16,power_law_1.01,0.09133440256118774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,16,power_law_1.01,0.10956159830093384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,16,power_law_1.01,0.11306240558624267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,16,power_law_1.01,0.14191999435424804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,16,power_law_1.01,0.16580480337142944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,16,power_law_1.01,0.20996479988098143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,16,power_law_1.01,0.2634943962097168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,16,power_law_1.01,0.36423680782318113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,16,power_law_1.01,0.44627838134765624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,16,power_law_1.01,0.6106815814971924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,2,power_law_1.2,0.2673856019973755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,16,power_law_1.01,0.8342080116271973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,2,power_law_1.01,0.12591999769210815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,2,power_law_1.01,0.1903488039970398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,2,power_law_1.01,0.22915201187133788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,2,power_law_1.01,0.35613439083099363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,2,power_law_1.01,0.4457024097442627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,2,power_law_1.01,0.6612927913665771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,2,power_law_1.01,0.7110079765319824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,2,power_law_1.01,0.7036096096038819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,2,power_law_1.01,0.7181951999664307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,2,power_law_1.01,0.7608831882476806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,2,power_law_1.01,0.7751423835754394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,2,power_law_1.01,0.810431957244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,2,power_law_1.01,0.8044992446899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,2,power_law_1.01,0.8198335647583008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,2,power_law_1.01,0.8382656097412109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,2,power_law_1.01,0.8835136413574218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,2,power_law_1.01,0.9192447662353516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,2,power_law_1.01,0.9984640121459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,2,power_law_1.01,1.0439999580383301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,2,power_law_1.01,1.2233216285705566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,2,power_law_1.01,1.367244815826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,2,power_law_1.01,1.6931648254394531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,2,power_law_1.01,1.7093824386596679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,16,power_law_1.01,0.059699201583862306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,16,power_law_1.01,0.0717248022556305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,2,power_law_1.01,2.109132766723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,16,power_law_1.01,0.08161919713020324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,16,power_law_1.01,0.08556159734725952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,2,power_law_1.01,2.8126272201538085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,32,power_law_1.2,0.11811200380325318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,32,power_law_1.2,0.10978560447692871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,32,power_law_1.2,0.1088703989982605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,32,power_law_1.2,0.10448640584945679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,32,power_law_1.2,0.10725760459899902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,16,power_law_1.01,0.08635519742965699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,32,power_law_1.2,0.10764800310134888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,32,power_law_1.2,0.10914560556411743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,32,power_law_1.2,0.11272959709167481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,2,power_law_1.01,3.7021953582763674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,16,power_law_1.01,0.08510079979896545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,16,power_law_1.01,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,16,power_law_1.01,0.09137920141220093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,2,power_law_1.01,4.2307392120361325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,16,power_law_1.01,0.08654720187187195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,16,power_law_1.01,0.08529279828071594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,16,power_law_1.01,0.08565760254859925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,32,power_law_1.2,0.11326719522476196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,16,power_law_1.01,0.09038720130920411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,32,power_law_1.2,0.12214399576187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,16,power_law_1.01,0.09355520009994507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,16,power_law_1.01,0.09223039746284485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,16,power_law_1.01,0.10028799772262573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,16,power_law_1.01,0.1019711971282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,16,power_law_1.01,0.10826239585876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,16,power_law_1.01,0.12383359670639038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,32,power_law_1.2,0.1257472038269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,16,power_law_1.01,0.13193600177764891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,32,power_law_1.2,0.12865920066833497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,16,power_law_1.01,0.1585536003112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,32,power_law_1.2,0.12524800300598143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,16,power_law_1.01,0.18272000551223755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,32,power_law_1.2,0.1269503951072693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,32,power_law_1.2,0.141593599319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,16,power_law_1.01,0.23772799968719482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,32,power_law_1.2,0.145305597782135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,16,power_law_1.01,0.26922240257263186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,32,power_law_1.2,0.15507839918136596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,16,power_law_1.01,0.35380480289459226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,16,power_law_1.01,0.4403711795806885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,16,power_law_1.01,0.6474239826202393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,16,power_law_1.01,0.8504575729370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,32,power_law_1.2,0.18810880184173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,32,power_law_1.2,0.22465920448303223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,32,power_law_1.2,0.26347520351409914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,32,power_law_1.2,0.2837440013885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,32,power_law_1.2,0.39667201042175293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,32,power_law_1.2,0.4103871822357178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,32,power_law_1.2,0.5381184101104737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,32,power_law_1.2,0.7460224151611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,32,power_law_1.2,1.0446720123291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,32,power_law_1.01,0.023827199637889863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,32,power_law_1.01,0.02508159875869751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,32,power_law_1.01,0.02114560008049011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,32,power_law_1.01,0.02215680032968521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,32,power_law_1.01,0.022495999932289124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,32,power_law_1.2,1.3559295654296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,32,power_law_1.01,0.023020799458026885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,32,power_law_1.01,0.02401279956102371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,32,power_law_1.01,0.025222399830818178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,32,power_law_1.01,0.02568320035934448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,32,power_law_1.01,0.025651198625564576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,32,power_law_1.01,0.03031040132045746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,128,power_law_1.2,0.05443840026855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,32,power_law_1.01,0.030508801341056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,128,power_law_1.2,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,128,power_law_1.2,0.04530560076236725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,128,power_law_1.2,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,128,power_law_1.2,0.04856959879398346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,128,power_law_1.2,0.04822399914264679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,128,power_law_1.2,0.04991999864578247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,32,power_law_1.01,0.030451199412345885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,128,power_law_1.2,0.049542400240898135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,32,power_law_1.01,0.05270400047302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,32,power_law_1.01,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,32,power_law_1.01,0.05630080103874206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,32,power_law_1.01,0.05333120226860046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,32,power_law_1.01,0.05400320291519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,32,power_law_1.01,0.0602944016456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,32,power_law_1.01,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,128,power_law_1.2,0.050297600030899045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,128,power_law_1.2,0.05084800124168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,128,power_law_1.2,0.052211201190948485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,128,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,128,power_law_1.2,0.05583999752998352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,128,power_law_1.2,0.057068800926208495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,128,power_law_1.2,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,32,power_law_1.01,0.08035200238227844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,32,power_law_1.01,0.1043328046798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,32,power_law_1.01,0.1258239984512329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,32,power_law_1.01,0.16284799575805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,32,power_law_1.01,0.2008768081665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,32,power_law_1.01,0.291756796836853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,128,power_law_1.2,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,32,power_law_1.01,0.3707648038864136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,128,power_law_1.2,0.0711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,128,power_law_1.2,0.08986240029335021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,128,power_law_1.2,0.10019840002059936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,128,power_law_1.2,0.12983039617538453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,128,power_law_1.2,0.14384000301361083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,128,power_law_1.2,0.1975808024406433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,128,power_law_1.2,0.24476799964904786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,128,power_law_1.2,0.3557760000228882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,128,power_law_1.2,0.47477121353149415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,128,power_law_1.2,0.7272192001342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,128,power_law_1.2,0.8162943840026855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,16,power_law_1.01,0.04562560021877289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,16,power_law_1.01,0.06832640171051026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,16,power_law_1.01,0.06858879923820496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,16,power_law_1.01,0.06870399713516236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,16,power_law_1.01,0.06907520294189454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,16,power_law_1.01,0.06828160285949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,16,power_law_1.01,0.07416319847106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,16,power_law_1.01,0.07676799893379212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,16,power_law_1.01,0.1039423942565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,16,power_law_1.01,0.10760960578918458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,16,power_law_1.01,0.09964799880981445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,16,power_law_1.01,0.12721920013427734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,16,power_law_1.01,0.1382848024368286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,16,power_law_1.01,0.14158079624176026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,16,power_law_1.01,0.14268159866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,16,power_law_1.01,0.15010559558868408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,16,power_law_1.01,0.149619197845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,16,power_law_1.01,0.1586624026298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,16,power_law_1.01,0.14585599899291993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,16,power_law_1.01,0.14670079946517944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,16,power_law_1.01,0.15590399503707886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,16,power_law_1.01,0.19801599979400636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,16,power_law_1.01,0.25868160724639894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,16,power_law_1.01,0.3295232057571411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,16,power_law_1.01,0.4040639877319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,16,power_law_1.01,0.5887807846069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,16,power_law_1.01,0.7582464218139648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,8,power_law_1.2,0.023289600014686586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,8,power_law_1.2,0.028140801191329955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,8,power_law_1.2,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,8,power_law_1.2,0.041152000427246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,8,power_law_1.2,0.042259201407432556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,8,power_law_1.2,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,8,power_law_1.2,0.052211201190948485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,8,power_law_1.2,0.049420800805091855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,8,power_law_1.2,0.05687040090560913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,8,power_law_1.2,0.05317760109901428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,8,power_law_1.2,0.054028797149658206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,8,power_law_1.2,0.06556159853935242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,8,power_law_1.2,0.06473600268363952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,8,power_law_1.2,0.06308479905128479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,8,power_law_1.2,0.09887999892234803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,8,power_law_1.2,0.10077439546585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,8,power_law_1.2,0.10076800584793091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,8,power_law_1.2,0.11081600189208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.2,0.11686400175094605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.2,0.12028800249099732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.2,0.13569920063018798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.2,0.16769920587539672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.2,0.1889024019241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.2,0.2585983991622925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.2,0.3114432096481323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.2,0.4685952186584473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.2,0.5872255802154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,4,balanced,0.07045333087444305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,4,balanced,0.07052266597747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,4,balanced,0.09588799873987834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,4,balanced,0.09925867120424907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,4,balanced,0.10079999764760335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,4,balanced,0.0999679962793986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,4,balanced,0.10142399867375691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,4,balanced,0.10213866829872131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,4,balanced,0.10105066498120625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,4,balanced,0.10135466853777568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,4,balanced,0.10157333811124165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,4,balanced,0.1092800001303355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,4,balanced,0.10879466931025188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,4,balanced,0.1095306674639384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,4,balanced,0.12964266538619995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,4,balanced,0.130213330189387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,256,balanced,0.04663999875386556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,4,balanced,0.11944533387819926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,256,balanced,0.046997333566347756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,4,balanced,0.15663466850916544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,256,balanced,0.047322665651639305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,256,balanced,0.050101334849993386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,256,balanced,0.04967466493447622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,4,balanced,0.15189866224924722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,4,balanced,0.1946773330370585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,4,balanced,0.256714661916097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,4,balanced,0.3309760093688965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,4,balanced,0.41978665192921955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,4,balanced,0.5830399990081787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,4,balanced,0.7375199794769287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,256,balanced,0.052000001072883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,4,balanced,1.0920586585998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,4,balanced,1.4146186510721843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,256,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,256,balanced,0.05607999861240387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,256,balanced,0.062218666076660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,256,balanced,0.06016000111897787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,256,balanced,0.05804799993832906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,256,balanced,0.0622026671965917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,256,balanced,0.06442666550477345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,256,balanced,0.06407999992370605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,256,balanced,0.07446399827798207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,256,balanced,0.07438933352629344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,256,balanced,0.07653866708278656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,256,balanced,0.08462400237719218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,256,balanced,0.08878399928410848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,256,balanced,0.10114666819572449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,256,balanced,0.11474133531252544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,256,balanced,0.1467359960079193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,256,balanced,0.1750826636950175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,256,balanced,0.2266719937324524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,256,balanced,0.2747093240420024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,2,balanced,0.10206400354703267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,256,balanced,0.3749493360519409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,2,balanced,0.15052266915639242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,2,balanced,0.25548799832661945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,256,balanced,0.4700479904810588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,2,balanced,0.47221867243448895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,2,balanced,0.8873546918233236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,2,balanced,1.102463960647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,2,balanced,1.1069706281026204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,2,balanced,0.023168000082174938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,2,balanced,0.023472001155217487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,2,balanced,0.025450666745503742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,2,balanced,0.03148266673088074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,2,balanced,0.04669866462548574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,2,balanced,0.06594666838645935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,2,balanced,0.06968000034491222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,2,balanced,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,2,balanced,0.07417066891988118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,2,balanced,0.07666666805744171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,2,balanced,0.08241066833337148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,2,balanced,0.07965866724650066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,2,balanced,0.08278400202592213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,2,balanced,0.087909330924352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,2,balanced,0.08919999996821086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,2,balanced,0.09168000022570293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,2,balanced,0.09904000163078308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,2,balanced,1.1083786487579346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,2,balanced,0.1253600021203359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,2,balanced,0.1328159968058268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,2,balanced,1.1105066935221355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,2,balanced,0.1918613314628601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,2,balanced,0.20457067092259726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,2,balanced,0.2621866663297017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,64,power_law_1.2,0.06053119897842407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,2,balanced,0.29121599594751996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,2,balanced,0.34310932954152423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,64,power_law_1.2,0.05743359923362732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,64,power_law_1.2,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,64,power_law_1.2,0.059308797121047974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,64,power_law_1.2,0.05864959955215454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,2,balanced,0.3924373388290405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,2,balanced,0.6242239872614542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,2,balanced,0.7322346369425455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,2,balanced,1.1092053254445393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,2,balanced,1.1180000305175781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,2,balanced,1.1192959944407146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,64,power_law_1.2,0.059961599111557004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,2,balanced,1.1303733189900715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,64,power_law_1.2,0.05864959955215454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,64,power_law_1.2,0.06092159748077393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,8,balanced,0.037834666669368744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,8,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,64,power_law_1.2,0.06322559714317322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,8,balanced,0.03993066648642222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,8,balanced,0.043978666265805565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,64,power_law_1.2,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,64,power_law_1.2,0.06599680185317994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,64,power_law_1.2,0.0694208025932312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,64,power_law_1.2,0.0750656008720398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,64,power_law_1.2,0.07731199860572815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,64,power_law_1.2,0.08524799942970276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,8,balanced,0.04785599807898203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,64,power_law_1.2,0.09128959774971009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,8,balanced,0.055999999245007835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,8,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,8,balanced,0.05473599831263224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,64,power_law_1.2,0.09479680061340331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,8,balanced,0.05522133409976959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,8,balanced,0.05599466462930044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,64,power_law_1.2,0.10811519622802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,2,balanced,1.1350986957550049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,64,power_law_1.2,0.11723519563674926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,2,balanced,1.1459306875864665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,64,power_law_1.2,0.15201280117034913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,2,balanced,1.1532586415608723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,64,power_law_1.2,0.1701311945915222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,2,balanced,1.1624373594919841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,64,power_law_1.2,0.23739519119262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,2,balanced,1.3590186436971028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,64,power_law_1.2,0.25011839866638186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,8,balanced,0.055957332253456116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,8,balanced,0.05801066756248474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,64,power_law_1.2,0.3342400074005127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,2,balanced,1.2206400235493977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,2,balanced,1.5775465965270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,64,power_law_1.2,0.41743998527526854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,2,balanced,1.3354345957438152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,2,balanced,1.7189599672953289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,8,balanced,0.05927466849486033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,8,balanced,0.05806399881839752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,8,balanced,0.06585066517194112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,8,balanced,0.06655466556549072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,2,balanced,1.5757919947306316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,8,balanced,0.0684799998998642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,8,balanced,0.07680533329645793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,8,balanced,0.08286933104197185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,8,balanced,0.10123200217882793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,8,balanced,0.1135040024916331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,2,balanced,2.603247960408529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,8,balanced,0.1569866637388865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,8,balanced,0.18698134024937949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,8,balanced,0.2586666742960612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,2,balanced,2.4779200553894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,64,power_law_1.2,0.6220928192138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,2,balanced,4.465738614400228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,8,balanced,0.31621867418289185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,8,balanced,0.4397173325220744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,2,balanced,4.644288063049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,8,balanced,0.5684586763381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,64,power_law_1.2,0.7398911952972412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,1,power_law_1.01,0.077811199426651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,1,power_law_1.01,0.08949120044708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,1,power_law_1.01,0.11541759967803955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,1,power_law_1.01,0.15896960496902465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,1,power_law_1.01,0.2050623893737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,1,power_law_1.01,0.2522495985031128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,1,power_law_1.01,0.32371840476989744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,1,power_law_1.01,0.33902080059051515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,1,power_law_1.01,0.35704960823059084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,1,power_law_1.01,0.3577215909957886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,1,power_law_1.01,0.36637439727783205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,1,power_law_1.01,0.3788480043411255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,1,power_law_1.01,0.39451520442962645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,1,power_law_1.01,0.4052159786224365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,1,power_law_1.01,0.41006717681884763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,1,power_law_1.01,0.41914877891540525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,1,power_law_1.01,0.44147839546203616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,1,power_law_1.01,0.5020800113677979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.01,0.5650239944458008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.01,0.671673583984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.01,0.7359936237335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.01,0.9389120101928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.01,1.1357952117919923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.01,1.5287424087524415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.01,1.962495994567871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.01,2.7282304763793945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.01,3.6228736877441405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,4,balanced,0.07293866574764252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,4,balanced,0.0897653301556905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,4,balanced,0.12708800037701926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,4,balanced,0.18726933002471924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,4,balanced,0.29475732644399005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,4,balanced,0.494432012240092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,4,balanced,0.48502933979034424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,4,balanced,0.4728320042292277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,4,balanced,0.4816319942474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,4,balanced,0.4592213233311971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,4,balanced,0.45774932702382404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,4,balanced,0.4440373182296753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,4,balanced,0.4423626661300659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,4,balanced,0.4468373457590739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,4,balanced,0.4630133310953776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,4,balanced,0.4612213373184204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,4,balanced,0.4747733275095622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,4,balanced,0.47853867212931317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,4,balanced,0.4737386703491211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,4,balanced,0.5011786619822184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,4,balanced,0.5114773511886597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,4,balanced,0.5394560098648071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,4,balanced,0.5839626789093018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,4,balanced,0.9772533575693766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,4,balanced,1.0039146741231282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,4,balanced,1.3736906051635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,4,balanced,1.7749706904093425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,1,power_law_1.01,0.07910400032997131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,1,power_law_1.01,0.11236480474472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,1,power_law_1.01,0.13655680418014526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,1,power_law_1.01,0.21318399906158447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,1,power_law_1.01,0.2022847890853882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,1,power_law_1.01,0.21337599754333497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,1,power_law_1.01,0.2143359899520874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,1,power_law_1.01,0.2203007936477661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,1,power_law_1.01,0.2224575996398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,1,power_law_1.01,0.21839361190795897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,1,power_law_1.01,0.2344831943511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,1,power_law_1.01,0.24350719451904296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,1,power_law_1.01,0.24865920543670655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,1,power_law_1.01,0.2585983991622925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,1,power_law_1.01,0.2855295896530151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,1,power_law_1.01,0.2987328052520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,1,power_law_1.01,0.3009727954864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,1,power_law_1.01,0.3497728109359741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,4,power_law_1.01,0.01690240055322647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,4,power_law_1.01,0.016659200191497803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,1,power_law_1.01,0.3864255905151367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,1,power_law_1.01,0.49153919219970704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,4,power_law_1.01,0.016601599752902985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,4,power_law_1.01,0.017510400712490083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,4,power_law_1.01,0.017849600315093993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,4,power_law_1.01,0.018822400271892546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,4,power_law_1.01,0.021184000372886657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,1,power_law_1.01,0.5906943798065185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,4,power_law_1.01,0.018700799345970152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,4,power_law_1.01,0.021529600024223328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,4,power_law_1.01,0.021964800357818604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,4,power_law_1.01,0.021984000504016877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,4,power_law_1.01,0.023231999576091768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,4,power_law_1.01,0.023622399568557738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,1,power_law_1.01,0.7905344009399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,4,power_law_1.01,0.030579200387001036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,4,power_law_1.01,0.031673601269721983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,4,power_law_1.01,0.03223679959774017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,4,power_law_1.01,0.033395200967788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,1,power_law_1.01,0.9880640029907226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,4,power_law_1.01,0.03875199854373932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.01,0.04078719913959503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,8,power_law_1.01,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,8,power_law_1.01,0.08284159898757934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.01,0.05399680137634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,8,power_law_1.01,0.09380480051040649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,8,power_law_1.01,0.10269440412521362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,4,power_law_1.2,0.05449600219726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,4,power_law_1.2,0.05507199764251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.01,0.05940480232238769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,4,power_law_1.2,0.058278399705886844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,1,power_law_1.01,1.3716416358947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,8,power_law_1.01,0.10355199575424194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,8,power_law_1.01,0.10899840593338013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,8,power_law_1.01,0.12106879949569702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,4,power_law_1.2,0.06726400256156921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,8,power_law_1.01,0.12369279861450196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,4,power_law_1.2,0.07024639844894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,8,power_law_1.01,0.12768640518188476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,4,power_law_1.2,0.08780800104141236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,4,power_law_1.2,0.08761600255966187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,4,power_law_1.2,0.09048320055007934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,4,power_law_1.2,0.09285759925842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,4,power_law_1.2,0.09203839898109437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,4,power_law_1.2,0.09546239972114563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,4,power_law_1.2,0.09868159890174866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,1,balanced,0.035775999228159584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,4,power_law_1.2,0.10189440250396728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.01,0.0797760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,4,power_law_1.2,0.10572160482406616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,4,power_law_1.2,0.1205631971359253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,4,power_law_1.2,0.12252160310745239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,4,power_law_1.2,0.12373759746551513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,4,power_law_1.2,0.13571840524673462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,4,power_law_1.2,0.15274879932403565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,1,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,1,balanced,0.04996266464392344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,1,balanced,0.08086400230725606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,1,balanced,0.12285332878430684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,1,balanced,0.12786133090655008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,1,balanced,0.13105600078900656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,1,balanced,0.12711999813715616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,1,balanced,0.12774933377901712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,1,balanced,0.12922666470209757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,1,balanced,0.13131733735402426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,1,balanced,0.13426132996877035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,1,balanced,0.13488533099492392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,1,balanced,0.14058666427930197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,1,balanced,0.14851733048756918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,1,balanced,0.15173866351445517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,8,power_law_1.01,0.1283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,4,power_law_1.2,0.18815360069274903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,8,power_law_1.01,0.12262400388717651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,8,power_law_1.01,0.1347264051437378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,8,power_law_1.01,0.14339200258255005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,8,power_law_1.01,0.1463871955871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,8,power_law_1.01,0.16309759616851807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,8,power_law_1.01,0.16669440269470215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,1,balanced,0.1600320041179657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,1,balanced,0.17299199104309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,1,balanced,0.18621333440144858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,4,power_law_1.2,0.2294912099838257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,1,balanced,0.22748267650604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,1,balanced,0.2480106751124064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,4,power_law_1.2,0.30459520816802976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,1,balanced,0.32656532526016235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,4,power_law_1.2,0.39077761173248293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,4,power_law_1.2,0.5214399814605712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,8,power_law_1.01,0.1471168041229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,8,power_law_1.01,0.16068480014801026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,8,power_law_1.01,0.17201919555664064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,8,power_law_1.01,0.19716479778289794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,1,balanced,0.3905973434448242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,4,power_law_1.2,0.6742527961730957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,8,power_law_1.01,0.22337920665740968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,8,power_law_1.01,0.2733247995376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,8,power_law_1.01,0.31880319118499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,8,power_law_1.01,0.41208319664001464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,8,power_law_1.01,0.5190400123596192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,1,balanced,0.5306453307469686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,1,balanced,0.6577066580454508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,8,power_law_1.01,0.6795519828796387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,1,balanced,0.9376373291015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,8,power_law_1.01,0.8783679962158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,1,balanced,1.2344000339508057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.01,0.09567360281944275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,4,power_law_1.2,0.9790847778320313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,1,power_law_1.01,1.8217279434204101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.01,0.12524160146713256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,1,power_law_1.01,2.601094436645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.01,0.15838719606399537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,4,power_law_1.2,1.3449472427368163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,1,power_law_1.01,3.346144104003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.01,0.22197120189666747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,16,power_law_1.2,0.03930880129337311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,16,power_law_1.2,0.03436160087585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,16,power_law_1.2,0.03361279964447021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,16,power_law_1.2,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,16,power_law_1.2,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,16,power_law_1.2,0.03492479920387268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,16,power_law_1.2,0.03544960021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,16,power_law_1.2,0.03697920143604279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,16,power_law_1.2,0.038867199420928956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,16,power_law_1.2,0.03829120099544525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,16,power_law_1.2,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,16,power_law_1.2,0.040627199411392215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,16,power_law_1.2,0.04184960126876831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,16,power_law_1.2,0.044435200095176694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,16,power_law_1.2,0.05097600221633911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,16,power_law_1.2,0.05573760271072388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,16,power_law_1.2,0.054022401571273804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,16,power_law_1.2,0.06541439890861511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,8,balanced,0.0384853333234787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,8,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,8,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,8,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,8,balanced,0.059157331784566246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,8,balanced,0.08478933572769165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,8,balanced,0.08563199639320374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,8,balanced,0.08479467034339905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,8,balanced,0.08462933699289958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,8,balanced,0.08527466654777527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,8,balanced,0.0848640004793803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,8,balanced,0.08789333701133728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,8,balanced,0.09004799524943034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,8,balanced,0.08987200260162354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,8,balanced,0.09827199578285217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,16,power_law_1.2,0.07233279943466187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,8,balanced,0.09917866190274556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,8,balanced,0.10188266634941101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,8,balanced,0.1114026705423991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,16,power_law_1.2,0.09178239703178406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,16,power_law_1.2,0.10871039628982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,16,power_law_1.2,0.1410048007965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,8,balanced,0.11955199639002483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,8,balanced,0.1372106671333313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,8,balanced,0.1483733355998993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,8,balanced,0.18794665733973184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,8,balanced,0.2125813364982605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,8,balanced,0.2860106627146403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,8,balanced,0.3550399939219157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,8,balanced,0.4989493290583293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,8,balanced,0.6315466562906901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,16,power_law_1.2,0.17677439451217652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,16,power_law_1.2,0.23381121158599855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.01,0.2857408046722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,16,power_law_1.2,0.31781759262084963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,16,power_law_1.2,0.43075199127197267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,8,power_law_1.01,0.06772480010986329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,8,power_law_1.01,0.09473279714584351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,8,power_law_1.01,0.10585600137710571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,8,power_law_1.01,0.12352639436721802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,8,power_law_1.01,0.12665599584579468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,8,power_law_1.01,0.11660799980163575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,8,power_law_1.01,0.14228479862213134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,8,power_law_1.01,0.13084160089492797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,8,power_law_1.01,0.14090240001678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,8,power_law_1.01,0.14075520038604736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,8,power_law_1.01,0.14713599681854247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,8,power_law_1.01,0.1451200008392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,8,power_law_1.01,0.1530303955078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,8,power_law_1.01,0.1558527946472168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,8,power_law_1.01,0.1625856041908264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,8,power_law_1.01,0.1681920051574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,8,power_law_1.01,0.1764032006263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,8,power_law_1.01,0.1944640040397644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,8,power_law_1.01,0.20787200927734376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,8,power_law_1.01,0.2599168062210083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,8,power_law_1.01,0.2816512107849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,8,power_law_1.01,0.3554176092147827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,8,power_law_1.01,0.42087678909301757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,8,power_law_1.01,0.5875455856323242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,8,power_law_1.01,0.7210048198699951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,8,power_law_1.01,1.0211263656616212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,8,power_law_1.01,1.272755241394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,16,power_law_1.2,0.5595392227172852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,2,power_law_1.01,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,2,power_law_1.01,0.0674560010433197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,2,power_law_1.01,0.09205759763717651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,2,power_law_1.01,0.15918079614639283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,2,power_law_1.01,0.1950976014137268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,2,power_law_1.01,0.22829439640045165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,2,power_law_1.01,0.3055295944213867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,2,power_law_1.01,0.32332160472869875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,2,power_law_1.01,0.3280128002166748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,2,power_law_1.01,0.33958399295806885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,2,power_law_1.01,0.3725055932998657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,2,power_law_1.01,0.35628159046173097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,2,power_law_1.01,0.36129920482635497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,2,power_law_1.01,0.38788480758666993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,2,power_law_1.01,0.3791743993759155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,2,power_law_1.01,0.3789504051208496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,2,power_law_1.01,0.39451520442962645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,2,power_law_1.01,0.427507209777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,2,power_law_1.01,0.46166400909423827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,2,power_law_1.01,1.0372735977172851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,2,power_law_1.01,1.178617572784424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,16,power_law_1.2,0.032767999172210696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,2,power_law_1.01,0.7561535835266113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,16,power_law_1.2,0.0336896002292633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,16,power_law_1.2,0.03043839931488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,2,power_law_1.01,0.835654354095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,16,power_law_1.2,0.032307198643684386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,16,power_law_1.2,0.031667199730873105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,2,power_law_1.01,0.9783488273620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,16,power_law_1.2,0.03386879861354828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,16,power_law_1.2,0.03464959859848023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,16,power_law_1.2,0.03560959994792938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,2,power_law_1.01,1.1453503608703612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,16,power_law_1.2,0.036057600378990175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,16,power_law_1.2,0.03697920143604279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,16,power_law_1.2,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,2,power_law_1.01,1.5016127586364747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,16,power_law_1.2,0.03852800130844116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,16,power_law_1.2,0.0395904004573822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,16,power_law_1.2,0.04089600145816803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,2,power_law_1.01,1.8687871932983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,16,power_law_1.2,0.04405120015144348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,16,power_law_1.2,0.04824320077896118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,16,power_law_1.2,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,16,power_law_1.2,0.055641597509384154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,16,power_law_1.2,0.06338559985160827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,16,power_law_1.2,0.07498239874839782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,16,power_law_1.2,0.09021440148353577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,16,power_law_1.2,0.12490880489349365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,2,power_law_1.01,0.12343679666519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,2,power_law_1.01,0.1926911950111389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,2,power_law_1.01,0.23272318840026857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,16,power_law_1.2,0.15671039819717408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,2,power_law_1.01,0.33276801109313964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,2,power_law_1.01,0.44218878746032714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,2,power_law_1.01,0.683519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,2,power_law_1.01,0.7023551940917969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,2,power_law_1.01,0.7519743919372559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,16,power_law_1.2,0.23038721084594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,2,power_law_1.01,0.7250944137573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,2,power_law_1.01,0.7634751796722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,2,power_law_1.01,0.7758848190307617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,16,power_law_1.2,0.30034561157226564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,2,power_law_1.01,0.8004416465759278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,2,power_law_1.01,0.7976704120635987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,2,power_law_1.01,0.8367615699768066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,2,power_law_1.01,0.867801570892334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,2,power_law_1.01,0.8772416114807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,2,power_law_1.01,0.9442303657531739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,2,power_law_1.01,1.0478848457336425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,2,power_law_1.01,1.0490495681762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,2,power_law_1.01,1.2039360046386718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,2,power_law_1.01,1.3002367973327638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,2,power_law_1.01,1.5458815574645997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,2,power_law_1.01,1.6699712753295899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,2,power_law_1.01,2.0713727951049803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,2,power_law_1.01,2.549817657470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,2,power_law_1.01,3.503046417236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,2,power_law_1.01,4.216576004028321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,16,power_law_1.2,0.4199488162994385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,2,power_law_1.01,0.0442111998796463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,2,power_law_1.01,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,2,power_law_1.01,0.05058559775352478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,2,power_law_1.01,0.06488959789276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,2,power_law_1.01,0.06335359811782837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,2,power_law_1.01,0.06622719764709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,2,power_law_1.01,0.06599680185317994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,2,power_law_1.01,0.06744959950447083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,2,power_law_1.01,0.0701632022857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,2,power_law_1.01,0.07133439779281617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,2,power_law_1.01,0.07115520238876342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,2,power_law_1.01,0.07332479953765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,2,power_law_1.01,0.07337599992752075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,2,power_law_1.01,0.07772160172462464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,2,power_law_1.01,0.08603519797325135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,2,power_law_1.01,0.08798080086708068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,2,power_law_1.01,0.107315194606781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,2,power_law_1.01,0.12108160257339477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.01,0.11992319822311401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.01,0.15523200035095214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.01,0.19315840005874635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,16,power_law_1.2,0.5012671947479248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.01,0.2748480081558228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.01,0.28906879425048826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.01,0.462553596496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,4,power_law_1.01,0.15957759618759154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,4,power_law_1.01,0.21689600944519044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,4,power_law_1.01,0.25633280277252196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,4,power_law_1.01,0.3214848041534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,4,power_law_1.01,0.37318398952484133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,4,power_law_1.01,0.4619647979736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,4,power_law_1.01,0.48816637992858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,4,power_law_1.01,0.5343039989471435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,4,power_law_1.01,0.5231872081756592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,4,power_law_1.01,0.5620672225952148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,4,power_law_1.01,0.5279488086700439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,4,power_law_1.01,0.5546624183654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,4,power_law_1.01,0.567961597442627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.01,0.54170241355896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,4,power_law_1.01,0.5701504230499268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,4,power_law_1.01,0.6230527877807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,4,power_law_1.01,0.6136191844940185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,4,power_law_1.01,0.6530176162719726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,4,power_law_1.01,0.7300928115844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,4,power_law_1.01,0.7598336219787598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,4,power_law_1.01,0.8864128112792968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,4,power_law_1.01,0.970803165435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.01,0.7887167930603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,4,power_law_1.01,1.1863360404968262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,4,power_law_1.01,1.241331195831299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,4,power_law_1.01,1.4983231544494628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,4,power_law_1.01,2.140224075317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,4,power_law_1.01,2.7809471130371093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,4,power_law_1.01,3.411347198486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.01,1.0880576133728028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,4,balanced,0.04364799956480662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,4,balanced,0.04394666850566864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,4,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,4,balanced,0.07640000184377034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,4,balanced,0.11209066708882649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,4,balanced,0.17493333419164023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,4,balanced,0.17673067251841226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,4,balanced,0.17797333002090454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,4,balanced,0.17849600315093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,4,balanced,0.17880000670750937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,4,balanced,0.18102933963139853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,4,balanced,0.18248534202575684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,4,balanced,0.18274666865666708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,4,balanced,0.18458133935928345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,4,balanced,0.18856000900268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,4,balanced,0.19340266784032187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,4,balanced,0.1951626737912496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,4,balanced,0.20594666401545206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,4,balanced,0.21458667516708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,4,balanced,0.23185600837071738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,4,balanced,0.24780799945195517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,4,balanced,0.29739733537038165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,4,balanced,0.31835732857386273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,4,balanced,0.4086666504542033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,4,balanced,0.4777439832687378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,4,balanced,0.6575093269348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,4,balanced,0.7952106793721517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,128,balanced,0.05808533231417338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,128,balanced,0.06017066538333893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,128,balanced,0.047770669062932335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,128,balanced,0.05049600203831991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,128,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,128,balanced,0.052101333936055504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,128,balanced,0.06866133213043213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,128,balanced,0.06862399975458781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,128,balanced,0.06824000179767609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,128,balanced,0.06947200000286102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,128,balanced,0.06977066894372304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,128,balanced,0.07010133564472198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,128,balanced,0.0721013347307841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,128,balanced,0.07453866799672444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,128,balanced,0.07905599971612294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,128,balanced,0.08024533092975616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,128,balanced,0.08447999755541484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,128,balanced,0.09047999978065491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,128,balanced,0.094842662413915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,128,balanced,0.10736533006032307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,128,balanced,0.11799466609954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,128,balanced,0.14147733648618063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,128,balanced,0.16567466656366983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,128,balanced,0.2116960088411967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,128,balanced,0.25685866673787433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,128,balanced,0.355679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,128,balanced,0.43589333693186444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,2,balanced,0.056176001826922096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,2,balanced,0.060032000144322716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,2,balanced,0.0660159985224406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,2,balanced,0.08896533648173015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,2,balanced,0.11888532837231953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,2,balanced,0.13979732990264893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,2,balanced,0.14011733730634054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,2,balanced,0.1386186679204305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,2,balanced,0.1383519967397054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,2,balanced,0.142767995595932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,2,balanced,0.14154133200645447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,2,balanced,0.14569600423177084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,2,balanced,0.14940266807874045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,2,balanced,0.15171200037002563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,2,balanced,0.15847466389338175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,2,balanced,0.16099199652671814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,2,balanced,0.16828266779581705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,2,balanced,0.1916266679763794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,2,balanced,0.2127359906832377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,2,balanced,0.25778667132059735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,2,balanced,0.2885439991950989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,2,balanced,0.4192266861597697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,2,balanced,0.4898453156153361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,2,balanced,0.7373653252919515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,2,balanced,0.9078346888224283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,2,balanced,1.3366986910502117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,2,balanced,1.691941261291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,1,balanced,0.09515200058619182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,1,balanced,0.10322133700052898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,1,balanced,0.1316426694393158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,1,balanced,0.1876479983329773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,1,balanced,0.2697920004526774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,1,balanced,0.42981334527333576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,1,balanced,0.43728001912434894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,1,balanced,0.4255359967549642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,1,balanced,0.43545599778493244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,1,balanced,0.42846401532491046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,1,balanced,0.43486400445302326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,1,balanced,0.4325600067774455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,1,balanced,0.4393226703008016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,1,balanced,0.4487733443578084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,1,balanced,0.45182931423187256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,1,balanced,0.4580693244934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,1,balanced,0.4768426815668742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,1,balanced,0.5065279801686605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,1,balanced,0.5318506558736166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,1,balanced,0.601418654123942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,1,balanced,0.6523893276850382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,1,balanced,0.797205368677775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,1,balanced,0.9268373648325602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,1,balanced,1.433370590209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,1,balanced,1.691482702891032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,1,balanced,2.420970598856608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,1,balanced,3.1614131927490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,16,power_law_1.01,0.09100800156593322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,16,power_law_1.01,0.12023680210113526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,16,power_law_1.01,0.11835520267486573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,16,power_law_1.01,0.11851520538330078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,16,power_law_1.01,0.12135679721832275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,16,power_law_1.01,0.10579839944839478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,16,power_law_1.01,0.1103551983833313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,16,power_law_1.01,0.11397759914398194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,16,power_law_1.01,0.11756800413131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,16,power_law_1.01,0.11278079748153687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,16,power_law_1.01,0.10774400234222412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,16,power_law_1.01,0.11395200490951538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,16,power_law_1.01,0.11660799980163575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,16,power_law_1.01,0.11418880224227905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,16,power_law_1.01,0.12396160364151002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,16,power_law_1.01,0.12548480033874512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,16,power_law_1.01,0.12452479600906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,16,power_law_1.01,0.13966720104217528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,16,power_law_1.01,0.1476096034049988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,16,power_law_1.01,0.1732095956802368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,16,power_law_1.01,0.19868799448013305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,16,power_law_1.01,0.24547200202941893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,16,power_law_1.01,0.28217599391937254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,16,power_law_1.01,0.3812096118927002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,16,power_law_1.01,0.4961984157562256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,16,power_law_1.01,0.657535982131958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,16,power_law_1.01,0.9093055725097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,1,power_law_1.01,0.13747199773788452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,1,power_law_1.01,0.2203455924987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,1,power_law_1.01,0.38083839416503906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,1,power_law_1.01,0.7243584156036377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,1,power_law_1.01,1.0937151908874512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,1,power_law_1.01,1.3830975532531737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,1,power_law_1.01,1.9927871704101563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,1,power_law_1.01,2.0912256240844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,1,power_law_1.01,2.1139392852783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,1,power_law_1.01,2.2170688629150392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,1,power_law_1.01,2.292038345336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,1,power_law_1.01,2.3134016036987304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,1,power_law_1.01,2.407040023803711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,1,power_law_1.01,2.4476736068725584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,1,power_law_1.01,2.506598472595215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,1,power_law_1.01,2.5288000106811523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,1,power_law_1.01,2.6962047576904298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,1,power_law_1.01,2.879347229003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,1,power_law_1.01,2.9771583557128904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,1,power_law_1.01,3.257126235961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,1,balanced,0.0876586635907491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,1,balanced,0.10225066542625427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,1,balanced,0.13306666413942972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,1,power_law_1.01,3.1067583084106447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,1,power_law_1.01,3.5433727264404298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,1,balanced,0.18932799498240152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,1,balanced,0.2836853265762329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,1,balanced,0.4446186621983846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,1,balanced,0.4411253531773885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,1,power_law_1.01,4.215564727783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,1,power_law_1.01,4.938860702514648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,1,balanced,0.419594685236613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,1,balanced,0.4108426570892334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,1,balanced,0.41139201323191327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,1,balanced,0.4141013224919637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,1,balanced,0.41732267538706463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,1,balanced,0.42209601402282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,1,balanced,0.42709867159525555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,1,balanced,0.42579734325408936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,1,power_law_1.01,5.503776168823242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,1,balanced,0.43352532386779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,1,balanced,0.44580801328023273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,1,balanced,0.4804106553395589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,1,balanced,0.510042667388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,1,balanced,0.5618933439254761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,1,balanced,0.612394650777181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,1,balanced,0.7295253276824951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,1,balanced,0.8552213509877523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,64,power_law_1.01,0.0833728015422821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,1,power_law_1.01,7.0816703796386715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,1,balanced,1.3335572878519695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,1,balanced,1.5835946400960286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,1,balanced,2.317728042602539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,64,power_law_1.01,0.08041599988937378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,64,power_law_1.01,0.08017280101776122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,64,power_law_1.01,0.08247039914131164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,1,balanced,3.0210132598876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,64,power_law_1.01,0.0828927993774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,64,power_law_1.01,0.08223999738693237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,64,power_law_1.01,0.08985599875450134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,64,power_law_1.01,0.0878271996974945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,1,power_law_1.01,8.703091430664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,64,power_law_1.01,0.08613759875297547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,64,power_law_1.01,0.0878271996974945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,64,power_law_1.01,0.09005439877510071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,64,power_law_1.01,0.09115520119667053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,64,power_law_1.01,0.08815360069274902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,64,power_law_1.01,0.08952320218086243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,64,power_law_1.01,0.09444479942321778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,64,power_law_1.01,0.09516800045967103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,64,power_law_1.01,0.09671679735183716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,64,power_law_1.01,0.10458240509033204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,64,power_law_1.01,0.1133247971534729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,64,power_law_1.01,0.12476160526275634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,64,power_law_1.01,0.1530943989753723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,64,power_law_1.01,0.18798719644546508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,64,power_law_1.01,0.24558720588684083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,64,power_law_1.01,0.29522559642791746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,64,power_law_1.01,0.3661760091781616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,64,power_law_1.01,0.5099520206451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,64,power_law_1.01,0.689299201965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,16,power_law_1.01,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,16,power_law_1.01,0.031065601110458373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,16,power_law_1.01,0.02988159954547882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,16,power_law_1.01,0.03047040104866028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,16,power_law_1.01,0.031539198756217954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,16,power_law_1.01,0.03208959996700287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,16,power_law_1.01,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,16,power_law_1.01,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,16,power_law_1.01,0.03310079872608185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,16,power_law_1.01,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,16,power_law_1.01,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,16,power_law_1.01,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,16,power_law_1.01,0.035795199871063235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,16,power_law_1.01,0.03653759956359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,16,power_law_1.01,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,16,power_law_1.01,0.04257279932498932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,16,power_law_1.01,0.04387840032577515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,16,power_law_1.01,0.04681600034236908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,16,power_law_1.01,0.05130879878997803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,16,power_law_1.01,0.05674880146980286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,16,power_law_1.01,0.06689280271530151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,16,power_law_1.01,0.08232319951057435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,16,power_law_1.01,0.09423360228538513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,16,power_law_1.01,0.13537919521331787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,16,power_law_1.01,0.1755328059196472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,16,power_law_1.01,0.2527935981750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,16,power_law_1.01,0.3224384069442749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,16,power_law_1.2,0.06741759777069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,128,power_law_1.2,0.07975680232048035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,16,power_law_1.2,0.08848000168800355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,128,power_law_1.2,0.08803200125694274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,16,power_law_1.2,0.09907199740409851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,128,power_law_1.2,0.06408960223197938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,16,power_law_1.2,0.10063999891281128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,128,power_law_1.2,0.06847360134124755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,16,power_law_1.2,0.10093439817428589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,128,power_law_1.2,0.06330239772796631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,16,power_law_1.2,0.10087679624557495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,128,power_law_1.2,0.06480000019073487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,16,power_law_1.2,0.08981119990348815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,128,power_law_1.2,0.06371200084686279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,128,power_law_1.2,0.06865280270576476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,16,power_law_1.2,0.09838079810142517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,128,power_law_1.2,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,16,power_law_1.2,0.09901440143585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,128,power_law_1.2,0.06765440106391907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,16,power_law_1.2,0.09758080244064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,128,power_law_1.2,0.0699072003364563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,16,power_law_1.2,0.09957759976387023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,128,power_law_1.2,0.06996480226516724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,16,power_law_1.2,0.10291839838027954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,16,power_law_1.2,0.10224000215530396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,16,power_law_1.2,0.10631040334701539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,128,power_law_1.2,0.07221760153770447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,16,power_law_1.2,0.10918400287628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,128,power_law_1.2,0.07447680234909057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,128,power_law_1.2,0.08362240195274354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,128,power_law_1.2,0.0797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,128,power_law_1.2,0.08972160220146179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,128,power_law_1.2,0.10014079809188843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,16,power_law_1.2,0.11511679887771606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,16,power_law_1.2,0.12520320415496827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,16,power_law_1.2,0.14487680196762084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,16,power_law_1.2,0.1538815975189209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,16,power_law_1.2,0.17815040349960326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,128,power_law_1.2,0.11568640470504761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,16,power_law_1.2,0.20454399585723876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,128,power_law_1.2,0.15103360414505004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,16,power_law_1.2,0.2776063919067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,128,power_law_1.2,0.1510975956916809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,16,power_law_1.2,0.3037760019302368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,128,power_law_1.2,0.20142719745635987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,16,power_law_1.2,0.42644481658935546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,128,power_law_1.2,0.2384768009185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,128,power_law_1.2,0.3377919912338257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,128,power_law_1.2,0.4248191833496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,128,power_law_1.2,0.7124800205230712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,128,power_law_1.2,0.7415167808532714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,16,power_law_1.2,0.5443456172943115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,16,power_law_1.2,0.7939263820648194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,16,power_law_1.2,0.9695743560791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,4,power_law_1.01,0.02590720057487488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,4,power_law_1.01,0.025824001431465148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,4,power_law_1.01,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,4,power_law_1.01,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,4,power_law_1.01,0.0438975989818573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,4,power_law_1.01,0.04760960042476654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,4,power_law_1.01,0.047788798809051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,4,power_law_1.01,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,4,power_law_1.01,0.05126399993896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,4,power_law_1.01,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,4,power_law_1.01,0.053491199016571046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,4,power_law_1.01,0.05580800175666809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,4,power_law_1.01,0.05775359869003296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,4,power_law_1.01,0.06531839966773986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,4,power_law_1.01,0.06877440214157104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,4,power_law_1.01,0.07143679857254029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,4,power_law_1.01,0.07215359807014465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,4,power_law_1.01,0.074726402759552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,4,power_law_1.01,0.09015679955482483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,4,power_law_1.01,0.10409599542617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,4,power_law_1.01,0.10604159832000733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,4,power_law_1.01,0.1347839951515198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,4,power_law_1.01,0.15406719446182252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,4,power_law_1.01,0.20665600299835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,4,power_law_1.01,0.25207679271697997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,4,power_law_1.01,0.3817728042602539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,4,power_law_1.01,0.46448001861572263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,16,4,power_law_1.2,0.03247359991073608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,16,4,power_law_1.2,0.0326335996389389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,16,4,power_law_1.2,0.033555200695991515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,16,4,power_law_1.2,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,16,4,power_law_1.2,0.03651840090751648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,16,4,power_law_1.2,0.037747201323509214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,16,4,power_law_1.2,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,16,4,power_law_1.2,0.03873279988765717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,16,4,power_law_1.2,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,16,4,power_law_1.2,0.03988479971885681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,16,4,power_law_1.2,0.040275201201438904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,16,4,power_law_1.2,0.043289598822593686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,16,4,power_law_1.2,0.04599039852619171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,16,4,power_law_1.2,0.04926080107688904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,16,4,power_law_1.2,0.058316802978515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,16,4,power_law_1.2,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,16,4,power_law_1.2,0.07407360076904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,16,4,power_law_1.2,0.09816319942474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,16,4,power_law_1.2,0.10727039575576783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,16,4,power_law_1.2,0.1500864028930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,16,4,power_law_1.2,0.18227839469909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,16,4,power_law_1.2,0.2616447925567627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,16,4,power_law_1.2,0.32512640953063965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,16,4,power_law_1.2,0.5420095920562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,32,balanced,0.04921066761016846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,32,balanced,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,32,balanced,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,32,balanced,0.05630933245023092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,32,balanced,0.056176001826922096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,32,balanced,0.06232533355553945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,32,balanced,0.06623999774456024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,32,balanced,0.06437333424886067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,32,balanced,0.06607999900976817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,32,balanced,0.06589866677920024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,1,balanced,0.07948266466458638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,1,balanced,0.11749866604804993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,1,balanced,0.1871839960416158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,1,balanced,0.18579200903574625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,1,balanced,0.18686934312184653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,1,balanced,0.17663466930389404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,1,balanced,0.19316800435384116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,1,balanced,0.18269866704940796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,1,balanced,0.18054932355880737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,1,balanced,0.18516800800959268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,1,balanced,0.1805973251660665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,32,balanced,0.06619733572006226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,32,balanced,0.0661599983771642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,32,balanced,0.06836799780527751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,32,balanced,0.06855999926726024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,32,balanced,0.07517866790294647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,32,balanced,0.07442666590213776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,1,power_law_1.2,0.11737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,32,balanced,0.0765173335870107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,32,balanced,0.08477333188056946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,32,balanced,0.08919999996821086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,32,balanced,0.103301336367925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,32,balanced,0.11332799990971883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,32,balanced,0.14689600467681885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,32,balanced,0.17082667350769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,1,balanced,0.16173866391181946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,1,balanced,0.17321600516637167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,1,balanced,0.16982932885487875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,1,balanced,0.16475199659665427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,16,4,power_law_1.2,0.6694719791412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,1,balanced,0.15359999736150107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,1,power_law_1.2,0.16357760429382323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,1,balanced,0.15450132886568704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,1,power_law_1.2,0.24216320514678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,1,balanced,0.2690933346748352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,32,balanced,0.22657066583633423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,32,balanced,0.2779093384742737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,32,balanced,0.38302401701609295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,1,balanced,0.22638932863871256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,32,balanced,0.48786131540934247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,1,balanced,0.30423466364542645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,1,balanced,0.37692801157633465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,1,balanced,0.5343733231226603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,1,power_law_1.2,0.3663232088088989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,1,power_law_1.2,0.36382079124450684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,1,balanced,0.6870293617248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,1,power_law_1.2,0.3840255975723267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,1,power_law_1.2,0.38575360774993894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,1,balanced,0.9938186804453532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,1,power_law_1.2,0.39025919437408446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,1,balanced,1.3075306415557861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,1,balanced,1.9148747126261394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,1,balanced,2.5466720263163247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,1,power_law_1.2,0.3975744009017944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,1,power_law_1.2,0.39669759273529054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,1,power_law_1.2,0.4127999782562256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,1,power_law_1.2,0.42783360481262206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,1,power_law_1.2,0.4265024185180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,1,power_law_1.2,0.42598400115966795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,1,power_law_1.2,0.47292160987854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,1,power_law_1.2,0.4809855937957764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,1,power_law_1.2,0.541926383972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,1,power_law_1.2,0.598476791381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,1,power_law_1.2,0.703436803817749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,1,power_law_1.2,0.8283072471618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,1,power_law_1.2,1.0026240348815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,1,power_law_1.2,1.301030445098877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,16,4,power_law_1.2,1.0524415969848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,8,power_law_1.01,0.12577919960021972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,8,power_law_1.01,0.17498879432678222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,8,power_law_1.01,0.19993599653244018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,8,power_law_1.01,0.2336127996444702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,8,power_law_1.01,0.23384320735931396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,8,power_law_1.01,0.23399040699005128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,8,power_law_1.01,0.2531008005142212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,8,power_law_1.01,0.24605441093444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,8,power_law_1.01,0.24778239727020263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,8,power_law_1.01,0.25948801040649416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,8,power_law_1.01,0.2504319906234741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,8,power_law_1.01,0.24851839542388915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,8,power_law_1.01,0.2637439966201782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,8,power_law_1.01,0.2732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,8,power_law_1.01,0.28394880294799807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,8,power_law_1.01,0.2669248104095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,8,power_law_1.01,0.28556160926818847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,8,power_law_1.01,0.29690239429473875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,8,power_law_1.01,0.3091007947921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,8,power_law_1.01,0.3560703992843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,8,power_law_1.01,0.38826239109039307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,8,power_law_1.01,0.49429759979248045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,1,power_law_1.2,1.6769792556762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,8,power_law_1.01,0.5786367893218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,1,power_law_1.2,2.4068864822387694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,16,4,power_law_1.2,1.391481590270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,1,power_law_1.2,3.0834495544433596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,1,power_law_1.01,0.056704002618789676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,1,power_law_1.01,0.07150080204010009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,1,power_law_1.01,0.08793600201606751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,1,power_law_1.01,0.13135360479354857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,1,power_law_1.2,4.37017593383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,1,power_law_1.01,0.13349119424819947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,1,power_law_1.01,0.13549439907073973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,1,power_law_1.01,0.13768320083618163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,1,power_law_1.01,0.13608319759368898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,1,power_law_1.01,0.1398848056793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,1,power_law_1.01,0.1405311942100525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,1,power_law_1.01,0.14183679819107056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,1,power_law_1.01,0.1501952052116394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,1,power_law_1.01,0.1512511968612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,1,power_law_1.01,0.16062719821929933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,1,power_law_1.01,0.18098560571670533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,1,power_law_1.01,0.18234879970550538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,1,power_law_1.01,0.18355200290679932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,32,power_law_1.2,0.02242559939622879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,32,power_law_1.2,0.02245119959115982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,32,power_law_1.2,0.021740800142288207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,32,power_law_1.2,0.022111999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,32,power_law_1.2,0.02374400049448013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,1,power_law_1.01,0.21595520973205568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,32,power_law_1.2,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,1,power_law_1.01,0.24196479320526124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,32,power_law_1.2,0.050886398553848265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,32,power_law_1.2,0.05017600059509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,32,power_law_1.2,0.04999679923057556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,32,power_law_1.2,0.0465472012758255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,32,power_law_1.2,0.04718720018863678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,32,power_law_1.2,0.04642559885978699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,1,power_law_1.01,0.29404799938201903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,32,power_law_1.2,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,1,power_law_1.01,0.3447743892669678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,32,power_law_1.2,0.047200000286102294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,32,power_law_1.2,0.04570879936218262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,32,power_law_1.2,0.04519680142402649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,32,power_law_1.2,0.047679999470710756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,32,power_law_1.2,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.2,0.05424000024795532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.2,0.069651198387146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.2,0.0856383979320526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.2,0.10915199518203736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.2,0.12961280345916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.2,0.18744319677352905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,32,power_law_1.2,0.06702719926834107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,32,power_law_1.2,0.06515840291976929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.2,0.24978559017181395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,32,power_law_1.2,0.06532480120658875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.2,0.3278847932815552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,32,power_law_1.2,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,32,power_law_1.2,0.07090560197830201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,32,power_law_1.2,0.06518399715423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,32,power_law_1.2,0.06968320012092591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,1,power_law_1.01,0.44897918701171874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,32,power_law_1.2,0.07336320281028748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,32,power_law_1.2,0.07408000230789184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,32,power_law_1.2,0.07501440048217774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,32,power_law_1.2,0.07793279886245727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,32,power_law_1.2,0.07953280210494995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,32,power_law_1.2,0.08238080143928528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,32,power_law_1.2,0.08027520179748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,32,power_law_1.2,0.08761600255966187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,32,power_law_1.2,0.0975488007068634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,32,power_law_1.2,0.10168960094451904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,32,power_law_1.2,0.12038400173187255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,32,power_law_1.2,0.12698880434036255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,32,power_law_1.2,0.1581760048866272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,32,power_law_1.2,0.18759039640426636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,32,power_law_1.2,0.25005440711975097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,32,power_law_1.2,0.28884479999542234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,1,power_law_1.01,0.5286655902862549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,32,power_law_1.2,0.3797247886657715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,32,power_law_1.2,0.5128704071044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,32,power_law_1.2,0.7317376136779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,32,power_law_1.2,0.9324543952941895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.2,0.4267136096954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,1,power_law_1.01,0.7633088111877442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,1,power_law_1.2,5.701958465576172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,1,power_law_1.01,0.9496576309204101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,1,power_law_1.01,1.3523839950561523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,4,power_law_1.01,0.05377920269966126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,1,power_law_1.01,1.7649663925170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,4,power_law_1.01,0.060121601819992064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,4,power_law_1.01,0.06931840181350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,4,power_law_1.01,0.08170880079269409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,4,power_law_1.01,0.09060479998588562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,4,power_law_1.01,0.11605759859085082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,4,power_law_1.01,0.11841280460357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,4,power_law_1.01,0.11662080287933349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,4,power_law_1.01,0.11498240232467652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,4,power_law_1.01,0.11623040437698365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,4,power_law_1.01,0.1211583971977234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,4,power_law_1.01,0.11991679668426514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,4,power_law_1.01,0.12715519666671754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,4,power_law_1.01,0.12930560111999512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,4,power_law_1.01,0.13490560054779052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,4,power_law_1.01,0.14337919950485228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,4,power_law_1.01,0.14569599628448487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,4,power_law_1.01,0.16284159421920777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,4,power_law_1.01,0.16954879760742186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,4,power_law_1.01,0.20097920894622803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,4,power_law_1.01,0.2291264057159424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,4,power_law_1.01,0.2660223960876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,4,power_law_1.01,0.32899200916290283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,4,power_law_1.01,0.40524802207946775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,4,power_law_1.01,0.503276777267456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,4,power_law_1.01,0.6635968208312988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,4,power_law_1.01,0.9322048187255859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,256,power_law_1.2,0.021209600567817687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,256,power_law_1.2,0.020908799767494202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,256,power_law_1.2,0.02186879962682724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,256,power_law_1.2,0.026080000400543212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,256,power_law_1.2,0.04924159944057464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,256,power_law_1.2,0.03829120099544525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,256,power_law_1.2,0.03873279988765717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,256,power_law_1.2,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,256,power_law_1.2,0.039340800046920775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,256,power_law_1.2,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,256,power_law_1.2,0.03903360068798065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,256,power_law_1.2,0.03961600065231323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,256,power_law_1.2,0.03827199935913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,256,power_law_1.2,0.03898879885673523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,256,power_law_1.2,0.03873920142650604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,256,power_law_1.2,0.04035840034484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,256,power_law_1.2,0.04195840060710907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,256,power_law_1.2,0.045311999320983884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,256,power_law_1.2,0.051500797271728516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,256,power_law_1.2,0.06679040193557739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,256,power_law_1.2,0.07559679746627808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,256,power_law_1.2,0.10574079751968384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,256,power_law_1.2,0.12372479438781739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,256,power_law_1.2,0.18077440261840821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,256,power_law_1.2,0.2365567922592163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,256,power_law_1.2,0.3293951988220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,256,power_law_1.2,0.42879362106323243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,8,power_law_1.2,0.18480000495910645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,8,power_law_1.2,0.2480639934539795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,8,power_law_1.2,0.29109759330749513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,8,power_law_1.2,0.32678399085998533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,8,power_law_1.2,0.39310081005096437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,8,power_law_1.2,0.42557439804077146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,8,power_law_1.2,0.3927040100097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,8,power_law_1.2,0.4134208202362061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,8,power_law_1.2,0.4205376148223877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,8,power_law_1.2,0.4000063896179199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,8,power_law_1.2,0.41169281005859376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,8,power_law_1.2,0.43221120834350585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,8,power_law_1.2,0.4308288097381592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,8,power_law_1.2,0.47005438804626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,8,power_law_1.2,0.46883840560913087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,8,power_law_1.2,0.4905792236328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,8,power_law_1.2,0.5417791843414307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,8,power_law_1.2,0.616319990158081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,8,power_law_1.2,0.5910079956054688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,8,power_law_1.2,0.7235392093658447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,8,power_law_1.2,0.7559232234954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,8,power_law_1.2,0.9642751693725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,8,power_law_1.2,1.0067456245422364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,8,power_law_1.2,1.2725760459899902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,8,power_law_1.2,1.606585693359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,8,power_law_1.2,2.354252815246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,8,power_law_1.2,2.5488000869750977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,1,balanced,0.07075199981530507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,1,balanced,0.0900320013364156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,1,balanced,0.12950399518013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,1,balanced,0.20617065827051798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,1,balanced,0.3435680071512858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,1,balanced,0.4761279821395874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,1,balanced,0.48130667209625244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,1,balanced,0.4811520179112752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,1,balanced,0.4850986798604329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,1,balanced,0.4886133273442586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,1,balanced,0.4915359814961751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,1,balanced,0.5006080071131388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,1,balanced,0.5026880105336508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,1,balanced,0.5105546712875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,1,balanced,0.5218506654103597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,1,balanced,0.5313760042190552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,1,balanced,0.5570346514383951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,1,balanced,0.6026240189870199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,1,balanced,0.6402399937311808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,64,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,1,balanced,0.9218826293945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,1,balanced,0.8920533657073975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,1,balanced,1.422229290008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,1,balanced,1.4392852783203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,64,balanced,0.04969066878159841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,64,balanced,0.04994666576385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,64,balanced,0.053504000107447304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,64,balanced,0.0544106662273407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,64,balanced,0.062021334966023765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,1,balanced,2.173802693684896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,64,balanced,0.06225599845250448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,64,balanced,0.06020266811052958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,64,balanced,0.06198399762312571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,64,balanced,0.06228800117969513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,64,balanced,0.06364266574382782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,64,balanced,0.06394133468468984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,64,balanced,0.06645866731802623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,64,balanced,0.06634666522343953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,64,balanced,0.07181866466999054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,64,balanced,0.0702400008837382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,64,balanced,0.07425599793593089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,64,balanced,0.08066666622956593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,64,balanced,0.08469333251317342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,64,balanced,0.09718400239944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,64,balanced,0.10745066404342651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,64,balanced,0.1376533309618632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,64,balanced,0.16446933150291443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,64,balanced,0.21552000443140665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,64,balanced,0.26473067204157513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,64,balanced,0.36210131645202637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,64,balanced,0.4604213237762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,1,balanced,2.47106138865153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,1,balanced,3.722026824951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,4,balanced,0.07018666466077168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,4,balanced,0.09475732843081157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,4,balanced,0.14656000336011252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,4,balanced,0.2511199911435445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,4,balanced,0.4445706605911255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,4,balanced,0.8347626527150472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,4,balanced,1.2302453517913818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,4,balanced,1.228432019551595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,1,balanced,4.577269236246745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,4,balanced,1.225552002588908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,4,balanced,1.2278772989908855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,4,balanced,1.237829367319743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,4,balanced,1.2454133033752441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,4,balanced,1.2464853127797444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,4,balanced,1.2439146836598713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,4,balanced,1.2535839875539143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,4,balanced,1.2668053309122722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,4,balanced,1.2723039786020915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,4,balanced,1.2947253386179607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,4,balanced,1.3048160076141357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,4,balanced,1.3411787350972493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,4,balanced,1.385632038116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,4,balanced,1.4642079671223958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,4,balanced,1.525045394897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,4,balanced,1.6474080085754395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,4,balanced,1.8406559626261394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,4,balanced,2.068277359008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,4,balanced,2.8983627955118814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,8,power_law_1.2,0.03054080009460449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,8,power_law_1.2,0.028467199206352232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,8,power_law_1.2,0.028531199693679808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,8,power_law_1.2,0.030214399099349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,8,power_law_1.2,0.030796799063682555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,8,power_law_1.2,0.03139840066432953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,8,power_law_1.2,0.03166080117225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,8,power_law_1.2,0.03258239924907684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,8,power_law_1.2,0.033478400111198424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,8,power_law_1.2,0.03402880132198334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,8,power_law_1.2,0.03340159952640533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,8,power_law_1.2,0.03586559891700745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,8,power_law_1.2,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,8,power_law_1.2,0.03736959993839264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,8,power_law_1.2,0.044972801208496095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,8,power_law_1.2,0.04508799910545349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,2,power_law_1.01,0.02449280023574829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,2,power_law_1.01,0.038022398948669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,8,power_law_1.2,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,2,power_law_1.01,0.05013120174407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,2,power_law_1.01,0.06811519861221313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,8,power_law_1.2,0.05904639959335327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,2,power_law_1.01,0.0877951979637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,2,power_law_1.01,0.10431360006332398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,2,power_law_1.01,0.12902400493621827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,2,power_law_1.01,0.12837120294570922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,2,power_law_1.01,0.14312959909439088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,2,power_law_1.01,0.13678079843521118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,2,power_law_1.01,0.14755840301513673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,2,power_law_1.2,0.06376320123672485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,8,power_law_1.2,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,2,power_law_1.2,0.07039999961853027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,2,power_law_1.2,0.08303999900817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,2,power_law_1.2,0.1021183967590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,2,power_law_1.01,0.15011199712753295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,2,power_law_1.2,0.12065279483795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,2,power_law_1.01,0.14996479749679564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,2,power_law_1.2,0.14006400108337402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,2,power_law_1.2,0.1792255997657776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,2,power_law_1.01,0.16447999477386474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,2,power_law_1.2,0.1803712010383606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,2,power_law_1.01,0.15722880363464356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,2,power_law_1.01,0.15782400369644164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,8,power_law_1.2,0.07491840124130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,2,power_law_1.01,0.16889599561691285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,2,power_law_1.01,0.18435839414596558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,2,power_law_1.01,0.19668480157852172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,2,power_law_1.01,0.4054272174835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,2,power_law_1.2,0.1868288040161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,2,power_law_1.2,0.19522559642791748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,2,power_law_1.2,0.19612799882888793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,2,power_law_1.01,0.4787712097167969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,2,power_law_1.2,0.1990399956703186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,2,power_law_1.2,0.20577919483184814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,2,power_law_1.01,0.3401535987854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,2,power_law_1.2,0.2116863965988159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,2,power_law_1.01,0.3479552030563354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,2,power_law_1.2,0.2267008066177368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,2,power_law_1.01,0.4156352043151855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,2,power_law_1.2,0.22537600994110107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,2,power_law_1.01,0.4831679821014404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,2,power_law_1.2,0.24190080165863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,2,power_law_1.01,0.6430335998535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,2,power_law_1.2,0.2656192064285278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,2,power_law_1.01,0.8152895927429199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,2,power_law_1.2,0.29748480319976806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,8,power_law_1.2,0.08535040020942689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,2,power_law_1.2,0.36129279136657716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,2,power_law_1.2,0.40227198600769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,2,power_law_1.2,0.5083456039428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,2,power_law_1.2,0.6372543811798096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,2,power_law_1.2,0.8628543853759766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,8,power_law_1.2,0.12122880220413208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,2,power_law_1.2,1.089139175415039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,2,power_law_1.2,1.4473983764648437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,2,power_law_1.2,2.08287353515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,8,power_law_1.2,0.1464959979057312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,8,power_law_1.2,0.2003711938858032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,4,power_law_1.01,0.018374399840831758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,4,power_law_1.01,0.017696000635623932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,4,power_law_1.01,0.017484800517559053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,4,power_law_1.01,0.01823360025882721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,4,power_law_1.01,0.018611200153827667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,4,power_law_1.01,0.019193600118160247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,4,power_law_1.01,0.02253440022468567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,4,power_law_1.01,0.021670399606227873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,4,power_law_1.01,0.039743998646736146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,4,power_law_1.01,0.03969280123710632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,4,power_law_1.01,0.03989120125770569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,4,power_law_1.01,0.03462400138378143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,4,power_law_1.01,0.03380480110645294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,4,power_law_1.01,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,4,power_law_1.01,0.0347135990858078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,4,power_law_1.01,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,4,power_law_1.01,0.03585279881954193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,4,power_law_1.01,0.046035200357437134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.01,0.0482367992401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.01,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,8,power_law_1.2,0.2710335969924927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.01,0.076665598154068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.01,0.09729920029640197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.01,0.12003200054168701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.01,0.1641983985900879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.01,0.20675840377807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,8,power_law_1.2,0.37112319469451904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,32,power_law_1.2,0.04160000085830688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,32,1,power_law_1.2,0.021695999801158904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,32,power_law_1.2,0.04035199880599975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,32,1,power_law_1.2,0.021913599967956544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,32,power_law_1.2,0.038278400897979736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,32,1,power_law_1.2,0.02418559938669205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,32,1,power_law_1.2,0.029100799560546876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,32,1,power_law_1.2,0.036627200245857236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,32,1,power_law_1.2,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,32,1,power_law_1.2,0.06392959952354431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,32,1,power_law_1.2,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,32,power_law_1.2,0.041145598888397215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,32,1,power_law_1.2,0.07028480172157288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,32,power_law_1.2,0.040300801396369934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,32,1,power_law_1.2,0.07265920042991639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,32,power_law_1.2,0.04192639887332916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,32,1,power_law_1.2,0.07863039970397949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,32,power_law_1.2,0.04270080029964447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,32,1,power_law_1.2,0.08387839794158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,32,power_law_1.2,0.04385280013084412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,32,1,power_law_1.2,0.09130240082740784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,32,1,power_law_1.2,0.1003648042678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,32,1,power_law_1.2,0.10584959983825684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,32,1,power_law_1.2,0.11055999994277954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,32,power_law_1.2,0.04431999921798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,32,1,power_law_1.2,0.1270848035812378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,32,power_law_1.2,0.04446719884872437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,32,1,power_law_1.2,0.12373119592666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,32,power_law_1.2,0.05022720098495483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,32,1,power_law_1.2,0.1434816002845764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,32,power_law_1.2,0.051558399200439455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,32,power_law_1.2,0.0525056004524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,32,power_law_1.2,0.05297279953956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,32,power_law_1.2,0.057766401767730714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,32,power_law_1.2,0.05829120278358459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,32,power_law_1.2,0.05751680135726929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,32,power_law_1.2,0.06431360244750976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,32,power_law_1.2,0.06850559711456299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,32,power_law_1.2,0.08056319952011108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,32,power_law_1.2,0.09387519955635071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,32,power_law_1.2,0.11544320583343506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,32,power_law_1.2,0.14602880477905272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,32,power_law_1.2,0.19537919759750366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,32,power_law_1.2,0.22832000255584717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,32,1,power_law_1.2,0.18676480054855346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,32,power_law_1.2,0.3579391956329346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,32,1,power_law_1.2,0.2033087968826294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,32,power_law_1.2,0.4519680023193359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,32,1,power_law_1.2,0.296729588508606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,32,1,power_law_1.2,0.3507456064224243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,32,1,power_law_1.2,0.38880000114440916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,32,1,power_law_1.2,0.46274561882019044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,32,1,power_law_1.2,0.6104767799377442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,32,1,power_law_1.2,0.7644032001495361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.01,0.2945856094360352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,64,power_law_1.01,0.0794368028640747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,64,power_law_1.01,0.07690240144729614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,64,power_law_1.01,0.0763647973537445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,64,power_law_1.01,0.08085119724273682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,64,power_law_1.01,0.08120959997177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,64,power_law_1.01,0.07760639786720276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,64,power_law_1.01,0.08490239977836608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,64,power_law_1.01,0.08668799996376038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,64,power_law_1.01,0.0870464026927948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,64,power_law_1.01,0.08419200181961059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,64,power_law_1.01,0.08751999735832214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,64,power_law_1.01,0.08900480270385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,64,power_law_1.01,0.08874880075454712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,64,power_law_1.01,0.08736000061035157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,64,power_law_1.01,0.0979200005531311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,64,power_law_1.01,0.09616000056266785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,64,power_law_1.01,0.09964159727096558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,64,power_law_1.01,0.11578240394592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,64,power_law_1.01,0.1291264057159424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,64,power_law_1.01,0.1506943941116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,64,power_law_1.01,0.18299520015716553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,64,power_law_1.01,0.2244096040725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,64,power_law_1.01,0.28641281127929685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,64,power_law_1.01,0.39083518981933596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,64,power_law_1.01,0.4772480010986328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,64,power_law_1.01,0.6828735828399658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,64,power_law_1.01,0.8952383995056152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,8,power_law_1.2,0.4881919860839844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.01,0.38158719539642333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,4,balanced,0.04590400060017904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,4,balanced,0.04651733239491781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,4,balanced,0.06427200138568878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,4,balanced,0.09272533655166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,4,balanced,0.13563733299573263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,4,balanced,0.22485333681106567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,4,balanced,0.22394667069117227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,4,balanced,0.22614934047063193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,4,balanced,0.22544533014297485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,4,balanced,0.22750399510065714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,4,balanced,0.22773865858713785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,4,balanced,0.23196800549825033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,4,balanced,0.23541865746180216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,4,balanced,0.23718933264414468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,4,balanced,0.24623467524846396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,4,balanced,0.24759999910990396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,4,balanced,0.25512532393137616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,4,balanced,0.2722666660944621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,4,balanced,0.28803199529647827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,4,balanced,0.3173653284708659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,4,balanced,0.3628480037053426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,4,balanced,0.4216373364130656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,4,balanced,0.47310400009155273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,4,balanced,0.6361546516418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,4,balanced,0.7477173010508219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,4,balanced,1.061626672744751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,4,balanced,1.3171466986338298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,2,power_law_1.2,0.05118079781532288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,2,power_law_1.2,0.05619199872016907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,2,power_law_1.2,0.07486079931259156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,2,power_law_1.2,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,2,power_law_1.2,0.11245440244674683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,2,power_law_1.2,0.13728640079498292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,2,power_law_1.2,0.1507904052734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,2,power_law_1.2,0.18605439662933348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,2,power_law_1.2,0.1932096004486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,2,power_law_1.2,0.19792640209197998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,2,power_law_1.2,0.20376958847045898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,2,power_law_1.2,0.2173504114151001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,2,power_law_1.2,0.2158207893371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,2,power_law_1.2,0.2373055934906006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,2,power_law_1.2,0.25260798931121825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,2,power_law_1.2,0.2548543930053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,2,power_law_1.2,0.282092809677124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,2,power_law_1.2,0.32410879135131837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,2,power_law_1.2,0.3408639907836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,2,power_law_1.2,0.4176767826080322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,2,power_law_1.2,0.4837183952331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,2,power_law_1.2,0.626035213470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,2,power_law_1.2,0.7582911968231201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,2,power_law_1.2,0.9888447761535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,2,power_law_1.2,1.362604808807373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,2,power_law_1.2,1.918841552734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,2,power_law_1.2,2.3237695693969727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,4,balanced,0.021040000021457672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,4,balanced,0.021349333226680756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,4,balanced,0.023205332458019257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,4,balanced,0.027424000203609467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,4,balanced,0.02956266701221466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,4,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,4,balanced,0.0432533323764801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,4,balanced,0.04141333450873693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,4,balanced,0.04354133208592733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,4,balanced,0.04554666578769684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,4,balanced,0.048997332652409874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,4,balanced,0.05205333232879639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,4,balanced,0.05530133346716563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,4,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,4,balanced,0.05840533475081126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,4,balanced,0.06230400005976359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,4,balanced,0.0739519993464152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,4,balanced,0.08515733480453491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,4,balanced,0.09104532996813457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,4,balanced,0.1104800005753835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,4,balanced,0.14667200048764548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,4,balanced,0.17390400171279907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,4,balanced,0.1983733375867208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,4,balanced,0.2502666711807251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,4,balanced,0.2851840058962504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,4,balanced,0.4461173216501872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,4,balanced,0.5382399956385294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,1,balanced,0.02402133246262868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,1,balanced,0.02926933268706004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,1,balanced,0.04170133173465729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,1,balanced,0.057578667998313904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,1,balanced,0.08888000249862671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,1,balanced,0.10698666175206502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,1,balanced,0.11340799927711487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,1,balanced,0.11351466178894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,1,balanced,0.11583466331164043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,1,balanced,0.11960533261299133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,1,balanced,0.12611732880274454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,1,balanced,0.13529599706331888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,1,balanced,0.13983466227849325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,1,balanced,0.13635200262069702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,1,balanced,0.14314132928848267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,1,balanced,0.1511146624883016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,1,balanced,0.16752000649770102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,1,balanced,0.18731200695037842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,1,balanced,0.21418132384618124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,1,balanced,0.2859999934832255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,1,balanced,0.27030932903289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,1,balanced,0.37086931864420575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,1,balanced,0.3387306531270345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,1,balanced,0.5519946813583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,1,balanced,0.607754667599996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,1,balanced,0.8822453022003174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,1,balanced,1.1451626618703206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,4,power_law_1.01,0.15411200523376464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,4,power_law_1.01,0.23864960670471191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,4,power_law_1.01,0.2335808038711548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,4,power_law_1.01,0.31227519512176516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,4,power_law_1.01,0.28768000602722166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,4,power_law_1.01,0.39000959396362306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,4,power_law_1.01,0.3621376037597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,4,power_law_1.01,0.361023998260498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,4,power_law_1.01,0.369215989112854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,4,power_law_1.01,0.3885632038116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,4,power_law_1.01,0.3917248010635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,4,power_law_1.01,0.37336320877075196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,4,power_law_1.01,0.38431999683380125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,4,power_law_1.01,0.3955391883850098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,4,power_law_1.01,0.39880321025848386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,4,power_law_1.01,0.4006527900695801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,4,power_law_1.01,0.41130881309509276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,4,power_law_1.01,0.4517375946044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.01,0.481657600402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.01,0.5571328163146972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.01,0.5973120212554932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.01,0.832646369934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.01,0.9181183815002442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.01,1.2141311645507813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.01,1.473459243774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.01,2.0269695281982423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.01,2.6164928436279298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,16,balanced,0.029989334444204967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,16,balanced,0.031311998764673867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,16,balanced,0.031285333136717476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,16,balanced,0.03322133421897888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,16,balanced,0.036277333895365395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,16,balanced,0.03787733366092046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,16,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,16,balanced,0.03812800099452337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,16,balanced,0.0391146664818128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,16,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,16,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,16,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,16,balanced,0.04146133363246918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,16,balanced,0.04159466673930486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,16,balanced,0.04411733150482178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,16,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,16,balanced,0.04595200220743815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,16,balanced,0.04986133178075155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,16,balanced,0.054005334774653115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,16,balanced,0.060549333691596985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,16,balanced,0.07062399884064992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,16,balanced,0.10198400417963664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,16,balanced,0.11565333604812622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,16,balanced,0.15436800320943198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,16,balanced,0.19552532831827799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,16,balanced,0.26257065931955975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,16,balanced,0.33032000064849854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,2,power_law_1.2,0.028224000334739686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,2,power_law_1.2,0.029407998919486998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,2,power_law_1.2,0.028326401114463808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,2,power_law_1.2,0.030643200874328612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,2,power_law_1.2,0.030822399258613586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,2,power_law_1.2,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,2,power_law_1.2,0.033190399408340454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,2,power_law_1.2,0.03482879996299744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,2,power_law_1.2,0.036051198840141296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,2,power_law_1.2,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,2,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,2,power_law_1.2,0.03644160032272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,2,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,2,power_law_1.2,0.03943679928779602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,2,power_law_1.2,0.0452672004699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,2,power_law_1.2,0.04629760086536407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,2,power_law_1.2,0.050342398881912234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,2,power_law_1.2,0.06082559823989868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,2,power_law_1.2,0.065830397605896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,2,power_law_1.2,0.08303359746932984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,2,power_law_1.2,0.10908160209655762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,2,power_law_1.2,0.14219520092010499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,2,power_law_1.2,0.16168320178985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,2,power_law_1.2,0.2420543909072876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,8,power_law_1.2,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,8,power_law_1.2,0.03651840090751648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,8,power_law_1.2,0.03642880022525787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,8,power_law_1.2,0.036908799409866334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,8,power_law_1.2,0.03780480027198792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,8,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,8,power_law_1.2,0.03935360014438629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,8,power_law_1.2,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,8,power_law_1.2,0.04181120097637177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,8,power_law_1.2,0.04348160028457641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,8,power_law_1.2,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,8,power_law_1.2,0.04832639992237091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,8,power_law_1.2,0.04904960095882416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,8,power_law_1.2,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,8,power_law_1.2,0.061401599645614625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,8,power_law_1.2,0.06280959844589233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,8,power_law_1.2,0.06788480281829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,8,power_law_1.2,0.08427519798278808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.2,0.09305599927902222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,2,power_law_1.2,0.30411520004272463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.2,0.11795200109481811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,32,2,power_law_1.01,0.0797760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,32,2,power_law_1.01,0.08889600038528442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,32,2,power_law_1.01,0.0976256012916565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,32,2,power_law_1.01,0.11447679996490479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,32,2,power_law_1.01,0.13507200479507447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,32,2,power_law_1.01,0.1507840037345886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,32,2,power_law_1.01,0.1898751974105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,32,2,power_law_1.01,0.19122560024261476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,32,2,power_law_1.01,0.19032319784164428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.2,0.1313472032546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,32,2,power_law_1.01,0.19052799940109252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,32,2,power_law_1.01,0.19340159893035888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,32,2,power_law_1.01,0.20162560939788818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,32,2,power_law_1.01,0.19783040285110473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,32,2,power_law_1.01,0.20435841083526612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,32,2,power_law_1.01,0.2094719886779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,32,2,power_law_1.01,0.21647360324859619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,32,2,power_law_1.01,0.22990078926086427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,32,2,power_law_1.01,0.24380159378051758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,32,2,power_law_1.01,0.2733247995376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,32,2,power_law_1.01,0.32175359725952146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,32,2,power_law_1.01,0.3746687889099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,32,2,power_law_1.01,0.4911935806274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,32,2,power_law_1.01,0.6006912231445313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,32,2,power_law_1.01,0.7890880107879639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,32,2,power_law_1.01,0.9937472343444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.2,0.18692480325698851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,32,2,power_law_1.01,1.46561918258667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,32,2,power_law_1.01,1.7873088836669921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.2,0.21144959926605225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,2,power_law_1.2,0.4279488086700439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.2,0.3198400020599365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,16,power_law_1.01,0.05114240050315857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,2,power_law_1.2,0.5436927795410156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,16,power_law_1.01,0.06570240259170532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,16,power_law_1.01,0.06513919830322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,16,power_law_1.01,0.05702400207519531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,16,power_law_1.01,0.05600000023841858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,16,power_law_1.01,0.05678079724311828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,16,power_law_1.01,0.05935360193252563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,16,power_law_1.01,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,16,power_law_1.01,0.06209920048713684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,16,power_law_1.01,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,16,power_law_1.01,0.06033920049667359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,16,power_law_1.01,0.0638592004776001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,16,power_law_1.01,0.06550400257110596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,16,power_law_1.01,0.06858239769935608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,16,power_law_1.01,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,16,power_law_1.01,0.08156160116195679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,16,power_law_1.01,0.08965759873390197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,16,power_law_1.01,0.11066880226135253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,16,power_law_1.01,0.12111999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,16,power_law_1.01,0.14737919569015503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,16,power_law_1.01,0.17269760370254517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,16,power_law_1.01,0.2285759925842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,16,power_law_1.01,0.2912832021713257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,16,power_law_1.01,0.3894272089004517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,16,power_law_1.01,0.4887231826782227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,16,power_law_1.01,0.7076608180999756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,16,power_law_1.01,0.9020928382873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.2,0.36625280380249026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.2,0.5230016231536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,4,power_law_1.01,0.06099839806556702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,4,power_law_1.01,0.08170239925384522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,4,power_law_1.01,0.0654591977596283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,4,power_law_1.01,0.08907520174980163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,4,power_law_1.01,0.08910719752311706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,4,power_law_1.01,0.09178239703178406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,4,power_law_1.01,0.0939136028289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,4,power_law_1.01,0.09710720181465149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,4,power_law_1.01,0.09556480050086975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,4,power_law_1.01,0.1053887963294983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,4,power_law_1.01,0.10341759920120239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,4,power_law_1.01,0.11242879629135132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,4,power_law_1.01,0.11960959434509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,4,power_law_1.01,0.13288320302963258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,4,power_law_1.01,0.1566975951194763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,4,power_law_1.01,0.16965119838714598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,4,power_law_1.01,0.16562559604644775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,4,power_law_1.01,0.18528000116348267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,4,power_law_1.01,0.23865599632263185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,4,power_law_1.01,0.2921472072601318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,16,power_law_1.2,0.051545602083206174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,4,power_law_1.01,0.38817920684814455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,16,power_law_1.2,0.054553598165512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,16,power_law_1.2,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,16,power_law_1.2,0.05100799798965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,16,power_law_1.2,0.04933120012283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,16,power_law_1.2,0.05307520031929016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,16,power_law_1.2,0.05262079834938049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,16,power_law_1.2,0.05434240102767944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,1,power_law_1.01,0.09770879745483399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,16,power_law_1.2,0.054092800617218016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,1,power_law_1.01,0.12268799543380737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,16,power_law_1.2,0.05600000023841858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,1,power_law_1.01,0.16814080476760865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,16,power_law_1.2,0.05602560043334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,1,power_law_1.01,0.22943360805511476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,16,power_law_1.2,0.0577023983001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,16,power_law_1.2,0.06039040088653565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,16,power_law_1.2,0.06225280165672302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,16,power_law_1.2,0.06877440214157104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,16,power_law_1.2,0.07149440050125122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,16,power_law_1.2,0.07708799839019775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,16,power_law_1.2,0.08911359906196595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,16,power_law_1.2,0.09106559753417968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,16,power_law_1.2,0.11336959600448608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,1,power_law_1.01,0.27864320278167726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,16,power_law_1.2,0.13419519662857055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,1,power_law_1.01,0.37614080905914304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,16,power_law_1.2,0.15911680459976196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,1,power_law_1.01,0.39352319240570066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,16,power_law_1.2,0.20495998859405518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,4,power_law_1.01,0.49662079811096194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,16,power_law_1.2,0.2886591911315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,16,power_law_1.2,0.3898688077926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,16,power_law_1.2,0.5706560134887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,16,power_law_1.2,0.6497471809387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,1,power_law_1.01,0.4014336109161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,1,power_law_1.01,0.411846399307251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,1,power_law_1.01,0.4170368194580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,1,power_law_1.01,0.42914562225341796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,1,power_law_1.01,0.4386943817138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,1,power_law_1.01,0.4434815883636475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,1,power_law_1.01,0.4553535938262939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,1,power_law_1.01,0.4602560043334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,1,power_law_1.01,0.4636672019958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,1,power_law_1.01,0.49723520278930666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,1,power_law_1.01,0.5632256031036377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.01,0.6070208072662353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.01,0.712063980102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.01,0.8248384475708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,2,power_law_1.01,0.04799999892711639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,4,power_law_1.01,0.5995007991790772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.01,1.051097583770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,2,power_law_1.01,0.051148802042007446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.01,1.2697983741760255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.01,1.7227584838867187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.2,0.7587135791778564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.01,2.173049545288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,2,power_law_1.01,0.0553600013256073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,2,power_law_1.01,0.06835839748382569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,2,power_law_1.01,0.07843199968338013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.01,3.066694450378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,2,power_law_1.01,0.09322879910469055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,2,power_law_1.01,0.11063679456710815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,2,power_law_1.01,0.11110399961471558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.01,3.950143814086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,2,power_law_1.01,0.11128959655761719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,2,power_law_1.01,0.11312639713287354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,2,power_law_1.01,0.11991679668426514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,2,power_law_1.01,0.127347195148468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,2,power_law_1.01,0.1333631992340088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,2,power_law_1.01,0.1399168014526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,2,power_law_1.01,0.15802240371704102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,2,power_law_1.01,0.1627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,2,power_law_1.01,0.16488959789276122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,2,power_law_1.01,0.18899840116500854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,2,power_law_1.01,0.22145280838012696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,2,power_law_1.01,0.2792639970779419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,2,power_law_1.01,0.33704319000244143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,2,power_law_1.01,0.43955202102661134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,4,power_law_1.01,0.8895359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,2,power_law_1.01,0.5629183769226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,2,power_law_1.01,0.7796927928924561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,2,power_law_1.01,0.9837311744689942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,4,power_law_1.01,1.1083328247070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,2,power_law_1.01,1.4131903648376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,2,power_law_1.01,1.9043840408325194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,4,power_law_1.01,1.7708032608032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,4,power_law_1.01,2.3716991424560545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,4,power_law_1.2,0.0639680027961731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,4,power_law_1.2,0.07207679748535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,4,power_law_1.2,0.06823040246963501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,4,power_law_1.2,0.09035519957542419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,4,power_law_1.2,0.08960639834403991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,4,power_law_1.2,0.09237120151519776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,4,power_law_1.2,0.09210879802703857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,4,power_law_1.2,0.09455360174179077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,4,power_law_1.2,0.10623359680175781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,4,power_law_1.2,0.10560640096664428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,4,power_law_1.2,0.10894720554351807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,4,power_law_1.2,0.11712640523910522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,4,power_law_1.2,0.11422719955444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,4,power_law_1.2,0.13594239950180054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,4,power_law_1.2,0.15883519649505615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,4,power_law_1.2,0.17637120485305785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,4,power_law_1.2,0.1563647985458374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,4,power_law_1.2,0.21870079040527343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,4,power_law_1.2,0.23407359123229982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,4,power_law_1.2,0.3335103988647461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,4,power_law_1.2,0.3864192008972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,4,power_law_1.2,0.48621439933776855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,8,power_law_1.01,0.05467519760131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,8,power_law_1.01,0.055340802669525145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,8,power_law_1.01,0.054451197385787964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,8,power_law_1.01,0.056569600105285646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,8,power_law_1.01,0.057657599449157715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,8,power_law_1.01,0.059334397315979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,2,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,8,power_law_1.01,0.05997440218925476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,2,balanced,0.04267199834187826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,2,balanced,0.06402666866779327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,2,balanced,0.06685866912206014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,2,balanced,0.06816533207893372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,2,balanced,0.0680213322242101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,2,balanced,0.06814399858315785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,2,balanced,0.06942399839560191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,2,balanced,0.06931733091672261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,2,balanced,0.06830933193365733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,2,balanced,0.0701333334048589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,2,balanced,0.07036266724268596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,4,power_law_1.2,0.684607982635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,8,power_law_1.01,0.06056960225105286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,8,power_law_1.01,0.06352639794349671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,8,power_law_1.01,0.0652671992778778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,8,power_law_1.01,0.06735360026359558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,8,power_law_1.01,0.06866559982299805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,8,power_law_1.01,0.070278400182724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,8,power_law_1.01,0.07824640274047852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,8,power_law_1.01,0.09137279987335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,8,power_law_1.01,0.09594879746437072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,8,power_law_1.01,0.09720960259437561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,2,balanced,0.07222400108973186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,2,balanced,0.07446933289368947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,2,balanced,0.08186133205890656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,2,balanced,0.08475200335184734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,2,balanced,0.09682666261990865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,2,balanced,0.11409067114194234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,2,balanced,0.10966933767000835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,2,balanced,0.13385599851608276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,2,balanced,0.15904532869656882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,2,balanced,0.2049013376235962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,2,balanced,0.2519413431485494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,2,balanced,0.33875731627146405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,2,balanced,0.4302239815394084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,8,power_law_1.01,0.1250175952911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,2,balanced,0.6010719935099283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,2,balanced,0.7811840375264486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.01,0.1373247981071472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.01,0.18007680177688598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.01,0.2198784112930298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.01,0.2828864097595215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.01,0.33715200424194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,4,power_law_1.2,0.8890751838684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.01,0.5819968223571778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,4,power_law_1.2,1.1614975929260254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.01,0.7330495834350585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,4,power_law_1.2,1.8272575378417968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,2,power_law_1.01,0.05763840079307556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,2,power_law_1.01,0.06673920154571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,2,power_law_1.01,0.07319679856300354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,2,power_law_1.01,0.08443520069122315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,2,power_law_1.01,0.0939903974533081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,2,power_law_1.01,0.11383039951324463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,2,power_law_1.01,0.11656320095062256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,2,power_law_1.01,0.11815680265426635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,2,power_law_1.01,0.12070399522781372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,2,power_law_1.01,0.12015999555587768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,2,power_law_1.01,0.12263679504394531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,2,power_law_1.01,0.12560640573501586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,2,power_law_1.01,0.12761600017547609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,2,power_law_1.01,0.13372160196304322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,2,power_law_1.01,0.13461120128631593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,2,power_law_1.01,0.1385983943939209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,2,power_law_1.01,0.1463744044303894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,2,power_law_1.01,0.16300159692764282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,2,power_law_1.01,0.19028480052948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,2,power_law_1.01,0.22704000473022462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,2,power_law_1.01,0.2799743890762329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,2,power_law_1.01,0.3779903888702393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,2,power_law_1.01,0.4650879859924316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.01,1.2077247619628906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,2,power_law_1.01,0.6778751850128174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,2,power_law_1.01,0.8569087982177734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,1,power_law_1.01,0.09872639775276185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,2,power_law_1.01,1.2339520454406738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,2,power_law_1.01,1.633407974243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,1,power_law_1.01,0.11082240343093872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,1,power_law_1.01,0.13955199718475342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,1,power_law_1.01,0.1965183973312378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,1,power_law_1.01,0.24846079349517822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,1,power_law_1.01,0.2952127933502197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,1,power_law_1.01,0.38883199691772463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,1,power_law_1.01,0.39397759437561036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,1,power_law_1.01,0.404863977432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,1,power_law_1.01,0.4060031890869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,1,power_law_1.01,0.42693119049072265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,64,balanced,0.030613332986831665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,64,balanced,0.029290666182835896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,64,balanced,0.02958933264017105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,64,balanced,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,64,balanced,0.03498133271932602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,64,balanced,0.04370133578777313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,64,balanced,0.06505600114663442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,64,balanced,0.06638399759928386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,64,balanced,0.07399466633796692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,64,balanced,0.07187200089295705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,64,balanced,0.07047466437021892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,64,balanced,0.06840000053246816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,64,balanced,0.060965334375699363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,64,balanced,0.05300266544024149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,64,balanced,0.09066133697827657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,1,power_law_1.01,0.4333439826965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,64,balanced,0.08676266670227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,64,balanced,0.07227199772993724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,1,power_law_1.01,0.44739837646484376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,1,power_law_1.01,0.4524672031402588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,1,power_law_1.01,0.4651455879211426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,1,power_law_1.01,0.4694528102874756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,1,power_law_1.01,0.4871039867401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,64,balanced,0.09513599673906963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,64,balanced,0.11563733220100403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,1,power_law_1.01,0.5425663948059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,64,balanced,0.15550399820009866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.01,0.5898431777954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,64,balanced,0.18876800934473673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,64,balanced,0.2718133330345154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.01,0.6982399940490722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,64,balanced,0.34815998872121173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.01,0.7791615962982178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.01,0.9902463912963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.01,1.5802495956420899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.01,1.1885503768920898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,64,balanced,0.5115413268407186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,64,balanced,0.6813759803771973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.01,1.5855680465698243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,64,balanced,1.0156266689300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.01,2.000729560852051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,64,balanced,1.3493760426839192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,4,power_law_1.2,2.573600006103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.01,2.795174407958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.01,3.597119903564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,4,power_law_1.2,0.11765120029449463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,4,power_law_1.2,0.16532479524612426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,4,power_law_1.2,0.17868800163269044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,4,power_law_1.2,0.22101759910583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,4,power_law_1.2,0.2682431936264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,4,power_law_1.2,0.2996799945831299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,4,power_law_1.2,0.32717440128326414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,4,power_law_1.2,0.3536639928817749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,4,power_law_1.2,0.38988161087036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,4,power_law_1.2,0.38852479457855227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,4,power_law_1.2,0.3920192003250122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,32,power_law_1.2,0.14551039934158325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,32,power_law_1.2,0.13595520257949828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,32,power_law_1.2,0.13701119422912597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,32,power_law_1.2,0.13540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,32,power_law_1.2,0.1296895980834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,32,power_law_1.2,0.11515519618988038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,32,power_law_1.2,0.1173248052597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,32,power_law_1.2,0.12417919635772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,32,power_law_1.2,0.1285823941230774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,32,power_law_1.2,0.12400640249252319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,32,power_law_1.2,0.12517759799957276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,32,power_law_1.2,0.1260607957839966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,32,power_law_1.2,0.12511359453201293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,32,power_law_1.2,0.1287616014480591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,32,power_law_1.2,0.1294592022895813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,32,power_law_1.2,0.13374719619750977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,32,power_law_1.2,0.143231999874115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,32,power_law_1.2,0.16485120058059693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.2,0.17357439994812013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.2,0.2021631956100464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.2,0.22394239902496338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.2,0.3049407958984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.2,0.3533184051513672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.2,0.449510383605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.2,0.6103871822357178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.2,0.929747200012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.2,1.1649791717529296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,64,power_law_1.01,0.026643198728561402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,2,balanced,0.10321600238482158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,64,power_law_1.01,0.025990399718284606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,4,power_law_1.2,0.05429760217666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,64,power_law_1.01,0.026316800713539125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,4,power_law_1.2,0.05763840079307556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,64,power_law_1.01,0.029337599873542786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,4,power_law_1.2,0.06149759888648987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,64,power_law_1.01,0.03020800054073334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,4,power_law_1.2,0.07541120052337646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,64,power_law_1.01,0.03162879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,4,power_law_1.2,0.08164479732513427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,2,balanced,0.10942400495211284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,2,balanced,0.11893866459528606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,2,balanced,0.13959466417630514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,2,balanced,0.17130666971206665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,2,balanced,0.22632000843683878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,2,balanced,0.22657066583633423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,4,power_law_1.2,0.08732159733772278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,2,balanced,0.22564266125361124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,4,power_law_1.2,0.1060479998588562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,2,balanced,0.23090134064356485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,4,power_law_1.2,0.10257279872894287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,2,balanced,0.23047999540964761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,2,balanced,0.23437867561976114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,16,power_law_1.2,0.06954879760742187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,2,balanced,0.23940267165501913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,16,power_law_1.2,0.09801599979400635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,2,balanced,0.24230400721232095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,2,balanced,0.24664533138275146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,64,power_law_1.01,0.031430399417877196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,2,balanced,0.25075199206670123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,2,balanced,0.2537386616071065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,64,power_law_1.01,0.03210879862308502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,2,balanced,0.26337067286173504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,2,balanced,0.29179733991622925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,4,power_law_1.2,0.09943680167198181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,4,power_law_1.2,0.09996160268783569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,1,power_law_1.01,0.02969599962234497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,4,power_law_1.2,0.11016960144042968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,16,power_law_1.2,0.09028480052947999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,1,power_law_1.01,0.030476799607276915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,4,power_law_1.2,0.10914560556411743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,1,power_law_1.01,0.030931198596954347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,16,power_law_1.2,0.09751679897308349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,1,power_law_1.01,0.034995201230049136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,64,power_law_1.01,0.032179200649261476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,16,power_law_1.2,0.09740800261497498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,1,power_law_1.01,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,64,power_law_1.01,0.03223679959774017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,16,power_law_1.2,0.09253119826316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,2,balanced,0.31189332405726117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,2,balanced,0.36737600962320965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,16,power_law_1.2,0.09537919759750366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,4,power_law_1.2,0.1110975980758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,2,balanced,0.41766401131947833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,4,power_law_1.2,0.11411839723587036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,16,power_law_1.2,0.09578880071640014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,2,balanced,0.5195519924163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,4,power_law_1.2,0.1208575963973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,1,power_law_1.01,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,16,power_law_1.2,0.09303039908409119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,1,power_law_1.01,0.0381056010723114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,4,power_law_1.2,0.12658560276031494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,16,power_law_1.2,0.09630720019340515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,1,power_law_1.01,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,4,power_law_1.2,0.134879994392395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,16,power_law_1.2,0.10332159996032715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,4,power_law_1.2,0.15187840461730956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,4,power_law_1.2,0.15801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,4,power_law_1.2,0.18921600580215453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,2,balanced,0.61899201075236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,4,power_law_1.2,0.21150720119476318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,2,balanced,0.9103413422902426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,1,power_law_1.01,0.03870719969272614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,1,power_law_1.01,0.03980799913406372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,16,power_law_1.2,0.09995520114898682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,2,balanced,1.1121813456217449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,1,power_law_1.01,0.03953920006752014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,16,power_law_1.2,0.10453120470046998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,1,power_law_1.01,0.04292480051517487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,16,power_law_1.2,0.10804480314254761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,2,balanced,1.606730620066325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,16,power_law_1.2,0.1171455979347229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,16,power_law_1.2,0.12414079904556274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,2,balanced,2.1142187118530273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,4,power_law_1.2,0.2712064027786255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,64,power_law_1.01,0.03369599878787995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,4,power_law_1.2,0.3230720043182373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,4,power_law_1.2,0.45703678131103515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,1,power_law_1.01,0.04384639859199524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,1,power_law_1.01,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,4,power_law_1.2,0.5713024139404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,16,power_law_1.2,0.1331007957458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,1,power_law_1.01,0.055404800176620486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,16,power_law_1.2,0.15244799852371216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,4,power_law_1.2,0.783680009841919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,16,power_law_1.2,0.1589311957359314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,1,power_law_1.01,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,16,power_law_1.2,0.19314559698104858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,4,power_law_1.2,1.0876735687255858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,16,power_law_1.2,0.22775681018829347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,1,power_law_1.01,0.058246397972106935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,16,power_law_1.2,0.2881279945373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,1,power_law_1.01,0.07437440156936645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,16,power_law_1.2,0.3575295925140381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,16,power_law_1.2,0.5077631950378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,64,power_law_1.01,0.03466239869594574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,16,power_law_1.2,0.5905856132507324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,16,power_law_1.2,0.8835712432861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,16,power_law_1.2,1.1566847801208495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,1,power_law_1.01,0.08253440260887146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,64,power_law_1.01,0.03643519878387451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,64,power_law_1.01,0.0377344012260437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,64,power_law_1.01,0.042803201079368594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,64,power_law_1.01,0.04133760035037994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,64,power_law_1.01,0.04531840085983276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,64,power_law_1.01,0.0536191999912262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,64,power_law_1.01,0.057132798433303836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,1,power_law_1.01,0.10687999725341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,64,power_law_1.01,0.06871039867401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,1,power_law_1.01,0.12506239414215087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,64,power_law_1.01,0.0853056013584137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,1,power_law_1.01,0.16591999530792237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,64,power_law_1.01,0.10563839673995971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,64,power_law_1.01,0.14415359497070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,64,power_law_1.01,0.21884799003601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,64,power_law_1.01,0.2596479892730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,64,power_law_1.01,0.3537600040435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,64,power_law_1.01,0.559507179260254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,1,power_law_1.01,0.2003648042678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,1,power_law_1.01,0.2846400022506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,1,power_law_1.01,0.3592447996139526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,1,power_law_1.01,0.5206143856048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,128,power_law_1.01,0.07698559761047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,128,power_law_1.01,0.07352319955825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,128,power_law_1.01,0.0732159972190857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,128,power_law_1.01,0.07498239874839782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,128,power_law_1.01,0.07692800164222717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,128,power_law_1.01,0.08119040131568908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,128,power_law_1.01,0.07874559760093688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,128,power_law_1.01,0.0795199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,128,power_law_1.01,0.08083199858665466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,128,power_law_1.01,0.07994880080223084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,128,power_law_1.01,0.08112000226974488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,128,power_law_1.01,0.08448640108108521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,128,power_law_1.01,0.08842880129814149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,128,power_law_1.01,0.08950399756431579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,128,power_law_1.01,0.09827839732170104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,8,power_law_1.2,0.03802880048751831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,8,power_law_1.2,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,8,power_law_1.2,0.03678080141544342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,8,power_law_1.2,0.038790398836135866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,8,power_law_1.2,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,8,power_law_1.2,0.04714879989624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,8,power_law_1.2,0.050457602739334105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,8,power_law_1.2,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,8,power_law_1.2,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,128,power_law_1.01,0.1004480004310608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,8,power_law_1.2,0.0590399980545044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,128,power_law_1.01,0.10496640205383301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,8,power_law_1.2,0.05567359924316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,128,power_law_1.01,0.12223999500274658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,8,power_law_1.2,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,128,power_law_1.01,0.12791680097579955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,128,power_law_1.01,0.1600000023841858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,8,power_law_1.2,0.06567680239677429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,128,power_law_1.01,0.17669119834899902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,8,power_law_1.2,0.06549760103225707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,128,power_law_1.01,0.2237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,1,power_law_1.01,0.6807231903076172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,128,power_law_1.01,0.25365118980407714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,128,power_law_1.01,0.3105792045593262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,128,power_law_1.01,0.3803391933441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,128,power_law_1.01,0.548748779296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,8,power_law_1.2,0.0779263973236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,128,power_law_1.01,0.6484352111816406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,8,power_law_1.2,0.08087679743766785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,8,power_law_1.2,0.0918079972267151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,8,power_law_1.2,0.1107200026512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,8,power_law_1.2,0.12553600072860718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,8,power_law_1.2,0.16773760318756104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,8,power_law_1.2,0.19705599546432495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,8,power_law_1.2,0.26019840240478515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,8,power_law_1.2,0.3361407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,8,power_law_1.2,0.4516160011291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,8,power_law_1.2,0.5970304012298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,8,power_law_1.2,0.8199423789978028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,4,power_law_1.01,0.020479999482631683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,4,power_law_1.01,0.022681599855422972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,4,power_law_1.01,0.024902400374412537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,4,power_law_1.01,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,8,power_law_1.2,1.0590720176696777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,4,power_law_1.01,0.0377023994922638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,4,power_law_1.01,0.044921600818634035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,4,power_law_1.01,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,4,power_law_1.01,0.04755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,4,power_law_1.01,0.047654399275779726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,4,power_law_1.01,0.04955520033836365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,4,power_law_1.01,0.050783997774124144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,4,power_law_1.01,0.055264002084732054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,4,power_law_1.01,0.05867519974708557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,4,power_law_1.01,0.054073601961135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,4,power_law_1.01,0.06177279949188232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,4,power_law_1.01,0.06406400203704835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,8,balanced,0.11807466546694438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,8,balanced,0.12316800157229106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,8,balanced,0.12040000160535176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,8,balanced,0.12106133500734965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,8,balanced,0.12210133671760559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,8,balanced,0.12133333086967468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,8,balanced,0.12177600463231404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,8,balanced,0.12037332852681477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,8,balanced,0.12957866986592612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,8,balanced,0.13104533155759177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,8,balanced,0.1328213314215342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,8,balanced,0.1309866706530253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,8,balanced,0.13166399796803793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,8,balanced,0.14258666833241782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,8,balanced,0.14418133099873862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,8,balanced,0.14697066942850748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,8,balanced,0.15481066703796387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,8,balanced,0.18010665973027548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,2,power_law_1.01,0.0335999995470047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,8,balanced,0.18419732650121054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,8,balanced,0.22075732549031576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,2,power_law_1.01,0.03678719997406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,2,power_law_1.01,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,8,balanced,0.2634773254394531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,8,balanced,0.3285973270734151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,4,power_law_1.01,0.0753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,8,balanced,0.41951465606689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,4,power_law_1.01,0.08592000007629394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,4,power_law_1.01,0.0880832016468048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,4,power_law_1.01,0.10877439975738526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,4,power_law_1.01,0.1373311996459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,4,power_law_1.01,0.16232320070266723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,64,power_law_1.01,0.0315775990486145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,2,power_law_1.01,0.058361601829528806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,4,power_law_1.01,0.17723519802093507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,64,power_law_1.01,0.031948798894882204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,2,power_law_1.01,0.060755199193954466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,4,power_law_1.01,0.22692480087280273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,64,power_law_1.01,0.0271807998418808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,2,power_law_1.01,0.07992320060729981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,64,power_law_1.01,0.02852480113506317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,64,power_law_1.01,0.029068800806999206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,64,power_law_1.01,0.030048000812530517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,64,power_law_1.01,0.02996479868888855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,64,power_law_1.01,0.030316799879074097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,64,power_law_1.01,0.02999039888381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,64,power_law_1.01,0.03023360073566437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,8,balanced,0.5885813236236572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,2,power_law_1.01,0.08256000280380249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,2,power_law_1.01,0.08723840117454529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,2,power_law_1.01,0.08974080085754395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,2,power_law_1.01,0.09009280204772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,2,power_law_1.01,0.09550079703330994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,64,power_law_1.01,0.03086079955101013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,2,power_law_1.01,0.10114560127258301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,64,power_law_1.01,0.03131519854068756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,2,power_law_1.01,0.10129280090332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,64,power_law_1.01,0.032179200649261476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,2,power_law_1.01,0.10496000051498414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,2,power_law_1.01,0.11338239908218384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,2,power_law_1.01,0.11762559413909912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,2,power_law_1.01,0.12535680532455445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,2,power_law_1.01,0.14876799583435057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,2,power_law_1.01,0.16634880304336547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,8,balanced,0.7393333117167155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,64,power_law_1.01,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,4,power_law_1.01,0.28418560028076173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,64,power_law_1.01,0.03628160059452057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,64,power_law_1.01,0.03713279962539673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,4,power_law_1.01,0.383955192565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,64,power_law_1.01,0.03937279880046844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,64,power_law_1.01,0.043136000633239746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,64,power_law_1.01,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,64,power_law_1.01,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,64,power_law_1.01,0.05138559937477112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,64,power_law_1.01,0.06279680132865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,2,power_law_1.01,0.19870079755783082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,64,power_law_1.01,0.06728320121765137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,8,balanced,1.0499733289082844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,2,power_law_1.01,0.23941121101379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,64,power_law_1.01,0.08623999953269959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,2,power_law_1.01,0.31775360107421874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,4,power_law_1.01,0.48290557861328126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,2,power_law_1.01,0.3780992031097412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,2,power_law_1.01,0.5360511779785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,2,power_law_1.01,0.6943679809570312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,2,power_law_1.01,0.9784192085266114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,2,power_law_1.01,1.2678208351135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,8,balanced,1.4488587379455566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,64,power_law_1.01,0.10252799987792968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,64,power_law_1.01,0.13825279474258423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,64,power_law_1.01,0.1580288052558899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,8,balanced,0.021498667697111767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,8,balanced,0.02109866589307785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,8,balanced,0.021007999777793884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,8,balanced,0.02107200026512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,16,power_law_1.01,0.07301759719848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,8,balanced,0.021370666722456615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,16,power_law_1.01,0.07354239821434021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,16,power_law_1.01,0.069651198387146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,8,balanced,0.023498666783173878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,8,balanced,0.04781866570313772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,8,balanced,0.048021331429481506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,8,balanced,0.03921066721280416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,16,power_law_1.01,0.07189760208129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,16,power_law_1.01,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,8,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,16,power_law_1.01,0.06977919936180114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,8,balanced,0.03640533238649368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,8,balanced,0.04376000165939331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,8,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,8,balanced,0.037802666425704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,8,balanced,0.04375466704368591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,8,balanced,0.039813332259655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,8,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,8,balanced,0.056741332014401756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,8,balanced,0.06192000210285187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,8,balanced,0.08349866668383281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,8,balanced,0.10354666908582051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,8,balanced,0.13829867045084634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,8,balanced,0.16761600971221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,8,balanced,0.23939200242360434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,8,balanced,0.3014880021413167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,8,balanced,0.4386666615804036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,8,balanced,0.5729120175043741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,16,power_law_1.01,0.07136639952659607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,16,power_law_1.01,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,16,power_law_1.01,0.07370880246162415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,16,power_law_1.01,0.07389439940452576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,16,power_law_1.01,0.07252479791641235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,16,power_law_1.01,0.07352319955825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,16,power_law_1.01,0.07539839744567871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,16,power_law_1.01,0.07426559925079346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,16,power_law_1.01,0.07986559867858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,16,power_law_1.01,0.08055040240287781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,16,power_law_1.01,0.08353279829025269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,16,power_law_1.01,0.09464319944381713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,16,power_law_1.01,0.10647679567337036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,16,power_law_1.01,0.12640000581741334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,16,power_law_1.01,0.14547200202941896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,16,power_law_1.01,0.18636800050735475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,16,power_law_1.01,0.22278399467468263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,16,power_law_1.01,0.30271360874176023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,16,power_law_1.01,0.3864703893661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,16,power_law_1.01,0.5406015872955322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,16,power_law_1.01,0.7365952014923096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,1,power_law_1.01,0.03371520042419433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,1,power_law_1.01,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,1,power_law_1.01,0.047916799783706665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,1,power_law_1.01,0.0656000018119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,1,power_law_1.01,0.07363200187683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,1,power_law_1.01,0.10945279598236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,1,power_law_1.01,0.11269760131835938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,1,power_law_1.01,0.11651840209960937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,1,power_law_1.01,0.11779199838638306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,1,power_law_1.01,0.11895040273666382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,1,power_law_1.01,0.12369279861450196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,1,power_law_1.01,0.12400640249252319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,1,power_law_1.01,0.12647680044174195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,1,power_law_1.01,0.1300480008125305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,1,power_law_1.01,0.1377408027648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,1,power_law_1.01,0.1428096055984497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,1,power_law_1.01,0.15898239612579346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,1,power_law_1.01,0.1825279951095581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,1,power_law_1.01,0.17352319955825807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,1,power_law_1.01,0.2057663917541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,1,power_law_1.01,0.25118720531463623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,1,power_law_1.01,0.305516791343689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,1,power_law_1.01,0.3741247892379761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,1,power_law_1.01,0.49062399864196776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,1,power_law_1.01,0.557478380203247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,4,power_law_1.2,0.0176704004406929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,1,power_law_1.01,0.7788479804992676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,4,power_law_1.2,0.018886399269104005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,4,power_law_1.2,0.01912959963083267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,4,power_law_1.2,0.019545599818229675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,1,power_law_1.01,0.9954943656921387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,4,power_law_1.2,0.020390400290489198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,4,power_law_1.2,0.022336000204086305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,4,power_law_1.2,0.02375040054321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,4,power_law_1.2,0.023948800563812257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,4,power_law_1.2,0.024134400486946105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,2,balanced,0.07298666735490163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,4,power_law_1.2,0.024985599517822265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,2,balanced,0.0942133367061615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,2,balanced,0.12357866764068604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,4,power_law_1.2,0.026188799738883974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,2,balanced,0.18981866041819254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,4,power_law_1.2,0.02723200023174286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,4,power_law_1.2,0.02871679961681366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,4,power_law_1.2,0.027699199318885804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,4,power_law_1.2,0.02948479950428009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,4,power_law_1.2,0.03110400140285492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,4,power_law_1.2,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,4,power_law_1.2,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,2,balanced,0.30930133660634357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,2,balanced,0.5578506787618002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,2,balanced,0.7986666361490885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,2,balanced,0.8006933530171713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,2,balanced,0.8023200035095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,2,balanced,0.8054560025533041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,2,balanced,0.8088906606038412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,4,power_law_1.2,0.06037120223045349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,2,balanced,0.8152053356170654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,4,power_law_1.2,0.07285119891166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,2,balanced,0.822330633799235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,4,power_law_1.2,0.06487680077552796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,2,balanced,0.82369065284729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,4,power_law_1.2,0.07643520236015319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,2,balanced,0.8392693201700846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,1,power_law_1.2,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,2,balanced,0.8437600135803223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,1,power_law_1.2,0.04803839921951294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,2,balanced,0.8597813447316488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,1,power_law_1.2,0.06906239986419678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,2,balanced,0.8836639722188314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,1,power_law_1.2,0.0877568006515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,2,balanced,0.9102293650309244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,1,power_law_1.2,0.10985599756240845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,2,balanced,0.9596532980600992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,1,power_law_1.2,0.17486079931259155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,1,power_law_1.2,0.1901695966720581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,4,power_law_1.2,0.08945279717445373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,1,power_law_1.2,0.1900287985801697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,1,power_law_1.2,0.19711359739303588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,1,power_law_1.2,0.19959039688110353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,1,power_law_1.2,0.20111360549926757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,1,power_law_1.2,0.2065727949142456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,1,power_law_1.2,0.21136000156402587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,1,power_law_1.2,0.21870079040527343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,2,balanced,1.008565346399943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,1,power_law_1.2,0.23306241035461425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,1,power_law_1.2,0.23797121047973632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,2,balanced,1.1309066613515217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,1,power_law_1.2,0.24039039611816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,1,power_law_1.2,0.26795520782470705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,1,power_law_1.2,0.28767359256744385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,1,power_law_1.2,0.3323967933654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,1,power_law_1.2,0.36206080913543703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,1,power_law_1.2,0.45250558853149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,1,power_law_1.2,0.5370751857757569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,2,balanced,1.2388213475545247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,1,power_law_1.2,0.715993595123291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,2,balanced,1.453770637512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,4,power_law_1.2,0.11741440296173096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,2,balanced,1.8736054102579753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,8,power_law_1.01,0.05135359764099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,2,balanced,2.2970080375671387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,8,power_law_1.01,0.0685375988483429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,8,power_law_1.01,0.0945855975151062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,2,balanced,3.1039040883382163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,8,power_law_1.01,0.09311360120773315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,1,power_law_1.2,0.8798784255981446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,8,power_law_1.01,0.11199359893798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,8,power_law_1.01,0.11975680589675904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,8,power_law_1.01,0.11722240447998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,8,power_law_1.01,0.12346880435943604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,8,power_law_1.01,0.11520639657974244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,8,power_law_1.01,0.11747839450836181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,8,power_law_1.01,0.11969920396804809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,8,power_law_1.01,0.12224639654159546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,8,power_law_1.01,0.12092159986495972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,8,power_law_1.01,0.12138240337371826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,8,power_law_1.01,0.13690880537033082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,8,power_law_1.01,0.13864320516586304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,8,power_law_1.01,0.14925440549850463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,8,power_law_1.01,0.1610751986503601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,8,power_law_1.01,0.18240640163421631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,8,power_law_1.01,0.20775039196014405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,8,power_law_1.01,0.23007359504699706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,8,power_law_1.01,0.27667200565338135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,8,power_law_1.01,0.3472192049026489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,8,power_law_1.01,0.4720128059387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,8,power_law_1.01,0.5795072078704834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,8,power_law_1.01,0.8339967727661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,8,power_law_1.01,0.9697664260864258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,2,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,2,balanced,0.03526400029659271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,2,balanced,0.037248000502586365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,2,balanced,0.04420800010363261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,2,balanced,0.06935466825962067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,2,balanced,0.0724480003118515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,2,balanced,0.07286933561166127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,2,balanced,0.0740533322095871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,2,balanced,0.07418133318424225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,2,balanced,0.07539199789365132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,2,balanced,0.07730133334795634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,2,balanced,0.07984533409277599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,2,balanced,0.0794239987929662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,2,balanced,0.08437333504358928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,2,balanced,0.0906826655069987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,4,power_law_1.2,0.14454400539398193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,2,balanced,0.09525332848230998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,2,balanced,0.09669333696365356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,2,balanced,0.11269866426785786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,2,balanced,0.131221334139506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,1,power_law_1.2,1.211359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,2,balanced,0.15821866194407144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,2,balanced,0.1799573302268982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,2,balanced,0.26233067115147907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,2,balanced,0.3120959997177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,2,balanced,0.45634667078653973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,2,balanced,0.5924853483835856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,2,balanced,0.8656000296274821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,2,balanced,1.1158453623453777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,4,power_law_1.2,0.1931007981300354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,1,power_law_1.2,1.5472000122070313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,4,power_law_1.2,0.2505408048629761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,8,power_law_1.01,0.04275839924812317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,8,power_law_1.01,0.04136959910392761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,8,power_law_1.01,0.04351359903812409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,8,power_law_1.01,0.04547199904918671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,8,power_law_1.01,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,8,power_law_1.01,0.0500927984714508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,8,power_law_1.01,0.04999040067195892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,8,power_law_1.01,0.050963199138641356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,8,power_law_1.01,0.051801598072052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,8,power_law_1.01,0.050886398553848265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,8,power_law_1.01,0.05243520140647888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,8,power_law_1.01,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,8,power_law_1.01,0.05569919943809509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,8,power_law_1.01,0.06046079993247986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,8,power_law_1.01,0.06679040193557739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,8,power_law_1.01,0.06893439888954163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,8,power_law_1.01,0.0702015995979309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,8,power_law_1.01,0.08496639728546143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.01,0.1004032015800476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.01,0.12298239469528198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.01,0.12826240062713623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.01,0.16132479906082153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.01,0.1838655948638916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.01,0.27254400253295896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.01,0.3328768014907837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.01,0.4967936038970947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.01,0.6246143817901612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,16,power_law_1.2,0.0699455976486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,16,power_law_1.2,0.08711680173873901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,16,power_law_1.2,0.08145920038223267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,16,power_law_1.2,0.0887167990207672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,16,power_law_1.2,0.08544639945030212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,16,power_law_1.2,0.08768640160560608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,16,power_law_1.2,0.08949120044708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,16,power_law_1.2,0.09244800209999085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,16,power_law_1.2,0.09219840168952942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,16,power_law_1.2,0.09205120205879211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,16,power_law_1.2,0.0936896026134491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,16,power_law_1.2,0.09609599709510804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,16,power_law_1.2,0.09775999784469605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,16,power_law_1.2,0.09777920246124268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,16,power_law_1.2,0.10202879905700683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,16,power_law_1.2,0.10406399965286255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,16,power_law_1.2,0.10766719579696656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,16,power_law_1.2,0.11964800357818603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,16,power_law_1.2,0.1356927990913391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,16,power_law_1.2,0.15813119411468507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,16,power_law_1.2,0.18042240142822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,16,power_law_1.2,0.22623999118804933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,16,power_law_1.2,0.28317439556121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,16,power_law_1.2,0.37304320335388186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,16,power_law_1.2,0.5160128116607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,16,power_law_1.2,0.7202688217163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,16,power_law_1.2,1.0036928176879882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,16,power_law_1.2,0.05280640125274658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,16,power_law_1.2,0.0702015995979309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,16,power_law_1.2,0.06599680185317994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,16,power_law_1.2,0.07220479846000671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,16,power_law_1.2,0.07466880083084107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,16,power_law_1.2,0.0652288019657135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,16,power_law_1.2,0.0722495973110199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,16,power_law_1.2,0.07363200187683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,16,power_law_1.2,0.0762943983078003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,16,power_law_1.2,0.07750399708747864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,16,power_law_1.2,0.0775168001651764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,16,power_law_1.2,0.08031359910964966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,16,power_law_1.2,0.08366720080375671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,16,power_law_1.2,0.0880895972251892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,16,power_law_1.2,0.09916800260543823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,16,power_law_1.2,0.10180480480194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,16,power_law_1.2,0.10790400505065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,16,power_law_1.2,0.12097920179367065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,16,power_law_1.2,0.1376512050628662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,16,power_law_1.2,0.16821119785308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,16,power_law_1.2,0.18762240409851075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,16,power_law_1.2,0.23826560974121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,16,power_law_1.2,0.3106623888015747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,16,power_law_1.2,0.40458240509033205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,16,power_law_1.2,0.5750016212463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,2,power_law_1.2,0.052288001775741576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,16,power_law_1.2,0.7877376079559326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,2,power_law_1.2,0.05578240156173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,16,power_law_1.2,1.0077504158020019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,2,power_law_1.2,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,2,power_law_1.2,0.08967040181159973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,2,power_law_1.2,0.10983680486679077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,2,power_law_1.2,0.13746559619903564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,2,power_law_1.2,0.14212479591369628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,2,power_law_1.2,0.18942079544067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,2,power_law_1.2,0.18850560188293458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,2,power_law_1.2,0.1936959981918335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,2,power_law_1.2,0.2003391981124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,2,power_law_1.2,0.20511999130249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,2,power_law_1.2,0.20700159072875976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,2,power_law_1.2,0.21806080341339112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,2,power_law_1.2,0.23042559623718262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,2,power_law_1.2,0.23746559619903565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,2,power_law_1.2,0.2632256031036377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,2,power_law_1.2,0.288588809967041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,2,power_law_1.2,0.31536641120910647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,2,power_law_1.2,0.38894081115722656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,2,power_law_1.2,0.45520639419555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,2,power_law_1.2,0.588921594619751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,2,power_law_1.2,0.667468786239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,2,power_law_1.2,0.8716352462768555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,2,power_law_1.2,1.1348928451538085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,2,power_law_1.2,1.6057727813720704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,8,balanced,0.046367997924486794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,8,balanced,0.04788800080617269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,2,power_law_1.2,1.9140607833862304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,8,balanced,0.06730666756629944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,8,balanced,0.09851732850074768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,8,balanced,0.14963199694951376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,8,balanced,0.24560532967249551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,8,balanced,0.34833598136901855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,8,balanced,0.34695998827616376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,8,balanced,0.34721068541208905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,8,balanced,0.34886399904886883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,8,balanced,0.3500639994939168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,8,balanced,0.3511360088984172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,8,balanced,0.35358933607737225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,8,balanced,0.35709865887959796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,8,balanced,0.3594133456548055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,8,balanced,0.36497068405151367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,8,balanced,0.3689546585083008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,8,balanced,0.3820426861445109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,8,balanced,0.3917706807454427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,8,balanced,0.4147040049235026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,8,balanced,0.43959466616312665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,8,balanced,0.4886560042699178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,8,balanced,0.5214986801147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,8,balanced,0.5988693237304688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,8,balanced,0.721349318822225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,8,balanced,0.8746506373087565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,8,balanced,1.133519967397054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,32,power_law_1.01,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,32,power_law_1.01,0.02933120131492615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,32,power_law_1.01,0.028198400139808656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,32,power_law_1.01,0.0302592009305954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,32,power_law_1.01,0.030604800581932066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,32,power_law_1.01,0.03158400058746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,32,power_law_1.01,0.031404799222946166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,32,power_law_1.01,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,32,power_law_1.01,0.032678401470184325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,32,power_law_1.01,0.032620799541473386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,32,power_law_1.01,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,32,power_law_1.01,0.03409920036792755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,32,power_law_1.01,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,32,power_law_1.01,0.03550080060958862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,32,power_law_1.01,0.038700801134109494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,32,power_law_1.01,0.04116480052471161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,32,power_law_1.01,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,32,power_law_1.01,0.0438400000333786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,32,power_law_1.01,0.04677119851112366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,32,power_law_1.01,0.051641601324081424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,32,power_law_1.01,0.059596800804138185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,32,power_law_1.01,0.07157760262489318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,1,balanced,0.04651199777921041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,32,power_law_1.01,0.07962239980697632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,32,power_law_1.01,0.12033280134201049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,32,power_law_1.01,0.1444991946220398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,32,power_law_1.01,0.22421760559082032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,1,balanced,0.05657066901524862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,32,power_law_1.01,0.2622720003128052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,1,balanced,0.07428800066312154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,1,balanced,0.07884266475836436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,1,balanced,0.07331199944019318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,1,balanced,0.07036800185839336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,1,balanced,0.08454400300979614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,32,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,32,balanced,0.03230399886767069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,32,balanced,0.02958400050799052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,32,balanced,0.033733333150545754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,32,balanced,0.035487999518712364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,1,balanced,0.06817600131034851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,32,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,32,balanced,0.036506667733192444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,32,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,32,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,32,balanced,0.037647999823093414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,32,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,32,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,32,balanced,0.03758399933576584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,32,balanced,0.04064533362785975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,32,balanced,0.04204266766707102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,32,balanced,0.04180799921353658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,32,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,32,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,32,balanced,0.049925332268079124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,32,balanced,0.05316799879074097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,32,balanced,0.0609493354956309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,32,balanced,0.06979733208815257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,32,balanced,0.07576000193754832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,32,balanced,0.08929066856702168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,32,balanced,0.10749333103497823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,32,balanced,0.13124799728393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,32,balanced,0.16029333074887595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,1,balanced,0.07588266829649608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,1,balanced,0.07358933488527934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,1,balanced,0.07368533313274384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,1,balanced,0.07050133248170216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,1,balanced,0.07654400169849396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,1,balanced,0.07257600128650665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,1,balanced,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,1,balanced,0.075914666056633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,1,balanced,0.07975466549396515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,1,balanced,0.11343466242154439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,1,balanced,0.11066133777300517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,1,balanced,0.14291733503341675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,1,balanced,0.16766933600107828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,1,balanced,0.2243893345197042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,1,balanced,0.27900266647338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,1,balanced,0.3922613461812337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,1,balanced,0.502784013748169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,1,balanced,0.7275466918945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,1,balanced,0.953279972076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,8,balanced,0.05017599960168203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,8,balanced,0.05186133086681366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,8,balanced,0.05269866685072581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,8,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,8,balanced,0.06914133330186208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,8,balanced,0.09783466657002766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,8,balanced,0.09828266501426697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,8,balanced,0.0990666647752126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,8,balanced,0.09982400139172871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,8,balanced,0.10130666693051656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,8,balanced,0.10228799780209859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,8,balanced,0.10322667161623637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,8,balanced,0.10744000474611919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,8,balanced,0.10640000303586324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,8,balanced,0.11342400312423706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,8,balanced,0.1150933305422465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,8,balanced,0.12000532944997151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,8,balanced,0.13033599654833475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,8,balanced,0.14198933045069376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,8,balanced,0.16078933080037436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,8,balanced,0.1727893352508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,8,balanced,0.21145067612330118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,8,balanced,0.2508479952812195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,8,balanced,0.34586668014526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,8,balanced,0.4211893479029338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,8,balanced,0.5914346774419149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,8,balanced,0.7603680292765299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,16,power_law_1.2,0.05211520195007324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,16,power_law_1.2,0.061689597368240354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,16,power_law_1.2,0.05975040197372437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,16,power_law_1.2,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,16,power_law_1.2,0.057792001962661745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,16,power_law_1.2,0.058208000659942624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,16,power_law_1.2,0.05910400152206421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,16,power_law_1.2,0.0604095995426178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,16,power_law_1.2,0.06044160127639771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,16,power_law_1.2,0.061222398281097413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,16,power_law_1.2,0.06149759888648987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,16,power_law_1.2,0.06399359703063964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,16,power_law_1.2,0.06528000235557556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,16,power_law_1.2,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,16,power_law_1.2,0.07352319955825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,16,power_law_1.2,0.07477759718894958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,16,power_law_1.2,0.08234239816665649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,16,power_law_1.2,0.09159680008888245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,16,power_law_1.2,0.0992255985736847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,16,power_law_1.2,0.1302016019821167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,16,power_law_1.2,0.1452288031578064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,16,power_law_1.2,0.19411840438842773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,16,power_law_1.2,0.23452160358428956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,16,power_law_1.2,0.3150207996368408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,16,power_law_1.2,0.4260223865509033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,16,power_law_1.2,0.5350592136383057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,16,power_law_1.2,0.7883520126342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,4,power_law_1.2,0.022284799814224245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,4,power_law_1.2,0.02797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,4,power_law_1.2,0.040057599544525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,4,power_law_1.2,0.050400000810623166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,4,power_law_1.2,0.05533440113067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,4,power_law_1.2,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,4,power_law_1.2,0.07611520290374756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,4,power_law_1.2,0.07459840178489685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,4,power_law_1.2,0.0726207971572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,4,power_law_1.2,0.07741439938545228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,4,power_law_1.2,0.0770687997341156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,4,power_law_1.2,0.08178560137748718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,4,power_law_1.2,0.08058879971504211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,4,power_law_1.2,0.08849920034408569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,4,power_law_1.2,0.09214079976081849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,4,power_law_1.2,0.09781759977340698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,4,power_law_1.2,0.0991487979888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,4,power_law_1.2,0.11137280464172364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.2,0.12410880327224731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.2,0.15445760488510132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.2,0.20208640098571778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.2,0.22955520153045655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.2,0.25319681167602537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.2,0.3349951982498169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.2,0.40568318367004397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.2,0.5445824146270752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.2,0.734284782409668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,2,balanced,0.047882666190465294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,2,balanced,0.04978133241335551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,2,balanced,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,2,balanced,0.0651039977868398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,2,balanced,0.09321066737174988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,2,balanced,0.13896532853444418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,2,balanced,0.14411200086275736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,2,balanced,0.1362506647904714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,2,balanced,0.13699199755986533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,2,balanced,0.13943466544151306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,2,balanced,0.14033599694569907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,2,balanced,0.14416000247001648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,2,balanced,0.1469439963499705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,2,balanced,0.151119997104009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,2,balanced,0.15468266606330872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,2,balanced,0.16011733810106912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,2,balanced,0.1683573325475057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,2,balanced,0.18888533115386963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,2,balanced,0.20410666863123575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,2,balanced,0.24665600061416626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,2,balanced,0.28276799122492474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,2,balanced,0.3949386676152547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,2,balanced,0.4536213477452596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,2,balanced,0.6662186781565348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,2,balanced,0.8172372976938883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,2,balanced,1.199562629063924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,2,balanced,1.5509653091430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,1,power_law_1.01,0.04465279877185822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,1,power_law_1.01,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,1,power_law_1.01,0.052134400606155394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,1,power_law_1.01,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,1,power_law_1.01,0.06512640118598938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,1,power_law_1.01,0.06682239770889283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,1,power_law_1.01,0.06731520295143127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,1,power_law_1.01,0.06730239987373351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,1,power_law_1.01,0.06728960275650024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,1,power_law_1.01,0.06815360188484192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,1,power_law_1.01,0.06878719925880432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,1,power_law_1.01,0.07340160012245178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,1,power_law_1.01,0.0731328010559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,1,power_law_1.01,0.07813760042190551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,1,power_law_1.01,0.09625599980354309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,1,power_law_1.01,0.09728639721870422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,1,power_law_1.01,0.09883520007133484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,1,power_law_1.01,0.11658240556716919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,4,balanced,0.05194133520126343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,4,balanced,0.05406400064627329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.01,0.14032000303268433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,4,balanced,0.07043200234572093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,4,balanced,0.09928533434867859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,4,balanced,0.13554666439692178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,4,balanced,0.16837332646052042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,4,balanced,0.16908266146977743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,4,balanced,0.16851733128229776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,4,balanced,0.1701386570930481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,4,balanced,0.17069866259892783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,4,balanced,0.17292799552281699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,4,balanced,0.17633066574732462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,4,balanced,0.17864000797271729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,4,balanced,0.1795039971669515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,4,balanced,0.18756800889968872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,4,balanced,0.19138665994008383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,4,balanced,0.20082666476567587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,4,balanced,0.22501333554585776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,4,balanced,0.23627734184265137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,4,balanced,0.3160586754480998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.01,0.18195199966430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,4,balanced,0.3333066701889038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,4,balanced,0.5205866495768229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,4,balanced,0.4950133164723714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,4,balanced,0.715008020401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,4,balanced,0.8194239934285482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,4,balanced,1.2080106735229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,4,balanced,1.47324800491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.01,0.21825919151306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.01,0.29852800369262694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.01,0.3826368093490601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.01,0.5492544174194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.01,0.7183807849884033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.01,1.0355072021484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,1,power_law_1.2,0.10248960256576538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,1,power_law_1.2,0.14963200092315673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,1,power_law_1.2,0.24546558856964112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,1,power_law_1.2,0.3990080118179321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,1,power_law_1.2,0.5834559917449951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,1,power_law_1.2,0.8017919540405274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,1,power_law_1.2,1.1772607803344726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,1,power_law_1.2,1.239731216430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,256,power_law_1.01,0.16720000505447388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,1,power_law_1.2,1.249824047088623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,256,power_law_1.01,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,1,power_law_1.2,1.3172672271728516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,256,power_law_1.01,0.05788159966468811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,256,power_law_1.01,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,1,power_law_1.2,1.3361663818359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,256,power_law_1.01,0.05541120171546936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,1,power_law_1.2,1.4403840065002442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,256,power_law_1.01,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,256,power_law_1.01,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,1,power_law_1.2,1.4048640251159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,256,power_law_1.01,0.05545600056648255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,256,power_law_1.01,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,1,power_law_1.2,1.510035228729248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,256,power_law_1.01,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,256,power_law_1.01,0.05939840078353882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,256,power_law_1.01,0.06552960276603699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,256,power_law_1.01,0.0691968023777008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,256,power_law_1.01,0.07625600099563598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,256,power_law_1.01,0.08346880078315735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,256,power_law_1.01,0.08490239977836608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,256,power_law_1.01,0.09227520227432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,256,power_law_1.01,0.10704640150070191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,1,power_law_1.2,1.5678976058959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,256,power_law_1.01,0.11151360273361206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,256,power_law_1.01,0.14206720590591432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,1,power_law_1.2,1.5475647926330567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,256,power_law_1.01,0.15207040309906006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,256,power_law_1.01,0.19956480264663695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,1,power_law_1.2,1.6841535568237305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,256,power_law_1.01,0.20527360439300538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,256,power_law_1.01,0.2835776090621948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,1,power_law_1.2,1.864313507080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,256,power_law_1.01,0.3427839994430542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,1,power_law_1.2,1.892460823059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,1,power_law_1.2,2.0842239379882814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,256,power_law_1.01,0.5175615787506104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,1,power_law_1.2,2.137759971618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,256,power_law_1.01,0.5956672191619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,1,power_law_1.2,2.429747200012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,1,power_law_1.2,2.7954559326171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,1,power_law_1.2,3.5042110443115235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,1,power_law_1.2,4.057292938232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,1,power_law_1.2,5.306297683715821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,1,power_law_1.2,6.450592041015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.01,1.3848064422607422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,4,balanced,0.04811733464399973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,4,balanced,0.06267199913660686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,4,balanced,0.08594666918118794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,4,balanced,0.13435733318328857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,4,balanced,0.21521600087483725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,4,balanced,0.3849173386891683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,4,balanced,0.3875466585159302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,4,balanced,0.3867199818293254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,4,balanced,0.3881760040918986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,4,balanced,0.38929065068562824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,4,balanced,0.39060266812642414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,4,balanced,0.39337066809336346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,4,balanced,0.3970400094985962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,4,balanced,0.40087465445200604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,4,balanced,0.4058506488800049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,4,balanced,0.4098186492919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,4,balanced,0.41420265038808185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,4,balanced,0.4321066538492839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,4,balanced,0.44441068172454834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,4,balanced,0.4721599817276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,4,balanced,0.499781330426534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,4,balanced,0.5418026844660441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,4,balanced,0.601306676864624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,4,balanced,0.7121547063191732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,4,balanced,0.7957066694895426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,4,balanced,1.1712853113810222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,4,balanced,1.2756373087565105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,8,power_law_1.2,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,8,power_law_1.2,0.05820159912109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,8,power_law_1.2,0.05942400097846985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,8,power_law_1.2,0.0649728000164032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,8,power_law_1.2,0.06807039976119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,8,power_law_1.2,0.07184640169143677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,8,power_law_1.2,0.07297919988632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,8,power_law_1.2,0.07144960165023803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,8,power_law_1.2,0.07065600156784058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,8,power_law_1.2,0.07299839854240417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,8,power_law_1.2,0.07504640221595764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,8,power_law_1.2,0.07668480277061462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,8,power_law_1.2,0.0782975971698761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,8,power_law_1.2,0.08091520071029663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,8,power_law_1.2,0.0916159987449646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,8,power_law_1.2,0.09210240244865417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,8,power_law_1.2,0.10097279548645019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,8,power_law_1.2,0.12189439535140992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,8,power_law_1.2,0.13704960346221923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,8,power_law_1.2,0.17549439668655395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,8,power_law_1.2,0.1730239987373352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,8,power_law_1.2,0.23033599853515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,8,power_law_1.2,0.30320000648498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,8,power_law_1.2,0.39353599548339846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,8,power_law_1.2,0.4967167854309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,8,power_law_1.2,0.6959616184234619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,8,power_law_1.2,1.0013055801391602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,16,power_law_1.2,0.03699199855327606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,16,power_law_1.2,0.03189760148525238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,16,power_law_1.2,0.030451199412345885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,16,power_law_1.2,0.030969598889350893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,16,power_law_1.2,0.03057279884815216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,16,power_law_1.2,0.032627201080322264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,16,power_law_1.2,0.033606401085853575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,16,power_law_1.2,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,16,power_law_1.2,0.03406080007553101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,16,power_law_1.2,0.0348800003528595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,16,power_law_1.2,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,16,power_law_1.2,0.036723199486732486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,16,power_law_1.2,0.03726080060005188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,16,power_law_1.2,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,16,power_law_1.2,0.04464640021324158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,16,power_law_1.2,0.04702079892158508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,16,power_law_1.2,0.04780159890651703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,16,power_law_1.2,0.05095040202140808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,16,power_law_1.2,0.05371519923210144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,16,power_law_1.2,0.06496639847755432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,16,power_law_1.2,0.07280640006065368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,16,power_law_1.2,0.0927295982837677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,16,power_law_1.2,0.11645439863204957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,4,power_law_1.2,0.025472000241279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,4,power_law_1.2,0.04424319863319397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,4,power_law_1.2,0.051123201847076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,16,power_law_1.2,0.1811776041984558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,4,power_law_1.2,0.06481279730796814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,4,power_law_1.2,0.07541760206222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,4,power_law_1.2,0.08104320168495179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,4,power_law_1.2,0.10023679733276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,4,power_law_1.2,0.1089792013168335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,4,power_law_1.2,0.11117440462112427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,4,power_law_1.2,0.10907520055770874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,4,power_law_1.2,0.11352319717407226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,4,power_law_1.2,0.11383680105209351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,4,power_law_1.2,0.12319359779357911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,4,power_law_1.2,0.12095359563827515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,4,power_law_1.2,0.14803839921951295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,8,power_law_1.2,0.026144000887870788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,4,power_law_1.2,0.14025599956512452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,8,power_law_1.2,0.03741439878940582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,8,power_law_1.2,0.039110401272773744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,8,power_law_1.2,0.04673280119895935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,8,power_law_1.2,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,8,power_law_1.2,0.04744960069656372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,8,power_law_1.2,0.0531391978263855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,8,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,4,power_law_1.2,0.13836159706115722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,8,power_law_1.2,0.06258559823036194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,4,power_law_1.2,0.27411839962005613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,8,power_law_1.2,0.06364160180091857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,4,power_law_1.2,0.28582398891448973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,8,power_law_1.2,0.05793920159339905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,8,power_law_1.2,0.06106240153312683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,4,power_law_1.2,0.24173440933227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,8,power_law_1.2,0.06262400150299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,4,power_law_1.2,0.23633279800415039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,8,power_law_1.2,0.0643392026424408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,4,power_law_1.2,0.29071359634399413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,16,power_law_1.2,0.20356481075286864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,4,power_law_1.2,0.31808640956878664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,4,power_law_1.2,0.4121664047241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,4,power_law_1.2,0.48131837844848635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,8,power_law_1.2,0.0680191993713379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,4,power_law_1.2,0.703327989578247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,4,power_law_1.2,0.8632063865661621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,8,power_law_1.2,0.07121919989585876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,8,power_law_1.2,0.07319679856300354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,8,power_law_1.2,0.07866880297660828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,8,power_law_1.2,0.14140160083770753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,8,power_law_1.2,0.17989759445190429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,8,power_law_1.2,0.12814079523086547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,8,power_law_1.2,0.15123200416564941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,8,power_law_1.2,0.1593791961669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,8,power_law_1.2,0.19967999458312988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,8,power_law_1.2,0.26428799629211425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,8,power_law_1.2,0.37985920906066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,8,power_law_1.2,0.45397119522094725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,16,power_law_1.2,0.3119744062423706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,16,power_law_1.2,0.42485761642456055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,2,balanced,0.049685334165891014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,8,power_law_1.01,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,8,power_law_1.01,0.06094080209732056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,8,power_law_1.01,0.05950719714164734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,8,power_law_1.01,0.061260801553726194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,8,power_law_1.01,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,8,power_law_1.01,0.07253760099411011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,8,power_law_1.01,0.07489920258522034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,2,balanced,0.050794666012128196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,2,balanced,0.05589333176612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,2,balanced,0.06821866830190022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,8,power_law_1.01,0.07313920259475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,2,balanced,0.093231995900472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,2,balanced,0.14249599973360697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,2,balanced,0.19081600507100424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,2,balanced,0.17460266749064127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,2,balanced,0.17371733983357748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,2,balanced,0.17671465873718262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,2,balanced,0.17806400855382284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,2,balanced,0.18181333939234415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,2,balanced,0.18388799826304117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,2,balanced,0.18935465812683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,2,balanced,0.19618666172027588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,8,power_law_1.01,0.07388160228729249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,2,balanced,0.20106667280197144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,8,power_law_1.01,0.07601919770240784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,2,balanced,0.21090133984883627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,8,power_law_1.01,0.07390720248222352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,8,power_law_1.01,0.07560319900512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,8,power_law_1.01,0.07887359857559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,8,power_law_1.01,0.08231679797172546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,8,power_law_1.01,0.09497600197792053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,8,power_law_1.01,0.09468160271644592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,8,power_law_1.01,0.09932159781455993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,8,power_law_1.01,0.11349760293960572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,8,power_law_1.01,0.11025279760360718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,8,power_law_1.01,0.14670720100402831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,8,power_law_1.01,0.12599680423736573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,2,balanced,0.23373866081237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,2,balanced,0.25248533487319946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,8,power_law_1.01,0.15495680570602416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,2,balanced,0.29848533868789673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,2,balanced,0.33507732550303143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,8,power_law_1.01,0.16278400421142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,2,balanced,0.4243786732355754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,8,power_law_1.01,0.21258881092071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,2,balanced,0.5336800018946329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,8,power_law_1.01,0.2840768098831177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,2,balanced,0.7054826418558756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,8,power_law_1.01,0.3238271951675415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,8,power_law_1.01,0.4211584091186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,2,balanced,0.9188746611277262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,2,balanced,1.3209653695424397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,2,balanced,1.7377173105875652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,1,power_law_1.01,0.10774400234222412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,1,power_law_1.01,0.11290240287780762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,1,power_law_1.01,0.1498047947883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,1,power_law_1.01,0.1713088035583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,1,power_law_1.01,0.18691200017929077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,1,power_law_1.01,0.21006081104278565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,1,power_law_1.01,0.24263041019439696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,1,power_law_1.01,0.24872961044311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,1,power_law_1.01,0.2596415996551514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,1,power_law_1.01,0.262175989151001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,1,power_law_1.01,0.2674623966217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,1,power_law_1.01,0.27457280158996583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,1,power_law_1.01,0.28185598850250243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,1,power_law_1.01,0.28806400299072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,1,power_law_1.01,0.29462399482727053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,1,power_law_1.01,0.2979520082473755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,1,power_law_1.01,0.3076544046401978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,1,power_law_1.01,0.3428544044494629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,1,power_law_1.01,0.3739840030670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,1,power_law_1.01,0.447603178024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,1,power_law_1.01,0.5199679851531982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,1,power_law_1.2,0.1041599988937378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,1,power_law_1.2,0.10787839889526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,1,power_law_1.2,0.11665279865264892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,1,power_law_1.2,0.14742399454116822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,1,power_law_1.2,0.1843008041381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,1,power_law_1.2,0.21894400119781493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,1,power_law_1.2,0.2426367998123169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,1,power_law_1.2,0.30138239860534666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,1,power_law_1.2,0.3138751983642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,1,power_law_1.2,0.31598079204559326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,1,power_law_1.2,0.3271104097366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,1,power_law_1.2,0.341593599319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,1,power_law_1.2,0.34985599517822263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,1,power_law_1.2,0.3721024036407471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,1,power_law_1.2,0.3863552093505859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,1,power_law_1.2,0.3889280080795288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,1,power_law_1.2,0.41994237899780273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,1,power_law_1.2,0.4807487964630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.2,0.5251264095306396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.2,0.6339776039123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.2,0.7202047824859619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.2,0.9155839920043946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.2,1.114367961883545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.2,1.5105664253234863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,1,power_law_1.2,2.060972785949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,1,power_law_1.2,2.8529983520507813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,1,power_law_1.2,3.491993713378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,32,balanced,0.05330666899681091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,32,balanced,0.043552001317342125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,32,balanced,0.04205866654713949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,32,balanced,0.047685335079828896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,32,balanced,0.05003199974695841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,32,balanced,0.06206933160622915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,32,balanced,0.06340266764163971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,32,balanced,0.06402666866779327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,32,balanced,0.06406933565934499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,32,balanced,0.06582933167616527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,32,balanced,0.06655466556549072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,32,balanced,0.06749333441257477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,32,balanced,0.06810133159160614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,32,balanced,0.07020266850789388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,32,balanced,0.07268799841403961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,32,balanced,0.07478933533032735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,32,balanced,0.07525333265463512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,32,balanced,0.08489066362380981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,32,balanced,0.08251200119654338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,32,balanced,0.09641066193580627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,32,balanced,0.0995253324508667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,32,balanced,0.1234933336575826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,32,balanced,0.13777599732081094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,32,balanced,0.18418665726979574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,32,balanced,0.19890666007995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,32,balanced,0.29305599133173627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,32,balanced,0.3327146569887797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,128,balanced,0.05194133520126343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,128,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,128,balanced,0.05400000015894572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,128,balanced,0.05624533196290334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,128,balanced,0.05388799806435903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,128,balanced,0.06206933160622915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,128,balanced,0.062037333846092224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,128,balanced,0.06230400005976359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,128,balanced,0.06238399942715963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,128,balanced,0.06201066573460897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,128,balanced,0.0606826643149058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,128,balanced,0.06413866579532623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,128,balanced,0.06613333523273468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,128,balanced,0.06438933312892914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,128,balanced,0.07055999835332234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,128,balanced,0.07021866738796234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,128,balanced,0.07628799974918365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,128,balanced,0.07871466875076294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,128,balanced,0.0843946635723114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,128,balanced,0.09585600097974141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,128,balanced,0.10547199845314026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,128,balanced,0.135861337184906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,128,balanced,0.16088533401489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,128,balanced,0.21017066637674967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,128,balanced,0.2583306630452474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,128,balanced,0.3551093339920044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,128,balanced,0.4490346511205037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,64,power_law_1.2,0.0219200000166893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,64,power_law_1.2,0.0220223993062973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,64,power_law_1.2,0.02232320010662079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,64,power_law_1.2,0.022393600642681123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,64,power_law_1.2,0.025235199928283693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,64,power_law_1.2,0.03130879998207092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,64,power_law_1.2,0.05381760001182556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,64,power_law_1.2,0.05550720095634461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,64,power_law_1.2,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,64,power_law_1.2,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,64,power_law_1.2,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,64,power_law_1.2,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,64,power_law_1.2,0.05235199928283692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,64,power_law_1.2,0.05141760110855102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,64,power_law_1.2,0.05102720260620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,64,power_law_1.2,0.05311999917030334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,64,power_law_1.2,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,64,power_law_1.2,0.05292159914970398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.2,0.05948160290718078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.2,0.07933440208435058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.2,0.09198079705238342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.2,0.12881920337677003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.2,0.1581760048866272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.2,0.21034879684448243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.2,0.27853438854217527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.2,0.40717439651489257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.2,0.5703360080718994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,16,power_law_1.2,0.044761601090431216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,16,power_law_1.2,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,16,power_law_1.2,0.0456063985824585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,16,power_law_1.2,0.0483711987733841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,16,power_law_1.2,0.048563200235366824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,16,power_law_1.2,0.0484607994556427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,16,power_law_1.2,0.048767998814582825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,16,power_law_1.2,0.050367999076843264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,16,power_law_1.2,0.04996480047702789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,16,power_law_1.2,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,16,power_law_1.2,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,16,power_law_1.2,0.053235197067260744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,16,power_law_1.2,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,16,power_law_1.2,0.0574400007724762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,16,power_law_1.2,0.06257920265197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,16,power_law_1.2,0.06558719873428345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,16,power_law_1.2,0.06562560200691223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,1,balanced,0.07532266775767009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,16,power_law_1.2,0.07470080256462097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,1,balanced,0.10871466994285583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,1,balanced,0.17192000150680542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,16,power_law_1.2,0.0798143982887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,1,balanced,0.3024266759554545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,16,power_law_1.2,0.09859840273857116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,1,balanced,0.5522186756134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,1,balanced,1.0571146806081135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,1,balanced,1.0640426476796467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,1,balanced,1.0678133169809978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,1,balanced,1.0728533267974854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,1,balanced,1.0790773232777913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,1,balanced,1.073632001876831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,1,power_law_1.2,0.05898879766464234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,1,balanced,1.0829866727193196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,1,power_law_1.2,0.06302719712257385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,1,balanced,1.0865493615468342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,1,power_law_1.2,0.08200960159301758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,1,balanced,1.0971840222676594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,1,power_law_1.2,0.11530239582061767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,1,balanced,1.1049173672993977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,1,balanced,1.1110933621724446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,1,power_law_1.2,0.15459200143814086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,1,balanced,1.1258347034454346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,16,power_law_1.2,0.1109760046005249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,1,balanced,1.1493546962738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,16,power_law_1.2,0.1480064034461975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,1,balanced,1.18885334332784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,16,power_law_1.2,0.19381120204925537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,1,balanced,1.2391573588053386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,1,balanced,1.3102773030598958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,1,power_law_1.2,0.19560960531234742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,1,power_law_1.2,0.2287168025970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,16,power_law_1.2,0.2600255966186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,1,power_law_1.2,0.2794624090194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,16,power_law_1.2,0.3167360067367554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,1,power_law_1.2,0.3020416021347046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,16,power_law_1.2,0.4853759765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,1,balanced,1.4053173065185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,16,power_law_1.2,0.6234303951263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,1,balanced,1.5003466606140137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,1,balanced,1.8881759643554688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,1,balanced,1.9419573148091633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,1,power_law_1.2,0.30115199089050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,1,power_law_1.2,0.32057600021362304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,1,power_law_1.2,0.3372159957885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,1,power_law_1.2,0.3589567899703979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,1,power_law_1.2,0.38222079277038573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,1,power_law_1.2,0.39723520278930663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,1,power_law_1.2,0.4222911834716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,1,power_law_1.2,0.46002559661865233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,1,power_law_1.2,0.5521728038787842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,1,balanced,2.829312006632487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,1,power_law_1.2,0.5651584148406983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,1,power_law_1.2,0.6894847869873046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,1,balanced,3.1645758946736655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,1,power_law_1.2,0.7690688133239746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,1,power_law_1.2,0.9825152397155762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,1,power_law_1.2,1.209830379486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,1,power_law_1.2,1.6144704818725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,1,power_law_1.2,2.0705087661743162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,1,power_law_1.2,2.8056768417358398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,1,power_law_1.2,3.567705535888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,32,balanced,0.050186668833096824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,32,balanced,0.032826667030652366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,32,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,32,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,32,balanced,0.03344533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,32,balanced,0.038058665891488395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,32,balanced,0.03596800069014231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,32,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,32,balanced,0.03745600084463755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,32,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,32,balanced,0.035973332822322845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,32,balanced,0.03739733248949051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,32,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,32,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,32,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,32,balanced,0.04182399809360504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,32,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,32,balanced,0.05586666862169901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,32,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,32,balanced,0.057909334699312844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,128,power_law_1.2,0.030393600463867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,32,balanced,0.05275199810663859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,32,balanced,0.0663679987192154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,128,power_law_1.2,0.03041279911994934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,32,balanced,0.06673066814740498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,32,balanced,0.08587200442949931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,128,power_law_1.2,0.0315775990486145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,32,balanced,0.08455999692281087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,32,balanced,0.11741333206494649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,32,balanced,0.11964799960454305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,128,power_law_1.2,0.03728640079498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,128,power_law_1.2,0.03592959940433502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,128,power_law_1.2,0.03224959969520569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,128,power_law_1.2,0.03187200129032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,128,power_law_1.2,0.03194240033626557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,128,power_law_1.2,0.03219839930534363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,128,power_law_1.2,0.032716798782348636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,128,power_law_1.2,0.03375360071659088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,128,power_law_1.2,0.03484799861907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,128,power_law_1.2,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,128,power_law_1.2,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,128,power_law_1.2,0.0406143993139267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,128,power_law_1.2,0.042182400822639465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,128,power_law_1.2,0.04398080110549927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,128,power_law_1.2,0.05175679922103882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,128,power_law_1.2,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,128,power_law_1.2,0.07033600211143494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,128,power_law_1.2,0.08131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,128,power_law_1.2,0.11080319881439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,128,power_law_1.2,0.12666239738464355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,128,power_law_1.2,0.18795520067214966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,128,power_law_1.2,0.22195839881896973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,128,power_law_1.2,0.36433920860290525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,16,power_law_1.2,0.04750719964504242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,16,power_law_1.2,0.05930240154266357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,16,power_law_1.2,0.04864639937877655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,16,power_law_1.2,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,16,power_law_1.2,0.05048959851264954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,16,power_law_1.2,0.049856001138687135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,16,power_law_1.2,0.051283198595047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,16,power_law_1.2,0.05237759947776795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,16,power_law_1.2,0.053529602289199826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,16,power_law_1.2,0.05466880202293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,16,power_law_1.2,0.05548160076141358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,16,power_law_1.2,0.05722879767417908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,16,power_law_1.2,0.060236799716949466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,16,power_law_1.2,0.06108160018920898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,16,power_law_1.2,0.06815999746322632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,16,power_law_1.2,0.07136639952659607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,16,power_law_1.2,0.07471359968185425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,128,power_law_1.2,0.3674367904663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,16,power_law_1.2,0.0846783995628357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,16,power_law_1.2,0.09784319996833801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,16,power_law_1.2,0.12549760341644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,16,power_law_1.2,0.15015679597854614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,16,power_law_1.2,0.20422399044036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,16,power_law_1.2,0.25111680030822753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,16,power_law_1.2,0.35164799690246584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,16,power_law_1.2,0.44313597679138184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,16,power_law_1.2,0.6484543800354003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,16,power_law_1.2,0.8751423835754395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,8,power_law_1.2,0.04280959963798523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,8,power_law_1.2,0.040038400888442995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,8,power_law_1.2,0.043507200479507444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,8,power_law_1.2,0.04865919947624207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,8,power_law_1.2,0.04794879853725433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,8,power_law_1.2,0.05294079780578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,8,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,8,power_law_1.2,0.05380480289459229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,8,power_law_1.2,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,8,power_law_1.2,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,8,power_law_1.2,0.05674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,8,power_law_1.2,0.05631999969482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,8,power_law_1.2,0.05927039980888367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,8,power_law_1.2,0.06087039709091187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,8,power_law_1.2,0.07167999744415283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,8,power_law_1.2,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,8,power_law_1.2,0.07596799731254578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,8,power_law_1.2,0.08548480272293091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,8,power_law_1.2,0.09791359901428223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,8,power_law_1.2,0.12035200595855713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,8,power_law_1.2,0.1522047996520996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,8,power_law_1.2,0.19731199741363525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,8,power_law_1.2,0.2734528064727783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,8,power_law_1.2,0.4028480052947998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,8,power_law_1.2,0.5467584133148193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,8,power_law_1.2,0.7697216033935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,64,power_law_1.01,0.018367999792099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,64,power_law_1.01,0.01841920018196106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,64,power_law_1.01,0.017888000607490538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,64,power_law_1.01,0.0176704004406929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,64,power_law_1.01,0.017632000148296356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,64,power_law_1.01,0.019072000682353974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,64,power_law_1.01,0.01996160000562668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,64,power_law_1.01,0.02407039999961853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,64,power_law_1.01,0.024294400215148927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,64,power_law_1.01,0.023372800648212434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,64,power_law_1.01,0.03416320085525513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,64,power_law_1.01,0.03501439988613129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,64,power_law_1.01,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,64,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,64,power_law_1.01,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,64,power_law_1.01,0.03338240087032318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,64,power_law_1.01,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,64,power_law_1.01,0.03712640106678009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,64,power_law_1.01,0.03866879940032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,64,power_law_1.01,0.04444159865379334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,64,power_law_1.01,0.053299200534820554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,64,power_law_1.01,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,64,power_law_1.01,0.07871999740600585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,64,power_law_1.01,0.1027519941329956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,64,power_law_1.01,0.12609280347824098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,64,power_law_1.01,0.18991999626159667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,64,power_law_1.01,0.23712000846862794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,8,power_law_1.2,1.0956095695495605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,4,balanced,0.021397332350413006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,4,balanced,0.0201706662774086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,4,balanced,0.02128000060717265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,4,balanced,0.023157333334287006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,4,balanced,0.029605334003766377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,4,balanced,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,4,balanced,0.03339733431736628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,4,balanced,0.03583999971548716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,4,balanced,0.03645866612593333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,4,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,4,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,4,balanced,0.04163199911514918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,4,balanced,0.04675200084845225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,4,balanced,0.047050664822260536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,4,balanced,0.046581332882245384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,4,balanced,0.0507893313964208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,4,balanced,0.05898666878541311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,4,balanced,0.06234666705131531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,4,balanced,0.10026133060455322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,4,balanced,0.12450133760770161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,4,balanced,0.11289599537849426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,4,balanced,0.1325706640879313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,4,balanced,0.14038399855295816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,4,balanced,0.22004799048105875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,4,balanced,0.24686932563781738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,4,balanced,0.35547200838724774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,4,balanced,0.46268800894419354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,1,8,balanced,0.08745066324869792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,1,8,balanced,0.08699733018875122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,1,8,balanced,0.0879306693871816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,1,8,balanced,0.09109866619110107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,1,8,balanced,0.09302933017412822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,1,8,balanced,0.09303466478983562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,1,8,balanced,0.09324799974759419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,1,8,balanced,0.09617066383361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,1,8,balanced,0.0953439970811208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,1,8,balanced,0.09597333272298177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,1,8,balanced,0.09572266538937886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,1,8,balanced,0.1029919981956482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,1,8,balanced,0.10338667035102844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,1,8,balanced,0.11913599570592244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,1,8,balanced,0.12824533383051553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,1,8,balanced,0.11211199561754863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,1,8,balanced,0.13063466548919678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,1,8,balanced,0.14866666992505392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,1,8,balanced,0.15455466508865356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,1,8,balanced,0.1841920018196106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,1,8,balanced,0.208021342754364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,1,8,balanced,0.31088000535964966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,1,8,balanced,0.36524800459543866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,1,8,balanced,0.5243253310521444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,1,8,balanced,0.6649973392486572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,1,8,balanced,0.9517973264058431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,1,8,balanced,1.2462506294250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,4,power_law_1.2,0.06464639902114869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,4,power_law_1.2,0.08495360016822814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,4,power_law_1.2,0.09543039798736572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,4,power_law_1.2,0.11987199783325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,4,power_law_1.2,0.15089279413223267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,4,power_law_1.2,0.15837440490722657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,4,power_law_1.2,0.20220799446105958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,4,power_law_1.2,0.19502079486846924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,4,power_law_1.2,0.20021119117736816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,4,power_law_1.2,0.20682239532470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,4,power_law_1.2,0.203385591506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,4,power_law_1.2,0.2082495927810669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,4,power_law_1.2,0.20903680324554444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,4,power_law_1.2,0.21869440078735353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,4,power_law_1.2,0.23543040752410888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,4,power_law_1.2,0.24208641052246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,4,power_law_1.2,0.2523711919784546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,4,power_law_1.2,0.2919487953186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,4,power_law_1.2,0.301311993598938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,4,power_law_1.2,0.37049601078033445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,4,power_law_1.2,0.4052608013153076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,4,power_law_1.2,0.49933438301086425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,4,power_law_1.2,0.5722367763519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,4,power_law_1.2,0.7634751796722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,4,power_law_1.2,0.9964287757873536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,4,power_law_1.2,1.3073920249938964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,4,power_law_1.2,1.6728191375732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,2,power_law_1.01,0.10241279602050782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,32,power_law_1.01,0.02460159957408905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,2,power_law_1.01,0.14402559995651246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,32,power_law_1.01,0.02099200040102005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,32,power_law_1.01,0.02099840044975281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,2,power_law_1.01,0.1717311978340149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,32,power_law_1.01,0.022592000663280487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,2,power_law_1.01,0.2585088014602661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,32,power_law_1.01,0.022873599827289582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,32,power_law_1.01,0.02542720139026642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,32,power_law_1.01,0.025363200902938844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,32,power_law_1.01,0.02502399981021881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,32,power_law_1.01,0.02831999957561493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,32,power_law_1.01,0.028729599714279175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,32,power_law_1.01,0.030976000428199767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,32,power_law_1.01,0.03113600015640259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,32,power_law_1.01,0.03110400140285492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,32,power_law_1.01,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,2,power_law_1.01,0.30500481128692625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,2,power_law_1.01,0.5016511917114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,2,power_law_1.01,0.5063488006591796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,2,power_law_1.01,0.5128640174865723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,2,power_law_1.01,0.5464255809783936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,2,power_law_1.01,0.5421504020690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,32,power_law_1.01,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,32,power_law_1.01,0.034457600116729735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,2,power_law_1.01,0.5533567905426026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,32,power_law_1.01,0.0349375993013382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,32,power_law_1.01,0.036748799681663516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,32,power_law_1.01,0.04092159867286682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,32,power_law_1.01,0.04434559941291809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,32,power_law_1.01,0.04850560128688812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,32,power_law_1.01,0.05782399773597717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,32,power_law_1.01,0.06840320229530335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,32,power_law_1.01,0.09253119826316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,2,power_law_1.01,0.5544960021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,32,power_law_1.01,0.11389440298080444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,2,power_law_1.01,0.5826176166534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,32,power_law_1.01,0.14172799587249757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,2,power_law_1.01,0.5781248092651368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,32,power_law_1.01,0.16876800060272218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,2,power_law_1.01,0.5963200092315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,2,power_law_1.01,0.6177984237670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,2,power_law_1.01,0.6734335899353028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,2,power_law_1.01,0.7584447860717773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,2,power_law_1.01,0.7435904026031495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,2,power_law_1.01,0.8789119720458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,2,power_law_1.01,0.9309439659118652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,2,power_law_1.01,1.1304320335388183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,2,power_law_1.01,1.2873536109924317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,2,power_law_1.01,1.6159744262695312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,2,power_law_1.01,2.226367950439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,2,power_law_1.01,2.8242368698120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,2,power_law_1.01,3.4845375061035155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,8,power_law_1.01,0.04749439954757691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,8,power_law_1.01,0.04650880098342895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,8,power_law_1.01,0.04519039988517761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,8,power_law_1.01,0.04679040014743805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,8,power_law_1.01,0.04847359955310822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,8,power_law_1.01,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,8,power_law_1.01,0.07345920205116271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,8,power_law_1.01,0.0657920002937317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,8,power_law_1.01,0.08826239705085755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,8,power_law_1.01,0.07891839742660522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,8,power_law_1.01,0.07459200024604798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,8,power_law_1.01,0.13255679607391357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,8,power_law_1.01,0.12510720491409302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,8,power_law_1.01,0.09955840110778809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,8,power_law_1.01,0.14496639966964722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,8,power_law_1.01,0.12503039836883545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,8,power_law_1.01,0.16097919940948485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,8,power_law_1.01,0.21115520000457763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.01,0.2718463897705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.01,0.3868096113204956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.01,0.5189311981201172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.01,0.7438720226287842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.01,0.9850496292114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.01,1.4541119575500487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.01,1.9153919219970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,1,balanced,0.05190399785836538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,1,balanced,0.07220800220966339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,1,balanced,0.10129066308339436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.01,2.8695999145507813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,1,balanced,0.16012266278266907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,1,balanced,0.264080007870992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,1,balanced,0.3230773409207662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,1,balanced,0.32496533791224164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,1,balanced,0.3308853308359782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,1,balanced,0.3309760093688965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,1,balanced,0.33397865295410156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,2,power_law_1.2,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,1,balanced,0.3382026751836141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,1,balanced,0.34082667032877606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,2,power_law_1.2,0.11790080070495605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,1,balanced,0.3461493253707886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,1,balanced,0.35683198769887287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,1,balanced,0.3625173171361287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,1,balanced,0.37059199810028076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,1,balanced,0.38489067554473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,1,balanced,0.42019200325012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,1,balanced,0.4471199909845988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,1,balanced,0.5849706729253134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,1,balanced,0.5906080007553101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,1,balanced,0.9075466791788737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,2,power_law_1.2,0.132915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,1,balanced,0.8834133148193359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,2,power_law_1.2,0.17096960544586182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,2,power_law_1.2,0.20008320808410646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,1,balanced,1.4829920132954915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,2,power_law_1.2,0.2237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,1,balanced,1.5472426414489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,2,power_law_1.2,0.3012991905212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,2,power_law_1.2,0.32445440292358396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,1,balanced,2.495525360107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,2,power_law_1.2,0.2914560079574585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,2,power_law_1.2,0.3122560024261475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,1,balanced,2.893712043762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,2,power_law_1.2,0.33145599365234374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,2,power_law_1.2,0.32953600883483886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,32,balanced,0.049839998284975685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,2,power_law_1.2,0.3368832111358643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,32,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,32,balanced,0.05132266879081726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,2,power_law_1.2,0.33886721134185793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,32,balanced,0.05473066866397858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,32,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,2,power_law_1.2,0.34262399673461913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,32,balanced,0.07211733361085255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,32,balanced,0.07247999807198842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,2,power_law_1.2,0.35674879550933836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,32,balanced,0.07230933507283528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,32,balanced,0.07520000139872234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,32,balanced,0.07422400017579396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,32,balanced,0.0758133331934611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,2,power_law_1.2,0.37828478813171384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,2,power_law_1.2,0.40956802368164064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.2,0.4444096088409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.2,0.5088704109191895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.2,0.5771584033966064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.2,0.7070720195770264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,32,balanced,0.08075733482837677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,32,balanced,0.08036800225575765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,32,balanced,0.08247999846935272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.2,0.8786496162414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.2,1.186847972869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.2,1.3505023956298827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,32,balanced,0.09699199597040813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,32,balanced,0.09715200463930766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,32,balanced,0.09500267108281453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.2,1.9525568008422851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,32,balanced,0.10822400450706482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,32,balanced,0.11284266908963521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,32,balanced,0.12892799576123556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,32,balanced,0.1439359982808431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,32,balanced,0.17497066656748453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.2,2.413862419128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,32,balanced,0.20016000668207803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,32,balanced,0.25748799244562787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,32,balanced,0.3109653393427531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,32,balanced,0.4270240068435669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,32,balanced,0.5615466833114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.01,3.7993854522705077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,32,power_law_1.2,0.03404799997806549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,32,power_law_1.2,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,32,power_law_1.2,0.028409600257873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,32,power_law_1.2,0.030105599761009218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,32,power_law_1.2,0.030105599761009218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,32,power_law_1.2,0.031948798894882204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,32,power_law_1.2,0.03194240033626557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,32,power_law_1.2,0.03271040022373199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,32,power_law_1.2,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,32,power_law_1.2,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,32,power_law_1.2,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,32,power_law_1.2,0.03589119911193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,32,power_law_1.2,0.0371071994304657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,32,power_law_1.2,0.03703039884567261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,32,power_law_1.2,0.04159359931945801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,32,power_law_1.2,0.04249599874019623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,32,power_law_1.2,0.0471807986497879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,32,power_law_1.2,0.05030400156974792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,32,power_law_1.2,0.05626879930496216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,32,power_law_1.2,0.062028801441192626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,32,power_law_1.2,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,32,power_law_1.2,0.10607999563217163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,32,power_law_1.2,0.11327359676361085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,32,power_law_1.2,0.16569600105285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,32,power_law_1.2,0.2094655990600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,16,balanced,0.048165331284205117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,16,balanced,0.04987733562787374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,16,balanced,0.05018133421738943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,16,balanced,0.05457599957784017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,16,balanced,0.05415999889373779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,16,balanced,0.06463466584682465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,16,balanced,0.06493866443634033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,16,balanced,0.06648000081380208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,16,balanced,0.06677866478761037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,16,balanced,0.06740800042947133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,16,balanced,0.06817600131034851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,16,balanced,0.07044266661008199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,32,balanced,0.04659200211366018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,16,balanced,0.07039999961853027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,16,balanced,0.07133333384990692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,16,balanced,0.07898666461308797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,16,balanced,0.07816533247629802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,16,balanced,0.0835093359152476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,16,balanced,0.09501333038012187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,16,balanced,0.10151466727256775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,16,balanced,0.11683733264605205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,16,balanced,0.1381173332532247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,16,balanced,0.16235733032226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,16,balanced,0.18483734130859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,16,balanced,0.24425599972407022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,16,balanced,0.3022186756134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,32,balanced,0.049365331729253135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,16,balanced,0.4114186763763428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,32,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,32,balanced,0.07222933570543925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,32,balanced,0.08306666711966197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,32,balanced,0.084197332461675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,32,balanced,0.08509866396586101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,32,balanced,0.08522666494051616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,32,balanced,0.08498133222262065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,32,balanced,0.08711466193199158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,32,balanced,0.08693333466847737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,32,balanced,0.09065066774686177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,32,balanced,0.09170666337013245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,32,balanced,0.09317866961161296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,16,balanced,0.5248213211695353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,32,balanced,0.09886399904886882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,32,balanced,0.0995786686738332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,32,balanced,0.10470400253931682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,32,balanced,0.11997866630554199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,32,balanced,0.12584533294041952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,32,balanced,0.1532960037390391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,32,balanced,0.17538666725158691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,32,balanced,0.23566399017969766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,32,balanced,0.22421866655349731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,32,balanced,0.2959199945131938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,32,balanced,0.3397066593170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,32,balanced,0.47246400515238446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,32,balanced,0.5667413473129272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,32,power_law_1.2,0.28813440799713136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,32,power_law_1.2,0.02622080147266388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,32,power_law_1.2,0.024646399915218352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,32,power_law_1.2,0.024268800020217897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,32,power_law_1.2,0.024966399371623992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,32,power_law_1.2,0.02563199996948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,32,power_law_1.2,0.026796799898147584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,32,power_law_1.2,0.026815998554229736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,32,power_law_1.2,0.028415998816490172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,32,power_law_1.2,0.03030399978160858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,32,power_law_1.2,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,32,power_law_1.2,0.03113600015640259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,32,power_law_1.2,0.03359360098838806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,32,power_law_1.2,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,32,power_law_1.2,0.03298560082912445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,32,power_law_1.2,0.0365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,32,power_law_1.2,0.03641600012779236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,32,power_law_1.2,0.03803519904613495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,32,power_law_1.2,0.04425599873065948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,32,power_law_1.2,0.04766719937324524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,32,power_law_1.2,0.05311999917030334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,32,power_law_1.2,0.05889279842376709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,32,power_law_1.2,0.0737600028514862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,32,power_law_1.2,0.09191039800643921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,32,power_law_1.2,0.10962560176849365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,32,power_law_1.2,0.13533439636230468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,32,power_law_1.2,0.39682559967041015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,32,power_law_1.2,0.18583680391311647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,16,power_law_1.2,0.01855359971523285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,16,power_law_1.2,0.018822400271892546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,16,power_law_1.2,0.018681600689888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,16,power_law_1.2,0.018195199966430663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,16,power_law_1.2,0.020294399559497835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,16,power_law_1.2,0.02222079932689667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,64,power_law_1.01,0.0319680005311966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,16,power_law_1.2,0.02510719895362854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,64,power_law_1.01,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,64,power_law_1.01,0.030406400561332703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,64,power_law_1.01,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,64,power_law_1.01,0.032313600182533264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,16,power_law_1.2,0.02535040080547333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,16,power_law_1.2,0.035078400373458864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,16,power_law_1.2,0.035283198952674864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,16,power_law_1.2,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,16,power_law_1.2,0.03795199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,16,power_law_1.2,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,16,power_law_1.2,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,64,power_law_1.01,0.03278720080852508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,16,power_law_1.2,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,16,power_law_1.2,0.03470079898834229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,16,power_law_1.2,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,16,power_law_1.2,0.03917439877986908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,16,power_law_1.2,0.04219520092010498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,16,power_law_1.2,0.049593600630760196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,16,power_law_1.2,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,16,power_law_1.2,0.06718720197677612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,16,power_law_1.2,0.0815168023109436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,64,power_law_1.01,0.0324864000082016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,64,power_law_1.01,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,64,power_law_1.01,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,64,power_law_1.01,0.033318400382995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,64,power_law_1.01,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,64,power_law_1.01,0.03492479920387268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,64,power_law_1.01,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,64,power_law_1.01,0.038950398564338684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,64,power_law_1.01,0.044787201285362246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,64,power_law_1.01,0.04723840057849884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,64,power_law_1.01,0.04490880072116852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,64,power_law_1.01,0.05333120226860046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,64,power_law_1.01,0.05952640175819397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,64,power_law_1.01,0.06975359916687011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,16,power_law_1.2,0.11218559741973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,64,power_law_1.01,0.07361279726028443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,64,power_law_1.01,0.09423360228538513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,64,power_law_1.01,0.11106560230255128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,64,power_law_1.01,0.13150080442428588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,64,power_law_1.01,0.16568959951400758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,64,power_law_1.01,0.24077439308166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,64,power_law_1.01,0.31027200222015383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,16,balanced,0.044154668847719826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,16,balanced,0.042266666889190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,16,balanced,0.044266665975252785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,16,balanced,0.048351998130480446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,16,balanced,0.0524586687485377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,16,balanced,0.06685866912206014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,16,balanced,0.06765866776307423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,16,balanced,0.06718400120735168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,16,balanced,0.06835199892520905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,16,balanced,0.06885866820812225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,16,balanced,0.07045333087444305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,16,balanced,0.07157333195209503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,16,balanced,0.07473599910736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,16,balanced,0.07587733368078868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,16,balanced,0.08125866452852885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,16,balanced,0.0832586685816447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,16,balanced,0.08654399712880452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,16,balanced,0.1011840005715688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,16,balanced,0.10346666971842448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,16,balanced,0.12805333733558655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,32,power_law_1.2,0.2502336025238037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,16,balanced,0.14123200376828512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,16,balanced,0.18728532393773398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,16,balanced,0.2022506594657898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,16,power_law_1.2,0.1376255989074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,16,balanced,0.28752533594767254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,16,balanced,0.34786665439605713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,16,balanced,0.47434135278066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,16,balanced,0.605189323425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,16,power_law_1.2,0.19441920518875122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,16,power_law_1.2,0.2562688112258911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,16,balanced,0.019589333484570186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,16,balanced,0.020010666300853092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,16,balanced,0.01945066700379054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,16,balanced,0.019007999449968338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,16,balanced,0.021168000996112823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,128,power_law_1.01,0.10364160537719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,16,balanced,0.025205334027608235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,128,power_law_1.01,0.10467840433120727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,128,power_law_1.01,0.10261119604110717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,16,balanced,0.027664000789324444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,128,power_law_1.01,0.1068992018699646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,128,power_law_1.01,0.10511360168457032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,16,balanced,0.0271573339899381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,128,power_law_1.01,0.10771839618682862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,16,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,16,balanced,0.03311999887228012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,16,balanced,0.03136533250411352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,16,balanced,0.044938668608665466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,16,balanced,0.044341335693995156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,16,balanced,0.04376000165939331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,16,balanced,0.04933866858482361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,16,balanced,0.04886933167775472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,16,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,16,balanced,0.05067733426888784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,128,power_law_1.01,0.11252479553222657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,128,power_law_1.01,0.11861120462417603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,16,balanced,0.05694933235645294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,16,balanced,0.06756266454855601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,16,balanced,0.07539199789365132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,16,balanced,0.09416000048319499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,16,balanced,0.1086133321126302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,16,balanced,0.14009066422780356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,16,balanced,0.17086933056513467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,16,balanced,0.2507200042406718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,16,balanced,0.31385600566864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,128,power_law_1.01,0.11512320041656494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,128,power_law_1.01,0.11712000370025635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,128,power_law_1.01,0.11381759643554687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,128,power_law_1.01,0.11310720443725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,128,power_law_1.01,0.1171839952468872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,128,power_law_1.01,0.11562880277633666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,128,power_law_1.01,0.1254271984100342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,128,power_law_1.01,0.12363519668579101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,128,power_law_1.01,0.12007039785385132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,128,power_law_1.01,0.12859519720077514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,128,power_law_1.01,0.13719040155410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,128,power_law_1.01,0.18165119886398315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,128,power_law_1.01,0.20042879581451417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,128,power_law_1.01,0.24785919189453126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,128,power_law_1.01,0.301363205909729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,128,power_law_1.01,0.405075216293335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,128,power_law_1.01,0.49465599060058596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,128,power_law_1.01,0.7347263813018798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,128,power_law_1.01,0.9030655860900879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,1,power_law_1.2,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,1,power_law_1.2,0.03587839901447296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,1,power_law_1.2,0.0394239991903305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,1,power_law_1.2,0.047167998552322385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,1,power_law_1.2,0.05605120062828064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,1,power_law_1.2,0.07779200077056884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,1,power_law_1.2,0.08153600096702576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,1,power_law_1.2,0.08457599878311158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,1,power_law_1.2,0.08599680066108703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,1,power_law_1.2,0.0876800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,1,power_law_1.2,0.09160320162773132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,1,power_law_1.2,0.09541119933128357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,1,power_law_1.2,0.09882879853248597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,1,power_law_1.2,0.10712319612503052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,1,power_law_1.2,0.11553280353546143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,1,power_law_1.2,0.12136319875717164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,1,power_law_1.2,0.12899839878082275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,1,power_law_1.2,0.1516543984413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,1,power_law_1.2,0.17649279832839965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,1,power_law_1.2,0.22072958946228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,1,power_law_1.2,0.2910527944564819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,1,power_law_1.2,0.3748159885406494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,1,power_law_1.2,0.45783681869506837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,1,power_law_1.2,0.6223807811737061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,1,power_law_1.2,0.7982336044311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,1,power_law_1.2,1.161023998260498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,8,1,power_law_1.01,0.05660799741744995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,8,1,power_law_1.01,0.07455999851226806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,1,power_law_1.2,1.498380756378174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,8,1,power_law_1.01,0.09765120148658753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,8,1,power_law_1.01,0.15445120334625245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,8,1,power_law_1.01,0.20931200981140136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,8,1,power_law_1.01,0.259552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,8,1,power_law_1.01,0.3412031888961792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,32,power_law_1.01,0.05923200249671936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,8,1,power_law_1.01,0.366214394569397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,32,power_law_1.01,0.05745919942855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,8,1,power_law_1.01,0.37581439018249513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,32,power_law_1.01,0.057811200618743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,8,1,power_law_1.01,0.38551039695739747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,32,power_law_1.01,0.06110079884529114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,8,1,power_law_1.01,0.4020991802215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,32,power_law_1.01,0.05702400207519531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,8,1,power_law_1.01,0.4142591953277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,32,power_law_1.01,0.053439998626708986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,8,1,power_law_1.01,0.4297920227050781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,32,power_law_1.01,0.06220800280570984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,8,1,power_law_1.01,0.4453887939453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,32,power_law_1.01,0.05717759728431702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,8,1,power_law_1.01,0.46346240043640136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,32,power_law_1.01,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,8,1,power_law_1.01,0.46526079177856444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,32,power_law_1.01,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,8,1,power_law_1.01,0.5004799842834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,32,power_law_1.01,0.06098560094833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,32,power_law_1.01,0.06144639849662781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,32,power_law_1.01,0.06751999855041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,32,power_law_1.01,0.06872959733009339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,32,power_law_1.01,0.07411199808120728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,32,power_law_1.01,0.07715200185775757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,32,power_law_1.01,0.0801472008228302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,32,power_law_1.01,0.09127680063247681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,8,1,power_law_1.01,0.5606912136077881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,8,1,power_law_1.01,0.6042304039001465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,8,1,power_law_1.01,0.7091904163360596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,8,power_law_1.2,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,8,1,power_law_1.01,0.806719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,8,power_law_1.2,0.035955199599266054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,8,power_law_1.2,0.0353408008813858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,8,power_law_1.2,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,32,power_law_1.01,0.09683840274810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,8,power_law_1.2,0.03654400110244751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,32,power_law_1.01,0.12009600400924683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,32,power_law_1.01,0.1283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,32,power_law_1.01,0.15535999536514283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,32,power_law_1.01,0.19104000329971313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,32,power_law_1.01,0.25055999755859376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,8,1,power_law_1.01,1.016652774810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,32,power_law_1.01,0.3247999906539917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,8,power_law_1.2,0.03749760091304779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,32,power_law_1.01,0.4439295768737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,8,power_law_1.2,0.037503999471664426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,8,1,power_law_1.01,1.162502384185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,8,power_law_1.2,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,8,power_law_1.2,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,8,power_law_1.2,0.0394239991903305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,8,1,power_law_1.01,1.517484760284424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,8,power_law_1.2,0.0398719996213913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,16,power_law_1.2,0.021766400337219237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,8,power_law_1.2,0.0443583995103836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,16,power_law_1.2,0.025337600708007814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,8,1,power_law_1.01,1.8609472274780274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,16,power_law_1.2,0.02542079985141754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,16,power_law_1.2,0.025472000241279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,32,power_law_1.01,0.5765376091003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,16,power_law_1.2,0.02540160119533539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,8,1,power_law_1.01,2.5531583786010743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,16,power_law_1.2,0.029023998975753786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,16,power_law_1.2,0.03227519989013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,16,power_law_1.2,0.03221119940280914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,16,power_law_1.2,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,16,power_law_1.2,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,16,power_law_1.2,0.04529919922351837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,16,power_law_1.2,0.06033279895782471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,16,power_law_1.2,0.06061440110206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,16,power_law_1.2,0.062003201246261595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,8,power_law_1.2,0.04356479942798615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,16,power_law_1.2,0.07181439995765686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,8,1,power_law_1.01,3.2275070190429687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,16,power_law_1.2,0.07246720194816589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,16,power_law_1.2,0.07405440211296081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,16,power_law_1.2,0.0737600028514862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,16,power_law_1.2,0.08563200235366822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,16,power_law_1.2,0.09325439929962158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,16,power_law_1.2,0.09663360118865967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,16,power_law_1.2,0.11974400281906128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,8,power_law_1.2,0.04561919867992401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,16,power_law_1.2,0.14582400321960448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,8,power_law_1.2,0.05356159806251526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,16,power_law_1.2,0.1909183979034424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,16,power_law_1.2,0.2410111904144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,8,power_law_1.2,0.055936002731323244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,16,power_law_1.2,0.3361471891403198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,16,power_law_1.2,0.4547327995300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,8,power_law_1.2,0.057817602157592775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,8,power_law_1.2,0.06601600050926208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.2,0.07217280268669128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.2,0.08913919925689698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.2,0.11014399528503419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.2,0.14451839923858642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.2,0.16858880519866942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.2,0.2403968095779419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.2,0.3042304039001465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,2,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,2,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,2,balanced,0.05444266895453135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,2,balanced,0.0800906668106715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,2,balanced,0.12730133533477783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,2,balanced,0.1274079978466034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,2,balanced,0.1281013290087382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,2,balanced,0.13145066301027933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,2,balanced,0.13292266925175986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,2,balanced,0.13411200046539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,2,balanced,0.13449600338935852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,2,balanced,0.1366933286190033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,2,balanced,0.1416906714439392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,2,balanced,0.14388799667358398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,2,balanced,0.1527466674645742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,2,balanced,0.15803733468055725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,2,balanced,0.16328533490498862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,2,balanced,0.17859200636545816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,2,balanced,0.19567465782165527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,2,balanced,0.23759466409683228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,2,balanced,0.26734934250513714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,2,balanced,0.34593598047892254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,2,balanced,0.4102666775385539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,2,balanced,0.5833226839701334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,2,balanced,0.7268266677856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,2,balanced,1.0421706835428874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,2,balanced,1.3575894037882488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,128,balanced,0.038191998998324074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,128,balanced,0.03629333277543386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,128,balanced,0.03793599953254064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,128,balanced,0.03823466598987579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,128,balanced,0.04001066585381826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,128,balanced,0.039962666730086006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,128,balanced,0.03955733279387156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,128,balanced,0.04146133363246918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,128,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,128,balanced,0.041696002086003624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,128,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,128,balanced,0.04186666508515676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,128,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,128,balanced,0.043749332427978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,128,balanced,0.04572799801826477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,128,balanced,0.046442667643229164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,128,balanced,0.04776533444722494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,128,balanced,0.04980800052483877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,128,balanced,0.05077333251635233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,128,balanced,0.056261335810025535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,128,balanced,0.058320000767707825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,128,balanced,0.06730133295059204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,128,balanced,0.07541866600513458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.2,0.4358719825744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,128,balanced,0.09955199559529622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,128,balanced,0.11754666765530904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,128,balanced,0.15217066804567972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,128,balanced,0.18895999590555826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,4,power_law_1.01,0.1347584009170532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,4,power_law_1.01,0.1910591959953308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,4,power_law_1.01,0.23512959480285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,4,power_law_1.01,0.30106239318847655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,4,power_law_1.01,0.307859206199646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,4,power_law_1.01,0.373471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,4,power_law_1.01,0.4471295833587646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,4,power_law_1.01,0.4230720043182373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,4,power_law_1.01,0.4277632236480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,4,power_law_1.01,0.4410367965698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,4,power_law_1.01,0.4376959800720215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,4,power_law_1.01,0.4240384101867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,4,power_law_1.01,0.4393919944763184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,4,power_law_1.01,0.4616703987121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,4,power_law_1.01,0.4640063762664795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,4,power_law_1.01,0.44780797958374025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,4,power_law_1.01,0.47614078521728515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,4,power_law_1.01,0.49005441665649413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,4,power_law_1.01,0.5060416221618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,4,power_law_1.01,0.5849728107452392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,4,power_law_1.01,0.6396224021911621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,4,power_law_1.01,0.7658688068389893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,4,power_law_1.01,0.9041088104248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,4,power_law_1.01,1.1526335716247558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,4,power_law_1.01,1.4164863586425782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,4,power_law_1.01,1.8996736526489257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,4,power_law_1.01,2.320057678222656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.2,0.5565631866455079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,16,power_law_1.01,0.15688960552215575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,16,power_law_1.01,0.16186239719390869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,16,power_law_1.2,0.0810368001461029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,16,power_law_1.01,0.1528447985649109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,16,power_law_1.2,0.09411839842796325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,16,power_law_1.01,0.14608000516891478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,16,power_law_1.2,0.09843199849128723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,16,power_law_1.2,0.10332800149917602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,16,power_law_1.2,0.10661120414733886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,16,power_law_1.2,0.10620160102844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,16,power_law_1.2,0.09707520008087159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,16,power_law_1.01,0.15162240266799926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,16,power_law_1.01,0.14099839925765992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,16,power_law_1.01,0.15263999700546266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,16,power_law_1.01,0.14264320135116576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,16,power_law_1.01,0.1526528000831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,16,power_law_1.01,0.13895679712295533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,16,power_law_1.01,0.15079679489135742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,16,power_law_1.01,0.15011839866638182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,16,power_law_1.01,0.1492735981941223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,16,power_law_1.2,0.10248960256576538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,16,power_law_1.01,0.1522495985031128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,16,power_law_1.2,0.10479999780654907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,16,power_law_1.01,0.15594880580902098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,16,power_law_1.2,0.10405119657516479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,16,power_law_1.01,0.15181440114974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,16,power_law_1.2,0.10629760026931763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,16,power_law_1.01,0.15902719497680665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,16,power_law_1.2,0.1075711965560913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,16,power_law_1.01,0.18032000064849854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,16,power_law_1.2,0.10970239639282227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.01,0.19919359683990479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,16,power_law_1.2,0.1164736032485962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.01,0.2308351993560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,16,power_law_1.2,0.12182400226593018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.01,0.281497597694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,16,power_law_1.2,0.12484480142593384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.01,0.32906880378723147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,16,power_law_1.2,0.13080960512161255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.01,0.396230411529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,16,power_law_1.2,0.15285120010375977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.01,0.5743103981018066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,16,power_law_1.2,0.1659327983856201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,16,power_law_1.2,0.20860159397125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.01,0.6650879859924317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,16,power_law_1.2,0.24710400104522706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.01,0.9475968360900879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,16,power_law_1.2,0.30769920349121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.01,1.1836544036865235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,16,power_law_1.2,0.3368704080581665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,128,power_law_1.01,0.04408960044384003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,128,power_law_1.01,0.043673598766326906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,16,power_law_1.2,0.4272448062896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,128,power_law_1.01,0.04332799911499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,16,power_law_1.2,0.5833471775054931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,128,power_law_1.01,0.04628480076789856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,128,power_law_1.01,0.04898560047149658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,128,power_law_1.01,0.0507968008518219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,16,power_law_1.2,0.8288064002990723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,16,power_law_1.2,1.0139583587646483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,128,power_law_1.01,0.05121920108795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,128,power_law_1.01,0.05308160185813904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,16,balanced,0.06479466458161671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,16,balanced,0.06438933312892914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,16,balanced,0.06859200199445088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,16,balanced,0.0824480007092158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,16,balanced,0.09404800335566203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,16,balanced,0.1367093324661255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,16,balanced,0.1565546691417694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,16,balanced,0.15665066242218018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,16,balanced,0.15901333093643188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,16,balanced,0.15690666437149048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,16,balanced,0.15849600235621134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,16,balanced,0.16190399726231894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,16,balanced,0.1625920037428538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,16,balanced,0.16475733121236166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,16,balanced,0.17467200756072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,16,balanced,0.1743519902229309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,128,power_law_1.01,0.0532800018787384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,128,power_law_1.01,0.05288959741592407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,16,balanced,0.18115200599034628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,16,balanced,0.1927573283513387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,16,balanced,0.2034506599108378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,16,balanced,0.21841599543889365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,128,power_law_1.01,0.055980801582336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,16,balanced,0.238154669602712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,16,balanced,0.27727999289830524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,16,balanced,0.315285325050354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,16,balanced,0.39616533120473224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,16,balanced,0.5250399907430013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,16,balanced,0.6932426293691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,16,balanced,0.8947306474049886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,128,power_law_1.01,0.05961599946022034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,128,power_law_1.01,0.06101120114326477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,128,power_law_1.01,0.05922560095787048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,128,power_law_1.01,0.07152640223503112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,128,power_law_1.01,0.07306240200996399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,128,power_law_1.01,0.07628160119056701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,128,power_law_1.01,0.08734080195426941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,128,power_law_1.01,0.09585279822349549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,128,power_law_1.01,0.11946879625320435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,128,power_law_1.01,0.13864959478378297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,128,power_law_1.01,0.17935999631881713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,128,power_law_1.01,0.21743359565734863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,128,power_law_1.01,0.30252799987792967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,128,power_law_1.01,0.39511680603027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,128,power_law_1.01,0.5668608188629151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,128,power_law_1.01,0.7929088115692139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,2,4,balanced,0.05229333539803823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,2,4,balanced,0.06709866722424825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,2,4,balanced,0.0865760048230489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,2,4,balanced,0.13174933195114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,2,4,balanced,0.20150399208068848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,2,4,balanced,0.26265599330266315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,2,4,balanced,0.2614026665687561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,2,4,balanced,0.2634613315264384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,2,4,balanced,0.2654719948768616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,2,4,balanced,0.26716800530751544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,2,4,balanced,0.2680373390515645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,2,4,balanced,0.2722666660944621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,2,4,balanced,0.27498666445414227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,2,4,balanced,0.2770666678746541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,2,4,balanced,0.28573334217071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,2,4,balanced,0.28993066151936847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,2,4,balanced,0.2975306709607442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,2,4,balanced,0.3232426643371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,2,4,balanced,0.3385813236236572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,2,4,balanced,0.4436426560084025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,2,4,balanced,0.4286773204803467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,2,4,balanced,0.6225226720174154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,2,4,balanced,0.589626669883728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,2,4,balanced,0.8661973476409912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,2,4,balanced,0.9591840108235677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,2,4,balanced,1.499381383260091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,2,4,balanced,1.6642026901245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,16,power_law_1.01,0.028940799832344054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,16,power_law_1.01,0.029190400242805482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,16,power_law_1.01,0.029228800535202028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,16,power_law_1.01,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,16,power_law_1.01,0.02757120132446289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,16,power_law_1.01,0.034176000952720643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,16,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,16,power_law_1.01,0.03402239978313446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,16,power_law_1.01,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,16,power_law_1.01,0.03697920143604279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,16,power_law_1.01,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,16,power_law_1.01,0.04031359851360321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,16,power_law_1.01,0.04439040124416351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,16,power_law_1.01,0.043315199017524716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,16,power_law_1.01,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,16,power_law_1.01,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,16,power_law_1.01,0.04960640072822571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,16,power_law_1.01,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,16,power_law_1.01,0.06170240044593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,16,power_law_1.01,0.0689791977405548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,16,power_law_1.01,0.07326080203056336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,16,power_law_1.01,0.09496960043907166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,16,power_law_1.01,0.0992255985736847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,16,power_law_1.01,0.13533439636230468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,16,power_law_1.01,0.15432319641113282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,16,power_law_1.01,0.2187903881072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,16,power_law_1.01,0.302239990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,16,power_law_1.01,0.025542399287223815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,16,power_law_1.01,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,16,power_law_1.01,0.04111360013484955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,16,power_law_1.01,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,16,power_law_1.01,0.03866879940032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,16,power_law_1.01,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,16,power_law_1.01,0.035155200958251955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,16,power_law_1.01,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,16,power_law_1.01,0.037920001149177554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,16,power_law_1.01,0.0411327987909317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,16,power_law_1.01,0.038848000764846805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,16,power_law_1.01,0.041555199027061465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,16,power_law_1.01,0.04378879964351654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,16,power_law_1.01,0.05148800015449524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,16,power_law_1.01,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,16,power_law_1.01,0.05496960282325745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,16,power_law_1.01,0.07610880136489868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,16,power_law_1.01,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,16,power_law_1.01,0.09975680112838745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,16,power_law_1.01,0.12432639598846436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,16,power_law_1.01,0.08661119937896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,16,power_law_1.01,0.10679680109024048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,16,power_law_1.01,0.11890560388565063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,16,power_law_1.01,0.16355839967727662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,16,power_law_1.01,0.1948799967765808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,16,power_law_1.01,0.28469760417938234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,16,power_law_1.01,0.3637824058532715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,16,power_law_1.2,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,16,power_law_1.2,0.03281919956207276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,16,power_law_1.2,0.03128960132598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,16,power_law_1.2,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,16,power_law_1.2,0.031814399361610415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,16,power_law_1.2,0.03409920036792755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,16,power_law_1.2,0.033529600501060484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,16,power_law_1.2,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,16,power_law_1.2,0.03317759931087494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,16,power_law_1.2,0.03296639919281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,16,power_law_1.2,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,16,power_law_1.2,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,16,power_law_1.2,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,16,power_law_1.2,0.03628160059452057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,16,power_law_1.2,0.03953920006752014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,16,power_law_1.2,0.04178560078144074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,16,power_law_1.2,0.04513919949531555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,16,power_law_1.2,0.049593600630760196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,16,power_law_1.2,0.051283198595047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,16,power_law_1.2,0.060844802856445314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,16,power_law_1.2,0.06941440105438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,16,power_law_1.2,0.08144639730453491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,16,power_law_1.2,0.0912447988986969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,16,power_law_1.2,0.15663360357284545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,16,power_law_1.2,0.12704639434814452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,16,power_law_1.2,0.19527039527893067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,16,power_law_1.2,0.2628351926803589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,1,power_law_1.01,0.09516159892082214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,64,power_law_1.01,0.07324159741401673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,64,power_law_1.01,0.07110400199890136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,64,power_law_1.01,0.06963840126991272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,64,power_law_1.01,0.07185919880867005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,64,power_law_1.01,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,64,power_law_1.01,0.0710207998752594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,64,power_law_1.01,0.06885759830474854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,1,power_law_1.01,0.1366528034210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,1,power_law_1.01,0.21785600185394288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,64,power_law_1.01,0.07294719815254211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,64,power_law_1.01,0.06740480065345764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,64,power_law_1.01,0.07097600102424621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,64,power_law_1.01,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,64,power_law_1.01,0.07192959785461425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,32,balanced,0.05387733379999796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,64,power_law_1.01,0.07222399711608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,32,balanced,0.05327466626962026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,32,balanced,0.053685332338015236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,32,balanced,0.05593066910902659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,32,balanced,0.058320000767707825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,1,power_law_1.01,0.39114880561828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,32,balanced,0.0749066670735677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,32,balanced,0.08401599526405334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,32,balanced,0.08303999900817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,1,power_law_1.01,0.5723392009735108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,32,balanced,0.08507733543713887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,32,balanced,0.08340799808502197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,32,balanced,0.08576533198356628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,1,power_law_1.01,0.7416831970214843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,32,balanced,0.08950933814048767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,64,power_law_1.01,0.07287039756774902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,32,balanced,0.08942400415738423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,64,power_law_1.01,0.07925760149955749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,1,power_law_1.01,1.035244846343994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,64,power_law_1.01,0.07900800108909607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,64,power_law_1.01,0.08488320112228394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,1,power_law_1.01,1.0745344161987305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,64,power_law_1.01,0.0991424024105072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.01,0.10743039846420288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,1,power_law_1.01,1.123468780517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.01,0.1250175952911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,32,balanced,0.09710933764775594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,1,power_law_1.01,1.1406720161437989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,1,power_law_1.01,1.1886207580566406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,1,power_law_1.01,1.2239680290222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,1,power_law_1.01,1.2380991935729981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.01,0.14098559617996215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.01,0.17469439506530762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,1,power_law_1.01,1.2598208427429198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.01,0.21139841079711913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.01,0.29585280418396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,1,power_law_1.01,1.3130880355834962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,32,balanced,0.10758399963378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,1,power_law_1.01,1.315328025817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,1,power_law_1.01,1.404627227783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,1,power_law_1.01,1.535212802886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.01,0.3732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,32,balanced,0.10770133137702942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.01,0.521292781829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,1,power_law_1.01,1.5525504112243653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.01,0.6496448040008544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,1,power_law_1.01,1.7631296157836913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,1,power_law_1.01,1.8664896011352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,1,power_law_1.01,2.153945541381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,32,balanced,0.10724799831708272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,32,balanced,0.12157332897186279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,32,balanced,0.12916266918182373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,32,balanced,0.1461066703001658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,32,balanced,0.16428266962369284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,1,power_law_1.01,2.316409683227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,32,balanced,0.20186134179433188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,32,balanced,0.23453867435455322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,32,balanced,0.29976532856623334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,32,balanced,0.3722720146179199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,1,power_law_1.01,2.938297653198242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,32,balanced,0.5122773249944051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,32,balanced,0.6745013395945231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,1,power_law_1.01,3.4762241363525392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,1,power_law_1.01,4.497273635864258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,1,power_law_1.01,5.590323257446289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,1,power_law_1.01,0.061324799060821535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,1,power_law_1.01,0.06266239881515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,1,power_law_1.01,0.07269759774208069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,1,power_law_1.01,0.09573119878768921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,1,power_law_1.01,0.13446400165557862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,1,power_law_1.01,0.1692415952682495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,1,power_law_1.01,0.18773119449615477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,1,power_law_1.01,0.22403841018676757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,1,power_law_1.01,0.23286399841308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,1,power_law_1.01,0.23927040100097657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,1,power_law_1.01,0.24706559181213378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,1,power_law_1.01,0.2550719976425171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,1,power_law_1.01,0.26460800170898435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,1,power_law_1.01,0.2785599946975708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,1,power_law_1.01,0.2884160041809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,4,balanced,0.06986133257548015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,4,balanced,0.07095466554164886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,1,power_law_1.01,0.29831039905548096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,4,balanced,0.07482133309046428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,4,balanced,0.09080533186594646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,1,power_law_1.01,0.314630389213562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,4,balanced,0.10710400342941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,1,power_law_1.01,0.3689471960067749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,1,power_law_1.01,0.4107776165008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,1,power_law_1.01,0.5043712139129639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,1,power_law_1.01,0.6102335929870606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,1,power_law_1.01,0.7852543830871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,4,balanced,0.14979733030001322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,4,balanced,0.1491146683692932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,1,power_law_1.01,0.9521599769592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,1,power_law_1.01,1.2727680206298828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,1,power_law_1.01,1.594649600982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,4,balanced,0.1483840048313141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,4,balanced,0.15175466736157736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,4,balanced,0.14941333731015524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,4,balanced,0.15241600076357523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,1,power_law_1.01,2.2851327896118163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,4,balanced,0.1564959983030955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,4,balanced,0.15626666943232217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,4,balanced,0.15870933731396994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,4,balanced,0.16192533572514853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,4,balanced,0.1623146633307139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,1,power_law_1.01,2.9718080520629884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,4,balanced,0.16832000017166138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,4,balanced,0.1772693395614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,4,balanced,0.1848693291346232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,4,balanced,0.20661866664886475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,4,balanced,0.22512000799179077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,4,balanced,0.265882670879364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,4,balanced,0.3075573245684306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,4,balanced,0.4315626621246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,4,balanced,0.5212159951527914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,4,balanced,0.735088030497233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,4,balanced,0.9607413609822592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,1,power_law_1.01,0.02627840042114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,1,power_law_1.01,0.04410240054130554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,1,power_law_1.01,0.06910079717636108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,1,power_law_1.01,0.10548479557037353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,1,power_law_1.01,0.15200639963150026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,1,power_law_1.01,0.2092992067337036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,1,power_law_1.01,0.27058560848236085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,1,power_law_1.01,0.2836287975311279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,1,power_law_1.01,0.30042879581451415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,1,power_law_1.01,0.2967168092727661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,1,power_law_1.01,0.31687040328979493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,1,power_law_1.01,0.3293312072753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,1,power_law_1.01,0.3276544094085693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,1,power_law_1.01,0.3388864040374756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,1,power_law_1.01,0.34528000354766847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,1,power_law_1.01,0.3473855972290039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,1,power_law_1.01,0.3721343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,1,power_law_1.01,0.41457281112670896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,1,power_law_1.01,0.4284224033355713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,1,power_law_1.01,0.4911488056182861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,1,power_law_1.01,0.5838272094726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,1,power_law_1.01,0.740934419631958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,1,power_law_1.01,0.7071872234344483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,1,power_law_1.01,0.9189951896667481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,1,power_law_1.01,1.125068759918213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,1,power_law_1.01,1.5421119689941407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,1,power_law_1.01,1.9528959274291993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,4,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,4,balanced,0.034117333590984344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,4,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,4,balanced,0.04182933270931244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,4,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,4,balanced,0.04397333165009817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,4,balanced,0.04391466577847799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,4,balanced,0.04576000074545542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,4,balanced,0.04422399898370107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,4,balanced,0.04478399952252706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,4,balanced,0.0443200021982193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,4,balanced,0.0524533341328303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,4,balanced,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,4,balanced,0.05403199791908264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,4,balanced,0.06225599845250448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,4,balanced,0.0643093337615331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,4,balanced,0.06607466439406078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,4,balanced,0.07460799813270569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,4,balanced,0.0786293347676595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,4,balanced,0.10073600212732951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,4,balanced,0.11762133240699768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,4,balanced,0.14945066968599954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,4,balanced,0.19040000438690186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,4,balanced,0.2664213379224141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,4,balanced,0.33637332916259766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,4,balanced,0.4861866633097331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,4,balanced,0.6386293172836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,4,power_law_1.01,0.0635968029499054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,4,power_law_1.01,0.06943359971046448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,4,power_law_1.01,0.07228800058364868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,4,power_law_1.01,0.08289920091629029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,4,power_law_1.01,0.08588160276412964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,4,power_law_1.01,0.08991360068321227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,4,power_law_1.01,0.09955840110778809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,4,power_law_1.01,0.09662079811096191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,4,power_law_1.01,0.1033471941947937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,4,power_law_1.01,0.09994239807128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,4,power_law_1.01,0.10412160158157349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,4,power_law_1.01,0.10250240564346313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,4,power_law_1.01,0.10717439651489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,4,power_law_1.01,0.11042560338973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,4,power_law_1.01,0.11639039516448975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,4,power_law_1.01,0.11769599914550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,4,power_law_1.01,0.12439680099487305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,4,power_law_1.01,0.13587839603424073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,4,power_law_1.01,0.14805760383605956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,4,power_law_1.01,0.17614079713821412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,4,power_law_1.01,0.19740159511566163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,4,power_law_1.01,0.24657919406890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,4,power_law_1.01,0.3051647901535034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,4,power_law_1.01,0.40264320373535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,4,power_law_1.01,0.5444672107696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,4,power_law_1.01,0.7184256076812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,4,power_law_1.01,0.8933055877685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,2,power_law_1.2,0.020576000213623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,2,power_law_1.2,0.02220800071954727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,2,power_law_1.2,0.026745599508285523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,2,power_law_1.2,0.03727999925613403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,2,power_law_1.2,0.046700799465179445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,2,power_law_1.2,0.05568000078201294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,2,power_law_1.2,0.0648959994316101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,2,power_law_1.2,0.07077119946479797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,2,power_law_1.2,0.07256320118904114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,2,power_law_1.2,0.07480319738388061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,2,power_law_1.2,0.07515519857406616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,2,power_law_1.2,0.07747840285301208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,2,power_law_1.2,0.08220160007476807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,2,power_law_1.2,0.0810368001461029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,2,power_law_1.2,0.08598399758338929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,2,power_law_1.2,0.08961920142173767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,2,power_law_1.2,0.09464319944381713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,2,power_law_1.2,0.11380480527877808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,2,power_law_1.2,0.1279744029045105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,2,power_law_1.2,0.15905920267105103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,2,power_law_1.2,0.1708799958229065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,2,power_law_1.2,0.21731200218200683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,2,power_law_1.2,0.2545792102813721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,2,power_law_1.2,0.3252351999282837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,2,power_law_1.2,0.39680640697479247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,2,power_law_1.2,0.5111936092376709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,2,power_law_1.2,0.6371456146240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,8,balanced,0.03639466563860575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,8,balanced,0.03571200122435888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,8,balanced,0.033557333052158356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,8,balanced,0.035802667339642845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,8,balanced,0.037946666280428566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,8,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,8,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,8,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,8,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,8,balanced,0.037989333271980286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,8,balanced,0.04170133173465729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,8,balanced,0.039818666875362396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,8,balanced,0.04182399809360504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,8,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,8,balanced,0.05619733532269796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,8,balanced,0.0580213318268458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,8,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,8,balanced,0.07454933226108551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,8,balanced,0.07473599910736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,8,balanced,0.09496000409126282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,8,balanced,0.10788800319035848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,8,balanced,0.1362613340218862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,8,balanced,0.1641439994176229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,8,balanced,0.2241386572519938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,8,balanced,0.27529066801071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,8,balanced,0.38230399290720624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,8,balanced,0.4870293140411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,2,power_law_1.2,0.07858560085296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,2,power_law_1.2,0.08627840280532836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,2,power_law_1.2,0.10337280035018921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,2,power_law_1.2,0.11427839994430541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,2,power_law_1.2,0.12309119701385499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,2,power_law_1.2,0.13972480297088624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,2,power_law_1.2,0.15356800556182862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,2,power_law_1.2,0.1603968024253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,2,power_law_1.2,0.1617535948753357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,2,power_law_1.2,0.1641088008880615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,2,power_law_1.2,0.16724480390548707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,2,power_law_1.2,0.17339520454406737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,2,power_law_1.2,0.1749119997024536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,2,power_law_1.2,0.1790336012840271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,2,power_law_1.2,0.18055039644241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,2,power_law_1.2,0.18344320058822633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,2,power_law_1.2,0.1966591954231262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,2,power_law_1.2,0.2133375883102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,2,power_law_1.2,0.23082880973815917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,2,power_law_1.2,0.278873610496521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,2,power_law_1.2,0.3150784015655518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,2,power_law_1.2,0.40687999725341795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,2,power_law_1.2,0.49735679626464846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,2,power_law_1.2,0.6740863800048829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,2,power_law_1.2,0.92042236328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,2,power_law_1.2,1.2180480003356933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,2,power_law_1.2,1.683091163635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,2,power_law_1.01,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,2,power_law_1.01,0.04970879852771759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,2,power_law_1.01,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,2,power_law_1.01,0.06640639901161194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,2,power_law_1.01,0.07653759717941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,2,power_law_1.01,0.09553279876708984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,2,power_law_1.01,0.09702399969100953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,2,power_law_1.01,0.10080640316009522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,2,power_law_1.01,0.10252799987792968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,2,power_law_1.01,0.10456960201263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,2,power_law_1.01,0.10698239803314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,1,2,balanced,0.08588266372680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,1,2,balanced,0.1379039982954661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,1,2,balanced,0.24335465828577676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,1,2,balanced,0.24480533599853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,1,2,balanced,0.24272000789642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,1,2,balanced,0.24276800950368246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,1,2,balanced,0.24202134211858115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,1,2,balanced,0.24449066321055093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,1,2,balanced,0.24513065814971924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,1,2,balanced,0.24794133504231772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,1,2,balanced,0.2511626680692037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,1,2,balanced,0.2532266577084859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,1,2,balanced,0.2518986662228902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,1,2,balanced,0.2685439984003703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,1,2,balanced,0.2762613296508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,1,2,balanced,0.27644266684850055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,1,2,balanced,0.2942453424135844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,1,2,balanced,0.34067201614379883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,1,2,balanced,0.36613333225250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,1,2,balanced,0.44974398612976074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,2,power_law_1.01,0.11085439920425415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,2,power_law_1.01,0.11028480529785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,2,power_law_1.01,0.11856640577316284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,2,power_law_1.01,0.1238144040107727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,2,power_law_1.01,0.13003519773483277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,2,power_law_1.01,0.14021120071411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,2,power_law_1.01,0.1647104024887085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,2,power_law_1.01,0.18995200395584105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,2,power_law_1.01,0.24177920818328857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,2,power_law_1.01,0.2975615978240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,2,power_law_1.01,0.3764480113983154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,2,power_law_1.01,0.47141118049621583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,2,power_law_1.01,0.6134335994720459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,2,power_law_1.01,0.8358016014099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,1,2,balanced,0.525493343671163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,1,2,balanced,0.7245333194732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,1,2,balanced,0.8862240314483643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,2,power_law_1.01,1.1222399711608886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,1,2,balanced,1.2826560338338215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,2,power_law_1.01,1.4529151916503906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,1,2,balanced,1.6873119672139485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,1,2,balanced,2.4254026412963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,1,2,balanced,3.261199951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,32,1,balanced,0.10275733470916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,32,1,balanced,0.11054399609565735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,32,1,balanced,0.1323306659857432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,32,1,balanced,0.17673067251841226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,32,1,balanced,0.2510613401730855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,32,1,balanced,0.4119786818822225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,32,1,balanced,0.41165868441263836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,32,1,balanced,0.40345601240793866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,32,1,balanced,0.43317333857218426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,32,1,balanced,0.40140267213185626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,32,1,balanced,0.41519999504089355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,32,1,balanced,0.4134933153788249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,32,1,balanced,0.4161386489868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,32,1,balanced,0.41247467199961346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,32,1,balanced,0.4301973183949788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,32,1,balanced,0.4376426537831624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,32,1,balanced,0.44680531819661456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,32,1,balanced,0.4934133291244507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,32,1,balanced,0.5279519955317179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,32,1,balanced,0.5990879933039347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,32,1,balanced,0.6660213470458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,32,1,balanced,0.8279573122660319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,32,1,balanced,0.9815359910329183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,32,1,balanced,1.4735093116760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,32,1,balanced,1.7878986994425456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,32,1,balanced,2.5946133931477866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,32,1,balanced,3.3752800623575845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,4,power_law_1.01,0.07899519801139832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,4,power_law_1.01,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,4,power_law_1.01,0.10846079587936401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,4,power_law_1.01,0.1189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,4,power_law_1.01,0.1288256049156189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,4,power_law_1.01,0.1410688042640686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,4,power_law_1.01,0.14263039827346802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,4,power_law_1.01,0.1579967975616455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,4,power_law_1.01,0.16446720361709594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,4,power_law_1.01,0.1625216007232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,4,power_law_1.01,0.16085760593414306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,4,power_law_1.01,0.16885759830474853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,4,power_law_1.01,0.16946560144424438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,4,power_law_1.01,0.1737663984298706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,4,power_law_1.01,0.17968000173568727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,4,power_law_1.01,0.18098560571670533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,4,power_law_1.01,0.19267840385437013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,4,power_law_1.01,0.20801920890808107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,4,power_law_1.01,0.22887680530548096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,4,power_law_1.01,0.27640318870544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,4,power_law_1.01,0.32724480628967284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,4,power_law_1.01,0.4261119842529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,4,power_law_1.01,0.5217152118682862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,4,power_law_1.01,0.7077311992645263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,4,power_law_1.01,0.8851455688476563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,4,power_law_1.01,1.300806427001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,4,power_law_1.01,1.6635967254638673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,4,power_law_1.2,0.0738752007484436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,4,power_law_1.2,0.08991360068321227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,4,power_law_1.2,0.08398720026016235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,4,power_law_1.2,0.09810559749603272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,4,power_law_1.2,0.09921280145645142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,4,power_law_1.2,0.09969279766082764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,4,power_law_1.2,0.09981439709663391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,4,power_law_1.2,0.10054399967193603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,4,power_law_1.2,0.1021183967590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,4,power_law_1.2,0.10189440250396728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,4,power_law_1.2,0.1034559965133667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,4,power_law_1.2,0.11779839992523193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,4,power_law_1.2,0.12098560333251954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,4,power_law_1.2,0.12466559410095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,4,power_law_1.2,0.1545215964317322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,4,power_law_1.2,0.15741440057754516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,4,power_law_1.2,0.1963711977005005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,4,power_law_1.2,0.23167359828948975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,4,power_law_1.2,0.26672000885009767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,4,power_law_1.2,0.405241584777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,4,power_law_1.2,0.4644608020782471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,4,power_law_1.2,0.7325503826141357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,4,power_law_1.2,0.8964735984802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,32,power_law_1.01,0.06454399824142457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,32,power_law_1.01,0.061926400661468504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,32,power_law_1.01,0.060684800148010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,32,power_law_1.01,0.06172800064086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,32,power_law_1.01,0.05863040089607239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,32,power_law_1.01,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,32,power_law_1.01,0.06227200031280518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,32,power_law_1.01,0.06162559986114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,32,power_law_1.01,0.06206079721450806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,32,power_law_1.01,0.06202239990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,32,power_law_1.01,0.06495360136032105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,32,power_law_1.01,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,32,power_law_1.01,0.06579840183258057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,32,power_law_1.01,0.06558079719543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,32,power_law_1.01,0.07056639790534973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,32,power_law_1.01,0.07111039757728577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,32,power_law_1.01,0.07355520129203796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,32,power_law_1.01,0.08284800052642823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,32,power_law_1.01,0.08517119884490967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,32,power_law_1.01,0.10098559856414795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,32,power_law_1.01,0.11427199840545654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,32,power_law_1.01,0.14266879558563234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,32,power_law_1.01,0.16891520023345946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,32,power_law_1.01,0.22030720710754395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,32,power_law_1.01,0.2803391933441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,32,power_law_1.01,0.40568318367004397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,4,power_law_1.2,1.2918335914611816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,32,power_law_1.01,0.5090176105499268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,32,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,32,balanced,0.033333333830038704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,32,balanced,0.03350933392842611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,32,balanced,0.03682133307059606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,32,balanced,0.03988266736268997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,32,balanced,0.0408693328499794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,32,balanced,0.040821333726247154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,32,balanced,0.04185600082079569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,32,balanced,0.04597333570321401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,32,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,32,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,4,balanced,0.03387733300526937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,4,power_law_1.2,1.6904256820678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,4,balanced,0.03162666658560435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,4,balanced,0.03203733265399933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,4,balanced,0.034128000338872276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,32,balanced,0.04772266745567322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,4,balanced,0.03339733431736628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,4,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,4,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,4,balanced,0.03375466664632162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,4,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,4,balanced,0.03552533437808355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,4,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,4,balanced,0.03482666611671448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,4,balanced,0.03331200033426285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,4,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,4,balanced,0.043493335445721946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,4,balanced,0.04161066561937332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,4,balanced,0.04366933306058248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,4,balanced,0.046298667788505554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,4,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,4,balanced,0.06070933242638906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,32,balanced,0.04701866706212362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,4,balanced,0.06243733565012614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,4,balanced,0.08498666683832805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,4,balanced,0.09576533238093059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,4,balanced,0.13025066256523132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,4,balanced,0.14666133125623068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,4,balanced,0.19615467389424643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,4,balanced,0.23811199267705283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,32,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,32,balanced,0.06025066475073496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,32,balanced,0.05983999868233999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,32,balanced,0.06025066475073496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,32,balanced,0.0668213317791621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,32,balanced,0.0670666644970576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,32,balanced,0.07486933469772339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,32,balanced,0.08054399987061818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,32,balanced,0.10633599758148193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,32,balanced,0.11690133810043335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,32,balanced,0.15202666322390238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,32,balanced,0.17889066537221274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,32,balanced,0.24498132864634195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,32,balanced,0.30930666128794354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,64,power_law_1.2,0.07582079768180847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,64,power_law_1.2,0.04695680141448975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,64,power_law_1.2,0.04505600035190582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,64,power_law_1.2,0.05013120174407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,64,power_law_1.2,0.049267199635505673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,64,power_law_1.2,0.05126399993896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,64,power_law_1.2,0.050419199466705325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,64,power_law_1.2,0.05119360089302063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,64,power_law_1.2,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,64,power_law_1.2,0.054816001653671266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,64,power_law_1.2,0.054067200422286986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,64,power_law_1.2,0.05735679864883423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,64,power_law_1.2,0.06356480121612548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,64,power_law_1.2,0.06913920044898987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,64,power_law_1.2,0.0763584017753601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,64,power_law_1.2,0.0797760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,64,power_law_1.2,0.08465920090675354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,64,power_law_1.2,0.10234880447387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,64,power_law_1.2,0.10865919589996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,64,power_law_1.2,0.13938560485839843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,64,power_law_1.2,0.16291199922561644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,64,power_law_1.2,0.224947190284729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,64,power_law_1.2,0.2700416088104248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,64,power_law_1.2,0.3945215940475464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,64,power_law_1.2,0.47712001800537107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,64,power_law_1.2,0.6843520164489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,64,power_law_1.2,0.8718015670776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,4,power_law_1.2,2.47644157409668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,4,balanced,0.04938666522502899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,4,balanced,0.05227733155091604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,4,balanced,0.05641599992911021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,4,balanced,0.07341333230336507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,4,balanced,0.10274133086204529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,4,balanced,0.10410666465759277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,4,balanced,0.10301333665847778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,4,balanced,0.1053600013256073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,4,balanced,0.10388799508412679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,4,balanced,0.10460266470909119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,4,balanced,0.10364266236623128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,4,balanced,0.10877333084742229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,4,balanced,0.10831999778747559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,4,balanced,0.11009066303571065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,4,balanced,0.11455466349919637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,4,balanced,0.11658666531244914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,4,balanced,0.11825066804885864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,4,balanced,0.1283253331979116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,4,balanced,0.1361120045185089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,4,balanced,0.15381866693496704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,4,balanced,0.16546133160591125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,4,balanced,0.23400533199310303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,4,balanced,0.2701866626739502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,4,balanced,0.3734613259633382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,4,balanced,0.4864586591720581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,4,balanced,0.6933226585388184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,4,balanced,0.9094879627227783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,32,power_law_1.01,0.07860479950904846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,32,power_law_1.01,0.08056319952011108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,32,power_law_1.01,0.07600640058517456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,32,power_law_1.01,0.07653120160102844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,32,power_law_1.01,0.07441920042037964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,32,power_law_1.01,0.07605760097503662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,32,power_law_1.01,0.0765504002571106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,32,power_law_1.01,0.07674239873886109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,32,power_law_1.01,0.07756800055503846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,32,power_law_1.01,0.07495679855346679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,32,power_law_1.01,0.07828480005264282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,32,power_law_1.01,0.07995520234107971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,32,power_law_1.01,0.07926399707794189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,32,power_law_1.01,0.08046079874038696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,32,power_law_1.01,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,32,power_law_1.01,0.08606079816818238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,32,power_law_1.01,0.09030399918556213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,32,power_law_1.01,0.1037824034690857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.01,0.11605759859085082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.01,0.132697594165802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.01,0.147705602645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.01,0.19841920137405394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.01,0.2300096035003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.01,0.312940788269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.01,0.39585280418395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.01,0.5688704013824463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.01,0.7697855949401855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,4,power_law_1.2,3.2088512420654296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,8,power_law_1.2,0.027564799785614012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,8,power_law_1.2,0.027564799785614012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,8,power_law_1.2,0.02789120078086853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,8,power_law_1.2,0.029657599329948426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,8,power_law_1.2,0.02863999903202057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,8,power_law_1.2,0.030329599976539612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,8,power_law_1.2,0.031513598561286923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,8,power_law_1.2,0.03080959916114807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,8,power_law_1.2,0.03123840093612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,8,power_law_1.2,0.0326335996389389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,8,power_law_1.2,0.032595199346542356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,8,power_law_1.2,0.03406080007553101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,8,power_law_1.2,0.03492479920387268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,8,power_law_1.2,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,8,power_law_1.2,0.04280959963798523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,8,power_law_1.2,0.04347519874572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,8,power_law_1.2,0.043968001008033754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,8,power_law_1.2,0.05196160078048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,8,power_law_1.2,0.05697280168533325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,8,power_law_1.2,0.06867200136184692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,8,power_law_1.2,0.0781440019607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,8,power_law_1.2,0.10291199684143067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,8,power_law_1.2,0.14307199716567992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,1,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,1,balanced,0.07525333265463512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,1,balanced,0.10974400242169698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,1,balanced,0.17695466677347818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,1,balanced,0.30259732405344647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,1,balanced,0.5574880043665568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,1,balanced,0.5599786837895712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,1,balanced,0.5607786575953165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,1,balanced,0.562117338180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,1,balanced,0.5673919916152954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,1,balanced,0.5687839984893799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,1,balanced,0.5753759940465292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,1,balanced,0.5808320045471191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,1,balanced,0.5866186618804932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,1,balanced,0.5893653233846029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,1,balanced,0.5959786574045817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,1,balanced,0.6069226662317911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,1,balanced,0.6304853359858195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,1,balanced,0.6607466538747152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,1,balanced,0.7093439896901449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,1,balanced,0.7707680066426595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,1,balanced,0.844650665918986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,1,balanced,0.9621760050455729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,1,balanced,1.154858668645223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,1,balanced,1.3406133651733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,1,balanced,1.9218613306681316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,1,balanced,2.305573304494222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,8,power_law_1.2,0.1965567946434021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,8,power_law_1.2,0.24785280227661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,1,power_law_1.2,0.04501760005950928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,1,power_law_1.2,0.05071359872817993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,1,power_law_1.2,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,1,power_law_1.2,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,1,power_law_1.2,0.07095680236816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,1,power_law_1.2,0.07337599992752075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,1,power_law_1.2,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,1,power_law_1.2,0.07306879758834839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,1,power_law_1.2,0.0717248022556305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,1,power_law_1.2,0.07187839746475219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,1,power_law_1.2,0.07478399872779846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,1,power_law_1.2,0.0770367980003357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,128,balanced,0.0317546675602595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,1,power_law_1.2,0.08044160008430482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,1,power_law_1.2,0.0807039976119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,1,power_law_1.2,0.09015039801597595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,1,power_law_1.2,0.08902400135993957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,1,power_law_1.2,0.0955456018447876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,128,balanced,0.03123733401298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,128,balanced,0.03331200033426285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,128,balanced,0.03955200066169103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,128,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,1,power_law_1.2,0.11438080072402954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,128,balanced,0.03332799921433131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,128,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,128,balanced,0.03194666653871536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,128,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,128,balanced,0.03326933334271113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,128,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,128,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,128,balanced,0.03429866582155228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,128,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,128,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,128,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.2,0.12303359508514404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.2,0.15493119955062867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,128,balanced,0.03735466549793879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,128,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,128,balanced,0.037952000896135964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,128,balanced,0.04166933397452036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,128,balanced,0.039962666730086006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,128,balanced,0.04433600107828776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,128,balanced,0.048767998814582825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,128,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,128,balanced,0.06457066535949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,8,power_law_1.2,0.3468800067901611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.2,0.1862015962600708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,128,balanced,0.07658133407433827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,128,balanced,0.08994666735331218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,32,2,power_law_1.2,0.017081600427627564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,128,power_law_1.01,0.01825280040502548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,32,2,power_law_1.2,0.016633599996566772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,32,2,power_law_1.2,0.016473600268363954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,128,power_law_1.01,0.019980800151824952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,32,2,power_law_1.2,0.017164799571037292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,32,2,power_law_1.2,0.01759359985589981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,128,power_law_1.01,0.01884160041809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,32,2,power_law_1.2,0.018303999304771425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,32,2,power_law_1.2,0.018559999763965607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,32,2,power_law_1.2,0.01950719952583313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,32,2,power_law_1.2,0.021222400665283202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,32,2,power_law_1.2,0.02131199985742569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,32,2,power_law_1.2,0.02110079973936081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,32,2,power_law_1.2,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,32,2,power_law_1.2,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,128,power_law_1.01,0.020204800367355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,32,2,power_law_1.2,0.039443200826644896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,128,power_law_1.01,0.023980799317359924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,32,2,power_law_1.2,0.03091840147972107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,128,power_law_1.01,0.037651199102401736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,128,power_law_1.01,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.2,0.23487999439239501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,128,power_law_1.01,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,128,power_law_1.01,0.03563520014286041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,128,power_law_1.01,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,128,power_law_1.01,0.03528960049152374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,32,2,power_law_1.2,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,32,2,power_law_1.2,0.032627201080322264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,128,power_law_1.01,0.03563520014286041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,32,2,power_law_1.2,0.037747201323509214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,128,power_law_1.01,0.0348800003528595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,32,2,power_law_1.2,0.040531200170516965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,128,power_law_1.01,0.034720000624656674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,128,power_law_1.01,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,128,power_law_1.01,0.034246399998664856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,128,power_law_1.01,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,128,power_law_1.01,0.03894400000572205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,128,power_law_1.01,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,128,power_law_1.01,0.04811519980430603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,32,2,power_law_1.2,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,128,power_law_1.01,0.05552639961242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,128,power_law_1.01,0.07062399983406067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,128,power_law_1.01,0.08450559973716736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,128,power_law_1.01,0.12447999715805054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,128,power_law_1.01,0.14967039823532105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,128,power_law_1.01,0.2234623908996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,128,power_law_1.01,0.28970239162445066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.2,0.2913536071777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,32,2,power_law_1.2,0.04991360008716583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,32,2,power_law_1.2,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.2,0.40894718170166017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,32,2,power_law_1.2,0.07916799783706666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.2,0.5161471843719483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,32,2,power_law_1.2,0.10714240074157715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,32,power_law_1.2,0.026393601298332216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,32,power_law_1.2,0.023974399268627166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,32,power_law_1.2,0.025472000241279602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,32,power_law_1.2,0.025190401077270507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,32,power_law_1.2,0.027219200134277345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,32,power_law_1.2,0.02842240035533905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,32,power_law_1.2,0.028428798913955687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,32,power_law_1.2,0.030796799063682555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,32,power_law_1.2,0.03172479867935181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,32,power_law_1.2,0.03175680041313171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,32,power_law_1.2,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,32,power_law_1.2,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,32,power_law_1.2,0.03537920117378235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,32,power_law_1.2,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,32,power_law_1.2,0.042438399791717527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,32,power_law_1.2,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,32,power_law_1.2,0.045535999536514285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,32,power_law_1.2,0.04733439981937408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,32,power_law_1.2,0.05294719934463501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,1,power_law_1.01,0.052236801385879515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,32,power_law_1.2,0.05950080156326294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,1,power_law_1.01,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,32,power_law_1.2,0.06855040192604064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,1,power_law_1.01,0.05625600218772888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,32,power_law_1.2,0.0866815984249115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,1,power_law_1.01,0.06598399877548218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,1,power_law_1.01,0.07292799949645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,1,power_law_1.01,0.09459199905395507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,1,power_law_1.01,0.09865599870681763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,1,power_law_1.01,0.10030080080032348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,1,power_law_1.01,0.10204800367355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,1,power_law_1.01,0.10271999835968018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,1,power_law_1.01,0.11015679836273193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,1,power_law_1.01,0.11092480421066284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,1,power_law_1.01,0.1148800015449524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,1,power_law_1.01,0.11751680374145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,1,power_law_1.01,0.12782080173492433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,1,power_law_1.01,0.13260159492492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,1,power_law_1.01,0.14634239673614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,1,power_law_1.01,0.17386239767074585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,1,power_law_1.01,0.20031359195709228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,1,power_law_1.01,0.25173120498657225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,1,power_law_1.01,0.30738561153411864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,1,power_law_1.01,0.42108798027038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,1,power_law_1.01,0.5371007919311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,32,power_law_1.2,0.10350719690322877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,1,power_law_1.01,0.7700928211212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,1,power_law_1.01,0.9990464210510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,1,power_law_1.01,1.4407679557800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,1,power_law_1.01,1.8851455688476562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,8,power_law_1.2,0.4551424026489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,32,power_law_1.2,0.12760319709777831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,32,power_law_1.2,0.15788160562515258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.2,0.7426688194274902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,32,power_law_1.2,0.22695040702819824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,32,2,power_law_1.2,0.13365119695663452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,32,2,power_law_1.2,0.18943359851837158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,32,power_law_1.2,0.2495743989944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,32,2,power_law_1.2,0.23529601097106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.2,0.9673215866088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,2,power_law_1.01,0.07378559708595275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,2,power_law_1.01,0.10168960094451904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,2,power_law_1.01,0.15028480291366578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,2,power_law_1.01,0.18172800540924072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,2,power_law_1.01,0.2711551904678345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,2,power_law_1.01,0.359500789642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,2,power_law_1.01,0.3766848087310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,2,power_law_1.01,0.48633599281311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,2,power_law_1.01,0.454585599899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,2,power_law_1.01,0.48497920036315917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,2,power_law_1.01,0.5347008228302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,2,power_law_1.01,0.5376768112182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,2,power_law_1.01,0.5521984100341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,2,power_law_1.01,0.5494016170501709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,2,power_law_1.01,0.5936704158782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,2,power_law_1.01,0.6124415874481202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,2,power_law_1.01,0.6092544078826905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,2,power_law_1.01,0.6613696098327637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,2,power_law_1.01,0.7176959991455079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,2,power_law_1.01,0.8316287994384766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,2,power_law_1.01,0.8782079696655274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,2,power_law_1.01,1.1012800216674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,2,power_law_1.01,1.1636351585388183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,2,power_law_1.01,1.444275188446045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,2,power_law_1.01,1.8594751358032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,2,power_law_1.01,2.434739112854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,2,power_law_1.01,2.956185531616211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,8,power_law_1.01,0.021958400309085847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,8,power_law_1.01,0.0236735999584198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,8,power_law_1.01,0.02746239900588989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,8,power_law_1.01,0.02770560085773468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,8,power_law_1.01,0.030950400233268737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,8,power_law_1.01,0.03225600123405457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,8,power_law_1.01,0.035776001214981076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,8,power_law_1.01,0.035231998562812804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,8,power_law_1.01,0.037011200189590455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,8,power_law_1.01,0.037171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,8,power_law_1.01,0.039750400185585025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,8,power_law_1.01,0.03993600010871887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,8,power_law_1.01,0.04186240136623383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,8,power_law_1.01,0.047712001204490664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,8,power_law_1.01,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,8,power_law_1.01,0.05448960065841675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,8,power_law_1.01,0.05687040090560913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,8,power_law_1.01,0.0674560010433197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.01,0.10965119600296021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,16,balanced,0.0421013335386912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.01,0.12446719408035278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.01,0.11356159448623657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.01,0.13443200588226317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.01,0.15079679489135742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,16,balanced,0.04278400043646494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,16,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,16,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,16,balanced,0.06095466514428457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,16,balanced,0.08927999933560689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,16,balanced,0.10248532891273499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,16,balanced,0.09582400321960449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,16,balanced,0.10366400082906087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,16,balanced,0.10827733079592387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,16,balanced,0.10588799913724263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,16,balanced,0.11479999621709187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.01,0.20747520923614501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,16,balanced,0.12541866302490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,16,balanced,0.13125333189964294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,16,balanced,0.13967466354370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,16,balanced,0.15678399801254272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,16,balanced,0.2588319977124532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,16,balanced,0.3102186719576518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,16,balanced,0.21692800521850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,16,balanced,0.248906672000885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.01,0.24919040203094484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.01,0.35576319694519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.01,0.45276799201965334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,16,balanced,0.30028265714645386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,16,balanced,0.37226665019989014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,16,balanced,0.4512639840443929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,16,balanced,0.6066453456878662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,16,balanced,0.7702293395996094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,16,balanced,1.18177064259847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,16,balanced,1.4976959228515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,8,balanced,0.03029866764942805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,8,balanced,0.031184000273545582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,8,balanced,0.027893332143624622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,8,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,8,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,8,balanced,0.03183466692765554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,8,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,8,balanced,0.03108799954255422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,8,balanced,0.03133866687615713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,8,balanced,0.03271999955177307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,8,balanced,0.03120533376932144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,8,balanced,0.033088001112143196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,8,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,8,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,8,balanced,0.03738666574160258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,8,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,8,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,8,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,8,balanced,0.0395359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,8,balanced,0.043605332573254905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,8,balanced,0.04460800190766653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,8,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,8,balanced,0.055205335219701133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,8,balanced,0.07448000212510426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,8,balanced,0.08916266759236653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,8,balanced,0.11825600266456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,8,balanced,0.13024000326792398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,8,2,power_law_1.01,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,8,2,power_law_1.01,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,8,2,power_law_1.01,0.038227200508117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,8,2,power_law_1.01,0.0419840008020401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,8,2,power_law_1.01,0.04227840006351471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,8,2,power_law_1.01,0.04261119961738587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,8,2,power_law_1.01,0.044512000679969785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,8,2,power_law_1.01,0.045286399126052854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,8,2,power_law_1.01,0.04445439875125885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,8,2,power_law_1.01,0.04520959854125976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,8,2,power_law_1.01,0.044633600115776065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,8,2,power_law_1.01,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,8,2,power_law_1.01,0.05454720258712768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,8,2,power_law_1.01,0.0639743983745575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,8,2,power_law_1.01,0.06968960165977478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,8,2,power_law_1.01,0.07254400253295898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,8,2,power_law_1.01,0.0760320007801056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,8,2,power_law_1.01,0.08974080085754395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,8,2,power_law_1.01,0.10945279598236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,8,2,power_law_1.01,0.14883840084075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,8,2,power_law_1.01,0.1587455987930298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,8,2,power_law_1.01,0.22794880867004394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,8,2,power_law_1.01,0.2935935974121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,64,power_law_1.01,0.11038080453872681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,64,power_law_1.01,0.08138880133628845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,64,power_law_1.01,0.08236160278320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,64,power_law_1.01,0.0868607997894287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,64,power_law_1.01,0.0834559977054596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,64,power_law_1.01,0.08781440258026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,64,power_law_1.01,0.08350719809532166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,8,2,power_law_1.01,0.44849281311035155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,64,power_law_1.01,0.08628479838371277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,64,power_law_1.01,0.0843392014503479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,64,power_law_1.01,0.08570240139961242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,64,power_law_1.01,0.0861631989479065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,64,power_law_1.01,0.08629119992256165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,64,power_law_1.01,0.08717439770698547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,64,power_law_1.01,0.09028480052947999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,64,power_law_1.01,0.096697598695755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,64,power_law_1.01,0.09748479723930359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,8,2,power_law_1.01,0.5545023918151856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,64,power_law_1.01,0.10111360549926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,64,power_law_1.01,0.11957119703292847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,64,power_law_1.01,0.12437759637832642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,64,power_law_1.01,0.14557440280914308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,64,power_law_1.01,0.16281599998474122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,64,power_law_1.01,0.2120192050933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,64,power_law_1.01,0.22853760719299315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,64,power_law_1.01,0.3023871898651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,64,power_law_1.01,0.3815808057785034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,1,balanced,0.1476640005906423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,64,power_law_1.01,0.5273344039916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,1,balanced,0.24382400512695312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,1,balanced,0.43884801864624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,64,power_law_1.01,0.6590208053588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,1,balanced,0.839903990427653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,1,balanced,1.6239892641703289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,1,balanced,3.1690985361735025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,8,balanced,0.06854933500289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,8,balanced,0.07833600044250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,8,balanced,0.08874133229255676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,1,balanced,4.734885215759277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,1,balanced,4.732432047526042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,1,balanced,4.738234519958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,1,balanced,4.753797213236491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,1,balanced,4.755914688110352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,1,balanced,4.793381373087565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,1,balanced,4.80020268758138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,1,balanced,4.82755184173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,1,balanced,4.86627737681071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,1,balanced,4.890522638956706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,1,balanced,4.896197319030762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,1,balanced,4.99673589070638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,1,balanced,5.039360046386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,1,balanced,5.1716000239054365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,1,balanced,5.334949493408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,1,balanced,5.5283997853597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,1,balanced,5.767077128092448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,1,balanced,6.233407974243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,8,2,power_law_1.01,0.7877056121826171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,1,balanced,6.657546361287435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,1,balanced,7.320869445800781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,16,balanced,0.039621333281199135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,16,balanced,0.03790933390458425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,16,balanced,0.0397119993964831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,16,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,16,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,16,balanced,0.04555733501911163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,16,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,16,balanced,0.047839999198913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,16,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,16,balanced,0.047685335079828896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,16,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,16,balanced,0.0481333335240682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,16,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,16,balanced,0.05179200073083242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,16,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,16,balanced,0.05598400036493937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,16,balanced,0.05614933371543884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,16,balanced,0.06237866481145223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,16,balanced,0.06609599788983662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,16,balanced,0.07542933523654938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,16,balanced,0.08319999774297078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,1,balanced,10.564170837402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,16,balanced,0.10322133700052898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,16,balanced,0.1241919994354248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,16,balanced,0.17053866386413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,16,balanced,0.2006346583366394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,16,balanced,0.27324267228444415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,16,balanced,0.34913067022959393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,8,2,power_law_1.01,0.9920319557189942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,8,power_law_1.2,0.04536960124969482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,8,power_law_1.2,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,8,power_law_1.2,0.05052800178527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,8,power_law_1.2,0.05314559936523437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,8,power_law_1.2,0.05741440057754517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,8,power_law_1.2,0.0573311984539032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,8,power_law_1.2,0.05867519974708557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,8,power_law_1.2,0.058899199962615965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,8,power_law_1.2,0.061484801769256595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,8,power_law_1.2,0.06257920265197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,8,power_law_1.2,0.06440320014953613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,8,power_law_1.2,0.0665727972984314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,8,power_law_1.2,0.07059199810028076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,8,power_law_1.2,0.07507200241088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,8,power_law_1.2,0.08354560136795045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,8,power_law_1.2,0.08791679739952088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,8,power_law_1.2,0.08835840225219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,1,power_law_1.2,0.01889919936656952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,1,power_law_1.2,0.019315199553966524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,1,power_law_1.2,0.02258560061454773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,1,power_law_1.2,0.02597759962081909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,1,power_law_1.2,0.033004799485206605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,1,power_law_1.2,0.04640640020370483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,1,power_law_1.2,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,1,power_law_1.2,0.04973439872264862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,1,power_law_1.2,0.05281280279159546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,8,power_law_1.2,0.1056447982788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,8,power_law_1.2,0.11857279539108276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,8,power_law_1.2,0.15375360250473022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,8,power_law_1.2,0.18368639945983886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,8,power_law_1.2,0.25938560962677004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,8,power_law_1.2,0.30981760025024413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,1,power_law_1.2,0.05454720258712768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,8,power_law_1.2,0.43985280990600584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,1,power_law_1.2,0.058259201049804685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,1,power_law_1.2,0.05749760270118713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,8,power_law_1.2,0.6075583934783936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,1,power_law_1.2,0.06110720038414001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,1,power_law_1.2,0.06598399877548218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,8,power_law_1.2,0.7896959781646729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,8,power_law_1.2,1.083622360229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,1,power_law_1.2,0.06195840239524841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,1,power_law_1.2,0.06556159853935242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,1,power_law_1.2,0.07436800003051758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,1,power_law_1.2,0.08470399975776673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.2,0.09541760087013244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.2,0.15772160291671752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.2,0.18542079925537108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.2,0.1672127962112427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.2,0.1959231972694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.2,0.2505728006362915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.2,0.3025791883468628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.2,0.4172095775604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,4,power_law_1.01,0.04376319944858551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,4,power_law_1.01,0.04808320105075836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,4,power_law_1.01,0.05260159969329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,4,power_law_1.01,0.0597055971622467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,4,power_law_1.01,0.06277120113372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,4,power_law_1.01,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,4,power_law_1.01,0.0709824025630951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,4,power_law_1.01,0.07055360078811646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,4,power_law_1.01,0.07013760209083557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,4,power_law_1.01,0.07098879814147949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,4,power_law_1.01,0.07370240092277527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,4,power_law_1.01,0.07604479789733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,4,power_law_1.01,0.0777728021144867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,4,power_law_1.01,0.07872639894485474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,4,power_law_1.01,0.08088319897651672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,4,power_law_1.01,0.0821120023727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,4,power_law_1.01,0.0868607997894287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,4,power_law_1.01,0.09342719912528992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,4,power_law_1.01,0.10453120470046998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,4,power_law_1.01,0.12431999444961547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,4,power_law_1.01,0.14276479482650756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,4,power_law_1.01,0.17722879648208617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,4,power_law_1.01,0.23511040210723877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.2,0.5369984149932862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,4,power_law_1.01,0.35427839756011964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,4,power_law_1.01,0.4602367877960205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,4,power_law_1.01,0.7094272136688232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,4,power_law_1.01,0.8620223999023438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,64,power_law_1.01,0.018956799805164338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,64,power_law_1.01,0.01847040057182312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,64,power_law_1.01,0.020710399746894835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,64,power_law_1.01,0.01889919936656952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,64,power_law_1.01,0.02111999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,4,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,4,balanced,0.045706664522488914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,4,balanced,0.04828266799449921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,4,balanced,0.056090667843818665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,4,balanced,0.06049066781997681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,4,balanced,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,4,balanced,0.10547733306884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,4,balanced,0.10126399993896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,4,balanced,0.1014400025208791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,4,balanced,0.10229866703351338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,4,balanced,0.10258666674296062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,4,balanced,0.10725333293279012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,4,balanced,0.11134933431943257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,4,balanced,0.11124266187349956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,4,balanced,0.12027733524640401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,4,balanced,0.12244266271591187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,4,balanced,0.13219733039538065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,4,balanced,0.15238933761914572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,4,balanced,0.16845333576202393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,4,balanced,0.20282665888468424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,4,balanced,0.23169066508611044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,4,balanced,0.301530659198761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,4,balanced,0.3832213481267293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,4,balanced,0.5199306805928549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,4,balanced,0.7094826698303223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,4,balanced,1.0050773620605469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,4,balanced,1.299349308013916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,8,balanced,0.040005333721637726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,8,balanced,0.03980266551176707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,8,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,8,balanced,0.04362666606903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,8,balanced,0.04455466568470001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,8,balanced,0.05009600023428599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,8,balanced,0.05188799897829691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,8,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,8,balanced,0.05148266752560934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,8,balanced,0.052015999952952065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,8,balanced,0.05392533540725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,8,balanced,0.05387733379999796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,8,balanced,0.0540533314148585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,8,balanced,0.05593066910902659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,8,balanced,0.057989334066708885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,8,balanced,0.06043200194835663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,8,balanced,0.062090665102005005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,8,balanced,0.07019199927647908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,8,balanced,0.07425066828727722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,8,balanced,0.08833066622416179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,8,balanced,0.09686932961146037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,8,balanced,0.1281760036945343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,8,balanced,0.15716266632080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,8,balanced,0.20734934012095133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,8,balanced,0.2547946572303772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,8,balanced,0.34938132762908936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,8,balanced,0.4413013458251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,16,balanced,0.04374399781227112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,16,balanced,0.043605332573254905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,16,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,16,balanced,0.04964800179004669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,16,balanced,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,16,balanced,0.05580799778302511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,16,balanced,0.058592001597086586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,16,balanced,0.060319999853769936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,16,balanced,0.0618453323841095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,16,balanced,0.06198399762312571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,16,balanced,0.06258133550484975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,16,balanced,0.06573866804440816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,16,balanced,0.06720533470312755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,16,balanced,0.06751466790835063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,16,balanced,0.07438933352629344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,16,balanced,0.07739733159542084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,16,balanced,0.0804319977760315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,16,balanced,0.09699733058611552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,16,balanced,0.1116426686445872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,16,balanced,0.1363200048605601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,16,balanced,0.15557333827018738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,16,balanced,0.20156800746917725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,16,balanced,0.2542080084482829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,16,balanced,0.341541330019633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,16,balanced,0.43254931767781574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,16,balanced,0.618885318438212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,16,balanced,0.7964266935984293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,4,8,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,4,8,balanced,0.029690665503342945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,4,8,balanced,0.03201066702604294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,4,8,balanced,0.035775999228159584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,4,8,balanced,0.05620799958705902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,4,8,balanced,0.059104000528653465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,4,8,balanced,0.059088001648585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,4,8,balanced,0.05792533357938131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,4,8,balanced,0.0598880002895991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,4,8,balanced,0.05914133290449778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,4,8,balanced,0.05916266640027364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,4,8,balanced,0.060778667529424034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,4,8,balanced,0.059994667768478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,4,8,balanced,0.06198399762312571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,4,8,balanced,0.06851733227570851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,4,8,balanced,0.06668800115585327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,4,8,balanced,0.0702453354994456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,4,8,balanced,0.07292266686757405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,4,8,balanced,0.08358400066693623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,4,8,balanced,0.09400000174840291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,4,8,balanced,0.09465066591898601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,4,8,balanced,0.11131200194358826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,4,8,balanced,0.1285706659158071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,4,8,balanced,0.16238933801651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,4,8,balanced,0.19715199867884317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,4,8,balanced,0.2633066574732463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,4,8,balanced,0.3246346712112427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,16,power_law_1.01,0.02115200012922287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,16,power_law_1.01,0.020691199600696562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,16,power_law_1.01,0.02078080028295517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,16,power_law_1.01,0.021267199516296388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,16,power_law_1.01,0.021619200706481934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,16,power_law_1.01,0.025856000185012818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,16,power_law_1.01,0.03187200129032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,16,power_law_1.01,0.030995199084281923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,16,power_law_1.01,0.04390400052070618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,16,power_law_1.01,0.04299519956111908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,16,power_law_1.01,0.04341759979724884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,16,power_law_1.01,0.04699519872665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,16,power_law_1.01,0.047679999470710756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,16,power_law_1.01,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,16,power_law_1.01,0.04647679924964905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,16,power_law_1.01,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,16,power_law_1.01,0.04831359982490539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,16,power_law_1.01,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.01,0.05475199818611145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,2,power_law_1.01,0.01713919937610626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.01,0.0643775999546051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,2,power_law_1.01,0.018483200669288637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.01,0.0732096016407013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,2,power_law_1.01,0.02099200040102005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.01,0.09145600199699402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,2,power_law_1.01,0.022227199375629426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.01,0.10712959766387939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,2,power_law_1.01,0.02369280010461807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,2,power_law_1.01,0.02655999958515167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.01,0.14766720533370972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,2,power_law_1.01,0.0271807998418808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.01,0.18680959939956665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.01,0.2520639896392822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.01,0.3335616111755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,2,power_law_1.01,0.028012800216674804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,2,power_law_1.01,0.030086401104927062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,2,power_law_1.01,0.03132160007953644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,2,power_law_1.01,0.03126400113105774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,2,power_law_1.01,0.03313280045986176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,2,power_law_1.01,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,2,power_law_1.01,0.036287999153137206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,2,power_law_1.01,0.03780480027198792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,2,power_law_1.01,0.04111360013484955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,2,power_law_1.01,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,2,power_law_1.01,0.04726400077342987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,2,power_law_1.01,0.05304960012435913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,2,power_law_1.01,0.06563839912414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,2,power_law_1.01,0.07968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,2,power_law_1.01,0.09559040069580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,2,power_law_1.01,0.1106943964958191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,2,power_law_1.01,0.13835519552230835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,2,8,power_law_1.2,0.057107198238372806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,2,8,power_law_1.2,0.05919359922409058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,2,8,power_law_1.2,0.05783680081367493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,2,8,power_law_1.2,0.0586624026298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,2,8,power_law_1.2,0.06033920049667359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,2,8,power_law_1.2,0.07010560035705567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,2,8,power_law_1.2,0.08042880296707153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,2,8,power_law_1.2,0.0761023998260498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,2,8,power_law_1.2,0.1435968041419983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,2,power_law_1.01,0.17568639516830445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,2,8,power_law_1.2,0.4137472152709961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,2,8,power_law_1.2,0.06910079717636108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,2,power_law_1.01,0.24160640239715575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,2,8,power_law_1.2,0.08074880242347718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,2,power_law_1.01,0.28250880241394044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,2,8,power_law_1.2,0.07992960214614868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,2,8,power_law_1.2,0.08684800267219543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,2,8,power_law_1.2,0.11548160314559937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,2,8,power_law_1.2,0.11762559413909912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,2,8,power_law_1.2,0.10870399475097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,8,power_law_1.01,0.038950398564338684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,2,8,power_law_1.2,0.13690880537033082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,8,power_law_1.01,0.03852800130844116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,8,power_law_1.01,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,8,power_law_1.01,0.04195840060710907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,8,power_law_1.01,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,8,power_law_1.01,0.043731200695037845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,8,power_law_1.01,0.04490880072116852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,8,power_law_1.01,0.04442239999771118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,8,power_law_1.01,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,8,power_law_1.01,0.045900800824165346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,8,power_law_1.01,0.046675199270248414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,2,8,power_law_1.2,0.17068159580230713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,8,power_law_1.01,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,8,power_law_1.01,0.048665601015090945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,8,power_law_1.01,0.05100160241127014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,8,power_law_1.01,0.05668479800224304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,8,power_law_1.01,0.058924800157546996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,8,power_law_1.01,0.06255360245704651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,8,power_law_1.01,0.07294080257415772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,8,power_law_1.01,0.08194559812545776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,8,power_law_1.01,0.097324800491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,8,power_law_1.01,0.10980479717254639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,8,power_law_1.01,0.1385151982307434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,8,power_law_1.01,0.17254400253295898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,8,power_law_1.01,0.26229119300842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,8,power_law_1.01,0.3077375888824463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,2,8,power_law_1.2,0.22119040489196778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,8,power_law_1.01,0.45418882369995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,8,power_law_1.01,0.5290304183959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,2,8,power_law_1.2,0.2618623971939087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,2,8,power_law_1.2,0.3754240036010742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,2,8,power_law_1.2,0.44382081031799314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,2,8,power_law_1.2,0.7404607772827149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,1,power_law_1.01,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,1,power_law_1.01,0.04186240136623383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,1,power_law_1.01,0.04523519873619079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,1,power_law_1.01,0.06295040249824524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,1,power_law_1.01,0.06441599726676941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,1,power_law_1.01,0.06605439782142639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,1,power_law_1.01,0.0662015974521637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,1,power_law_1.01,0.06657919883728028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,1,power_law_1.01,0.06736639738082886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,1,power_law_1.01,0.06718720197677612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,1,power_law_1.01,0.0695039987564087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,1,power_law_1.01,0.07333760261535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,1,power_law_1.01,0.07523199915885925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,1,power_law_1.01,0.07927680015563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,1,power_law_1.01,0.0870527982711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,1,power_law_1.01,0.09015679955482483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,1,power_law_1.01,0.09684479832649232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,2,power_law_1.01,0.03857280015945434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,2,power_law_1.01,0.03999359905719757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,2,power_law_1.01,0.04013440012931824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,2,power_law_1.01,0.044563201069831845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,2,power_law_1.01,0.04472959935665131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,2,power_law_1.01,0.04508160054683685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,2,power_law_1.01,0.045228800177574156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,2,power_law_1.01,0.045177599787712096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,2,power_law_1.01,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,2,power_law_1.01,0.04559360146522522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,2,power_law_1.01,0.04635519981384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,2,power_law_1.01,0.046700799465179445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,1,power_law_1.01,0.11239680051803588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.01,0.1282688021659851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,2,8,power_law_1.2,0.9343808174133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,2,power_law_1.01,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,2,power_law_1.01,0.05141119956970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,2,power_law_1.01,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,2,power_law_1.01,0.0562175989151001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.01,0.1598847985267639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,2,power_law_1.01,0.060159999132156375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,2,power_law_1.01,0.07391999959945679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,2,power_law_1.01,0.08729599714279175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.01,0.18626559972763063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,2,power_law_1.01,0.11350400447845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,2,power_law_1.01,0.13138560056686402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,2,power_law_1.01,0.18448640108108522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.01,0.24725759029388428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.01,0.3098752021789551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,2,power_law_1.01,0.23213438987731932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.01,0.4333631992340088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,2,power_law_1.01,0.3052031993865967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.01,0.5557759761810303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,2,8,power_law_1.2,1.3263872146606446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,2,8,power_law_1.2,1.4821056365966796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,2,power_law_1.01,0.40750718116760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.01,0.793126392364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.01,1.0354880332946776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,128,power_law_1.2,0.07401599884033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,128,power_law_1.2,0.07432960271835327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,128,power_law_1.2,0.07334399819374085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,128,power_law_1.2,0.07354879975318909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,128,power_law_1.2,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,128,power_law_1.2,0.08113920092582702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,128,power_law_1.2,0.08469120264053345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,128,power_law_1.2,0.08629119992256165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,128,power_law_1.2,0.08540160059928895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,128,power_law_1.2,0.08739200234413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,128,power_law_1.2,0.08764799833297729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,128,power_law_1.2,0.09126399755477906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,128,power_law_1.2,0.09149439930915833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,128,power_law_1.2,0.09312000274658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,2,power_law_1.01,0.5404160022735596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,128,power_law_1.2,0.10097279548645019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,128,power_law_1.2,0.10398080348968505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,128,power_law_1.2,0.11161600351333618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,128,power_law_1.2,0.14343680143356324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,128,power_law_1.2,0.15730559825897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,128,power_law_1.2,0.20254719257354736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,128,power_law_1.2,0.25287680625915526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,128,power_law_1.2,0.3475775957107544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,128,power_law_1.2,0.4307072162628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,128,power_law_1.2,0.5842688083648682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,128,power_law_1.2,0.7852479934692382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,128,power_law_1.2,1.2093631744384765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,128,power_law_1.2,1.606515121459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,2,power_law_1.01,0.7189119815826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,2,power_law_1.01,0.1312127947807312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,2,power_law_1.01,0.17442560195922852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,2,power_law_1.01,0.20279040336608886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,2,power_law_1.01,0.2696959972381592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,2,power_law_1.01,0.3208832025527954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,2,power_law_1.01,0.4951744079589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,2,power_law_1.01,0.527455997467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,2,power_law_1.01,0.530073595046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,2,power_law_1.01,0.5420735836029053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,2,power_law_1.01,0.5540287971496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,2,power_law_1.01,0.5588736057281494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,2,power_law_1.01,0.5911935806274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,2,power_law_1.01,0.6011648178100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,2,power_law_1.01,0.6249728202819824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,2,power_law_1.01,0.6298880100250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,2,power_law_1.01,0.6555903911590576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,2,power_law_1.01,0.6851456165313721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,2,power_law_1.01,0.772057580947876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,4,power_law_1.01,0.050367999076843264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,2,power_law_1.01,0.8500800132751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,4,power_law_1.01,0.06303359866142273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,2,power_law_1.01,1.0281663894653321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,4,power_law_1.01,0.06458879709243774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,4,power_law_1.01,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,2,power_law_1.01,1.1072959899902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,4,power_law_1.01,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,4,power_law_1.01,0.11110399961471558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,4,power_law_1.01,0.10678399801254272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,2,power_law_1.01,1.4333951950073243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,4,power_law_1.01,0.11128959655761719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,4,power_law_1.01,0.11301759481430054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,4,power_law_1.01,0.11416319608688355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,4,power_law_1.01,0.11883519887924195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,4,power_law_1.01,0.11894400119781494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,4,power_law_1.01,0.12006399631500245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,4,power_law_1.01,0.11909120082855225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,4,power_law_1.01,0.13022079467773437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,4,power_law_1.01,0.13225599527359008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,4,power_law_1.01,0.14164479970932006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,4,power_law_1.01,0.15158400535583497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,4,power_law_1.01,0.1550271987915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,2,power_law_1.01,1.6996416091918944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,4,power_law_1.01,0.1877951979637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,4,power_law_1.01,0.2123136043548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,4,power_law_1.01,0.2461632013320923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,2,power_law_1.01,2.297843170166016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,4,power_law_1.01,0.25880959033966067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,4,power_law_1.01,0.3247423887252808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,4,power_law_1.01,0.4112832069396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,2,power_law_1.01,2.841209602355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,2,power_law_1.01,3.9658432006835938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,4,power_law_1.01,0.5084224224090577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,4,power_law_1.01,0.6655488014221191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,2,power_law_1.01,5.113792037963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,2,power_law_1.01,0.02247679978609085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,2,power_law_1.01,0.02268799990415573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,2,power_law_1.01,0.025209599733352663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,2,power_law_1.01,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,2,power_law_1.01,0.04352000057697296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,2,power_law_1.01,0.05133439898490906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,2,power_law_1.01,0.057171201705932616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,2,power_law_1.01,0.06085759997367859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,2,power_law_1.01,0.06277120113372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,2,power_law_1.01,0.06726400256156921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,2,power_law_1.01,0.07248640060424805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,2,power_law_1.01,0.06883839964866638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,2,power_law_1.01,0.07141759991645813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,2,power_law_1.01,0.07951359748840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,2,power_law_1.01,0.07534080147743225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,2,power_law_1.01,0.08076800107955932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,2,power_law_1.01,0.08780800104141236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,2,power_law_1.01,0.10793600082397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,64,power_law_1.2,0.01640959978103638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.01,0.11857919692993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,64,power_law_1.2,0.016940799355506898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.01,0.18346879482269288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,64,power_law_1.2,0.016281600296497344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.01,0.19685759544372558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,64,power_law_1.2,0.01592320054769516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.01,0.2229311943054199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,64,power_law_1.2,0.01671680063009262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.01,0.24876799583435058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,64,power_law_1.2,0.019251200556755065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,64,power_law_1.2,0.019519999623298645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.01,0.31319680213928225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,64,power_law_1.2,0.02568320035934448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.01,0.3751744031906128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,64,power_law_1.2,0.0262719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,64,power_law_1.2,0.02598400115966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.01,0.5206463813781739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,64,power_law_1.2,0.026067200303077697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.01,0.6509312152862549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,64,power_law_1.2,0.02617599964141846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,64,power_law_1.2,0.026086398959159852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,64,power_law_1.2,0.027500799298286437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,64,power_law_1.2,0.02863999903202057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,64,power_law_1.2,0.029631999135017396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,64,power_law_1.2,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,64,power_law_1.2,0.03219839930534363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,64,power_law_1.2,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,64,power_law_1.2,0.038867199420928956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,64,power_law_1.2,0.04184960126876831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,64,power_law_1.2,0.04806399941444397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,64,power_law_1.2,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,64,power_law_1.2,0.07205119729042053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,64,power_law_1.2,0.09128320217132568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,64,power_law_1.2,0.12334719896316529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,128,balanced,0.021274665991465252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,128,balanced,0.023152001202106476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,128,balanced,0.021082667013009388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,128,balanced,0.0233599990606308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,32,power_law_1.01,0.04469119906425476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,128,balanced,0.02144533395767212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,32,power_law_1.01,0.036524799466133115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,32,power_law_1.01,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,32,power_law_1.01,0.03763200044631958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,128,balanced,0.02714666724205017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,32,power_law_1.01,0.03754880130290985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,32,power_law_1.01,0.03777920007705689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,32,power_law_1.01,0.03885439932346344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,128,balanced,0.025258667767047882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,128,balanced,0.0420959989229838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,128,balanced,0.04248000184694926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,128,balanced,0.04168533285458883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,128,balanced,0.04159466673930486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,128,balanced,0.0421066681543986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,128,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,128,balanced,0.03852800031503042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,128,balanced,0.04839999973773956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,128,balanced,0.045610666275024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,128,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,128,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,128,balanced,0.053445334235827126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,32,power_law_1.01,0.038540801405906676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,32,power_law_1.01,0.0383679986000061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,32,power_law_1.01,0.038950398564338684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,32,power_law_1.01,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,32,power_law_1.01,0.03917439877986908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,32,power_law_1.01,0.038873600959777835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,128,balanced,0.07492800056934357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,32,power_law_1.01,0.03969280123710632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,128,balanced,0.08481066425641377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,128,balanced,0.11687466502189636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,128,balanced,0.1455680032571157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,128,balanced,0.20414932568868002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,128,balanced,0.2576106588045756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,128,balanced,0.3731520175933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,32,power_law_1.01,0.04356479942798615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,128,balanced,0.49320534865061444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,32,power_law_1.01,0.04656000137329101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,32,power_law_1.01,0.047328001260757445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,32,power_law_1.01,0.05214080214500427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,32,power_law_1.01,0.055327999591827395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,32,power_law_1.01,0.062489598989486694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,32,power_law_1.01,0.07215359807014465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,32,power_law_1.01,0.08231679797172546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,32,power_law_1.01,0.09688959717750549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,32,power_law_1.01,0.12303359508514404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,32,power_law_1.01,0.1652351975440979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,32,power_law_1.01,0.21123199462890624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,32,power_law_1.01,0.2640255928039551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,1,balanced,0.021712000171343487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,1,balanced,0.021370666722456615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,1,balanced,0.0210506667693456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,1,balanced,0.02269333352645238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,1,balanced,0.025434667865435284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,1,balanced,0.03156266609827677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,1,balanced,0.030623999734719593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,1,balanced,0.031311998764673867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,1,balanced,0.028245332340399425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,1,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,1,balanced,0.03344533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,64,power_law_1.2,0.154310405254364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,1,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,1,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,1,balanced,0.037952000896135964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,1,balanced,0.06278400123119354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,1,balanced,0.07875733574231465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,1,balanced,0.0934879978497823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,1,balanced,0.05996799965699514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,1,balanced,0.06782933572928111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,1,balanced,0.0885599950949351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,1,balanced,0.10971732934315999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,1,balanced,0.1487573285897573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,1,balanced,0.18275733788808188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,1,balanced,0.25785066684087116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,1,balanced,0.3292693297068278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,1,balanced,0.4766933520634969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,1,balanced,0.6232959826787313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,2,power_law_1.2,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,2,power_law_1.2,0.03893760144710541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,2,power_law_1.2,0.044460800290107724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,2,power_law_1.2,0.05867519974708557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,2,power_law_1.2,0.0571008026599884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,2,power_law_1.2,0.059596800804138185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,2,power_law_1.2,0.06188160181045532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,2,power_law_1.2,0.06280320286750793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,2,power_law_1.2,0.06414080262184144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,2,power_law_1.2,0.06602240204811097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,2,power_law_1.2,0.06750079989433289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,2,power_law_1.2,0.06908159852027893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,2,power_law_1.2,0.07057279944419861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,2,power_law_1.2,0.07630079984664917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,2,power_law_1.2,0.08482559919357299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,2,power_law_1.2,0.08758400082588196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,2,power_law_1.2,0.09022079706192017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,2,power_law_1.2,0.10803840160369874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,2,power_law_1.2,0.12510720491409302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,2,power_law_1.2,0.1514240026473999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,2,power_law_1.2,0.18778239488601683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,2,power_law_1.2,0.24056320190429686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,2,power_law_1.2,0.3016256093978882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,4,power_law_1.01,0.08788480162620545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,4,power_law_1.01,0.1305343985557556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,4,power_law_1.01,0.1620800018310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,4,power_law_1.01,0.20982398986816406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,4,power_law_1.01,0.2500096082687378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,4,power_law_1.01,0.26990718841552735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,4,power_law_1.01,0.3302848100662231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,2,power_law_1.2,0.40694398880004884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,4,power_law_1.01,0.3105407953262329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,4,power_law_1.01,0.34250879287719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,4,power_law_1.01,0.3549247980117798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,4,power_law_1.01,0.3380160093307495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,4,power_law_1.01,0.3406399965286255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,4,power_law_1.01,0.3506175994873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,4,power_law_1.01,0.3640768051147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,4,power_law_1.01,0.3766527891159058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,1,power_law_1.01,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,4,power_law_1.01,0.3872384071350098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,1,power_law_1.01,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,4,power_law_1.01,0.4162879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,1,power_law_1.01,0.10164480209350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,4,power_law_1.01,0.45194239616394044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,1,power_law_1.01,0.13657599687576294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,16,power_law_1.01,0.07612800002098083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,1,power_law_1.01,0.17136640548706056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,4,power_law_1.01,0.45453438758850095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,16,power_law_1.01,0.0785215973854065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,1,power_law_1.01,0.24146559238433837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,16,power_law_1.01,0.07426559925079346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,4,power_law_1.01,0.533247995376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,2,power_law_1.2,0.4734464168548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,16,power_law_1.01,0.07882879972457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,4,power_law_1.01,0.5582464218139649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,4,power_law_1.01,0.6732992172241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,4,power_law_1.01,0.7874815940856934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,1,power_law_1.01,0.24892160892486573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,4,power_law_1.01,0.9797632217407226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,1,power_law_1.01,0.25622398853302003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,16,power_law_1.01,0.07416319847106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,1,power_law_1.01,0.2586047887802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,4,power_law_1.01,1.1839296340942382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,1,power_law_1.01,0.2686847925186157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,16,power_law_1.01,0.08210560083389282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,1,power_law_1.01,0.2782335996627808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,1,power_law_1.01,0.28273279666900636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,1,power_law_1.01,0.2877952098846436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,1,power_law_1.01,0.2999743938446045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,1,power_law_1.01,0.3122944116592407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,1,power_law_1.01,0.32614400386810305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,16,power_law_1.01,0.08135039806365967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,1,power_law_1.01,0.34754559993743894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,4,power_law_1.01,1.6401664733886718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,1,power_law_1.01,0.3994623899459839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.01,0.4287168025970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,4,power_law_1.01,2.0392831802368163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.01,0.5182591915130615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.01,0.6088895797729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.01,0.7909887790679931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,16,power_law_1.01,0.08023679852485657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,16,power_law_1.01,0.08414080142974853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.01,0.9637503623962402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,16,power_law_1.01,0.08531839847564697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,16,power_law_1.01,0.08721920251846313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.01,1.3306559562683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,16,power_law_1.01,0.08883839845657349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,16,power_law_1.01,0.08988800048828124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,16,power_law_1.01,0.09304959774017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.01,1.6715391159057618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,16,power_law_1.01,0.10135040283203126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,16,power_law_1.01,0.1019968032836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,16,power_law_1.01,0.11310720443725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.01,2.3554880142211916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.01,3.000160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,16,power_law_1.01,0.13063039779663085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,16,power_law_1.01,0.1359295964241028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,16,power_law_1.01,0.1692352056503296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,16,power_law_1.01,0.198470401763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,16,power_law_1.01,0.255679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,16,power_law_1.01,0.31748480796813966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,16,power_law_1.01,0.4262976169586182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,16,power_law_1.01,0.5155903816223144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,16,power_law_1.01,0.7015615940093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,16,power_law_1.01,0.9217920303344727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,2,power_law_1.2,0.8247679710388184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,2,power_law_1.2,0.9414336204528808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,8,power_law_1.2,0.03912320137023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,8,power_law_1.2,0.03871999979019165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,8,power_law_1.2,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,8,power_law_1.2,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,8,power_law_1.2,0.041094401478767396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,8,power_law_1.2,0.04160000085830688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,8,power_law_1.2,0.041510400176048276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,8,power_law_1.2,0.0416128009557724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,8,power_law_1.2,0.04167680144309997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,8,power_law_1.2,0.0419840008020401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,8,power_law_1.2,0.04242559969425201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,8,power_law_1.2,0.04479359984397888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,8,power_law_1.2,0.04405759871006012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,8,power_law_1.2,0.044870400428771974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,8,power_law_1.2,0.0513152003288269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,8,power_law_1.2,0.05187199711799621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,2,power_law_1.2,0.07491840124130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,2,power_law_1.2,0.101363205909729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,8,power_law_1.2,0.05556480288505554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,2,power_law_1.2,0.13470079898834228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,2,power_law_1.2,0.186080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,8,power_law_1.2,0.06209279894828797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,2,power_law_1.2,0.2510080099105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,8,power_law_1.2,0.07156479954719544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,2,power_law_1.2,0.3329024076461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,2,power_law_1.2,0.3082304000854492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,2,power_law_1.2,0.48446078300476075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,2,power_law_1.2,0.48625922203063965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,4,power_law_1.2,0.04867840111255646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,2,power_law_1.2,0.4627840042114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,4,power_law_1.2,0.05107839703559876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,2,power_law_1.2,0.5219583988189698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,4,power_law_1.2,0.0602944016456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,4,power_law_1.2,0.06839680075645446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,2,power_law_1.2,0.5463615894317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,8,power_law_1.2,0.10017919540405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,2,power_law_1.2,0.5487936019897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,2,power_law_1.2,0.5632256031036377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,4,power_law_1.2,0.07530239820480347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,2,power_law_1.2,0.5738111972808838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,4,power_law_1.2,0.08250240087509156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,2,power_law_1.2,0.6067135810852051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,4,power_law_1.2,0.09051520228385926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,2,power_law_1.2,0.6495039939880372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,2,power_law_1.2,0.6903039932250976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,2,power_law_1.2,0.7228032112121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,2,power_law_1.2,0.8787903785705566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,2,power_law_1.2,0.9516480445861817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,4,power_law_1.2,0.1094208002090454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,2,power_law_1.2,1.1319552421569825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,4,power_law_1.2,0.10297600030899048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,4,power_law_1.2,0.1098431944847107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,2,power_law_1.2,1.2226880073547364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,4,power_law_1.2,0.11052160263061524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,8,power_law_1.2,0.12345600128173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,4,power_law_1.2,0.11356159448623657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,4,power_law_1.2,0.11447679996490479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,4,power_law_1.2,0.11797120571136474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,4,power_law_1.2,0.12842880487442015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,4,power_law_1.2,0.12524800300598143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,4,power_law_1.2,0.140447998046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,2,power_law_1.2,1.5607616424560546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,4,power_law_1.2,0.1531775951385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,4,power_law_1.2,0.16995199918746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,4,power_law_1.2,0.2127232074737549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,2,power_law_1.2,1.9473920822143556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,4,power_law_1.2,0.24664320945739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,4,power_law_1.2,0.3115008115768433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,4,power_law_1.2,0.3922111988067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,4,power_law_1.2,0.5380544185638427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,4,power_law_1.2,0.7193215847015381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,8,power_law_1.2,0.16004480123519899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,4,power_law_1.2,1.0018239974975587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,2,power_law_1.2,2.5347904205322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,4,power_law_1.2,1.352524757385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,2,power_law_1.2,3.098713684082031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,1,power_law_1.2,0.07890560030937195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,1,power_law_1.2,0.11406079530715943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,1,power_law_1.2,0.1720128059387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,1,power_law_1.2,0.2812096118927002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,1,power_law_1.2,0.38102400302886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,1,power_law_1.2,0.511084794998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,1,power_law_1.2,0.7736000061035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,1,power_law_1.2,0.7921792030334472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,1,power_law_1.2,0.8350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,1,power_law_1.2,0.8690815925598144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,1,power_law_1.2,0.8836799621582031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,1,power_law_1.2,0.9054719924926757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,1,power_law_1.2,0.9410880088806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,1,power_law_1.2,0.9708479881286621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,1,power_law_1.2,0.9967807769775391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,1,power_law_1.2,1.02871675491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,1,power_law_1.2,1.086515235900879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,1,power_law_1.2,1.2015104293823242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,1,power_law_1.2,1.2011072158813476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,1,power_law_1.2,1.323788833618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,1,power_law_1.2,1.335807991027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,1,power_law_1.2,1.5287039756774903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,1,power_law_1.2,1.715660858154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,8,power_law_1.2,0.20699520111083985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,1,power_law_1.2,2.0986623764038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,1,power_law_1.2,2.4546112060546874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,1,power_law_1.2,3.11911678314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,1,power_law_1.2,3.7844287872314455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,8,power_law_1.2,0.26908159255981445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,8,power_law_1.2,0.3764928102493286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,8,power_law_1.2,0.5346943855285644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,16,power_law_1.2,0.017030400037765504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,16,power_law_1.2,0.016300800442695617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,16,power_law_1.2,0.016128000617027283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,16,power_law_1.2,0.016300800442695617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,16,power_law_1.2,0.0173567995429039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,16,power_law_1.2,0.01738879978656769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,16,power_law_1.2,0.018054400384426118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,16,power_law_1.2,0.01849599927663803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,16,power_law_1.2,0.019040000438690186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,16,power_law_1.2,0.01929599940776825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,16,power_law_1.2,0.02142080068588257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,16,power_law_1.2,0.02258560061454773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,16,power_law_1.2,0.022873599827289582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,16,power_law_1.2,0.029017600417137145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,16,power_law_1.2,0.02917119860649109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,16,power_law_1.2,0.02927359938621521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,16,power_law_1.2,0.030931198596954347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,16,power_law_1.2,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,16,power_law_1.2,0.03624959886074066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,16,power_law_1.2,0.042444801330566405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,16,power_law_1.2,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,8,power_law_1.2,0.7035840034484864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,16,power_law_1.2,0.053855997323989865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,16,power_law_1.2,0.06572160124778748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,16,power_law_1.2,0.08344320058822632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,1,power_law_1.2,0.2012415885925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,32,power_law_1.2,0.05797119736671448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,1,power_law_1.2,0.32245759963989257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,32,power_law_1.2,0.057183998823165896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,32,power_law_1.2,0.05625600218772888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,1,power_law_1.2,0.5925888061523438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,32,power_law_1.2,0.06032000184059143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,32,power_law_1.2,0.056524801254272464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,1,power_law_1.2,0.8985024452209472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,32,power_law_1.2,0.05722879767417908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,32,power_law_1.2,0.062054401636123656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,1,power_law_1.2,1.2779264450073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,32,power_law_1.2,0.06232960224151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,32,power_law_1.2,0.06108160018920898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,32,power_law_1.2,0.05681920051574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,1,power_law_1.2,2.3371456146240233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,32,power_law_1.2,0.059622400999069215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,32,power_law_1.2,0.06202239990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,32,power_law_1.2,0.06723200082778931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,1,power_law_1.2,2.476767921447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,32,power_law_1.2,0.06772480010986329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,32,power_law_1.2,0.07520639896392822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,32,power_law_1.2,0.07875840067863464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,32,power_law_1.2,0.08250880241394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,32,power_law_1.2,0.091839998960495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,32,power_law_1.2,0.10138880014419556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,32,power_law_1.2,0.12960000038146974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,32,power_law_1.2,0.13162239789962768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,32,power_law_1.2,0.1755136013031006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,32,power_law_1.2,0.20742399692535402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,1,power_law_1.2,2.6369344711303713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,32,power_law_1.2,0.2852864027023315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,16,power_law_1.2,0.10430079698562622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,32,power_law_1.2,0.35786240100860595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,1,power_law_1.2,2.6804351806640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,32,power_law_1.2,0.49344639778137206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,32,power_law_1.2,0.6234560012817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,1,power_law_1.2,2.7883968353271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,1,power_law_1.2,2.857260894775391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,1,power_law_1.2,2.921414375305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,1,power_law_1.2,2.989836883544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,1,power_law_1.2,3.17639045715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,1,power_law_1.2,3.2893310546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,1,power_law_1.2,3.3581825256347657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,1,power_law_1.2,3.4786750793457033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,1,power_law_1.2,3.8116798400878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,1,power_law_1.2,4.050809478759765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,1,power_law_1.2,4.455456161499024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,1,power_law_1.2,4.656620788574219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,1,power_law_1.2,5.4425407409667965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,1,power_law_1.2,5.812543869018555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,1,power_law_1.2,7.100691223144532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,1,power_law_1.2,10.026592254638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,16,power_law_1.2,0.14544639587402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,1,power_law_1.2,12.331756591796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,1,power_law_1.2,13.438720703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,16,4,balanced,0.04266666869322459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,16,4,balanced,0.043840001026789345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,16,4,balanced,0.048026666045188904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,16,4,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,16,4,balanced,0.06541866560777028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,16,4,balanced,0.07891199986139934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,16,4,balanced,0.0804746647675832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,16,4,balanced,0.0810346653064092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,16,4,balanced,0.08847999572753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,16,4,balanced,0.08251200119654338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,16,4,balanced,0.08310933411121368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,16,4,balanced,0.08684800068537395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,16,4,balanced,0.08960533142089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,16,4,balanced,0.09021866321563721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,16,4,balanced,0.09735999504725139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,16,4,balanced,0.1030399998029073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,16,4,balanced,0.10528533657391866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,16,4,balanced,0.12164800365765889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,16,4,balanced,0.13008532921473184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,16,4,balanced,0.17206400632858276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,16,4,balanced,0.1953493356704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,16,4,balanced,0.27449599901835126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,16,4,balanced,0.31036800146102905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,16,4,balanced,0.4514933427174886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,16,4,balanced,0.565829316775004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,16,4,balanced,0.7931466897328695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,16,4,balanced,1.0374826590220134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,16,power_law_1.2,0.19074560403823854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,32,balanced,0.022410665949185688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,32,balanced,0.02176533391078313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,32,balanced,0.021583999196688335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,32,balanced,0.021344001094500225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,32,balanced,0.025631998976071674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,32,balanced,0.029685333371162415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,32,balanced,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,32,balanced,0.03526400029659271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,32,balanced,0.05487999816735586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,32,balanced,0.05407466491063436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,32,balanced,0.05153599878152212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,32,balanced,0.05498133103052775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,32,balanced,0.05138133466243744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,32,balanced,0.04997866849104563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,32,balanced,0.06498133142789204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,32,balanced,0.06284800171852112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,32,balanced,0.058304001887639366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,32,balanced,0.0699839989344279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,32,balanced,0.07446933289368947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,32,balanced,0.08922132849693298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,32,balanced,0.10646933317184448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,32,balanced,0.13517866532007852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,32,balanced,0.1665386656920115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,32,balanced,0.22366400559743246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,1,power_law_1.01,0.11632000207901001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,32,balanced,0.2846986651420593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,32,balanced,0.41945068041483563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,1,power_law_1.01,0.1533504009246826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,32,balanced,0.539962649345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,1,power_law_1.01,0.22035839557647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,1,power_law_1.01,0.3219712018966675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,1,power_law_1.01,0.43259520530700685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,1,power_law_1.01,0.5422848224639892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,1,power_law_1.01,0.6857024192810058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,1,power_law_1.01,0.6969215869903564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,1,power_law_1.01,0.7309567928314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,1,power_law_1.01,0.7510272026062011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,1,power_law_1.01,0.7626368045806885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,1,power_law_1.01,0.7677824020385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,1,power_law_1.01,0.8040320396423339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,1,4,power_law_1.2,0.11991679668426514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,1,power_law_1.01,0.8058048248291015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,1,4,power_law_1.2,0.14885760545730592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,1,power_law_1.01,0.8049216270446777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,1,power_law_1.01,0.7995200157165527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,1,4,power_law_1.2,0.19422080516815185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,1,power_law_1.01,0.8470080375671387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,1,4,power_law_1.2,0.24056320190429686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,1,power_law_1.01,0.9086079597473145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,1,4,power_law_1.2,0.29786880016326905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,1,power_law_1.01,0.9714367866516114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,1,4,power_law_1.2,0.38369920253753664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,1,power_law_1.01,1.106118392944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,1,4,power_law_1.2,0.37455999851226807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,1,power_law_1.01,1.2145471572875977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,1,4,power_law_1.2,0.3969536066055298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,1,4,power_law_1.2,0.40471038818359373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,1,power_law_1.01,1.4908608436584472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,1,4,power_law_1.2,0.40512638092041015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,1,4,power_law_1.2,0.4184256076812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,1,power_law_1.01,1.7322303771972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,1,4,power_law_1.2,0.44359679222106935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,1,4,power_law_1.2,0.43907837867736815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,1,power_law_1.01,2.270438385009766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,1,4,power_law_1.2,0.45996799468994143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,1,4,power_law_1.2,0.48619518280029295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,1,4,power_law_1.2,0.46367359161376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,1,4,power_law_1.2,0.5417791843414307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,1,4,power_law_1.2,0.61910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,1,4,power_law_1.2,0.5979584217071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,1,power_law_1.01,2.773632049560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,1,power_law_1.01,3.7915775299072267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,1,4,power_law_1.2,0.7024576187133789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,1,4,power_law_1.2,0.7932223796844482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,1,power_law_1.01,4.807123184204102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,1,4,power_law_1.2,1.0061823844909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,1,4,power_law_1.2,1.0180159568786622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,1,4,power_law_1.2,1.2767807960510253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,1,4,power_law_1.2,1.6088447570800781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,1,4,power_law_1.2,2.3064064025878905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,1,4,power_law_1.2,2.56429443359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,1,power_law_1.2,0.05004159808158874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,1,power_law_1.2,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,1,power_law_1.2,0.062054401636123656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,1,power_law_1.2,0.08091520071029663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,1,power_law_1.2,0.10206719636917114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,1,power_law_1.2,0.1176192045211792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,1,power_law_1.2,0.1554111957550049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,1,power_law_1.2,0.1574463963508606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,1,balanced,0.11929066975911458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,1,power_law_1.2,0.16891520023345946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,1,balanced,0.21182399988174438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,1,power_law_1.2,0.16839679479598998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,1,power_law_1.2,0.17697279453277587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,1,power_law_1.2,0.18625919818878173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,1,power_law_1.2,0.19226880073547364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,1,power_law_1.2,0.20490880012512208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,1,power_law_1.2,0.21047680377960204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,1,balanced,0.38042132059733075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,1,power_law_1.2,0.2169408082962036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,1,balanced,0.3858720064163208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,1,balanced,0.38441065947214764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,1,balanced,0.38625065485636395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,1,balanced,0.38482666015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,1,balanced,0.3858666817347209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,1,balanced,0.3898719946543376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,1,balanced,0.3906506697336833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,1,balanced,0.39236799875895184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,1,power_law_1.2,0.22709760665893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,1,balanced,0.3963093360265096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,1,power_law_1.2,0.2573823928833008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,1,balanced,0.39948264757792157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,1,balanced,0.4199306567509969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,1,balanced,0.4211946725845337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,1,balanced,0.42506134510040283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,1,balanced,0.43935465812683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,1,balanced,0.5417600075403849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,1,balanced,0.5437599817911783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,1,balanced,0.7296266555786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,1,balanced,0.7932480176289877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,1,power_law_1.2,0.28499839305877683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,1,balanced,1.0937386353810628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,1,power_law_1.2,0.3416575908660889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,1,balanced,1.4612372716267903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,1,power_law_1.2,0.39027841091156007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,1,balanced,2.124074618021647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,1,power_law_1.2,0.489958381652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,1,balanced,2.754762649536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,1,power_law_1.2,0.5904640197753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,1,balanced,4.071226755777995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,1,power_law_1.2,0.7821184158325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,1,power_law_1.2,0.9722240447998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,1,balanced,5.401919682820638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,1,power_law_1.2,1.359334373474121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,1,power_law_1.2,1.7474752426147462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,32,power_law_1.01,0.02391040027141571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,32,power_law_1.01,0.02207999974489212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,32,power_law_1.01,0.022111999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,32,power_law_1.01,0.023520000278949738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,32,power_law_1.01,0.025119999051094057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,32,power_law_1.01,0.02592639923095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,32,power_law_1.01,0.025791999697685242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,32,power_law_1.01,0.02794240117073059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,32,power_law_1.01,0.029177600145339967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,32,power_law_1.01,0.02996479868888855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,32,power_law_1.01,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,32,power_law_1.01,0.031865599751472476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,32,power_law_1.01,0.032287999987602234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,32,power_law_1.01,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,32,power_law_1.01,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,32,power_law_1.01,0.036032000184059144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,32,power_law_1.01,0.0365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,32,power_law_1.01,0.04046719968318939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,4,power_law_1.2,0.04613119959831238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,32,power_law_1.01,0.045817598700523376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,4,power_law_1.2,0.05120000243186951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,32,power_law_1.01,0.05133439898490906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,4,power_law_1.2,0.051072001457214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,4,power_law_1.2,0.054745602607727054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,4,power_law_1.2,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,4,power_law_1.2,0.056883198022842404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,4,power_law_1.2,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,4,power_law_1.2,0.055232000350952146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,32,power_law_1.01,0.05367040038108826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,4,power_law_1.2,0.055302399396896365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,32,power_law_1.01,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,4,power_law_1.2,0.05363199710845947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,32,power_law_1.01,0.07293440103530884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,32,power_law_1.01,0.09918720126152039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,4,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,32,power_law_1.01,0.1152575969696045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,4,power_law_1.2,0.055225598812103274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,32,power_law_1.01,0.14975359439849853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,4,power_law_1.2,0.05552639961242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,32,power_law_1.01,0.21198079586029053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,4,power_law_1.2,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,4,power_law_1.2,0.07109119892120361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,4,power_law_1.2,0.06474879980087281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,4,power_law_1.2,0.07512320280075073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,4,power_law_1.2,0.09018239974975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.2,0.0956928014755249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,64,power_law_1.2,0.021721599996089934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,64,power_law_1.2,0.021484799683094025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,64,power_law_1.2,0.02069759964942932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,64,power_law_1.2,0.02309119999408722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.2,0.12679680585861205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,64,power_law_1.2,0.02237440049648285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.2,0.14218239784240722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,64,power_law_1.2,0.024672000110149382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,64,power_law_1.2,0.02481279969215393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,64,power_law_1.2,0.027340799570083618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,64,power_law_1.2,0.0286080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,64,power_law_1.2,0.029369598627090453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,64,power_law_1.2,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,64,power_law_1.2,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,64,power_law_1.2,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,64,power_law_1.2,0.029721599817276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,64,power_law_1.2,0.0342848002910614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,64,power_law_1.2,0.034380799531936644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,64,power_law_1.2,0.034643200039863584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,64,power_law_1.2,0.03758719861507416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,64,power_law_1.2,0.03993600010871887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,64,power_law_1.2,0.04547840058803558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,64,power_law_1.2,0.05055360198020935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,64,power_law_1.2,0.06088320016860962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,64,power_law_1.2,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.2,0.18950400352478028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.2,0.2388672113418579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,64,power_law_1.2,0.09859200119972229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.2,0.31636478900909426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,64,balanced,0.023344000180562336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,64,balanced,0.023157333334287006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,64,balanced,0.021386665602525074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,64,balanced,0.025050667424996693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,64,balanced,0.023344000180562336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,64,balanced,0.02554133286078771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,64,balanced,0.026821332673231762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,64,balanced,0.0269813338915507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,64,power_law_1.2,0.1154047966003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,64,balanced,0.025466665625572205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,64,balanced,0.025621332228183746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,64,balanced,0.027471999327341717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,64,balanced,0.02784000088771184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,64,balanced,0.027477333943049114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,64,balanced,0.02959999938805898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,64,balanced,0.033333333830038704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,64,balanced,0.032399999598662056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,64,balanced,0.03470933437347412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,64,balanced,0.03385066737731298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,64,balanced,0.03549333413441976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,64,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,64,balanced,0.037920000652472176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,64,balanced,0.03949866692225138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,64,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,64,balanced,0.04780800143877665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,64,balanced,0.04985066751639048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,64,balanced,0.05995733539263407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,64,balanced,0.07423999905586243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.2,0.40392317771911623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,1,balanced,0.1090666651725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,1,balanced,0.11142933368682861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,1,balanced,0.1325813333193461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,1,balanced,0.1809813380241394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,1,balanced,0.2855786681175232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,1,balanced,0.2881279985109965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,1,balanced,0.2916160027186076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,1,balanced,0.29280000925064087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,1,balanced,0.29579200347264606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,1,balanced,0.29969600836435956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,1,balanced,0.29917333523432416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,1,balanced,0.3092000087102254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,1,balanced,0.31016000111897785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,1,balanced,0.3150720000267029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,1,balanced,0.31674667199452716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,1,balanced,0.32597867647806805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,1,balanced,0.3362239996592204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,1,balanced,0.3664106527964274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,1,balanced,0.4012800057729085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,1,balanced,0.46756800015767414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,1,balanced,0.5248746474583944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,64,power_law_1.2,0.16565120220184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,1,balanced,0.7443573474884033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,1,balanced,0.8670720259348551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,1,balanced,1.2208853562672932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,1,balanced,1.6002240180969238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,1,balanced,2.3207200368245444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,1,balanced,3.0676533381144204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,4,power_law_1.01,0.030828800797462464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,4,power_law_1.01,0.029471999406814574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,4,power_law_1.01,0.029139199852943422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,4,power_law_1.01,0.031462401151657104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,4,power_law_1.01,0.031590399146080014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,4,power_law_1.01,0.033939200639724734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,4,power_law_1.01,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,4,power_law_1.01,0.03314560055732727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,4,power_law_1.01,0.033766400814056394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,4,power_law_1.01,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,4,power_law_1.01,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,4,power_law_1.01,0.034918400645256045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,4,power_law_1.01,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,4,power_law_1.01,0.03723520040512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,64,power_law_1.2,0.19528959989547728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,4,power_law_1.01,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,4,power_law_1.01,0.04455040097236633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,4,power_law_1.01,0.04524160027503967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,4,power_law_1.01,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.2,0.6451519966125489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,4,power_law_1.01,0.058303999900817874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,4,power_law_1.01,0.07237759828567505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,4,power_law_1.01,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,4,power_law_1.01,0.11504640579223632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,4,power_law_1.01,0.14839040040969848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,4,power_law_1.01,0.19591039419174194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,4,power_law_1.01,0.27654399871826174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,4,power_law_1.01,0.38774399757385253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.2,0.8333439826965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,4,power_law_1.01,0.46379518508911133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,4,4,balanced,0.04624533156553904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,4,4,balanced,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,4,4,balanced,0.05115733544031779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,4,4,balanced,0.07286400099595387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,4,4,balanced,0.09588266412417094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,4,4,balanced,0.14402666687965393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,4,4,balanced,0.1449013352394104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,4,4,balanced,0.14631999532381693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,4,4,balanced,0.14833066860834757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,4,4,balanced,0.15095999836921692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,4,4,balanced,0.149807999531428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,4,4,balanced,0.15340800086657205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,4,4,balanced,0.15683733423550925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,4,4,balanced,0.15851199626922607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,4,4,balanced,0.1672053337097168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,4,4,balanced,0.16839466492335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,4,4,balanced,0.17331733306248984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,4,4,balanced,0.19289066394170126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,4,4,balanced,0.20441599686940512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,4,4,balanced,0.23624000946680704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,4,4,balanced,0.26734934250513714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,4,4,balanced,0.3364959955215454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,4,4,balanced,0.3960053523381551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,4,4,balanced,0.5452853441238403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,4,4,balanced,0.6646026770273844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,4,4,balanced,0.9322933355967203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,4,4,balanced,1.1852266788482666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,2,power_law_1.01,0.06484479904174804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,2,power_law_1.01,0.06732159852981567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,2,power_law_1.01,0.07779840230941773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,2,power_law_1.01,0.08658559918403626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,2,power_law_1.01,0.08844799995422363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,2,power_law_1.01,0.08945279717445373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,2,power_law_1.01,0.09076480269432068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,2,power_law_1.01,0.09105920195579528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,2,power_law_1.01,0.09066240191459655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,2,power_law_1.01,0.09084799885749817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,2,power_law_1.01,0.09076480269432068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,2,power_law_1.01,0.10679680109024048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,2,power_law_1.01,0.09939200282096863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,2,power_law_1.01,0.11890560388565063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,2,power_law_1.01,0.12869759798049926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,2,power_law_1.01,0.12885119915008544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,2,power_law_1.01,0.1422271966934204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,2,power_law_1.01,0.17553919553756714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,2,power_law_1.01,0.1951167941093445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,2,power_law_1.01,0.2717375993728638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,2,power_law_1.01,0.3329983949661255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,2,power_law_1.01,0.43785600662231444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,2,power_law_1.01,0.5811200141906738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,8,power_law_1.01,0.034867200255393985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,8,power_law_1.01,0.034796801209449765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,8,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,8,power_law_1.01,0.036473599076271054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,8,power_law_1.01,0.03638400137424469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,8,power_law_1.01,0.036800000071525577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,8,power_law_1.01,0.03671039938926697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,8,power_law_1.01,0.036857599020004274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,8,power_law_1.01,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,8,power_law_1.01,0.03687680065631867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,8,power_law_1.01,0.037145599722862244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,8,power_law_1.01,0.03842560052871704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,8,power_law_1.01,0.038540801405906676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,8,power_law_1.01,0.03966079950332642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,8,power_law_1.01,0.04478079974651337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,8,power_law_1.01,0.045817598700523376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,8,power_law_1.01,0.04749439954757691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,8,power_law_1.01,0.05591679811477661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.01,0.06062080264091492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.01,0.07080960273742676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.01,0.07810559868812561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,2,power_law_1.01,0.8029120445251465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.01,0.10490880012512208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.01,0.1297919988632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,16,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,16,balanced,0.033589333295822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,16,balanced,0.03408533334732056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,16,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,16,balanced,0.04145599901676178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,16,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,16,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,16,balanced,0.041893333196640015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,16,balanced,0.04177066683769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,16,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,16,balanced,0.04333333174387614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,16,balanced,0.04333333174387614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,16,balanced,0.04398400088151296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,16,balanced,0.04443199932575226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,16,balanced,0.04600533346335093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,16,balanced,0.04764266808827718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,16,balanced,0.048021331429481506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,16,balanced,0.05147199829419454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,16,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,16,balanced,0.0580213318268458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,16,balanced,0.06001066664854685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,16,balanced,0.07028266787528992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,16,balanced,0.07666666805744171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,16,balanced,0.10174399614334106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,16,balanced,0.1251626710096995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,16,balanced,0.1623306671778361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,16,balanced,0.20164267222086588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,2,power_law_1.01,0.9427071571350097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.01,0.1697983980178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,2,power_law_1.01,1.6179008483886719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.01,0.22007040977478026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,2,power_law_1.01,2.1291391372680666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.01,0.32561280727386477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,4,balanced,0.02775999903678894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,4,balanced,0.02922133356332779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,4,balanced,0.029391999046007793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,4,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,4,balanced,0.03752533346414566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,4,balanced,0.03987200061480204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,4,balanced,0.0395359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,4,balanced,0.03807466725508372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,4,balanced,0.03967999915281931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,4,balanced,0.04011733333269755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,4,balanced,0.0397119993964831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,4,balanced,0.041477332512537636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,4,balanced,0.0414986660083135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,4,balanced,0.04180799921353658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,4,balanced,0.04782933493455251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,4,power_law_1.01,0.061887997388839724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,4,power_law_1.01,0.08098559975624084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,4,power_law_1.01,0.08810240030288696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,4,power_law_1.01,0.10532480478286743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,4,power_law_1.01,0.11568000316619872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,4,power_law_1.01,0.1315840005874634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,4,balanced,0.04606399933497111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,4,power_law_1.01,0.15000959634780883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,4,power_law_1.01,0.15829119682312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,4,power_law_1.01,0.15501439571380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,4,power_law_1.01,0.15545599460601806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,4,power_law_1.01,0.152128005027771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,4,power_law_1.01,0.16291840076446534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,4,power_law_1.01,0.1677567958831787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,4,balanced,0.053685332338015236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,4,balanced,0.056159997979799904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,4,balanced,0.062021334966023765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,4,balanced,0.08275733391443889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,4,balanced,0.09294933080673218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,4,balanced,0.12004799644152324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,4,balanced,0.156741331020991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,4,balanced,0.23468265930811563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,4,balanced,0.28121066093444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,4,balanced,0.4277120033899943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,4,balanced,0.5252960125605265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,4,power_law_1.01,0.1738816022872925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,4,power_law_1.01,0.1745792031288147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,4,power_law_1.01,0.18899199962615967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,4,power_law_1.01,0.1982848048210144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,4,power_law_1.01,0.22228479385375977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,4,power_law_1.01,0.24714241027832032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,4,power_law_1.01,0.28627839088439944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,4,power_law_1.01,0.32178559303283694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,4,power_law_1.01,0.39226880073547366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,4,power_law_1.01,0.48322558403015137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,4,power_law_1.01,0.6302591800689697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,4,power_law_1.01,0.8292863845825196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,4,power_law_1.01,1.1576704025268554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,4,power_law_1.01,1.5668928146362304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.01,0.4199679851531982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,8,balanced,0.04353066782156626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,8,balanced,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,8,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,8,balanced,0.05003199974695841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,8,balanced,0.053823997577031456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,8,balanced,0.07527466615041097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,8,balanced,0.07798400024573009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,8,balanced,0.07871466875076294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,8,balanced,0.07817600170771281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,8,balanced,0.07931200166543324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,8,balanced,0.07941866914431255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,8,balanced,0.08105066418647766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,8,balanced,0.08270933230717976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,8,balanced,0.08480532964070638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,8,balanced,0.0890880028406779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,8,balanced,0.09109333157539368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,8,balanced,0.0944160024325053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,8,balanced,0.10150399804115295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,8,balanced,0.10693867007891338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,8,balanced,0.12154666582743327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,8,balanced,0.13619732856750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,8,balanced,0.17292799552281699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,8,balanced,0.2032853364944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,8,balanced,0.2666880091031392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,8,balanced,0.33051733175913495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,8,balanced,0.4562559922536214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,8,balanced,0.5790026585261027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,32,power_law_1.2,0.0359360009431839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,32,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,32,power_law_1.2,0.03290880024433136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,32,power_law_1.2,0.035692799091339114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,32,power_law_1.2,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,32,power_law_1.2,0.03834240138530731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,32,power_law_1.2,0.0377023994922638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,32,power_law_1.2,0.03835520148277283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,32,power_law_1.2,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,32,power_law_1.2,0.03875199854373932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,32,power_law_1.2,0.038822400569915774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,8,balanced,0.06243733565012614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,8,balanced,0.062080000837643944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,8,balanced,0.06406400104363759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,8,balanced,0.07946666578451793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,8,balanced,0.0972160001595815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,8,balanced,0.14197333653767905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,8,balanced,0.14156267046928406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,8,balanced,0.14225600163141885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,8,balanced,0.14506666858990988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,8,balanced,0.1425973375638326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,8,balanced,0.1439253290494283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,8,balanced,0.14594133694966635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,32,power_law_1.2,0.0412992000579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,8,balanced,0.1449013352394104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,8,balanced,0.1474453310171763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,32,power_law_1.2,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,8,balanced,0.1513759990533193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,8,balanced,0.15202132860819498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,8,balanced,0.15607999761899313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,8,balanced,0.16331733266512552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,8,balanced,0.16884267330169678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,8,balanced,0.188426673412323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,8,balanced,0.20598934094111124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,8,balanced,0.23452266057332358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,8,balanced,0.26444800694783527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,8,balanced,0.3561866680781047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,8,balanced,0.4193386634190877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,8,balanced,0.5873013337453207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,8,balanced,0.7642880280812582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,32,power_law_1.2,0.04630399942398071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,32,power_law_1.2,0.05280640125274658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,32,power_law_1.2,0.05011839866638183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,32,power_law_1.2,0.05733759999275208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,32,power_law_1.2,0.06791679859161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,32,power_law_1.2,0.08099200129508972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,32,power_law_1.2,0.10292479991912842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,32,power_law_1.2,0.1341248035430908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,32,power_law_1.2,0.16837120056152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,32,power_law_1.2,0.20612480640411376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,32,power_law_1.2,0.33810560703277587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,32,power_law_1.2,0.4522047996520996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,64,power_law_1.2,0.050432002544403075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,32,power_law_1.2,0.7200064182281494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,64,power_law_1.2,0.04063999950885773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,64,power_law_1.2,0.040889599919319154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,64,power_law_1.2,0.04302720129489899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,64,power_law_1.2,0.042719998955726625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,64,power_law_1.2,0.04307839870452881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,64,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,64,power_law_1.2,0.045433598756790164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,64,power_law_1.2,0.04535680115222931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,64,power_law_1.2,0.045721599459648134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,64,power_law_1.2,0.04700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,64,power_law_1.2,0.048076799511909483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,64,power_law_1.2,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,64,power_law_1.2,0.050323200225830075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,64,power_law_1.2,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,64,power_law_1.2,0.05694720149040222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,64,power_law_1.2,0.05973119735717773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,64,power_law_1.2,0.06653439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,64,power_law_1.2,0.07080960273742676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,64,power_law_1.2,0.08550400137901307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,64,power_law_1.2,0.09704959988594056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,64,power_law_1.2,0.12961280345916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,64,power_law_1.2,0.1590016007423401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,64,power_law_1.2,0.22430078983306884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,64,power_law_1.2,0.30076799392700193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,64,power_law_1.2,0.4402048110961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,64,power_law_1.2,0.5661119937896728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,2,power_law_1.01,0.03150720000267029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,2,power_law_1.01,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,2,power_law_1.01,0.0374208003282547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,2,power_law_1.01,0.045440000295639035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,2,power_law_1.01,0.050271999835968015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,2,power_law_1.01,0.06477439999580384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,2,power_law_1.01,0.06574079990386963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,2,power_law_1.01,0.0680191993713379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,2,power_law_1.01,0.07244799733161926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,2,power_law_1.01,0.07269759774208069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,2,power_law_1.01,0.07453439831733703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,2,power_law_1.01,0.07822719812393189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,2,power_law_1.01,0.07989760041236878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,2,power_law_1.01,0.08588799834251404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,2,power_law_1.01,0.09162880182266235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,2,power_law_1.01,0.09653120040893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,2,power_law_1.01,0.1052672028541565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,2,power_law_1.01,0.12300159931182861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,2,power_law_1.01,0.14220800399780273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,2,power_law_1.01,0.1797824025154114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,2,power_law_1.01,0.20929279327392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,2,power_law_1.01,0.27098879814147947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,2,power_law_1.01,0.3630143880844116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,2,power_law_1.01,0.4995391845703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,2,power_law_1.01,0.690777587890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,2,power_law_1.01,0.9019840240478516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,2,power_law_1.01,1.228435230255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,32,power_law_1.2,0.8550080299377442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,1,power_law_1.01,0.05077760219573975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,1,power_law_1.01,0.06494719982147217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,1,power_law_1.01,0.07585279941558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,1,power_law_1.01,0.11477119922637939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,1,power_law_1.01,0.14805760383605956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,1,power_law_1.01,0.1844928026199341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,1,power_law_1.01,0.24008960723876954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,1,power_law_1.01,0.24788479804992675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,1,power_law_1.01,0.2543168067932129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,1,power_law_1.01,0.261843204498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,1,power_law_1.01,0.27614080905914307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,1,power_law_1.01,0.2791167974472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,1,power_law_1.01,0.28444159030914307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,1,power_law_1.01,0.2983871936798096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,1,power_law_1.01,0.30734078884124755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,1,power_law_1.01,0.31674239635467527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,1,power_law_1.01,0.330182409286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,1,power_law_1.01,0.36215040683746336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,1,power_law_1.01,0.3748863935470581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,1,power_law_1.01,0.4407616138458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,1,power_law_1.01,0.5001728057861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,1,power_law_1.01,0.6192255973815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,1,power_law_1.01,0.7039936065673829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,1,power_law_1.01,0.9161919593811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,1,power_law_1.01,1.1491264343261718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,8,balanced,0.030058667063713074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,8,balanced,0.029829333225886028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,1,power_law_1.01,1.5782272338867187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,8,balanced,0.030085332691669464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,8,balanced,0.03152533372243246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,8,balanced,0.03194666653871536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,8,balanced,0.032245332996050514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,8,balanced,0.03181333343187968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,8,balanced,0.03215466688076655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,8,balanced,0.032511999209721885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,8,balanced,0.0323786661028862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,8,balanced,0.03400533397992452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,8,balanced,0.034261333445707955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,8,balanced,0.035973332822322845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,8,balanced,0.03629866739114126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,8,balanced,0.038319999972979225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,8,balanced,0.04021333406368891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,8,balanced,0.04206933577855428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,8,balanced,0.042837331692377724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,8,balanced,0.04295999805132548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,8,balanced,0.05253866811593374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,1,power_law_1.01,2.00130558013916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,8,balanced,0.05481066803137461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,8,balanced,0.06733866532643636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,8,balanced,0.09244799613952637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,8,balanced,0.13523200154304504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,8,balanced,0.1585493286450704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,8,balanced,0.2198186715443929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,8,balanced,0.2638079921404521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,8,power_law_1.01,0.04958719909191132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,8,power_law_1.01,0.07713279724121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,8,power_law_1.01,0.0945792019367218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,8,power_law_1.01,0.10527360439300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,8,power_law_1.01,0.12212480306625366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,8,power_law_1.01,0.10861439704895019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,8,power_law_1.01,0.12343039512634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,8,power_law_1.01,0.11934080123901367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,8,power_law_1.01,0.12378879785537719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,8,power_law_1.01,0.13227519989013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,8,power_law_1.01,0.13659520149230958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,8,power_law_1.01,0.14224640130996705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,8,power_law_1.01,0.15040639638900757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,8,power_law_1.01,0.14588160514831544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,8,power_law_1.01,0.15199359655380248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,8,power_law_1.01,0.16550400257110595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,8,power_law_1.01,0.16176639795303344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,8,power_law_1.01,0.17495039701461793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.01,0.2497920036315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.01,0.25730559825897215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.01,0.29015040397644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.01,0.3196671962738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.01,0.3880192041397095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.01,0.4729663848876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.01,0.6053247928619385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.01,0.8704319953918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.01,1.0058239936828612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,32,balanced,0.04610133171081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,32,balanced,0.046037331223487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,32,balanced,0.047797332207361855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,32,balanced,0.052202666799227394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,32,balanced,0.0536106675863266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,32,balanced,0.06009600063165029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,32,balanced,0.058229332168896995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,32,balanced,0.06085333228111267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,32,balanced,0.060122668743133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,32,balanced,0.06031466523806254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,32,balanced,0.060005332032839455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,32,balanced,0.062090665102005005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,32,balanced,0.06373333434263866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,32,balanced,0.06213866670926412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,32,balanced,0.068271999557813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,32,balanced,0.06832533578077953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,32,balanced,0.07021333277225494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,32,balanced,0.07458133498827617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,32,balanced,0.08055999875068665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,32,balanced,0.08935999870300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,32,balanced,0.09862400094668071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,32,balanced,0.11910933256149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,32,balanced,0.13638933499654135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,32,balanced,0.18743467330932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,32,balanced,0.23035200436909994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,32,balanced,0.31193600098292035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,32,balanced,0.39631466070810956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,256,power_law_1.2,0.051571202278137204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,256,power_law_1.2,0.053939199447631835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,256,power_law_1.2,0.05260800123214722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,256,power_law_1.2,0.05550720095634461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,256,power_law_1.2,0.05546879768371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,256,power_law_1.2,0.056934398412704465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,256,power_law_1.2,0.05742080211639404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,256,power_law_1.2,0.05634559988975525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,256,power_law_1.2,0.05737599730491638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,256,power_law_1.2,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,256,power_law_1.2,0.0568448007106781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,8,balanced,0.04404800136884054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,256,power_law_1.2,0.057580798864364624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,8,balanced,0.04365866879622141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,8,balanced,0.049685334165891014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,256,power_law_1.2,0.058508801460266116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,8,balanced,0.07666666805744171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,8,balanced,0.10206933816274007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,256,power_law_1.2,0.05736960172653198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,8,balanced,0.1184213360150655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,8,balanced,0.11829333504041036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,256,power_law_1.2,0.06044800281524658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,256,power_law_1.2,0.06112639904022217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,256,power_law_1.2,0.06509439945220948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,256,power_law_1.2,0.0681984007358551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,8,balanced,0.12000532944997151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,8,balanced,0.12034133076667786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,8,balanced,0.1216319998105367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,8,balanced,0.12271466851234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,8,balanced,0.1237600048383077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,8,balanced,0.1271573305130005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,8,balanced,0.12739200393358865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,8,balanced,0.1365653375784556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,8,balanced,0.1374613344669342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,8,balanced,0.14016000429789224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,8,balanced,0.15652267138163248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,8,balanced,0.16265066464742026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,256,power_law_1.2,0.07491840124130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,8,balanced,0.1950613260269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,256,power_law_1.2,0.08632959723472595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,256,power_law_1.2,0.10394879579544067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,256,power_law_1.2,0.13461120128631593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,8,balanced,0.20819199085235596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,256,power_law_1.2,0.16160000562667848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,8,balanced,0.2711893320083618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,256,power_law_1.2,0.23280000686645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,8,balanced,0.2942240039507548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,8,balanced,0.4294826587041219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,8,balanced,0.5023946762084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,8,balanced,0.6945546468098959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,8,balanced,0.8899892965952555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,256,power_law_1.2,0.2894399881362915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,256,power_law_1.2,0.4388160228729248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,256,power_law_1.2,0.6006976127624511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,128,power_law_1.01,0.051123201847076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,128,power_law_1.01,0.05013759732246399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,128,power_law_1.01,0.050169599056243894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,128,power_law_1.01,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,128,power_law_1.01,0.05364480018615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,128,power_law_1.01,0.056441599130630495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,128,power_law_1.01,0.06321920156478882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,128,power_law_1.01,0.06310399770736694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,32,power_law_1.2,0.05793920159339905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,32,power_law_1.2,0.042483198642730716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,32,power_law_1.2,0.0430400013923645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,32,power_law_1.2,0.04478079974651337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,32,power_law_1.2,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,32,power_law_1.2,0.0471807986497879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,32,power_law_1.2,0.04837760031223297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,32,power_law_1.2,0.048793599009513855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,32,power_law_1.2,0.04921599924564361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,32,power_law_1.2,0.050860798358917235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,32,power_law_1.2,0.05060480237007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,128,power_law_1.01,0.06711040139198303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,32,power_law_1.2,0.05144320130348205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,128,power_law_1.01,0.06738560199737549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,32,power_law_1.2,0.05186560153961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,32,power_law_1.2,0.055980801582336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,32,power_law_1.2,0.0611519992351532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,32,power_law_1.2,0.06275200247764587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,32,power_law_1.2,0.06724479794502258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,32,power_law_1.2,0.0823743999004364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,32,power_law_1.2,0.09038079977035522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,32,power_law_1.2,0.11768959760665894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,32,power_law_1.2,0.11819519996643066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,32,power_law_1.2,0.15164159536361693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,128,power_law_1.01,0.06473600268363952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,32,power_law_1.2,0.1972480058670044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,32,power_law_1.2,0.23928959369659425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,2,power_law_1.2,0.08282880187034607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,32,power_law_1.2,0.31960320472717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,2,power_law_1.2,0.08584319949150085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,2,power_law_1.2,0.10587519407272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,2,power_law_1.2,0.13539199829101561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,2,power_law_1.2,0.14094719886779786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,128,power_law_1.01,0.06812160015106201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,128,power_law_1.01,0.07196159958839417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,128,power_law_1.01,0.07018880248069763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,128,power_law_1.01,0.08166400194168091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,128,power_law_1.01,0.08476160168647766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,128,power_law_1.01,0.09018239974975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,2,power_law_1.2,0.14164479970932006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,128,power_law_1.01,0.10017919540405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,2,power_law_1.2,0.14171520471572877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,2,power_law_1.2,0.14261759519577027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,2,power_law_1.2,0.14243199825286865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,2,power_law_1.2,0.14307199716567992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,2,power_law_1.2,0.14425599575042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,128,power_law_1.01,0.11025919914245605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,2,power_law_1.2,0.1722432017326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,128,power_law_1.01,0.13081599473953248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,2,power_law_1.2,0.17126400470733644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,128,power_law_1.01,0.15292160511016845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,2,power_law_1.2,0.17520639896392823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,2,power_law_1.2,0.1898368000984192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,2,power_law_1.2,0.18743679523468018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,2,power_law_1.2,0.2174015998840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,128,power_law_1.01,0.20523519515991212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,2,power_law_1.2,0.288921594619751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,128,power_law_1.01,0.24921600818634032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,128,power_law_1.01,0.35425920486450196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,2,power_law_1.2,0.3444927930831909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,128,power_law_1.01,0.455295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,128,power_law_1.01,0.6381440162658691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,128,power_law_1.01,0.9574591636657714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,2,power_law_1.2,0.40049281120300295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,2,power_law_1.2,0.5009280204772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,32,power_law_1.2,0.535807991027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,4,power_law_1.01,0.06117759943008423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,4,power_law_1.01,0.06821119785308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,4,power_law_1.01,0.08271359801292419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,4,power_law_1.01,0.09163519740104675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,4,power_law_1.01,0.10888960361480712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,4,power_law_1.01,0.12345600128173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,4,power_law_1.01,0.1282495975494385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,4,power_law_1.01,0.1417088031768799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,32,power_law_1.2,0.6055168151855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,4,power_law_1.01,0.1423424005508423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,4,power_law_1.01,0.1418239951133728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,4,power_law_1.01,0.1459391951560974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,2,power_law_1.2,0.6984767913818359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,4,power_law_1.01,0.14520959854125975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,4,power_law_1.01,0.15442559719085694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,2,power_law_1.2,0.8265727996826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,4,power_law_1.01,0.1500864028930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,4,power_law_1.01,0.1602944016456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,4,power_law_1.01,0.16462719440460205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,4,power_law_1.01,0.17694079875946045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,4,power_law_1.01,0.1877120018005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,4,power_law_1.01,0.2134783983230591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,4,power_law_1.01,0.24302079677581787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,4,power_law_1.01,0.27491838932037355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,4,power_law_1.01,0.35656960010528566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,4,power_law_1.01,0.44268159866333007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,4,power_law_1.01,0.5805503845214843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,4,power_law_1.01,0.7619071960449219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,4,power_law_1.01,1.062880039215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,4,power_law_1.01,1.3172543525695801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,2,power_law_1.2,1.4129023551940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,2,power_law_1.2,1.9210239410400392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,1,power_law_1.2,0.055225598812103274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,1,power_law_1.2,0.07401599884033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,1,power_law_1.2,0.10675200223922729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,1,power_law_1.2,0.15415680408477783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,1,power_law_1.2,0.19695359468460083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,1,power_law_1.2,0.32099199295043945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,1,power_law_1.2,0.33680000305175783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,1,power_law_1.2,0.3573184013366699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,1,power_law_1.2,0.3677119970321655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,1,power_law_1.2,0.3659456014633179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,1,power_law_1.2,0.37546238899230955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,1,power_law_1.2,0.3836159944534302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,1,power_law_1.2,0.3831295967102051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,1,power_law_1.2,0.39537920951843264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,1,power_law_1.2,0.40891518592834475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,1,power_law_1.2,0.4190207958221436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,1,power_law_1.2,0.4386943817138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,1,power_law_1.2,0.48427519798278806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,1,power_law_1.2,0.5264959812164307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,1,power_law_1.2,0.6095168113708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,1,power_law_1.2,0.6426815986633301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,1,power_law_1.2,0.7869760036468506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,1,power_law_1.2,0.9295552253723145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,1,power_law_1.2,1.2119872093200683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,1,power_law_1.2,1.5105343818664552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,2,power_law_1.2,2.6194496154785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,1,power_law_1.2,2.0635007858276366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,1,power_law_1.2,2.6308095932006834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,2,power_law_1.2,3.4336318969726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,2,power_law_1.2,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,2,power_law_1.2,0.05367680191993714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,2,power_law_1.2,0.06392959952354431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,2,power_law_1.2,0.0809216022491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,2,power_law_1.2,0.10106240510940552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,2,power_law_1.2,0.11221760511398315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,2,power_law_1.2,0.14872959852218628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,2,power_law_1.2,0.15203839540481567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,32,power_law_1.01,0.017612800002098083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,2,power_law_1.2,0.15134719610214234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,32,power_law_1.01,0.017132799327373504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,2,power_law_1.2,0.1565119981765747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,32,power_law_1.01,0.017158399522304534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,2,power_law_1.2,0.1712448000907898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,2,power_law_1.2,0.17505919933319092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,2,power_law_1.2,0.17608959674835206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,32,power_law_1.01,0.017740799486637114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,32,power_law_1.01,0.018592000007629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,2,power_law_1.2,0.19545600414276124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,1,power_law_1.01,0.04713599979877472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,2,power_law_1.2,0.19616639614105225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,32,power_law_1.01,0.02194560021162033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,2,power_law_1.2,0.2052608013153076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,32,power_law_1.01,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,2,power_law_1.2,0.23011839389801025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,2,power_law_1.2,0.26457600593566893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,2,power_law_1.2,0.282207989692688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,2,power_law_1.2,0.3412224054336548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,2,power_law_1.2,0.37061119079589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,2,power_law_1.2,0.46272640228271483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,32,power_law_1.01,0.0324864000082016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,32,power_law_1.01,0.0315200001001358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,2,power_law_1.2,0.5968704223632812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,32,power_law_1.01,0.03234559893608093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,2,power_law_1.2,0.7691008090972901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,1,power_law_1.01,0.051744002103805545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,1,power_law_1.01,0.05857279896736145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,1,power_law_1.01,0.07319039702415467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,1,power_law_1.01,0.06915839910507202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,32,power_law_1.01,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,32,power_law_1.01,0.0318015992641449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,2,power_law_1.2,0.9808256149291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,2,power_law_1.2,1.379923152923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,1,power_law_1.01,0.07166720032691956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,2,power_law_1.2,1.8089664459228516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,1,power_law_1.01,0.070278400182724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,1,power_law_1.01,0.06915199756622314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,32,power_law_1.01,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,1,power_law_1.01,0.06843519806861878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,32,power_law_1.01,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,1,power_law_1.01,0.06842240095138549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,32,power_law_1.01,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,1,power_law_1.01,0.07088000178337098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,32,power_law_1.01,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,32,power_law_1.01,0.03187839984893799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,32,power_law_1.01,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,32,power_law_1.01,0.03544960021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,32,power_law_1.01,0.04615679979324341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,32,power_law_1.01,0.04902400076389313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,32,power_law_1.01,0.05937280058860779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,32,power_law_1.01,0.07170559763908387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,1,power_law_1.01,0.07246720194816589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,32,power_law_1.01,0.092467200756073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,1,power_law_1.01,0.07511039972305297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,32,power_law_1.01,0.1184831976890564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,1,power_law_1.01,0.07967360019683838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,32,power_law_1.01,0.16264959573745727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,1,power_law_1.01,0.08704000115394592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,32,power_law_1.01,0.20095360279083252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,1,power_law_1.01,0.08886399865150452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,1,power_law_1.01,0.09323520064353943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,1,power_law_1.01,0.11047680377960205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.01,0.12432639598846436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.01,0.15502079725265502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.01,0.17667839527130128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.01,0.23759360313415528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.01,0.29400320053100587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.01,0.4062016010284424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.01,0.5265088081359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.01,0.7469823837280274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,8,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,8,balanced,0.03358400116364161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,8,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,8,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,8,balanced,0.0373333344856898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,8,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,8,balanced,0.035546667873859406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,8,balanced,0.03758399933576584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,8,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,8,balanced,0.03734933336575826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,8,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,8,balanced,0.03719999889532725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,8,balanced,0.03938666731119156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,8,balanced,0.039317332208156586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,8,balanced,0.04141333450873693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,8,balanced,0.04321066538492838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,8,balanced,0.04424533247947693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,8,balanced,0.04785599807898203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,8,balanced,0.04783466458320618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,8,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,8,balanced,0.05783466498057047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,8,balanced,0.06807466844717662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,8,balanced,0.0763679991165797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,8,balanced,0.10307199756304423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,8,balanced,0.12019200126330058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,8,balanced,0.16101333498954773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,8,balanced,0.2002293268839518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,1,power_law_1.01,0.1047808051109314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.01,0.9850303649902343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,1,power_law_1.01,0.15539200305938722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,1,power_law_1.01,0.25395200252532957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,1,power_law_1.01,0.44898557662963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,1,power_law_1.01,0.7601088047027588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,1,power_law_1.01,1.0894463539123536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,1,power_law_1.01,1.2573439598083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,32,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,1,power_law_1.01,1.6343360900878907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,1,power_law_1.01,1.7069440841674806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,1,power_law_1.01,1.746668815612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,32,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,32,balanced,0.0394400010506312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,32,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,32,balanced,0.043509334325790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,32,balanced,0.043791999419530235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,32,balanced,0.04550399879614512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,32,balanced,0.04584000011285146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,32,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,32,balanced,0.04582933088143667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,32,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,32,balanced,0.04796266555786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,32,balanced,0.04797333478927612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,32,balanced,0.05146133402983347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,32,balanced,0.05340266724427541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,32,balanced,0.054085334142049156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,32,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,32,balanced,0.061861331264177956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,32,balanced,0.06406400104363759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,32,balanced,0.07212266822655995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,32,balanced,0.07787199815114339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,32,balanced,0.09502399961153667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,1,power_law_1.01,1.8151615142822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,32,balanced,0.10917333761850993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,32,balanced,0.14437333742777506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,32,balanced,0.17829332749048868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,32,balanced,0.24231467644373575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,32,balanced,0.3096800049146016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,1,power_law_1.01,1.9046655654907227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,1,power_law_1.01,1.9502208709716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,1,power_law_1.01,1.9705535888671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,1,power_law_1.01,2.018489646911621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,1,power_law_1.01,2.1038272857666014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,1,power_law_1.01,2.1257856369018553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,1,power_law_1.01,2.261324882507324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,1,power_law_1.01,2.4842496871948243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,1,power_law_1.01,2.7543872833251952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,1,power_law_1.01,2.8703487396240233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,1,power_law_1.01,3.400953674316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,1,power_law_1.01,3.332051086425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,1,power_law_1.01,3.9578495025634766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,1,power_law_1.01,4.753529739379883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,1,power_law_1.01,6.017689514160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,1,power_law_1.01,8.002777862548829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,16,power_law_1.2,0.10309120416641235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,1,power_law_1.01,0.02011519968509674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,16,power_law_1.2,0.10131839513778687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,1,power_law_1.01,0.020326399803161622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,1,power_law_1.01,0.02022400051355362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,1,power_law_1.01,0.02134400010108948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,1,power_law_1.01,0.02455040067434311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,1,power_law_1.01,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,1,power_law_1.01,0.02861439883708954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,1,power_law_1.01,0.030291199684143066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,1,power_law_1.01,0.026739200949668883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,1,power_law_1.01,0.030323201417922975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,1,power_law_1.01,0.031999999284744264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,1,power_law_1.01,0.030419200658798218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,1,power_law_1.01,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,16,power_law_1.2,0.10848640203475952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,1,power_law_1.01,0.036601600050926206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,1,power_law_1.01,0.06170880198478699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,1,power_law_1.01,0.07664639949798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,1,power_law_1.01,0.09257599711418152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,16,power_law_1.2,0.10554879903793335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,1,power_law_1.01,0.05808640122413635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,16,power_law_1.2,0.09617919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,16,power_law_1.2,0.10122879743576049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.01,0.06639999747276307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,16,power_law_1.2,0.10894720554351807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,16,power_law_1.2,0.10807679891586304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,16,power_law_1.2,0.11306240558624267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,16,power_law_1.2,0.11276799440383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,16,power_law_1.2,0.11710079908370971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.01,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,16,power_law_1.2,0.1177791953086853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,16,power_law_1.2,0.11594239473342896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,16,power_law_1.2,0.12289919853210449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,16,power_law_1.2,0.13018239736557008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,16,power_law_1.2,0.1331712007522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.01,0.10810240507125854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,16,power_law_1.2,0.1446720004081726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,16,power_law_1.2,0.16885119676589966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,16,power_law_1.2,0.18044159412384034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,16,power_law_1.2,0.22866559028625488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,16,power_law_1.2,0.26800639629364015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,16,power_law_1.2,0.37214078903198244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,16,power_law_1.2,0.3521536111831665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,16,power_law_1.2,0.49642238616943357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.01,0.1451647996902466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,16,power_law_1.2,0.6754687786102295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,16,power_law_1.2,0.8254143714904785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.01,0.17964160442352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.01,0.25496959686279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.01,0.3266495943069458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.01,0.4727424144744873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,16,power_law_1.2,1.209331226348877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.01,0.6190080165863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,4,power_law_1.01,0.028140801191329955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,4,power_law_1.01,0.030041599273681642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,4,power_law_1.01,0.041606399416923526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,4,power_law_1.01,0.04479359984397888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,4,power_law_1.01,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,4,power_law_1.01,0.05745919942855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,4,power_law_1.01,0.057715201377868654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,4,power_law_1.01,0.05708799958229065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,4,power_law_1.01,0.060844802856445314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,4,power_law_1.01,0.06301440000534057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,4,power_law_1.01,0.06342399716377259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,4,power_law_1.01,0.0658623993396759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,4,power_law_1.01,0.06881279945373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,4,power_law_1.01,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,4,power_law_1.01,0.07867519855499268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,4,power_law_1.01,0.08236160278320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,4,power_law_1.01,0.08554239869117737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,4,power_law_1.01,0.0944703996181488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.01,0.1757823944091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.01,0.21542398929595946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.01,0.14720640182495118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.01,0.1786944031715393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.01,0.2109503984451294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.01,0.2745215892791748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.01,0.3460095882415771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,32,4,balanced,0.0386613334218661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.01,0.4868288040161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,32,4,balanced,0.03833599885304769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,32,4,balanced,0.039887999494870506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,32,4,balanced,0.04236799975236257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,32,4,balanced,0.043882668018341064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,32,4,balanced,0.043696001172065735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.01,0.6289279937744141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,32,4,balanced,0.043578664461771645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,32,4,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,32,4,balanced,0.043663998444875084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,32,4,balanced,0.043920000394185386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,32,4,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,32,4,balanced,0.04378666480382284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,32,4,balanced,0.04584000011285146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,32,4,balanced,0.04576000074545542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,32,4,balanced,0.04965866605440775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,32,4,balanced,0.0499946673711141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,32,4,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,32,4,balanced,0.06001600126425425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,32,4,balanced,0.06225066880385081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,32,4,balanced,0.07665599882602692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,32,4,balanced,0.09707199533780415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,32,4,balanced,0.12481066584587097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,32,4,balanced,0.14629333217938742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,32,4,balanced,0.2050079902013143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,32,4,balanced,0.25313599904378253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,32,4,balanced,0.3566133181254069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,32,4,balanced,0.4625973304112752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,4,power_law_1.2,0.04200319945812225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,4,power_law_1.2,0.03692800104618073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,4,power_law_1.2,0.04108160138130188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,4,power_law_1.2,0.04506239891052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,4,power_law_1.2,0.04650239944458008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,4,power_law_1.2,0.051577597856521606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,4,power_law_1.2,0.0525056004524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,4,power_law_1.2,0.05351679921150208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,4,power_law_1.2,0.05377920269966126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,4,power_law_1.2,0.05377280116081238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,4,power_law_1.2,0.05437440276145935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,4,power_law_1.2,0.059385597705841064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,4,power_law_1.2,0.05974400043487549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,4,power_law_1.2,0.06311039924621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,4,power_law_1.2,0.06974080204963684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,4,power_law_1.2,0.0721343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,4,power_law_1.2,0.07894399762153625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,4,power_law_1.2,0.09372159838676453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,4,power_law_1.2,0.09966080188751221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,4,power_law_1.2,0.1109120011329651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,4,power_law_1.2,0.12842880487442015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,4,power_law_1.2,0.17778559923171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,4,power_law_1.2,0.2184959888458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,4,power_law_1.2,0.2782655954360962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,4,power_law_1.2,0.37134079933166503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,64,power_law_1.01,0.11198079586029053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,64,power_law_1.01,0.10952960252761841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,64,power_law_1.01,0.10885759592056274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,64,power_law_1.01,0.1121791958808899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,64,power_law_1.01,0.10511360168457032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,64,power_law_1.01,0.10861439704895019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,64,power_law_1.01,0.11016960144042968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,64,power_law_1.01,0.11754239797592163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,64,power_law_1.01,0.11793919801712036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,64,power_law_1.01,0.11485439538955688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,4,power_law_1.2,0.5726912021636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,64,power_law_1.01,0.10913280248641968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,64,power_law_1.01,0.11460479497909545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,64,power_law_1.01,0.11599999666213989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,64,power_law_1.01,0.1158784031867981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,64,power_law_1.01,0.11897599697113037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,64,power_law_1.01,0.11756800413131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,64,power_law_1.01,0.12000000476837158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,64,power_law_1.01,0.14696320295333862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,64,power_law_1.01,0.16059520244598388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,64,power_law_1.01,0.1756351947784424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,64,power_law_1.01,0.1987391948699951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,64,power_law_1.01,0.25338881015777587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,64,power_law_1.01,0.30475521087646484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,64,power_law_1.01,0.3937920093536377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,64,power_law_1.01,0.49159040451049807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,64,power_law_1.01,0.7091328144073487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,64,power_law_1.01,0.9351360321044921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,4,power_law_1.2,0.7614528179168701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,4,power_law_1.01,0.04992640018463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,4,power_law_1.01,0.07106559872627258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,4,power_law_1.01,0.08120319843292237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,4,power_law_1.01,0.0976256012916565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,4,power_law_1.01,0.11652480363845825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,4,power_law_1.01,0.11070079803466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,4,power_law_1.01,0.13927680253982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,4,power_law_1.01,0.13197439908981323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,4,power_law_1.01,0.1490048050880432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,4,power_law_1.01,0.14075520038604736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,4,power_law_1.01,0.15202560424804687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,4,power_law_1.01,0.1555008053779602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,4,power_law_1.01,0.16108800172805787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,4,power_law_1.01,0.16684800386428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,4,power_law_1.01,0.17479039430618287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,4,power_law_1.01,0.18057600259780884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,4,power_law_1.01,0.18786560297012328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,4,power_law_1.01,0.21431679725646974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,4,power_law_1.01,0.2331455945968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,4,power_law_1.01,0.2743040084838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,4,power_law_1.01,0.32048640251159666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,4,power_law_1.01,0.43866238594055174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,4,power_law_1.01,0.49733757972717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,4,power_law_1.01,0.6661375999450684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,4,power_law_1.01,0.8482175827026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,4,power_law_1.01,1.2297984123229981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,4,power_law_1.01,1.619603157043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,32,2,balanced,0.07286933561166127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,32,2,balanced,0.07642666498819987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,32,2,balanced,0.08241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,32,2,balanced,0.09682133793830872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,32,2,balanced,0.1179253359635671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,32,2,balanced,0.17189866304397583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,32,2,balanced,0.2221013307571411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,32,2,balanced,0.2101973295211792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,32,2,balanced,0.21251734097798666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,32,2,balanced,0.21344000101089478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,32,2,balanced,0.2145973245302836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,32,2,balanced,0.2169706622759501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,32,2,balanced,0.22095467646916708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,32,2,balanced,0.22425599892934164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,32,2,balanced,0.23305066426595053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,32,2,balanced,0.23642132679621378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,2,balanced,0.044906665881474815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,2,balanced,0.047882666190465294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,2,balanced,0.04797333478927612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,2,balanced,0.053904001911481224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,2,balanced,0.07055999835332234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,2,balanced,0.09878933429718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,2,balanced,0.09944533308347066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,2,balanced,0.09962133566538493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,2,balanced,0.09910933176676433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,2,balanced,0.09916266798973083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,2,balanced,0.10010666648546855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,2,balanced,0.10365866621335347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,2,balanced,0.10289600491523743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,2,balanced,0.10702932874361674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,2,balanced,0.1111840009689331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,2,balanced,0.1134986678759257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,2,balanced,0.11930132905642192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,32,2,balanced,0.2458666761716207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,2,balanced,0.13150399923324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,32,2,balanced,0.27194132407506305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,32,2,balanced,0.2920266588528951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,32,2,balanced,0.34033600489298504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,32,2,balanced,0.38471468289693195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,32,2,balanced,0.4792746702829997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,32,2,balanced,0.5776533285776774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,2,balanced,0.1442453364531199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,2,balanced,0.16743467251459757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,2,balanced,0.1885706583658854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,2,balanced,0.2691466609636943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,2,balanced,0.3090239961942037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,2,balanced,0.4498186508814494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,2,balanced,0.563536008199056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,32,2,balanced,0.7880427042643229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,2,balanced,0.8092266718546549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,2,balanced,1.0540746847788494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,32,2,balanced,1.07424529393514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,32,balanced,0.020714666694402695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,32,2,balanced,1.4970026016235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,32,balanced,0.01904533306757609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,32,balanced,0.019023999571800232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,32,balanced,0.021087999145189922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,32,balanced,0.024773334463437397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,32,balanced,0.03126399964094162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,32,2,balanced,1.9729973475138347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,1,power_law_1.2,0.059680002927780154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,1,power_law_1.2,0.07956479787826538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,1,power_law_1.2,0.10191359519958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,1,power_law_1.2,0.1533951997756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,1,power_law_1.2,0.21427199840545655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,1,power_law_1.2,0.2893887996673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,1,power_law_1.2,0.33101439476013184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,1,power_law_1.2,0.4398784160614014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,1,power_law_1.2,0.4769728183746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,1,power_law_1.2,0.4841343879699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,1,power_law_1.2,0.510265588760376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,1,power_law_1.2,0.5292223930358887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,1,power_law_1.2,0.53951997756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,1,power_law_1.2,0.592736005783081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,1,power_law_1.2,0.6073023796081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,1,power_law_1.2,0.6158080101013184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,1,power_law_1.2,0.6559807777404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,1,power_law_1.2,0.7262144088745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,1,power_law_1.2,0.7904640197753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,1,power_law_1.2,0.9230655670166016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,1,power_law_1.2,1.0110848426818848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,1,power_law_1.2,1.2286656379699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,1,power_law_1.2,1.5443008422851563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,1,power_law_1.2,1.9828224182128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,1,power_law_1.2,2.2541824340820313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,1,power_law_1.2,2.9498048782348634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,1,power_law_1.2,3.8085697174072264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,2,balanced,0.041663999358812966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,2,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,2,balanced,0.05972266693909963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,2,balanced,0.06309333443641663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,2,balanced,0.0660159985224406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,2,balanced,0.06515199939409892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,2,balanced,0.06620799998442332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,2,balanced,0.06718400120735168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,2,balanced,0.06576533118883769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,2,balanced,0.06634133557478587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,2,balanced,0.06608533362547557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,2,balanced,0.06993066767851512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,2,balanced,0.06885333359241486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,2,balanced,0.06993066767851512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,2,balanced,0.07670933504899342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,2,balanced,0.0783679982026418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,2,balanced,0.081194669008255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,2,balanced,0.11326400438944499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,2,balanced,0.1090613305568695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,2,balanced,0.13370666901270548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,2,balanced,0.16944533586502075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,2,balanced,0.21961599588394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,2,balanced,0.27436800797780353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,2,balanced,0.38735465208689374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,2,balanced,0.499728004137675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,2,balanced,0.7207520008087158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,2,balanced,0.9432960351308187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,1,balanced,0.056885331869125366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,1,balanced,0.07714666426181793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,1,balanced,0.10979732871055603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,1,balanced,0.1803306738535563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,1,balanced,0.3035946687062581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,1,balanced,0.5613013505935669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,1,balanced,0.5662879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,1,balanced,0.5707146724065145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,1,balanced,0.5733493169148763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,1,balanced,0.5772906541824341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,1,balanced,0.578933318456014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,1,balanced,0.5854560136795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,1,balanced,0.5898186763127645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,1,balanced,0.5944053332010905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,1,balanced,0.6069813172022501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,1,balanced,0.6126453479131063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,1,balanced,0.6263306538263956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,1,balanced,0.6618666648864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,1,balanced,0.6909173329671224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,1,balanced,0.7577280203501383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,1,balanced,0.8306346734364828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,1,balanced,0.9347626368204752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,1,balanced,1.0555626551310222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,1,balanced,1.403231938680013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,1,balanced,1.5847573280334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,1,balanced,2.311253388722738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,1,balanced,2.7915627161661782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,8,power_law_1.2,0.15032960176467897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,8,power_law_1.2,0.24510719776153564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,8,power_law_1.2,0.30957438945770266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,8,power_law_1.2,0.379584002494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,8,power_law_1.2,0.5030335903167724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,8,power_law_1.2,0.5069568157196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,8,power_law_1.2,0.5012608051300049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,8,power_law_1.2,0.5854080200195313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,8,power_law_1.2,0.5417344093322753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,8,power_law_1.2,0.5547391891479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,8,power_law_1.2,0.5799871921539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,8,power_law_1.2,0.5875967979431153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,8,power_law_1.2,0.5895679950714111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,8,power_law_1.2,0.561356782913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,8,power_law_1.2,0.5527616024017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,8,power_law_1.2,0.5774720191955567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,8,power_law_1.2,0.6152895927429199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,8,power_law_1.2,0.6607232093811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,8,balanced,0.02349333216746648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,8,balanced,0.023498666783173878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,8,balanced,0.023423999547958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,8,power_law_1.2,0.7316544055938721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,8,balanced,0.02613866577545802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,8,balanced,0.029578665892283123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,8,balanced,0.03150933235883713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,8,balanced,0.031680000325044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,8,power_law_1.2,0.8790783882141113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,8,balanced,0.04752000172932943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,8,power_law_1.2,0.8643584251403809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,8,balanced,0.04706666866938273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,8,balanced,0.06471466521422069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,1,balanced,0.10422399640083313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,1,balanced,0.10786133011182149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,8,power_law_1.2,1.1326720237731933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,8,power_law_1.2,1.2094976425170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,8,balanced,0.02956799914439519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,8,balanced,0.029258665939172108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,8,balanced,0.029146666328112285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,8,balanced,0.03216533362865448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,8,balanced,0.033258666594823204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,8,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,8,power_law_1.2,1.542854404449463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,8,balanced,0.03708266715208689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,8,balanced,0.03953066716591517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,8,balanced,0.03956266740957896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,8,balanced,0.043925335009892784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,8,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,8,power_law_1.2,1.6411775588989257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,8,balanced,0.050479998191197716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,8,balanced,0.05418666700522105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,8,balanced,0.06543999910354614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,8,balanced,0.08083199958006541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,8,balanced,0.08822932839393616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,8,balanced,0.10977066556612651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,8,power_law_1.2,1.9233600616455078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,8,balanced,0.1336799959341685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,8,balanced,0.17672000328699747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,8,balanced,0.2075786590576172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,8,balanced,0.09231999516487122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,8,balanced,0.1349013348420461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,1,balanced,0.11981333295504253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,8,power_law_1.2,2.6144895553588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,1,balanced,0.14991466204325357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,1,balanced,0.20285866657892862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,1,balanced,0.30397866169611615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,1,balanced,0.3989866574605306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,1,balanced,0.4018933375676473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,1,balanced,0.40675731499989826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,1,balanced,0.4057600100835164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,1,balanced,0.41065065066019696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,1,balanced,0.41731735070546466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,1,balanced,0.41970666249593097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,8,balanced,0.21978666385014853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,1,balanced,0.42850132783253986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,8,balanced,0.22177600860595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,1,balanced,0.439029335975647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,1,balanced,0.4444693326950073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,1,balanced,0.46251734097798664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,1,balanced,0.5122720003128052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,1,balanced,0.5509706735610962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,1,balanced,0.633077343304952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,1,balanced,0.7078453699747721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,1,balanced,0.8838240305582682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,8,balanced,0.22241065899531046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,1,balanced,1.066426674524943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,1,balanced,1.4419679641723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,1,balanced,1.9446825981140137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,1,balanced,2.676410675048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,8,balanced,0.2241013248761495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,8,balanced,0.22477332750956217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,8,balanced,0.2258453369140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,8,balanced,0.2276266614596049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,8,balanced,0.23068799575169882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,8,balanced,0.2312426765759786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,8,balanced,0.23899734020233154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,8,balanced,0.2407360076904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,8,balanced,0.24610666433970133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,8,balanced,0.2616586685180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,8,balanced,0.2725493311882019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,8,balanced,0.2975200017293294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,8,balanced,0.32206932703653973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,8,balanced,0.3673226833343506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,8,balanced,0.40723733107248944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,8,balanced,0.5339839855829874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,8,balanced,0.6119146744410197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,8,balanced,0.8808053334554037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,8,balanced,1.0270133018493652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,1,balanced,3.593440055847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,256,power_law_1.01,0.04560000002384186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,256,power_law_1.01,0.045817598700523376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,256,power_law_1.01,0.046937599778175354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,256,power_law_1.01,0.05012480020523071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,256,power_law_1.01,0.049721598625183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,256,power_law_1.01,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,256,power_law_1.01,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,256,power_law_1.01,0.05143679976463318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,256,power_law_1.01,0.05189759731292724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,256,power_law_1.01,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,256,power_law_1.01,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,256,power_law_1.01,0.0530239999294281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,256,power_law_1.01,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,256,power_law_1.01,0.053932797908782956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,256,power_law_1.01,0.05744640231132507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,256,power_law_1.01,0.057740801572799684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,256,power_law_1.01,0.059935998916625974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,256,power_law_1.01,0.06793599724769592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,256,power_law_1.01,0.07010560035705567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,256,power_law_1.01,0.08326399922370911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,256,power_law_1.01,0.09297279715538025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,256,power_law_1.01,0.11610239744186401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,256,power_law_1.01,0.13798400163650512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,256,power_law_1.01,0.18562560081481932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,256,power_law_1.01,0.2327552080154419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,256,power_law_1.01,0.3240959882736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,256,power_law_1.01,0.417190408706665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,16,balanced,0.04167999823888143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,16,balanced,0.0440533310174942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,16,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,16,balanced,0.050010666251182556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,16,balanced,0.06157866617043813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,16,balanced,0.08443733056386311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,16,balanced,0.08475733796755473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,16,balanced,0.08475733796755473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,16,balanced,0.08523199955622356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,16,balanced,0.08648000160853068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,16,balanced,0.08627200126647949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,16,balanced,0.08844799796740214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,16,balanced,0.08994133273760478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,16,balanced,0.08865599830945332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,16,balanced,0.09481066465377808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,16,balanced,0.09640000263849895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,16,balanced,0.09779199957847595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,16,balanced,0.10678399602572124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,16,balanced,0.11236799756685893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,16,balanced,0.12480533123016357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,16,balanced,0.13987200458844504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,16,balanced,0.16523200273513794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,16,balanced,0.1904533306757609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,16,balanced,0.25353066126505536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,16,balanced,0.30223466952641803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,16,balanced,0.4217333396275838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,16,balanced,0.5338559945424398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,2,power_law_1.01,0.03201279938220978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,2,power_law_1.01,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,2,power_law_1.01,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,2,power_law_1.01,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,2,power_law_1.01,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,2,power_law_1.01,0.0757695972919464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,2,power_law_1.01,0.07798399925231933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,2,power_law_1.01,0.08035839796066284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,2,power_law_1.01,0.07802240252494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,2,power_law_1.01,0.08068479895591736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,2,power_law_1.01,0.08289920091629029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,2,power_law_1.01,0.08279680013656616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,2,power_law_1.01,0.08353279829025269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,2,power_law_1.01,0.08654720187187195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,2,power_law_1.01,0.09175040125846863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,2,power_law_1.01,0.09505280256271362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,2,power_law_1.01,0.0986303985118866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,2,power_law_1.01,0.1107200026512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,2,power_law_1.01,0.12039040327072144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,2,power_law_1.01,0.14455039501190187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,2,power_law_1.01,0.16379519701004028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,2,power_law_1.01,0.20413439273834227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,2,power_law_1.01,0.23324160575866698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,2,power_law_1.01,0.2959039926528931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,2,power_law_1.01,0.3654911994934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,2,power_law_1.01,0.5070208072662353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,2,power_law_1.01,0.667193603515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,32,power_law_1.2,0.055078399181365964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,32,power_law_1.2,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,32,power_law_1.2,0.05273600220680237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,128,power_law_1.01,0.03219200074672699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,32,power_law_1.2,0.049932798743247984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,128,power_law_1.01,0.03221760094165802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,32,power_law_1.2,0.05193600058555603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,128,power_law_1.01,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,32,power_law_1.2,0.05397760272026062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,32,power_law_1.2,0.05355520248413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,128,power_law_1.01,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,32,power_law_1.2,0.052748799324035645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,128,power_law_1.01,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,32,power_law_1.2,0.05258240103721619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,128,power_law_1.01,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,32,power_law_1.2,0.05438719987869263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,128,power_law_1.01,0.03191039860248566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,32,power_law_1.2,0.05258880257606506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,128,power_law_1.01,0.03255040049552917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,32,power_law_1.2,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,128,power_law_1.01,0.03237119913101196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,32,power_law_1.2,0.05400959849357605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,128,power_law_1.01,0.032953599095344545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,32,power_law_1.2,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,128,power_law_1.01,0.03473280072212219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,32,power_law_1.2,0.060575997829437254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,128,power_law_1.01,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,32,power_law_1.2,0.06130560040473938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,32,power_law_1.2,0.06659200191497802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,32,power_law_1.2,0.07362560033798218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.2,0.08435840010643006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.2,0.10229120254516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.2,0.11531519889831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.2,0.15210239887237548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,128,power_law_1.01,0.03787519931793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,128,power_law_1.01,0.03860479891300202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,128,power_law_1.01,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,128,power_law_1.01,0.04376319944858551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,128,power_law_1.01,0.044607999920845035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,128,power_law_1.01,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,128,power_law_1.01,0.059385597705841064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,128,power_law_1.01,0.06965759992599488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,128,power_law_1.01,0.07592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,128,power_law_1.01,0.09599360227584838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,128,power_law_1.01,0.11196160316467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,128,power_law_1.01,0.15841280221939086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,128,power_law_1.01,0.18696320056915283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,128,power_law_1.01,0.28371200561523435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.2,0.19404799938201905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,128,power_law_1.01,0.3913088083267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,64,balanced,0.05935466786225637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,64,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,64,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,64,balanced,0.045642669002215065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,64,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,64,balanced,0.06186666587988535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,64,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,64,balanced,0.0639573335647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,64,balanced,0.06427733103434245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,64,balanced,0.06428800026575725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,64,balanced,0.06537599861621857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,64,balanced,0.06577066580454509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,64,balanced,0.06573866804440816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,64,balanced,0.06701866785685222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,64,balanced,0.07246933380762736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,64,balanced,0.07204799850781758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,64,balanced,0.07419200241565704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,64,balanced,0.07836266855398814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,64,balanced,0.0814879983663559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,64,balanced,0.08928533395131429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,64,balanced,0.0988106628259023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,64,balanced,0.11620266238848369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,64,balanced,0.12773866454760233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,64,balanced,0.16269866625467935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,64,balanced,0.19172267119089761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,64,balanced,0.25784534215927124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,64,balanced,0.31260265906651813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,1,power_law_1.2,0.05196160078048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,1,power_law_1.2,0.05432320237159729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,1,power_law_1.2,0.055129599571228025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,1,power_law_1.2,0.06816639900207519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,1,power_law_1.2,0.06928640007972717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,1,power_law_1.2,0.0685375988483429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,1,power_law_1.2,0.07057920098304749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,1,power_law_1.2,0.06986240148544312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,1,power_law_1.2,0.06945919990539551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,1,power_law_1.2,0.07071359753608704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,1,power_law_1.2,0.07052159905433655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,1,power_law_1.2,0.07146239876747132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,1,power_law_1.2,0.07349119782447815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,1,power_law_1.2,0.07512320280075073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,1,power_law_1.2,0.08236799836158752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,1,power_law_1.2,0.08279680013656616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,1,power_law_1.2,0.08823040127754211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,1,power_law_1.2,0.10016640424728393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,1,power_law_1.2,0.12136319875717164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,1,power_law_1.2,0.14191360473632814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.2,0.2678591966629028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.2,0.3519167900085449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,1,power_law_1.2,0.16387200355529785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,1,power_law_1.2,0.21751039028167723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.2,0.49283199310302733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,1,power_law_1.2,0.26591360569000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,1,power_law_1.2,0.37372798919677735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,8,balanced,0.05020266771316528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,8,balanced,0.05401599903901418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,8,balanced,0.05735466877619425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,8,balanced,0.06680533289909363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,8,balanced,0.0743999977906545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,8,balanced,0.1067626674969991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,8,balanced,0.10372266173362732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,8,balanced,0.101583997408549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,8,balanced,0.10335466265678406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,8,balanced,0.10099200407663982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,8,balanced,0.10160533587137859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,8,balanced,0.1039680043856303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,8,balanced,0.10628267129262288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,8,balanced,0.10596799850463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,8,balanced,0.11387200156847636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,8,balanced,0.11482133467992146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,8,balanced,0.1197760005791982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,8,balanced,0.12781332929929098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,8,balanced,0.13596266508102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,8,balanced,0.15228266517321268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,8,balanced,0.16797866423924765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,8,balanced,0.20269866784413657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,8,balanced,0.2344320019086202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,8,balanced,0.3333866596221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,8,balanced,0.4012906551361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,8,balanced,0.5618133147557577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,8,balanced,0.7218186855316162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.2,0.6535552024841309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,1,power_law_1.2,0.4800384044647217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,1,power_law_1.2,0.676639986038208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,64,power_law_1.2,0.035078400373458864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,64,power_law_1.2,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,64,power_law_1.2,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,64,power_law_1.2,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,64,power_law_1.2,0.03557760119438171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,64,power_law_1.2,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,64,power_law_1.2,0.036595198512077334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,64,power_law_1.2,0.03648639917373657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,64,power_law_1.2,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,64,power_law_1.2,0.037299200892448425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,64,power_law_1.2,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,64,power_law_1.2,0.03811840116977692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,64,power_law_1.2,0.03804160058498383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,64,power_law_1.2,0.038675200939178464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,64,power_law_1.2,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,64,power_law_1.2,0.04359680116176605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,64,power_law_1.2,0.044409599900245664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,64,power_law_1.2,0.049267199635505673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,64,power_law_1.2,0.05247359871864319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,64,power_law_1.2,0.0623744010925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,64,power_law_1.2,0.06769919991493226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,64,power_law_1.2,0.08747519850730896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,64,power_law_1.2,0.10123519897460938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,64,power_law_1.2,0.12387200593948364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,64,power_law_1.2,0.15889919996261598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,64,power_law_1.2,0.22291200160980223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,1,power_law_1.2,0.883846378326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,4,power_law_1.01,0.04625279903411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,4,power_law_1.01,0.048102399706840514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,4,power_law_1.01,0.04970879852771759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,4,power_law_1.01,0.05416319966316223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,4,power_law_1.01,0.05256959795951843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,4,power_law_1.01,0.05475839972496033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,4,power_law_1.01,0.05310080051422119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,4,power_law_1.01,0.05304319858551025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,4,power_law_1.01,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,4,power_law_1.01,0.05482239723205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,4,power_law_1.01,0.050886398553848265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,4,power_law_1.01,0.0542464017868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,4,power_law_1.01,0.05750399827957153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,4,power_law_1.01,0.057843202352523805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,4,power_law_1.01,0.06312320232391358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,4,power_law_1.01,0.06847360134124755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,4,power_law_1.01,0.07575680017471313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,4,power_law_1.01,0.08579199910163879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.01,0.09596160054206848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.01,0.11743999719619751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.01,0.14806400537490844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.01,0.18052480220794678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.01,0.23637120723724364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,64,power_law_1.2,0.3273920059204102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.01,0.35640320777893064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.01,0.4077439785003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.01,0.5394368171691895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,32,balanced,0.03941866755485535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,32,balanced,0.03949866692225138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,32,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,32,balanced,0.04354666670163473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,32,balanced,0.04222933451334635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,32,balanced,0.04605866471926371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,32,balanced,0.04729066789150238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,32,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,32,balanced,0.04758933186531067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,32,balanced,0.048613334695498146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,32,balanced,0.047882666190465294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,32,balanced,0.04994133114814758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,32,balanced,0.051813334226608276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,32,balanced,0.052005335688591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,32,balanced,0.05618133147557577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,32,balanced,0.05690666536490122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,32,balanced,0.06022400160630544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,32,balanced,0.0643039991458257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,32,balanced,0.06836266815662384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,32,balanced,0.07881066699822743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,32,balanced,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,32,balanced,0.10732266306877136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,32,balanced,0.13179733355840048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,32,balanced,0.18755199511845908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,32,balanced,0.23348265886306763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,32,balanced,0.32232532898585003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,32,balanced,0.41758398214975995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,8,power_law_1.01,0.07329919934272766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,8,power_law_1.01,0.0943615972995758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,8,power_law_1.01,0.0898688018321991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,8,power_law_1.01,0.097952002286911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,8,power_law_1.01,0.10407040119171143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,8,power_law_1.01,0.11258879899978638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,8,power_law_1.01,0.10584319829940796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,8,power_law_1.01,0.10982400178909302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,8,power_law_1.01,0.11036800146102906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,8,power_law_1.01,0.11267199516296386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,8,power_law_1.01,0.11167360544204712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,8,power_law_1.01,0.11399680376052856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,8,power_law_1.01,0.11850880384445191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,8,power_law_1.01,0.12178560495376586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,8,power_law_1.01,0.12928639650344848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,8,power_law_1.01,0.13257600069046022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,8,power_law_1.01,0.14083199501037597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,8,power_law_1.01,0.16518399715423585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,8,power_law_1.01,0.18087040185928344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,8,power_law_1.01,0.2279423952102661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,8,power_law_1.01,0.2653696060180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,8,power_law_1.01,0.3442368030548096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,8,power_law_1.01,0.37506558895111086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,32,4,power_law_1.2,0.02945919930934906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,8,power_law_1.01,0.508735990524292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,32,4,power_law_1.2,0.02861439883708954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,32,4,power_law_1.2,0.027609598636627198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,8,power_law_1.01,0.6599167823791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,32,4,power_law_1.2,0.030323201417922975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,32,4,power_law_1.2,0.030156800150871278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,32,4,power_law_1.2,0.031097599864006044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,8,power_law_1.01,0.8808639526367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,8,power_law_1.01,1.0759552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,32,4,power_law_1.2,0.031916800141334536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,32,4,power_law_1.2,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,32,4,power_law_1.2,0.03381119966506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,32,4,power_law_1.2,0.03411200046539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,32,4,power_law_1.2,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,32,4,power_law_1.2,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,32,4,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,32,4,power_law_1.2,0.036748799681663516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,32,4,power_law_1.2,0.04323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,32,4,power_law_1.2,0.04529919922351837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,32,4,power_law_1.2,0.04708479940891266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,32,4,power_law_1.2,0.05143679976463318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,32,4,power_law_1.2,0.05941759943962097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,32,4,power_law_1.2,0.07130240201950074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.01,0.7589183807373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,32,4,power_law_1.2,0.0786624014377594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,32,4,power_law_1.2,0.11334400177001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,32,4,power_law_1.2,0.15489920377731323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,32,4,power_law_1.2,0.21215999126434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,64,power_law_1.01,0.026169601082801818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,64,power_law_1.01,0.025779199600219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,64,power_law_1.01,0.02531839907169342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,64,power_law_1.01,0.02803199887275696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,64,power_law_1.01,0.02876800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,64,power_law_1.01,0.030028799176216127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,32,power_law_1.01,0.023046399652957916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,32,power_law_1.01,0.026617598533630372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,32,power_law_1.01,0.023020799458026885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,32,power_law_1.01,0.02311040014028549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,32,power_law_1.01,0.023052799701690673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,32,power_law_1.01,0.02426239997148514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,32,power_law_1.01,0.02497279942035675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,32,power_law_1.01,0.02677760124206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,32,power_law_1.01,0.02685439884662628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,32,power_law_1.01,0.026649600267410277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,32,power_law_1.01,0.0324288010597229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,64,power_law_1.01,0.030611199140548707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,32,power_law_1.01,0.03252480030059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,64,power_law_1.01,0.030694401264190672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,32,power_law_1.01,0.032576000690460204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,32,power_law_1.01,0.053350400924682614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,32,power_law_1.01,0.051999998092651364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,32,power_law_1.01,0.05493760108947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,32,power_law_1.01,0.05546240210533142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,64,power_law_1.01,0.030969598889350893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,32,power_law_1.01,0.05985919833183288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,64,power_law_1.01,0.03111039996147156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.01,0.06241919994354248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.01,0.07262719869613647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,32,4,power_law_1.2,0.27582080364227296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.01,0.08295680284500122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.01,0.11454720497131347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.01,0.13059840202331544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.01,0.18526079654693603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.01,0.24089601039886474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.01,0.3404608011245728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,64,power_law_1.01,0.03212159872055054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.01,0.42383999824523927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,64,power_law_1.01,0.03278079926967621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,64,power_law_1.01,0.03386240005493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,64,power_law_1.01,0.03478400111198425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,64,power_law_1.01,0.0389055997133255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,64,power_law_1.01,0.04037120044231415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,64,power_law_1.01,0.04376960098743439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,64,power_law_1.01,0.04769279956817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,64,power_law_1.01,0.053862398862838744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,32,4,power_law_1.2,0.3874880075454712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,64,power_law_1.01,0.06229119896888733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,64,power_law_1.01,0.07788159847259521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,64,power_law_1.01,0.10406399965286255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,64,power_law_1.01,0.1340288043022156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,64,power_law_1.01,0.19129600524902343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,64,power_law_1.01,0.24281599521636962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,64,power_law_1.01,0.3336832046508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,64,power_law_1.01,0.5253439903259277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,32,4,power_law_1.2,0.49482879638671873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,2,power_law_1.01,0.04480000138282776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,2,power_law_1.01,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,2,power_law_1.01,0.05169919729232788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,2,power_law_1.01,0.06202239990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,2,power_law_1.01,0.0642687976360321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,2,power_law_1.01,0.06551039814949036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,2,power_law_1.01,0.0657472014427185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,2,power_law_1.01,0.06621440052986145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,2,power_law_1.01,0.06654080152511596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,2,power_law_1.01,0.06657919883728028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,2,power_law_1.01,0.06657919883728028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,2,power_law_1.01,0.0809984028339386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,2,power_law_1.01,0.08165760040283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,2,power_law_1.01,0.08380159735679626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,2,power_law_1.01,0.09482880234718323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,2,power_law_1.01,0.09587200284004212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,2,power_law_1.01,0.10104960203170776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,2,power_law_1.01,0.12558720111846924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,2,power_law_1.01,0.1416383981704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,2,power_law_1.01,0.18791040182113647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,2,power_law_1.01,0.22325758934020995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,16,power_law_1.2,0.025222399830818178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,16,power_law_1.2,0.023148800432682037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,16,power_law_1.2,0.02380799949169159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,2,power_law_1.01,0.3140223979949951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,16,power_law_1.2,0.02351360023021698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,16,power_law_1.2,0.025043201446533204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,16,power_law_1.2,0.025971201062202454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,16,power_law_1.2,0.02592639923095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,16,power_law_1.2,0.02974080145359039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,16,power_law_1.2,0.030899199843406677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,16,power_law_1.2,0.03204480111598969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,16,power_law_1.2,0.032620799541473386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,16,power_law_1.2,0.03364480137825012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,16,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,16,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,16,power_law_1.2,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,16,power_law_1.2,0.03804160058498383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,16,power_law_1.2,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,16,power_law_1.2,0.04279040098190308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,2,power_law_1.01,0.40724477767944334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,16,power_law_1.2,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,16,power_law_1.2,0.052985602617263795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,16,power_law_1.2,0.06114559769630432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,16,power_law_1.2,0.07351679801940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,16,power_law_1.2,0.08522239923477173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,16,power_law_1.2,0.12013440132141114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,2,power_law_1.01,0.6155327796936035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,8,power_law_1.01,0.07650560140609741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,8,power_law_1.01,0.08756480216979981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,8,power_law_1.01,0.08844159841537476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,8,power_law_1.01,0.11198079586029053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,8,power_law_1.01,0.10013439655303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,8,power_law_1.01,0.11471359729766846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,8,power_law_1.01,0.11374080181121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,8,power_law_1.01,0.11122560501098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,8,power_law_1.01,0.11747200489044189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,8,power_law_1.01,0.11887359619140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,8,power_law_1.01,0.11815040111541748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,8,power_law_1.01,0.12151039838790893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,8,power_law_1.01,0.12669440507888793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,8,power_law_1.01,0.12888959646224976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,8,power_law_1.01,0.13722879886627198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,8,power_law_1.01,0.13587199449539183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,16,power_law_1.2,0.14962559938430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,8,power_law_1.01,0.14149760007858275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,8,power_law_1.01,0.16177279949188234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,8,power_law_1.01,0.16829440593719483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,8,power_law_1.01,0.1958143949508667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,8,power_law_1.01,0.2031872034072876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,8,power_law_1.01,0.26574718952178955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,8,power_law_1.01,0.2956671953201294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,8,power_law_1.01,0.42602877616882323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,8,power_law_1.01,0.5336832046508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,8,power_law_1.01,0.7104320049285888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,2,power_law_1.01,0.7460095882415771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,8,power_law_1.01,0.8663167953491211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,1,power_law_1.01,0.04573439955711365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,1,power_law_1.01,0.04757120013237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,1,power_law_1.01,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,1,power_law_1.01,0.04705919921398163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,1,power_law_1.01,0.04890879988670349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,1,power_law_1.01,0.06024320125579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,1,power_law_1.01,0.07567999958992004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,1,power_law_1.01,0.06716799736022949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,1,power_law_1.01,0.08858879804611205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,1,power_law_1.01,0.0797760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,1,power_law_1.01,0.10401920080184937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,1,power_law_1.01,0.11945600509643554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,1,power_law_1.01,0.13432960510253905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,1,power_law_1.01,0.1004863977432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,1,power_law_1.01,0.14318079948425294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,1,power_law_1.01,0.12472319602966309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,1,power_law_1.01,0.1617408037185669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,1,power_law_1.01,0.21220479011535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.01,0.27400319576263427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,8,balanced,0.04966933528582255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,8,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,8,balanced,0.05402666827042898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,8,balanced,0.06420266628265381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,8,balanced,0.08658666412035625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,8,balanced,0.12446932991345723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,8,balanced,0.15713600317637125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,8,balanced,0.15777066349983215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,8,balanced,0.1590986649195353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,8,balanced,0.15954132874806723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,8,balanced,0.16080000003178915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,8,balanced,0.1642400026321411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,8,balanced,0.16772266228993735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,8,balanced,0.16555733482042947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,8,balanced,0.17348267634709677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,8,balanced,0.17908799648284912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,8,balanced,0.18422933419545492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,16,power_law_1.2,0.22551040649414061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.01,0.3871615886688232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,8,balanced,0.19358932971954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,8,balanced,0.2076693375905355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.01,0.5183680057525635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,8,balanced,0.23263466358184814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,2,power_law_1.01,1.2088000297546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,8,balanced,0.2518826723098755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,8,balanced,0.3047200043996175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,8,balanced,0.3543200095494588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.01,0.7437183856964111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,8,balanced,0.44175465901692706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,8,balanced,0.5745493173599243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,8,balanced,0.7429386774698893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,8,balanced,0.9666133721669515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.01,0.9875136375427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.01,1.4527935981750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.01,1.9160448074340821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,16,power_law_1.2,0.23034238815307617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.01,2.872313690185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,2,power_law_1.01,1.6055871963500976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,32,power_law_1.01,0.017798399925231932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,32,power_law_1.01,0.018457600474357606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,32,power_law_1.01,0.017612800002098083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,32,power_law_1.01,0.017632000148296356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,32,power_law_1.01,0.018105599284172057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,32,power_law_1.01,0.022310400009155275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,32,power_law_1.01,0.03857919871807099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,32,power_law_1.01,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,32,power_law_1.01,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,32,power_law_1.01,0.031718400120735166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,32,power_law_1.01,0.03160319924354553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,32,power_law_1.01,0.032076799869537355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,32,power_law_1.01,0.031788799166679385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,32,power_law_1.01,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,32,power_law_1.01,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,32,power_law_1.01,0.0342848002910614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.01,3.7992832183837892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,32,power_law_1.01,0.03703039884567261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,32,power_law_1.01,0.04090240001678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,32,power_law_1.01,0.04430719912052154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,32,power_law_1.01,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,32,power_law_1.01,0.05370879769325256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,32,power_law_1.01,0.06713600158691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,32,power_law_1.01,0.07884160280227662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,32,power_law_1.01,0.11093759536743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,32,power_law_1.01,0.13351039886474608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,32,power_law_1.01,0.1923840045928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,32,power_law_1.01,0.23368959426879882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,8,balanced,0.04370133578777313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,8,balanced,0.04394666850566864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,8,balanced,0.047482664386431374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,8,balanced,0.06806399921576183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,8,balanced,0.0890880028406779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,8,balanced,0.13310933113098145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,8,balanced,0.1351040005683899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,8,balanced,0.13780267039934793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,8,balanced,0.13794133067131042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,8,balanced,0.13818132877349854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,8,balanced,0.13979732990264893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,8,balanced,0.13896000385284424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,8,balanced,0.141594668229421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,8,balanced,0.1439893345038096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,8,balanced,0.14854933818181357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,8,balanced,0.14809067050615946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,8,balanced,0.15339199701944986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,8,balanced,0.1609440048535665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,8,balanced,0.16611733039220175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,8,balanced,0.18515199422836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,8,balanced,0.20645866791407266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,8,balanced,0.2368053396542867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,8,balanced,0.2639999985694885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,8,balanced,0.341536005338033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,8,balanced,0.3977866570154826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,8,balanced,0.5553919871648153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,8,balanced,0.6779786745707194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,8,balanced,0.027376001079877216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,8,balanced,0.02734400083621343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,8,balanced,0.029296000798543293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,8,balanced,0.03160000095764796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,8,balanced,0.03737066686153412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,8,balanced,0.042352000872294106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,8,balanced,0.03991466760635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,8,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,8,balanced,0.0407679999868075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,8,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,8,balanced,0.039818666875362396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,8,balanced,0.03986666599909464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,8,balanced,0.040991999208927155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,8,balanced,0.04182399809360504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,8,balanced,0.04580800235271454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,8,balanced,0.044853334625562034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,8,balanced,0.04783466458320618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,8,balanced,0.05114666620890299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,8,balanced,0.06648533542950948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,8,balanced,0.07551466425259908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,8,balanced,0.07877866427103679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,8,balanced,0.09654933214187622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,8,balanced,0.11664533615112305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,8,balanced,0.14396799604098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,8,balanced,0.17708800236384073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,8,balanced,0.23564799626668295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,8,balanced,0.29613866408665973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,8,balanced,0.02091199904680252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,8,balanced,0.021104000508785248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,8,balanced,0.023183998962243397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,8,balanced,0.02316266546646754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,8,balanced,0.025834667185942333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,8,balanced,0.027488000690937042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,8,balanced,0.029317334294319153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,8,balanced,0.03138133386770884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,8,balanced,0.03137599925200144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,8,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,8,balanced,0.03347733368476232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,8,balanced,0.035487999518712364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,8,balanced,0.03696000079313914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,8,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,8,balanced,0.04196266829967499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,8,balanced,0.048783997694651283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,8,balanced,0.05593599875768026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,8,balanced,0.06505600114663442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,8,balanced,0.08656533559163411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,8,balanced,0.11422399679819743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,8,balanced,0.11122666796048482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,8,balanced,0.1383786698182424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,8,balanced,0.1670773426691691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,8,balanced,0.20844799280166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,8,balanced,0.26310400168100995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,8,balanced,0.38416000207265216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,8,balanced,0.48281065622965497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,1,power_law_1.01,0.07927039861679078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,1,power_law_1.01,0.11834880113601684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,1,power_law_1.01,0.20211200714111327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,1,power_law_1.01,0.2873856067657471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,128,balanced,0.03990400085846583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,128,balanced,0.03972266614437103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,128,balanced,0.039818666875362396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,128,balanced,0.04225599765777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,128,balanced,0.041637333730856575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,128,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,1,power_law_1.01,0.38736639022827146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,1,power_law_1.01,0.5998464107513428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,1,power_law_1.01,0.6301184177398682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,1,power_law_1.01,0.645363187789917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,1,power_law_1.01,0.6577919960021973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,1,power_law_1.01,0.688640022277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,1,power_law_1.01,0.6875648021697998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,1,power_law_1.01,0.6943679809570312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,128,balanced,0.046021332343419395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,1,power_law_1.01,0.7200575828552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,1,power_law_1.01,0.7242432117462159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,1,power_law_1.01,0.7511295795440673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,1,power_law_1.01,0.760646390914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,1,power_law_1.01,0.7976319789886475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,128,balanced,0.04571733375390371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,1,power_law_1.01,0.8603455543518066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,1,power_law_1.01,0.9141568183898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,1,power_law_1.01,1.050540828704834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,1,power_law_1.01,1.1078207969665528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,1,power_law_1.01,1.2969599723815919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,1,power_law_1.01,1.5078144073486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,128,balanced,0.04983466863632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,128,balanced,0.0476800004641215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,1,power_law_1.01,1.8959936141967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,1,power_law_1.01,2.2243200302124024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,1,power_law_1.01,2.994911956787109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,1,power_law_1.01,4.015116882324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,128,balanced,0.05006400247414907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,128,balanced,0.05233600238958994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,128,balanced,0.049973333875338234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,128,balanced,0.055626665552457176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,128,balanced,0.06038400034109751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,128,balanced,0.0561653325955073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,128,balanced,0.06234666705131531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,128,balanced,0.07028266787528992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,128,balanced,0.07262933254241943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,128,balanced,0.08294933537642162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,128,balanced,0.09319466352462769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,128,balanced,0.1162506639957428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,128,balanced,0.13636799653371176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,128,balanced,0.17915199200312296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,128,balanced,0.2243679960568746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,128,balanced,0.3100586732228597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,128,balanced,0.39403732617696124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,4,balanced,0.020773333807786305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,4,balanced,0.019727999965349834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,4,balanced,0.019648000597953796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,4,balanced,0.01971199984351794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,4,balanced,0.021365332106749218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,4,balanced,0.021695998807748158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,4,balanced,0.024506665766239166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,4,balanced,0.02534399926662445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,4,balanced,0.04869333406289419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,4,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,4,balanced,0.0481333335240682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,4,balanced,0.03789866715669632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,4,balanced,0.03797333439191183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,4,balanced,0.03788800040880839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,4,balanced,0.0401706670721372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,4,balanced,0.038575999438762665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,4,balanced,0.043194666504859924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,4,balanced,0.0448586642742157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,4,balanced,0.0476800004641215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,4,balanced,0.06818133095900218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,4,balanced,0.07699733475844066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,4,balanced,0.10145599643389384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,4,balanced,0.12607466181119284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,4,balanced,0.17355199654897055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,4,balanced,0.21221866210301718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,4,balanced,0.3094186584154765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,4,balanced,0.3964853286743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,2,balanced,0.021695998807748158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,2,balanced,0.023152001202106476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,2,balanced,0.023989332218964893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,2,balanced,0.02752533306678136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,2,balanced,0.04046933352947235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,2,balanced,0.059845333298047386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,2,balanced,0.06233066817124685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,2,balanced,0.06427200138568878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,2,balanced,0.0681386689345042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,2,balanced,0.06865600248177846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,2,balanced,0.07235200206438701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,2,balanced,0.07216533521811168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,2,balanced,0.07446933289368947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,2,balanced,0.07941333452860515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,2,balanced,0.08135466774304707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,2,balanced,0.08231466511885326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,2,balanced,0.08922666311264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,2,balanced,0.11551466584205627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,2,balanced,0.12127466996510823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,2,balanced,0.16861865917841592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,2,balanced,0.18014933665593466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,2,balanced,0.23856000105539957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,2,balanced,0.26543466250101727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,2,balanced,0.3040906588236491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,2,balanced,0.34621866544087726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,2,balanced,0.5518453518549601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,2,balanced,0.6411306858062744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,2,balanced,0.031504000226656594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,2,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,2,balanced,0.03380800038576126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,2,balanced,0.04138666639725367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,2,balanced,0.04966933528582255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,2,balanced,0.04996266464392344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,2,balanced,0.05012799799442291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,2,balanced,0.050010666251182556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,2,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,2,balanced,0.05205333232879639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,2,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,2,balanced,0.05606399973233541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,2,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,2,balanced,0.06017066538333893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,2,balanced,0.06870933373769124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,2,balanced,0.07043200234572093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,2,balanced,0.07840000092983246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,2,balanced,0.09052266677220662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,2,balanced,0.10504000385602315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,2,balanced,0.11961066722869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,2,balanced,0.14455999930699667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,2,balanced,0.22395199537277222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,2,balanced,0.24988800287246704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,2,balanced,0.3518986701965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,2,balanced,0.4545546770095825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,2,balanced,0.6662986675898234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,2,balanced,0.8558932940165201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,32,balanced,0.0194560003777345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,32,balanced,0.01933866615096728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,32,balanced,0.019050666441520054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,32,balanced,0.01933866615096728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,32,balanced,0.020970667401949566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,32,balanced,0.023152001202106476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,32,balanced,0.039877332746982574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,32,balanced,0.04068266600370407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,32,balanced,0.03513599932193756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,32,balanced,0.034458667039871216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,32,balanced,0.03411199897527695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,32,balanced,0.03410666684309641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,32,balanced,0.03421333432197571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,32,balanced,0.03389333436886469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,32,balanced,0.04196266829967499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,32,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,32,balanced,0.0408693328499794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,32,balanced,0.04515199859937032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,32,balanced,0.04818666477998098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,32,balanced,0.05921066800753275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,32,balanced,0.07429333527882893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,32,balanced,0.09033067027727763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,32,balanced,0.11360533038775127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,32,balanced,0.15235732992490134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,32,balanced,0.1973653237024943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,32,balanced,0.2752000093460083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,32,balanced,0.35831467310587567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,2,power_law_1.01,0.0430400013923645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,2,power_law_1.01,0.044787201285362246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,2,power_law_1.01,0.056595200300216676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,2,power_law_1.01,0.07208319902420043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,2,power_law_1.01,0.07567999958992004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,2,power_law_1.01,0.07501440048217774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,2,power_law_1.01,0.07949439883232116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,2,power_law_1.01,0.08030719757080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,2,power_law_1.01,0.08117759823799134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,2,power_law_1.01,0.08232960104942322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,2,power_law_1.01,0.08568320274353028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,2,power_law_1.01,0.08694400191307068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,2,power_law_1.01,0.08675839900970458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,2,power_law_1.01,0.09698560237884521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,2,power_law_1.01,0.10505599975585937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,2,power_law_1.01,0.10609920024871826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,2,power_law_1.01,0.12910720109939575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,2,power_law_1.01,0.1388479948043823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,2,power_law_1.01,0.15267200469970704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,2,power_law_1.01,0.18705919981002808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,2,power_law_1.01,0.2072767972946167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,2,power_law_1.01,0.2770751953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,2,power_law_1.01,0.34923520088195803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,2,power_law_1.01,0.49930877685546876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,2,power_law_1.01,0.546124792098999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,128,power_law_1.2,0.06428160071372986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,128,power_law_1.2,0.06323199868202209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,128,power_law_1.2,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,128,power_law_1.2,0.06746240258216858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,128,power_law_1.2,0.07593600153923034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,128,power_law_1.2,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,128,power_law_1.2,0.06857600212097167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,128,power_law_1.2,0.07128959894180298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,128,power_law_1.2,0.06951040029525757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,128,power_law_1.2,0.0700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,128,power_law_1.2,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,128,power_law_1.2,0.07030400037765502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,128,power_law_1.2,0.07391999959945679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,128,power_law_1.2,0.07126399874687195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,128,power_law_1.2,0.07583360075950622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,128,power_law_1.2,0.07731199860572815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,128,power_law_1.2,0.085452800989151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,128,power_law_1.2,0.09233279824256897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,128,power_law_1.2,0.10163840055465698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,128,power_law_1.2,0.11989120244979859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,128,power_law_1.2,0.14473600387573243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,128,power_law_1.2,0.19201279878616334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,128,power_law_1.2,0.21024000644683838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,128,power_law_1.2,0.29333760738372805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,128,power_law_1.2,0.3803776025772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,128,power_law_1.2,0.551091194152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,128,power_law_1.2,0.8178303718566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,2,power_law_1.01,1.0181952476501466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,2,balanced,0.04433066646258036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,2,balanced,0.04596266647179922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,2,balanced,0.05547733108202616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,2,balanced,0.08313600222269694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,2,balanced,0.11734400192896526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,2,power_law_1.01,0.9390144348144531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,2,balanced,0.18657066424687704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,2,balanced,0.1857759952545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,2,balanced,0.1885439952214559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,2,balanced,0.18786134322484335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,2,balanced,0.18948266903559366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,2,balanced,0.19070400794347128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,2,balanced,0.19639466206232706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,2,balanced,0.19673067331314087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,2,balanced,0.19908267259597778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,2,balanced,0.20865599314371744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,2,balanced,0.21178666750590006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,2,balanced,0.21612799167633057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,2,balanced,0.23707733551661173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,2,balanced,0.25406400362650555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,8,power_law_1.2,0.03047040104866028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,8,power_law_1.2,0.030195200443267824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,8,power_law_1.2,0.030399999022483824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,8,power_law_1.2,0.03238399922847748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,8,power_law_1.2,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,8,power_law_1.2,0.03349120020866394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,8,power_law_1.2,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,2,balanced,0.2888373335202535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,8,power_law_1.2,0.03372159898281098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,2,balanced,0.3224479953447978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,2,balanced,0.40538668632507324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,2,balanced,0.46881067752838135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,2,balanced,0.6400320132573446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,2,balanced,0.7749280134836832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,2,balanced,1.1073066393534343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,2,balanced,1.394485314687093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,8,power_law_1.2,0.03427839875221252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,8,power_law_1.2,0.03547520041465759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,8,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,8,power_law_1.2,0.03775359988212586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,8,power_law_1.2,0.03868159949779511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,16,power_law_1.01,0.07507839798927307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,16,power_law_1.01,0.11676160097122193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,16,power_law_1.01,0.10019199848175049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,16,power_law_1.01,0.10551040172576905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,8,power_law_1.2,0.04108799993991852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,16,power_law_1.01,0.10305279493331909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,16,power_law_1.01,0.09111040234565734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,16,power_law_1.01,0.10065920352935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,16,power_law_1.01,0.10209280252456665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,16,power_law_1.01,0.10570240020751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,16,power_law_1.01,0.1027008056640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,8,power_law_1.2,0.04855040013790131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,16,power_law_1.01,0.10228480100631714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,16,power_law_1.01,0.10549759864807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,16,power_law_1.01,0.11045759916305542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,16,power_law_1.01,0.11153279542922974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,16,power_law_1.01,0.11792000532150268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,16,power_law_1.01,0.12086399793624877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,16,power_law_1.01,0.12767360210418702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,4,4,balanced,0.061717331409454346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,16,power_law_1.01,0.13705600500106813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,4,4,balanced,0.06611733138561249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,4,4,balanced,0.07853866616884868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,4,4,balanced,0.10761599739392598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,16,power_law_1.01,0.15031039714813232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,16,power_law_1.01,0.1720128059387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,16,power_law_1.01,0.19349759817123413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,16,power_law_1.01,0.236409592628479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,4,4,balanced,0.1548746625582377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,4,4,balanced,0.16926934321721396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,8,power_law_1.2,0.054790401458740236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,4,4,balanced,0.16947199900945029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,4,4,balanced,0.16634666919708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,4,4,balanced,0.16321067015329996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,4,4,balanced,0.16461867094039917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,4,4,balanced,0.15795733531316122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,4,4,balanced,0.1599999964237213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,4,4,balanced,0.16338133811950684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,4,4,balanced,0.16127467155456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,4,4,balanced,0.16848532358805338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,4,4,balanced,0.16961065928141275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,4,4,balanced,0.1742560068766276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,4,4,balanced,0.1834933360417684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,4,4,balanced,0.19298134247461954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,16,power_law_1.01,0.26857600212097166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,4,4,balanced,0.21358400583267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,4,4,balanced,0.2336746652921041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,16,power_law_1.01,0.38445439338684084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,4,4,balanced,0.3487253189086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,4,4,balanced,0.38344534238179523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,16,power_law_1.01,0.47150721549987795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,4,4,balanced,0.5277920166651408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,4,4,balanced,0.6900853315989176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,16,power_law_1.01,0.7390143871307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,8,power_law_1.2,0.05936639904975891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,4,4,balanced,0.9014293352762858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,16,power_law_1.01,0.8676287651062011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,4,4,balanced,1.21670929590861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,8,power_law_1.2,0.07390720248222352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,8,power_law_1.2,0.08883839845657349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,8,power_law_1.2,0.10953600406646728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,8,power_law_1.2,0.1417855978012085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,8,power_law_1.2,0.20480639934539796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,8,power_law_1.2,0.22476160526275635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,8,power_law_1.2,0.36294400691986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,2,power_law_1.2,0.029945600032806396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,2,power_law_1.2,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,2,power_law_1.2,0.031327998638153075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,2,power_law_1.2,0.03506560027599335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,2,power_law_1.2,0.03521920144557953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,2,power_law_1.2,0.039052799344062805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,2,power_law_1.2,0.040780800580978396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,2,power_law_1.2,0.04139519929885864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,2,power_law_1.2,0.04362240135669708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,2,power_law_1.2,0.04513919949531555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,2,power_law_1.2,0.04563199877738953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,2,power_law_1.2,0.04895359873771667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,2,power_law_1.2,0.05007359981536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,2,power_law_1.2,0.051801598072052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,2,power_law_1.2,0.06339840292930603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,2,power_law_1.2,0.06815360188484192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,2,power_law_1.2,0.07145599722862243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,2,power_law_1.2,0.0869055986404419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,2,power_law_1.2,0.09887999892234803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,2,power_law_1.2,0.12607359886169434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,2,power_law_1.2,0.15329920053482055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,8,power_law_1.2,0.45902080535888673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,2,power_law_1.2,0.2149120092391968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,2,power_law_1.2,0.23666560649871826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,2,power_law_1.2,0.322271990776062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,8,power_law_1.2,0.7388927936553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,2,power_law_1.2,0.43848958015441897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,8,power_law_1.2,1.116915225982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,2,power_law_1.2,0.5981567859649658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,4,power_law_1.2,0.07297279834747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,4,power_law_1.2,0.09018239974975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,4,power_law_1.2,0.1054527997970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,4,power_law_1.2,0.12428159713745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,4,power_law_1.2,0.14499839544296264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,4,power_law_1.2,0.16895999908447265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,4,power_law_1.2,0.1856063961982727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,4,power_law_1.2,0.20775680541992186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,4,power_law_1.2,0.2378432035446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,4,power_law_1.2,0.2286463975906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,4,power_law_1.2,0.20994560718536376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,4,power_law_1.2,0.2401792049407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,4,power_law_1.2,0.23032960891723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,4,power_law_1.2,0.23594880104064941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,4,power_law_1.2,0.25080959796905516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,4,power_law_1.2,0.25257599353790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,4,power_law_1.2,0.26462080478668215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,4,power_law_1.2,0.29845759868621824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,4,power_law_1.2,0.3295423984527588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,4,power_law_1.2,0.3817023992538452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,4,power_law_1.2,0.4504511833190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,4,power_law_1.2,0.542790412902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,4,power_law_1.2,0.6289663791656495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,4,power_law_1.2,0.820025634765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,4,power_law_1.2,1.0177663803100585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,4,power_law_1.2,1.4810496330261231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,4,power_law_1.2,1.7172159194946288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,16,balanced,0.03465066601832708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,16,balanced,0.033728001018365227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,16,balanced,0.03857066730658213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,16,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,16,balanced,0.05429333448410034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,2,balanced,0.024432001014550526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,2,balanced,0.023269332945346832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,2,balanced,0.025311999022960663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,2,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,2,balanced,0.06015466650327047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,2,balanced,0.08459200461705525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,2,balanced,0.08984532952308655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,2,balanced,0.09170132875442505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,2,balanced,0.09516266981760661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,2,balanced,0.09698667128880818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,2,balanced,0.10591466228167216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,2,balanced,0.10056533416112264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,2,balanced,0.10466667016347249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,2,balanced,0.11086933811505635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,2,balanced,0.1139359970887502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,2,balanced,0.11539199948310852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,2,balanced,0.12379200259844463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,2,balanced,0.1562933325767517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,2,balanced,0.1630826691786448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,2,balanced,0.2543626626332601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,2,balanced,0.2932800054550171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,2,balanced,0.30749332904815674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,2,balanced,0.336847980817159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,2,balanced,0.39263999462127686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,16,balanced,0.07588799794514973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,2,balanced,0.4384640057881673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,2,balanced,0.7162613073984782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,2,balanced,0.8315467039744059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,2,power_law_1.2,0.7690879821777343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,16,balanced,0.08737599849700928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,16,balanced,0.08444799979527791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,16,balanced,0.12574399511019388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,16,balanced,0.12100266416867574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,16,balanced,0.11699733138084412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,16,balanced,0.16343466440836588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,16,balanced,0.16454399625460306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,16,balanced,0.16714133818944296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,16,balanced,0.1715893348058065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,16,balanced,0.17368000745773315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,16,balanced,0.1656000018119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,16,balanced,0.1881600022315979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,16,balanced,0.19985065857569376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,16,balanced,0.2286346753438314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,16,balanced,0.26661332448323566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,16,balanced,0.33237866560618085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,16,balanced,0.39451201756795246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,16,balanced,0.522709329922994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,16,balanced,0.6489599943161011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,16,balanced,1.0036959648132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,1,2,power_law_1.01,0.13535360097885132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,16,balanced,1.2641440232594807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,1,2,power_law_1.01,0.21529600620269776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,1,2,power_law_1.01,0.30286080837249757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,1,2,power_law_1.01,0.4643263816833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,1,2,power_law_1.01,0.6131775856018067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,1,2,power_law_1.01,0.8212287902832032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,1,2,power_law_1.01,1.0684800148010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,1,2,power_law_1.01,1.0950783729553222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,1,2,power_law_1.01,1.117407989501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,1,2,power_law_1.01,1.1815296173095704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,1,2,power_law_1.01,1.1843520164489747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,1,2,power_law_1.01,1.2395520210266113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,1,2,power_law_1.01,1.209887981414795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,1,2,power_law_1.01,1.223686408996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,1,2,power_law_1.01,1.3262720108032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,1,2,power_law_1.01,1.297651195526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,1,2,power_law_1.01,1.3768575668334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,1,2,power_law_1.01,1.4597696304321288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,1,2,power_law_1.01,1.485593605041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,1,2,power_law_1.01,1.712019157409668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,1,2,power_law_1.01,1.6601919174194335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,1,2,power_law_1.01,1.8348031997680665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,1,2,power_law_1.01,2.2192127227783205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,1,2,power_law_1.01,2.669331169128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,16,balanced,0.04164266586303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,16,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,16,balanced,0.04195733368396759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,16,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,16,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,16,balanced,0.053957333167394005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,1,2,power_law_1.01,2.99619197845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,16,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,16,balanced,0.05588266750176748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,16,balanced,0.05595199763774872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,16,balanced,0.05613866448402405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,16,balanced,0.056159997979799904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,16,balanced,0.058005332946777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,16,balanced,0.05996266504128774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,16,balanced,0.060005332032839455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,1,2,power_law_1.01,3.7213695526123045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,16,balanced,0.06624533236026764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,16,balanced,0.06622399886449178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,16,balanced,0.06878933310508728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,16,balanced,0.07865599791208903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,16,balanced,0.0844586690266927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,16,balanced,0.10523200035095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,1,2,power_law_1.01,4.512793731689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,16,balanced,0.12471466263135274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,16,balanced,0.17118932803471884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,16,balanced,0.2005280057589213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,16,balanced,0.2735733389854431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,16,balanced,0.34832000732421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,16,balanced,0.4857706626256307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,16,balanced,0.6257226864496866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,32,balanced,0.02603733291228612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,32,balanced,0.02638400097688039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,32,balanced,0.025333332518736523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,32,balanced,0.025093334416548412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,32,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,32,balanced,0.04082666585842768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,32,balanced,0.04391466577847799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,32,balanced,0.04549333453178406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,32,balanced,0.048528000712394714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,32,balanced,0.04961066444714864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,32,balanced,0.05638400216897329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,32,balanced,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,32,balanced,0.055957332253456116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,32,balanced,0.09134933352470398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,32,balanced,0.09479999542236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,32,balanced,0.09508267045021057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,32,balanced,0.126202662785848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,32,balanced,0.15984533230463663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,32,balanced,0.1755253275235494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,32,balanced,0.2474293311436971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,32,balanced,0.1564853290716807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,32,balanced,0.20575465758641562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,32,balanced,0.26158400376637775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,32,balanced,0.3587413231531779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,32,balanced,0.4556106726328532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,32,balanced,0.7039679686228434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,32,balanced,0.9166399637858073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,8,power_law_1.2,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,8,power_law_1.2,0.046265599131584165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,8,power_law_1.2,0.04392319917678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,8,power_law_1.2,0.04595839977264404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,8,power_law_1.2,0.04524799883365631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,8,power_law_1.2,0.04527359902858734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,8,power_law_1.2,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,8,power_law_1.2,0.04407680034637451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,8,power_law_1.2,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,8,power_law_1.2,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,8,power_law_1.2,0.04230400025844574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,8,power_law_1.2,0.04573439955711365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,8,power_law_1.2,0.05062400102615357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,8,power_law_1.2,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,8,power_law_1.2,0.05555840134620667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,8,power_law_1.2,0.05750399827957153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,8,power_law_1.2,0.06218240261077881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,8,power_law_1.2,0.07139840126037597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.2,0.08266879916191101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.2,0.10377600193023681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.2,0.12727680206298828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.2,0.17238399982452393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,8,power_law_1.01,0.04439040124416351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,8,power_law_1.01,0.04423680007457733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,8,power_law_1.01,0.043750399351119997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,8,power_law_1.01,0.0441536009311676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,8,power_law_1.01,0.044582399725914004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,8,power_law_1.01,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,8,power_law_1.01,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,8,power_law_1.01,0.04659200012683869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,8,power_law_1.01,0.05304319858551025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,8,power_law_1.01,0.05568000078201294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,8,power_law_1.01,0.05868160128593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,8,power_law_1.01,0.06387199759483338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,8,power_law_1.01,0.06520959734916687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,8,power_law_1.01,0.06725760102272034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,8,power_law_1.01,0.08229119777679443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,8,power_law_1.01,0.08264319896697998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,8,power_law_1.01,0.08985599875450134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.2,0.2126847982406616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,8,power_law_1.01,0.1034432053565979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.01,0.1221951961517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.01,0.15427839756011963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,8,power_law_1.2,0.06161919832229614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,8,power_law_1.2,0.06195840239524841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,8,power_law_1.2,0.061715197563171384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,8,power_law_1.2,0.06517120003700257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,8,power_law_1.2,0.06483200192451477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,8,power_law_1.2,0.06616320013999939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,8,power_law_1.2,0.06728320121765137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,8,power_law_1.2,0.07169280052185059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,8,power_law_1.2,0.07390720248222352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.01,0.18688640594482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,8,power_law_1.2,0.08041599988937378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,8,power_law_1.2,0.07284479737281799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,8,power_law_1.2,0.07891839742660522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,8,power_law_1.2,0.08654720187187195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.01,0.25411200523376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,8,power_law_1.2,0.09166719913482665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,64,power_law_1.2,0.01791359931230545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,64,power_law_1.2,0.01749120056629181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,8,power_law_1.2,0.10772479772567749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,64,power_law_1.2,0.017926399409770966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,8,power_law_1.2,0.12354559898376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.01,0.3310080051422119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,8,power_law_1.2,0.12583680152893068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,8,power_law_1.2,0.1561344027519226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,64,power_law_1.2,0.018137599527835845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,64,power_law_1.2,0.020876799523830415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,64,power_law_1.2,0.03893119990825653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,64,power_law_1.2,0.02945919930934906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,64,power_law_1.2,0.029919999837875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,64,power_law_1.2,0.03000960052013397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,64,power_law_1.2,0.030060800909996032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,64,power_law_1.2,0.03031040132045746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,8,power_law_1.2,0.19254399538040162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,64,power_law_1.2,0.030777600407600404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,64,power_law_1.2,0.03080959916114807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,64,power_law_1.2,0.031225600838661195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,64,power_law_1.2,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,64,power_law_1.2,0.034041601419448855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,64,power_law_1.2,0.03482879996299744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,64,power_law_1.2,0.03919999897480011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.2,0.04428159892559051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.2,0.04922879934310913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.2,0.05306239724159241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.2,0.06520959734916687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.2,0.2869055986404419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.2,0.08647680282592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,8,power_law_1.2,0.25785601139068604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,8,power_law_1.2,0.3303168058395386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.01,0.46481919288635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.2,0.10981119871139526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,8,power_law_1.2,0.45664000511169434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.2,0.13782399892807007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.2,0.35151360034942625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.01,0.6556608200073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,2,balanced,0.031530665854612984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,2,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,2,balanced,0.03422400106986364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,2,balanced,0.035829332967599235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,2,balanced,0.037605332831541695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,2,balanced,0.037871999045213066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,2,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,2,balanced,0.039605334401130676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,2,balanced,0.03980266551176707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,2,balanced,0.03980266551176707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,2,balanced,0.041637333730856575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,2,balanced,0.04165333261092504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,8,power_law_1.2,0.5837823867797851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,2,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.2,0.19204479455947876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,2,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,2,balanced,0.06038400034109751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,2,balanced,0.06411199768384297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,2,balanced,0.06305600206057231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,2,balanced,0.07745600243409474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,2,balanced,0.08576533198356628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,2,balanced,0.1108746627966563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,2,balanced,0.13201600313186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,2,balanced,0.17227200667063394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,2,balanced,0.19778132438659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,2,balanced,0.27804799874623615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,2,balanced,0.35744531949361164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,2,balanced,0.5162026484807333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,2,balanced,0.6654880046844482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,64,power_law_1.2,0.05856639742851257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,64,power_law_1.2,0.054995197057724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,64,power_law_1.2,0.05434880256652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,64,power_law_1.2,0.05668479800224304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,64,power_law_1.2,0.05749120116233826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,64,power_law_1.2,0.059468799829483034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,64,power_law_1.2,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,64,power_law_1.2,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,64,power_law_1.2,0.06453120112419128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,64,power_law_1.2,0.06547840237617493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.01,0.9138367652893067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,64,power_law_1.2,0.06554239988327026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,64,power_law_1.2,0.06921600103378296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,64,power_law_1.2,0.07109760046005249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,64,power_law_1.2,0.075135999917984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,64,power_law_1.2,0.09102720022201538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,64,power_law_1.2,0.09066240191459655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,64,power_law_1.2,0.0913919985294342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,64,power_law_1.2,0.10945919752120972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,64,power_law_1.2,0.12086399793624877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,64,power_law_1.2,0.14257919788360596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,64,power_law_1.2,0.16228480339050294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,64,power_law_1.2,0.21269760131835938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,64,power_law_1.2,0.26226561069488524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,64,power_law_1.2,0.36702721118927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,64,power_law_1.2,0.49633278846740725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,64,power_law_1.2,0.717087984085083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,8,power_law_1.2,0.8641856193542481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,64,power_law_1.2,1.0038847923278809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.2,0.5544320106506347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,8,power_law_1.2,1.0538432121276855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.01,1.2703488349914551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.2,0.2535423994064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,8,power_law_1.2,1.6302656173706054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.2,0.6907199859619141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,8,power_law_1.2,2.1209535598754883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,64,power_law_1.2,0.026649600267410277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,64,power_law_1.2,0.026393601298332216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,64,power_law_1.2,0.02707200050354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,64,power_law_1.2,0.030015999078750612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,64,power_law_1.2,0.030348798632621764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,64,power_law_1.2,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,64,power_law_1.2,0.031430399417877196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,64,power_law_1.2,0.031667199730873105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,64,power_law_1.2,0.03207040131092072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,64,power_law_1.2,0.03188480138778686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,64,power_law_1.2,0.03307519853115082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,64,power_law_1.2,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,64,power_law_1.2,0.034457600116729735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,64,power_law_1.2,0.0369024008512497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,64,power_law_1.2,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,64,power_law_1.2,0.04206080138683319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,64,power_law_1.2,0.04800640046596527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,64,power_law_1.2,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,64,power_law_1.2,0.06069120168685913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,64,power_law_1.2,0.07927680015563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,64,power_law_1.2,0.10785919427871704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,64,power_law_1.2,0.14185600280761718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,64,power_law_1.2,0.1810495972633362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,64,power_law_1.2,0.2519551992416382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,2,balanced,0.02363733450571696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,2,balanced,0.0271573339899381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,2,balanced,0.04082666585842768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,2,balanced,0.0573226660490036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,2,balanced,0.08530132969220479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,2,balanced,0.10090667009353638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,2,balanced,0.1039680043856303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,2,balanced,0.11249066392580669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,2,balanced,0.11556800206502278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,2,balanced,0.1204853355884552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,2,balanced,0.11819733182589214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,64,power_law_1.2,0.358572793006897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,2,balanced,0.12678933143615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,2,balanced,0.13180266817410788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,2,balanced,0.12754133343696594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,2,balanced,0.13618666927019754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,2,balanced,0.1407786707083384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,2,balanced,0.17076265811920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,2,balanced,0.16781334082285562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,2,balanced,0.17299199104309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,2,balanced,0.277621328830719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,2,balanced,0.28902933994929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,2,balanced,0.315994660059611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,2,balanced,0.3291413386662801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,2,balanced,0.5455306768417358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,2,balanced,0.6076853275299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,2,balanced,0.867957353591919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,2,balanced,1.1374666690826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,1,power_law_1.01,0.09323520064353943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,1,power_law_1.01,0.10032000541687011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,1,power_law_1.01,0.11727360486984253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,1,power_law_1.01,0.15054080486297608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,1,power_law_1.01,0.20553600788116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,1,power_law_1.01,0.25143680572509763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,1,power_law_1.01,0.28954238891601564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,1,power_law_1.01,0.34011518955230713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,1,power_law_1.01,0.3510783910751343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,1,power_law_1.01,0.35795838832855226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,1,power_law_1.01,0.37163519859313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,1,power_law_1.01,0.38355839252471924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,1,power_law_1.01,0.38359038829803466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,1,power_law_1.01,0.4142591953277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,1,power_law_1.01,0.41422080993652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,1,power_law_1.01,0.43251838684082033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,1,power_law_1.01,0.4514624118804932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,1,power_law_1.01,0.5057727813720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.01,0.5494463920593262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.01,0.644377613067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.01,0.7292672157287597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.01,0.9300992012023925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.01,1.121286392211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.01,1.5171711921691895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,1,power_law_1.01,2.021772766113281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,1,power_law_1.01,2.791756820678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,1,power_law_1.01,3.4534976959228514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,64,power_law_1.2,0.5379007816314697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,8,1,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,8,1,power_law_1.2,0.05926399827003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,8,1,power_law_1.2,0.07845759987831116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,8,1,power_law_1.2,0.1090880036354065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,8,1,power_law_1.2,0.1398527979850769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,8,1,power_law_1.2,0.16721919775009156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,8,1,power_law_1.2,0.2354111909866333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,8,1,power_law_1.2,0.2390592098236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,8,1,power_law_1.2,0.25043840408325196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,8,1,power_law_1.2,0.2537983894348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,8,1,power_law_1.2,0.2708479881286621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,8,1,power_law_1.2,0.2848639965057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,8,1,power_law_1.2,0.2935231924057007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,8,1,power_law_1.2,0.3101952075958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,8,1,power_law_1.2,0.3342144012451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,8,1,power_law_1.2,0.35306880474090574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,8,1,power_law_1.2,0.36344959735870364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,8,1,power_law_1.2,0.4294464111328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,8,1,power_law_1.2,0.47207040786743165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,8,1,power_law_1.2,0.570636796951294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,8,1,power_law_1.2,0.6489727973937989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,8,1,power_law_1.2,0.8348992347717286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,8,1,power_law_1.2,1.056768035888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,8,1,power_law_1.2,1.4017344474792481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,8,1,power_law_1.2,1.760646438598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,8,1,power_law_1.2,2.4655296325683596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,8,1,power_law_1.2,3.2163455963134764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,4,power_law_1.01,0.02250239998102188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,4,power_law_1.01,0.02184319943189621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,4,power_law_1.01,0.021209600567817687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,4,power_law_1.01,0.02074880003929138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,4,power_law_1.01,0.021913599967956544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,4,power_law_1.01,0.025043201446533204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,4,power_law_1.01,0.03709439933300018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,4,power_law_1.01,0.029068800806999206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,4,power_law_1.01,0.03675520122051239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,4,power_law_1.01,0.03210879862308502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,4,power_law_1.01,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,4,power_law_1.01,0.04017280042171478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,4,power_law_1.01,0.04467839896678925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,4,power_law_1.01,0.04645760059356689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,4,power_law_1.01,0.06305279731750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,4,power_law_1.01,0.05541759729385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,64,power_law_1.2,0.6935423851013184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,4,power_law_1.01,0.06275839805603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,4,power_law_1.01,0.08312960267066956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.01,0.09903360009193421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.01,0.13684480190277098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.01,0.17144960165023804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.01,0.2384768009185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.01,0.2991872072219849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,128,power_law_1.01,0.03927040100097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,128,power_law_1.01,0.0392192006111145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,128,power_law_1.01,0.03873920142650604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,128,power_law_1.01,0.041510400176048276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,128,power_law_1.01,0.04273920059204102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,128,power_law_1.01,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.01,0.4355648040771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.01,0.5624576091766358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,128,power_law_1.01,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,128,power_law_1.01,0.04618239998817444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,128,power_law_1.01,0.04700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,128,power_law_1.01,0.04787200093269348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,128,power_law_1.01,0.04878720045089722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,8,balanced,0.05392533540725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,8,balanced,0.05590933561325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,8,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,8,balanced,0.06955733398596446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,8,balanced,0.07904533545176189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,8,balanced,0.12070399522781372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,8,balanced,0.11266133189201355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,8,balanced,0.11006933450698853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,128,power_law_1.01,0.050783997774124144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,8,balanced,0.11443733175595601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,8,balanced,0.11266666650772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,8,balanced,0.11409067114194234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,128,power_law_1.01,0.05002239942550659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,8,balanced,0.11425600449244182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,8,balanced,0.11633066336313884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,8,balanced,0.11580800016721089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,8,balanced,0.1220746636390686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,8,balanced,0.12713600198427835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,8,balanced,0.1274720033009847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,8,balanced,0.13805333773295084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,8,balanced,0.14752533038457236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,8,balanced,0.16416533788045248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,8,balanced,0.1783199906349182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,8,balanced,0.21617066860198975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,8,balanced,0.2519200046857198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,8,balanced,0.3558666706085205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,8,balanced,0.430186669031779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,8,balanced,0.6073333422342936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.01,0.8278016090393067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,1,power_law_1.2,0.016966399550437928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,128,power_law_1.01,0.054585599899291994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,1,power_law_1.2,0.01863040030002594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,1,power_law_1.2,0.021881599724292756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,128,power_law_1.01,0.05928320288658142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,1,power_law_1.2,0.024281600117683412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,128,power_law_1.01,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,1,power_law_1.2,0.02868480086326599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,1,power_law_1.2,0.038361600041389464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,1,power_law_1.2,0.04123519957065582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,1,power_law_1.2,0.04196479916572571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,1,power_law_1.2,0.042982399463653564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,1,power_law_1.2,0.044409599900245664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,8,balanced,0.7762933572133383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,1,power_law_1.2,0.047193598747253415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,128,power_law_1.01,0.0636352002620697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,1,power_law_1.2,0.04839679896831513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,128,power_law_1.01,0.0714303970336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,1,power_law_1.2,0.049856001138687135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,128,power_law_1.01,0.07691519856452941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,1,power_law_1.2,0.053395199775695804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,1,power_law_1.2,0.051072001457214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,1,power_law_1.2,0.05273600220680237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,1,power_law_1.2,0.057817602157592775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,1,power_law_1.2,0.06327679753303528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,1,power_law_1.2,0.06801279783248901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,1,power_law_1.2,0.11918079853057861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,128,power_law_1.01,0.09470720291137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,128,power_law_1.01,0.10635520219802856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,128,power_law_1.01,0.14046080112457277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,128,power_law_1.01,0.17737599611282348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,128,power_law_1.01,0.2426624059677124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,128,power_law_1.01,0.30587520599365237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,128,power_law_1.01,0.4549248218536377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,1,power_law_1.2,0.1363136053085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,128,power_law_1.01,0.5972864151000976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,1,power_law_1.2,0.13050240278244019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,1,power_law_1.2,0.1487552046775818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,1,power_law_1.2,0.1858240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,1,power_law_1.2,0.2264575958251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,1,power_law_1.2,0.30847361087799074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.01,1.0910335540771485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,1,power_law_1.2,0.3895488023757935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,8,power_law_1.01,0.1259775996208191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,8,power_law_1.01,0.1842303991317749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,8,power_law_1.01,0.22669439315795897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,8,power_law_1.01,0.23936638832092286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,8,power_law_1.01,0.26181759834289553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,8,power_law_1.01,0.2660991907119751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,8,power_law_1.01,0.2678911924362183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,8,power_law_1.01,0.24542078971862794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,8,power_law_1.01,0.24326400756835936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,8,power_law_1.01,0.25224320888519286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,8,power_law_1.01,0.27017600536346437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,8,power_law_1.01,0.25021440982818605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,8,power_law_1.01,0.26267518997192385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,8,power_law_1.01,0.26179840564727785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,8,power_law_1.01,0.2717375993728638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,8,power_law_1.01,0.27652480602264407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,8,power_law_1.01,0.2836992025375366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,8,power_law_1.01,0.30271360874176023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,8,power_law_1.01,0.31998720169067385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,8,power_law_1.01,0.3556544065475464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,8,power_law_1.01,0.40042881965637206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,8,power_law_1.01,0.5120255947113037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,8,power_law_1.01,0.5848703861236573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,8,power_law_1.01,0.7541120052337646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,8,power_law_1.01,0.9259072303771972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,8,power_law_1.01,1.191321563720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,8,power_law_1.01,1.6166719436645507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,2,power_law_1.2,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,2,power_law_1.2,0.0585536003112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,2,power_law_1.2,0.06381440162658691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,2,power_law_1.2,0.0738048017024994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,2,power_law_1.2,0.07537279725074768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,2,power_law_1.2,0.07811200022697448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,2,power_law_1.2,0.07657600045204163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,2,power_law_1.2,0.07466880083084107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,2,power_law_1.2,0.07375360131263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,2,power_law_1.2,0.0735360026359558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,2,power_law_1.2,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,2,power_law_1.2,0.07860479950904846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,2,power_law_1.2,0.07887359857559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,2,power_law_1.2,0.07994239926338195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,2,power_law_1.2,0.096697598695755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,2,power_law_1.2,0.0945472002029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,2,power_law_1.2,0.10597120523452759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,2,power_law_1.2,0.11470719575881957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.2,0.13026560544967652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,8,power_law_1.2,0.024172799289226533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,8,power_law_1.2,0.026342400908470155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.2,0.16117119789123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,8,power_law_1.2,0.032691198587417605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,8,power_law_1.2,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,8,power_law_1.2,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,8,power_law_1.2,0.0417279988527298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,8,power_law_1.2,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,8,power_law_1.2,0.044627198576927186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,8,power_law_1.2,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,8,power_law_1.2,0.056704002618789676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,8,power_law_1.2,0.05724800229072571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,8,power_law_1.2,0.05200639963150024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,8,power_law_1.2,0.08509439826011658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,8,power_law_1.2,0.08522239923477173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,8,power_law_1.2,0.08477439880371093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.2,0.1763584017753601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,8,power_law_1.2,0.08598399758338929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,8,power_law_1.2,0.08369920253753663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,8,power_law_1.2,0.08797439932823181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.2,0.08271999955177307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.2,0.10539519786834717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.2,0.11727999448776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.2,0.1425984025001526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.2,0.1740671992301941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.2,0.23038079738616943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.2,0.3020927906036377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.2,0.4182784080505371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.2,0.25994880199432374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.2,0.5435711860656738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.2,0.3516223907470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.2,0.41864957809448244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.2,0.5692927837371826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,8,power_law_1.2,0.030246400833129884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,8,power_law_1.2,0.031097599864006044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,8,power_law_1.2,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,8,power_law_1.2,0.0332863986492157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,8,power_law_1.2,0.03281919956207276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,8,power_law_1.2,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,8,power_law_1.2,0.039673599600791934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,8,power_law_1.2,0.03824639916419983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,8,power_law_1.2,0.040531200170516965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,8,power_law_1.2,0.04062080085277557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.2,0.7710464000701904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,8,power_law_1.2,0.04017280042171478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,8,power_law_1.2,0.046700799465179445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,8,power_law_1.2,0.04669440090656281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,8,power_law_1.2,0.047270399332046506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,8,power_law_1.2,0.056390398740768434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,8,power_law_1.2,0.0580672025680542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,8,power_law_1.2,0.06225919723510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,8,power_law_1.2,0.07601919770240784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,8,power_law_1.2,0.08586239814758301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,8,power_law_1.2,0.11670399904251098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,8,power_law_1.2,0.13932160139083863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,16,power_law_1.2,0.02456959933042526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,16,power_law_1.2,0.029606398940086365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,16,power_law_1.2,0.024710400402545928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,16,power_law_1.2,0.023270399868488313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,16,power_law_1.2,0.02359039932489395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,16,power_law_1.2,0.025542399287223815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,16,power_law_1.2,0.026726400852203368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,16,power_law_1.2,0.028543999791145323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,16,power_law_1.2,0.02837119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,16,power_law_1.2,0.02898559868335724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,8,power_law_1.2,0.18435839414596558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,16,power_law_1.2,0.03992320001125336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,16,power_law_1.2,0.04140160083770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,16,power_law_1.2,0.0418368011713028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,16,power_law_1.2,0.05761280059814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,16,power_law_1.2,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,16,power_law_1.2,0.060153597593307497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,16,power_law_1.2,0.07963520288467407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,16,power_law_1.2,0.08891519904136658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.2,0.07215359807014465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.2,0.08184319734573364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.2,0.09406080245971679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.2,0.1196287989616394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.2,0.1462272047996521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.2,0.20597119331359864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,8,power_law_1.2,0.2163072109222412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.2,0.26099200248718263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.2,0.37316479682922366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.2,0.456876802444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,8,balanced,0.02940800040960312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,8,balanced,0.029215998947620392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,8,balanced,0.031136001149813335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,8,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,8,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,8,balanced,0.0397173340121905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,8,balanced,0.0394400010506312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,8,balanced,0.041519999504089355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,8,balanced,0.041509332756201424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,8,balanced,0.04144533226887385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,8,balanced,0.04167999823888143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,8,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,8,balanced,0.04420800010363261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,8,balanced,0.04447466631730398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,8,balanced,0.048207998275756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,8,balanced,0.051370665431022644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,8,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,8,balanced,0.05402133365472158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,8,balanced,0.06217066446940104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,8,balanced,0.07658666869004567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,8,balanced,0.08468266328175862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,8,balanced,0.11730666955312093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,8,balanced,0.14074666301409403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,8,balanced,0.19216533501942953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,8,balanced,0.2437173326810201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,8,balanced,0.3396159807840983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,8,balanced,0.43937067190806073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,8,power_law_1.2,0.3108031988143921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.2,0.9140224456787109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,8,power_law_1.2,0.3837631940841675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,8,power_law_1.2,0.5545407772064209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,4,balanced,0.02548266698916753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,4,balanced,0.02698666602373123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,4,balanced,0.04009599983692169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,4,balanced,0.05438933273156484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,4,balanced,0.07627200086911519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,4,balanced,0.12300800283749898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,4,balanced,0.1280639966328939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,4,balanced,0.13432000080744425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,4,balanced,0.13089066743850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,4,balanced,0.13152533769607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,4,balanced,0.13773333032925925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,4,balanced,0.15099199612935385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,4,balanced,0.1527733306090037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,4,balanced,0.15691199898719788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,4,balanced,0.21468265851338705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,4,balanced,0.21289066473642984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,4,balanced,0.21766932805379233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,4,balanced,0.32442132631937665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,4,balanced,0.3607519865036011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,4,balanced,0.3203519980112712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,4,balanced,0.341599980990092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,4,balanced,0.37694935003916424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,4,balanced,0.41812264919281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,4,balanced,0.4808479944864909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,4,balanced,0.5493333339691162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,4,balanced,0.8917760054270426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,4,balanced,1.0276479721069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,8,power_law_1.2,0.7177279949188232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,16,power_law_1.2,0.09609599709510804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,16,power_law_1.2,0.1379520058631897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,16,power_law_1.2,0.14733439683914185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,16,power_law_1.2,0.14971519708633424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,16,power_law_1.2,0.1558079957962036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,16,power_law_1.2,0.1558784008026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,16,power_law_1.2,0.1568384051322937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,16,power_law_1.2,0.1391167998313904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,16,power_law_1.2,0.14542720317840577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,16,power_law_1.2,0.15407999753952026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,16,power_law_1.2,0.14951679706573487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,16,power_law_1.2,0.15544960498809815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,16,power_law_1.2,0.15485440492630004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,16,power_law_1.2,0.16073600053787232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,16,power_law_1.2,0.16624640226364135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,16,power_law_1.2,0.16657279729843139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,16,power_law_1.2,0.1786944031715393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,16,power_law_1.2,0.18655999898910522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,16,power_law_1.2,0.20245120525360108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,16,power_law_1.2,0.23980159759521485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,16,power_law_1.2,0.27729918956756594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,16,power_law_1.2,0.3434367895126343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,16,power_law_1.2,0.42717437744140624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,16,power_law_1.2,0.5657023906707763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,16,power_law_1.2,0.7051968097686767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,16,power_law_1.2,1.047532844543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,16,power_law_1.2,1.416870403289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,8,power_law_1.2,0.018540799617767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,8,power_law_1.2,0.018681600689888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,8,power_law_1.2,0.018918399512767792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,8,power_law_1.2,0.018745599687099455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,8,power_law_1.2,0.019219200313091277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,8,power_law_1.2,0.021695999801158904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,8,power_law_1.2,0.04472320079803467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,8,power_law_1.2,0.04524160027503967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,8,power_law_1.2,0.0321727991104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,8,power_law_1.2,0.03189760148525238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,8,power_law_1.2,0.03357439935207367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,8,power_law_1.2,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,8,power_law_1.2,0.035334399342536925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,8,power_law_1.2,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,4,balanced,0.05653333167235056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,4,balanced,0.05558399856090546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,4,balanced,0.07702399790287018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,16,balanced,0.029818666477998097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,4,balanced,0.08051200211048126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,16,balanced,0.027434666951497395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,8,power_law_1.2,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,16,balanced,0.027434666951497395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,16,balanced,0.029850666721661884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,16,balanced,0.03182933231194814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,8,power_law_1.2,0.03780480027198792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,16,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,16,balanced,0.031583999594052635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,16,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,16,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,16,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,16,balanced,0.035274667044480644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,16,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,8,power_law_1.2,0.04134399890899658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,8,power_law_1.2,0.043654400110244754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,16,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,16,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,16,balanced,0.039034667114416756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,16,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,16,balanced,0.03965866565704346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,16,balanced,0.043525333205858864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,16,balanced,0.043477331598599754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,16,balanced,0.04961066444714864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,16,balanced,0.0540533314148585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,16,balanced,0.06201600035031637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,16,balanced,0.06633600095907848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,16,balanced,0.09635733564694722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,16,balanced,0.12566933035850525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,16,balanced,0.16844266653060913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.2,0.04611839950084686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,16,balanced,0.21288534005482992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,4,balanced,0.08258133133252461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,4,balanced,0.08273600041866302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,4,balanced,0.08243200182914734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,4,balanced,0.0844053328037262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,4,balanced,0.08469866712888081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,4,balanced,0.08634133140246074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,4,balanced,0.08781333764394124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,4,balanced,0.08773333827654521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,4,balanced,0.09001066287358601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,4,balanced,0.09129066268603007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,4,balanced,0.10282133022944133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,4,balanced,0.10532266894976298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,4,balanced,0.10743467013041179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,4,balanced,0.12849600116411844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,4,balanced,0.13475733002026877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,4,balanced,0.17419199148813883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,4,balanced,0.19548267126083374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.2,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,4,balanced,0.25737067063649494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,4,balanced,0.30821865797042847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,4,balanced,0.43456534544626874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,4,balanced,0.5537333488464355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,4,balanced,0.7985813617706299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,4,balanced,1.0341546535491943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.2,0.075135999917984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.2,0.09887999892234803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,8,power_law_1.01,0.14586880207061767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,8,power_law_1.01,0.2410304069519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,8,power_law_1.01,0.29587841033935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.2,0.12390400171279907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,8,power_law_1.01,0.35041921138763427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,8,power_law_1.01,0.4137983798980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,8,power_law_1.01,0.3653439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,8,power_law_1.01,0.40609278678894045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,8,power_law_1.01,0.37947518825531007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,8,power_law_1.01,0.3803584098815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,8,power_law_1.01,0.4028927803039551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,8,power_law_1.01,0.3928256034851074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,8,power_law_1.01,0.4141183853149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,8,power_law_1.01,0.3870527982711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,8,power_law_1.01,0.40373759269714354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,8,power_law_1.01,0.4140927791595459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,8,power_law_1.01,0.45032320022583006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,8,power_law_1.01,0.4678783893585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,8,power_law_1.01,0.5130368232727051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,8,power_law_1.01,0.5298880100250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,8,power_law_1.01,0.6033215999603272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.2,0.17118719816207886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,8,power_law_1.01,0.5800896167755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,8,power_law_1.01,0.6830399990081787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,8,power_law_1.01,0.7599103927612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,8,power_law_1.01,1.0141504287719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,8,power_law_1.01,1.1313792228698731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,8,power_law_1.01,1.4284607887268066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,8,power_law_1.01,1.8569791793823243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.2,0.21020801067352296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.2,0.3068928003311157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,4,balanced,0.029994666576385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,4,balanced,0.02956799914439519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,4,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,4,balanced,0.04004266609748205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,4,balanced,0.060880000392595925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,4,balanced,0.06260799864927928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,4,balanced,0.06320533156394958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,4,balanced,0.06447466711203258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,4,balanced,0.06401066482067108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,4,balanced,0.06490133206049602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,4,balanced,0.06478933493296306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,4,balanced,0.06673599779605865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,4,balanced,0.07095466554164886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,4,balanced,0.06987200180689494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,4,balanced,0.07703466713428497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,4,balanced,0.07967466612656911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,4,balanced,0.08515733480453491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,4,balanced,0.09324799974759419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,4,balanced,0.10051199793815613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,4,balanced,0.12367467085520427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,4,balanced,0.13708266615867615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,4,balanced,0.1816213329633077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,4,balanced,0.21015999714533487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,4,balanced,0.2998080054918925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,4,balanced,0.3866879940032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,4,balanced,0.5540533463160197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,4,balanced,0.7133599917093912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,4,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,4,balanced,0.031701333820819855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,4,balanced,0.031632001201311745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,4,balanced,0.03530666728814443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,4,balanced,0.03937066594759623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,4,balanced,0.04114133367935816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,4,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,4,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,4,balanced,0.041519999504089355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,4,balanced,0.039919999738534294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,4,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,4,balanced,0.04189866781234741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,4,balanced,0.04388799766699473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,4,balanced,0.04381333291530609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,4,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,4,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,4,balanced,0.051872000098228455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,4,balanced,0.0561653325955073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,4,balanced,0.06439466774463654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,4,balanced,0.07656000057856242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,4,balanced,0.08803733189900716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,4,balanced,0.12185066938400269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,4,balanced,0.14389333128929138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,4,balanced,0.19800533850987753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,4,balanced,0.24496533473332724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,4,balanced,0.35100801785786945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,4,balanced,0.45580267906188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.2,0.39502720832824706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,8,power_law_1.2,0.06446719765663148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,8,power_law_1.2,0.07431039810180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,8,power_law_1.2,0.06786559820175171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,8,power_law_1.2,0.08381440043449402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,8,power_law_1.2,0.08981119990348815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,8,power_law_1.2,0.07937920093536377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,8,power_law_1.2,0.08902400135993957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,8,power_law_1.2,0.09095039963722229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,8,power_law_1.2,0.0944320023059845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,8,power_law_1.2,0.09176959991455078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,8,power_law_1.2,0.09355520009994507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,8,power_law_1.2,0.09610880017280579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,8,power_law_1.2,0.09802240133285522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,8,power_law_1.2,0.10038399696350098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,8,power_law_1.2,0.11248639822006226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,8,power_law_1.2,0.10986239910125732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,8,power_law_1.2,0.11818239688873292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,8,power_law_1.2,0.13570560216903688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,8,power_law_1.2,0.1495296001434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,8,power_law_1.2,0.17987200021743774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,8,power_law_1.2,0.2047424077987671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,8,power_law_1.2,0.28972160816192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,8,power_law_1.2,0.36231040954589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,8,power_law_1.2,0.5044159889221191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,8,power_law_1.2,0.6418943881988526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,8,power_law_1.2,1.0359423637390137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,8,power_law_1.2,1.3624383926391601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,32,2,power_law_1.2,0.07882879972457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,32,2,power_law_1.2,0.08794879913330078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,32,2,power_law_1.2,0.09279999732971192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,32,2,power_law_1.2,0.11227519512176513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,1,balanced,0.0767146646976471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,1,balanced,0.0828906645377477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,1,balanced,0.1076800028483073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,1,balanced,0.15094932913780212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,1,balanced,0.22100800275802612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,1,balanced,0.36683201789855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,1,balanced,0.3733280102411906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,1,balanced,0.3696586688359578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,1,balanced,0.37145066261291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,1,balanced,0.37619201342264813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,1,balanced,0.37628265221913654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,1,balanced,0.3826346794764201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,1,balanced,0.38841064771016437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,1,balanced,0.3914080063501994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,32,2,power_law_1.2,0.1274623990058899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,32,2,power_law_1.2,0.1457919955253601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,1,balanced,0.40030399958292645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,1,balanced,0.4085386594136556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,1,balanced,0.42364799976348877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,1,balanced,0.4718666474024455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,1,balanced,0.5004213253657023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,32,2,power_law_1.2,0.18517760038375855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,32,2,power_law_1.2,0.1991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,1,balanced,0.5798666477203369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,32,2,power_law_1.2,0.19736319780349731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,1,balanced,0.6496906677881876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,32,2,power_law_1.2,0.19370239973068237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,32,2,power_law_1.2,0.19733760356903077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,32,2,power_law_1.2,0.20239999294281005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,32,2,power_law_1.2,0.20067839622497557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,32,2,power_law_1.2,0.20718080997467042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,1,balanced,0.8091786702473959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,32,2,power_law_1.2,0.20880000591278075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,1,balanced,0.9580693244934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,32,2,power_law_1.2,0.21790080070495604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,32,2,power_law_1.2,0.22646400928497315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,1,balanced,1.40610138575236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,32,2,power_law_1.2,0.2579583883285522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,32,2,power_law_1.2,0.28526721000671384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,1,balanced,1.6985866228739421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,32,2,power_law_1.2,0.3387712001800537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,1,balanced,2.5007413228352866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,1,balanced,3.2041492462158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,32,2,power_law_1.2,0.38531200885772704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,32,2,power_law_1.2,0.5079552173614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,32,2,power_law_1.2,0.5993728160858154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,32,2,power_law_1.2,0.8152959823608399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,32,2,power_law_1.2,1.0666751861572266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,32,2,power_law_1.2,1.4326335906982421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,32,2,power_law_1.2,1.874732780456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,16,power_law_1.2,0.026368001103401185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,16,power_law_1.2,0.024409599602222443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,16,power_law_1.2,0.025811201333999632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,16,power_law_1.2,0.02346239984035492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,16,power_law_1.2,0.026073598861694337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,16,power_law_1.2,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,16,power_law_1.2,0.04314239919185638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,16,power_law_1.2,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,16,power_law_1.2,0.06357759833335877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,16,power_law_1.2,0.06286079883575439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,16,power_law_1.2,0.062067198753356936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,16,power_law_1.2,0.06630399823188782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,16,power_law_1.2,0.06684799790382386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,16,power_law_1.2,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,16,power_law_1.2,0.06069120168685913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,16,power_law_1.2,0.05885440111160278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,16,power_law_1.2,0.06314880251884461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,16,power_law_1.2,0.059411197900772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.2,0.06960639953613282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.2,0.08346880078315735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,32,balanced,0.046570668617884316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.2,0.09317119717597962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,32,balanced,0.04849066833655039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,32,balanced,0.05023466547330221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,32,balanced,0.05829866727193197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,32,balanced,0.0627040018637975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,32,balanced,0.06427200138568878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,32,balanced,0.06530666848023732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.2,0.12800639867782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,32,balanced,0.06425066788991292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,32,balanced,0.06407466530799866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,32,balanced,0.06477866570154826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,32,balanced,0.06615466872851054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,32,balanced,0.07018666466077168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,32,balanced,0.07010666529337566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.2,0.1461632013320923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,32,balanced,0.07222933570543925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,32,balanced,0.07629333436489105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.2,0.22004480361938478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,8,32,power_law_1.01,0.047270399332046506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,32,balanced,0.08268799881140391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,32,balanced,0.0862666666507721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,8,32,power_law_1.01,0.048563200235366824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,32,balanced,0.09675733248392741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,32,balanced,0.1088213324546814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,32,balanced,0.14040533701578775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,32,balanced,0.15985600153605142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,32,balanced,0.215503990650177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,32,balanced,0.2047306696573893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,32,balanced,0.2705333431561788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,32,balanced,0.32386134068171185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,32,balanced,0.4383466641108195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,32,balanced,0.5393493175506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,8,32,power_law_1.01,0.0470335990190506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,8,32,power_law_1.01,0.04945279955863953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,8,32,power_law_1.01,0.04981760084629059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,8,32,power_law_1.01,0.0496832013130188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,8,32,power_law_1.01,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,8,32,power_law_1.01,0.05190399885177612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,8,32,power_law_1.01,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,8,32,power_law_1.01,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.2,0.2894975900650024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,8,32,power_law_1.01,0.052288001775741576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,8,32,power_law_1.01,0.05428479909896851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,8,32,power_law_1.01,0.055257600545883176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,8,32,power_law_1.01,0.05592319965362549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,8,32,power_law_1.01,0.061222398281097413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,8,32,power_law_1.01,0.061414402723312375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,8,32,power_law_1.01,0.06402559876441956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,8,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,8,balanced,0.03573333223660787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,8,balanced,0.03548266738653183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,8,balanced,0.03735466549793879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,8,balanced,0.039173332353432976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,8,balanced,0.0384853333234787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,8,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,8,balanced,0.03923733284076055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,8,balanced,0.03897066662708918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,8,balanced,0.03809600075085958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,8,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,8,balanced,0.037776000797748566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,8,balanced,0.03993066648642222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,8,balanced,0.0399893323580424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,8,32,power_law_1.01,0.07077760100364686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,8,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,8,balanced,0.043568000197410583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,8,balanced,0.04456000030040741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,8,32,power_law_1.01,0.07858560085296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,8,balanced,0.047983999053637184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,8,balanced,0.0481279989083608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,8,32,power_law_1.01,0.09039999842643738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,8,balanced,0.055861334005991616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,8,balanced,0.05861866474151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,8,balanced,0.07247999807198842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,8,32,power_law_1.01,0.09953280091285706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,8,balanced,0.08356799681981404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,8,balanced,0.10949866970380147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,8,32,power_law_1.01,0.12232320308685303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,8,balanced,0.12928533554077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,8,balanced,0.17152533928553262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,8,32,power_law_1.01,0.15438719987869262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,8,balanced,0.21594132979710898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,8,32,power_law_1.01,0.20030720233917237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.2,0.38807680606842043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,8,32,power_law_1.01,0.24519679546356202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,8,32,power_law_1.01,0.34848001003265383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,8,32,power_law_1.01,0.4318336009979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.2,0.514847993850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,2,power_law_1.01,0.09002879858016968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,2,power_law_1.01,0.10451840162277222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,2,power_law_1.01,0.12092800140380859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,2,power_law_1.01,0.14071040153503417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,2,power_law_1.01,0.1627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,2,power_law_1.01,0.1791551947593689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,2,power_law_1.01,0.2054271936416626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,2,power_law_1.01,0.20325119495391847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,2,power_law_1.01,0.20943360328674315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,2,power_law_1.01,0.2195199966430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,2,power_law_1.01,0.22033278942108153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,2,power_law_1.01,0.22704639434814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,2,power_law_1.01,0.22545280456542968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,2,power_law_1.01,0.2373823881149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,2,power_law_1.01,0.24445440769195556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,2,power_law_1.01,0.24456319808959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,2,power_law_1.01,0.25829119682312013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,2,power_law_1.01,0.2892863988876343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,2,power_law_1.01,0.3133951902389526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,2,power_law_1.01,0.3738111972808838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,2,power_law_1.01,0.45795841217041017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,128,balanced,0.032144000132878624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,2,power_law_1.01,0.5785024166107178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,128,balanced,0.03138133386770884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,128,balanced,0.0316746657093366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,128,balanced,0.03377600014209747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,2,power_law_1.01,0.7246784210205078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,128,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,2,power_law_1.01,1.0019840240478515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,2,power_law_1.01,1.2598464012145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,2,power_law_1.01,1.8516735076904296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,2,power_law_1.01,2.386073684692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,64,power_law_1.01,0.031795200705528257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,64,power_law_1.01,0.03242239952087402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,64,power_law_1.01,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,64,power_law_1.01,0.03253760039806366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,64,power_law_1.01,0.033529600501060484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,64,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,64,power_law_1.01,0.03399679958820343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,64,power_law_1.01,0.0332863986492157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,64,power_law_1.01,0.03375360071659088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,64,power_law_1.01,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,64,power_law_1.01,0.035104000568389894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,64,power_law_1.01,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,64,power_law_1.01,0.03612160086631775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,64,power_law_1.01,0.03767040073871612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,64,power_law_1.01,0.04068480134010315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,64,power_law_1.01,0.04094719886779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,64,power_law_1.01,0.04184960126876831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,64,power_law_1.01,0.04353919923305512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,64,power_law_1.01,0.045952001214027406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,64,power_law_1.01,0.0497408002614975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,64,power_law_1.01,0.05978239774703979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,64,power_law_1.01,0.06558719873428345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,64,power_law_1.01,0.07339519858360291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,64,power_law_1.01,0.099891197681427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,64,power_law_1.01,0.1155519962310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,64,power_law_1.01,0.15360000133514404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,64,power_law_1.01,0.21679999828338622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,32,power_law_1.01,0.06583679914474487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,32,power_law_1.01,0.06114559769630432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,32,power_law_1.01,0.05859839916229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,32,power_law_1.01,0.05927039980888367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,32,power_law_1.01,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,32,power_law_1.01,0.06061440110206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,32,power_law_1.01,0.06295679807662964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,32,power_law_1.01,0.06358399987220764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,32,power_law_1.01,0.06375679969787598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,32,power_law_1.01,0.0636031985282898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,32,power_law_1.01,0.06525440216064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,32,power_law_1.01,0.06668800115585327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,32,power_law_1.01,0.06636800169944763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,32,power_law_1.01,0.06737279891967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,32,power_law_1.01,0.0732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,32,power_law_1.01,0.07337599992752075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,32,power_law_1.01,0.07701119780540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,32,power_law_1.01,0.08476799726486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,32,power_law_1.01,0.09507840275764465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,32,power_law_1.01,0.11173759698867798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,32,power_law_1.01,0.12700799703598023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,32,power_law_1.01,0.16317440271377565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,32,power_law_1.01,0.19408639669418334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,32,power_law_1.01,0.25861759185791017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,32,power_law_1.01,0.34438400268554686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,32,power_law_1.01,0.47422080039978026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,32,power_law_1.01,0.5990975856781006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,128,balanced,0.03721066564321518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,4,power_law_1.2,0.04090240001678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,4,power_law_1.2,0.04344959855079651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,4,power_law_1.2,0.04482559859752655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,4,power_law_1.2,0.048556798696517946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,4,power_law_1.2,0.050937598943710326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,4,power_law_1.2,0.05656319856643677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,4,power_law_1.2,0.06037759780883789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,4,power_law_1.2,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,4,power_law_1.2,0.06245120167732239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,4,power_law_1.2,0.06343680024147033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,4,power_law_1.2,0.06497920155525208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,4,power_law_1.2,0.06861439943313599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,4,power_law_1.2,0.07036799788475037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,4,power_law_1.2,0.07663999795913697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,4,power_law_1.2,0.08468480110168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,4,power_law_1.2,0.08949120044708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,4,power_law_1.2,0.09063040018081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,4,power_law_1.2,0.10194560289382934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,4,power_law_1.01,0.01889919936656952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,4,power_law_1.2,0.11490559577941895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,4,power_law_1.01,0.01918720006942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,4,power_law_1.2,0.1463487982749939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,4,power_law_1.2,0.1633792042732239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,4,power_law_1.2,0.21967999935150145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,4,power_law_1.2,0.2605760097503662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,4,power_law_1.2,0.3823168039321899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,4,power_law_1.01,0.020601600408554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,4,power_law_1.01,0.021011200547218323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,4,power_law_1.2,0.48140802383422854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,4,power_law_1.01,0.022041599452495574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,4,power_law_1.01,0.023871999979019166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,4,power_law_1.2,0.6875840187072754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,4,power_law_1.01,0.02503040134906769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,4,power_law_1.01,0.02707839906215668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,4,power_law_1.01,0.02770560085773468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,4,power_law_1.01,0.02812800109386444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,4,power_law_1.2,0.9825280189514161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,4,power_law_1.01,0.028403198719024657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,4,power_law_1.01,0.03585279881954193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,4,power_law_1.01,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,4,power_law_1.01,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,4,power_law_1.01,0.051648002862930295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,4,power_law_1.01,0.052697598934173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,4,power_law_1.01,0.06418560147285461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,4,power_law_1.01,0.06343680024147033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,4,power_law_1.01,0.06699519753456115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,4,power_law_1.01,0.07452160120010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,4,power_law_1.01,0.0814848005771637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,4,power_law_1.01,0.09581440091133117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,16,balanced,0.03169599920511246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,16,balanced,0.03137599925200144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,16,balanced,0.033674667278925575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,16,balanced,0.057162667314211525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,16,balanced,0.07849066456158955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,16,balanced,0.08026666442553203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,16,balanced,0.07976000010967255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,16,balanced,0.07874133189519246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,16,balanced,0.08037333190441132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,16,balanced,0.08066666622956593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,16,balanced,0.08060266574223836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,16,balanced,0.0820000022649765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,16,balanced,0.08271466692288716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,16,balanced,0.08216000099976857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,16,balanced,0.08747733632723491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,16,balanced,0.08801600337028503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,4,power_law_1.01,0.11679999828338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,4,power_law_1.01,0.15134079456329347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,4,power_law_1.01,0.18369280099868773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,4,power_law_1.01,0.2598335981369019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,4,power_law_1.01,0.3319360017776489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,16,balanced,0.08691733082135518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,16,balanced,0.08901333808898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,16,balanced,0.09335466225941975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,16,balanced,0.10518399874369304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,16,balanced,0.10604266325632732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,16,balanced,0.1220906674861908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,16,balanced,0.13495999574661255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,16,balanced,0.16709333658218384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,16,balanced,0.20106667280197144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,16,balanced,0.2529120047887166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,16,balanced,0.30291734139124554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,128,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,64,power_law_1.01,0.029420799016952513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,64,power_law_1.01,0.027577599883079527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,64,power_law_1.01,0.02686080038547516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,64,power_law_1.01,0.029209598898887634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,64,power_law_1.01,0.02900480031967163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,64,power_law_1.01,0.030752000212669373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,64,power_law_1.01,0.030847999453544616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,64,power_law_1.01,0.03084160089492798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,64,power_law_1.01,0.03136639893054962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,64,power_law_1.01,0.031769600510597226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,64,power_law_1.01,0.03250559866428375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,64,power_law_1.01,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,64,power_law_1.01,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,64,power_law_1.01,0.035180801153182985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,64,power_law_1.01,0.03806079924106598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,64,power_law_1.01,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,64,power_law_1.01,0.04087679982185364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,64,power_law_1.01,0.04559360146522522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,64,power_law_1.01,0.04475519955158234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,64,power_law_1.01,0.052908802032470705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,64,power_law_1.01,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,64,power_law_1.01,0.06724479794502258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,64,power_law_1.01,0.08739200234413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,64,power_law_1.01,0.10273920297622681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,64,power_law_1.01,0.14167040586471558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,64,power_law_1.01,0.18695679903030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,64,power_law_1.01,0.26477439403533937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,64,balanced,0.0768693337837855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,64,balanced,0.039605334401130676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,64,balanced,0.040896000961462654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,64,balanced,0.04035733391841253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,64,balanced,0.048165331284205117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,64,balanced,0.048010667165120445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,64,balanced,0.045968001087506614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,64,balanced,0.04797333478927612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,64,balanced,0.04931733508904775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,64,balanced,0.047930667797724404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,64,balanced,0.04782933493455251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,64,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,64,balanced,0.049679999550183616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,64,balanced,0.04938133557637533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,64,balanced,0.05205333232879639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,64,balanced,0.05287466446558634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,64,balanced,0.05332799752553304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,64,balanced,0.05509866774082184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,64,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,64,balanced,0.0601440022389094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,64,balanced,0.06406933565934499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,64,balanced,0.07866133252779643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,64,balanced,0.08891200025876363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,64,balanced,0.11362133423487346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,64,balanced,0.13370666901270548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,64,balanced,0.17529600858688354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,64,balanced,0.2193386753400167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,4,power_law_1.01,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,4,power_law_1.01,0.04986880123615265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,16,power_law_1.01,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,16,power_law_1.01,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,16,power_law_1.01,0.03852159976959228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,16,power_law_1.01,0.039366400241851805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,16,power_law_1.01,0.03893119990825653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,16,power_law_1.01,0.04126720130443573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,16,power_law_1.01,0.04129279851913452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,16,power_law_1.01,0.04137600064277649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,16,power_law_1.01,0.04178560078144074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,16,power_law_1.01,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,16,power_law_1.01,0.04312959909439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,16,power_law_1.01,0.043321600556373595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,16,power_law_1.01,0.04337919950485229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,16,power_law_1.01,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,16,power_law_1.01,0.04746879935264588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,4,power_law_1.01,0.05257599949836731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,16,power_law_1.01,0.048825600743293764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,4,power_law_1.01,0.06614400148391723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,16,power_law_1.01,0.04986880123615265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,4,power_law_1.01,0.06809599995613098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,16,power_law_1.01,0.05549439787864685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,4,power_law_1.01,0.08269439935684204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,4,power_law_1.01,0.09208959937095643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,4,power_law_1.01,0.0927232027053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,4,power_law_1.01,0.09397119879722596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,4,power_law_1.01,0.0983295977115631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,4,power_law_1.01,0.1017024040222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,4,power_law_1.01,0.10252799987792968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,16,power_law_1.01,0.05888640284538269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,16,power_law_1.01,0.07056000232696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,16,power_law_1.01,0.07608320116996765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,16,power_law_1.01,0.10374399423599243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,16,power_law_1.01,0.106278395652771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,16,power_law_1.01,0.1362944006919861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,16,power_law_1.01,0.1834879994392395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,16,power_law_1.01,0.24676480293273925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,16,power_law_1.01,0.3739135980606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,4,power_law_1.01,0.12426879405975341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,4,power_law_1.01,0.11571840047836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,4,power_law_1.01,0.1174847960472107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,4,power_law_1.01,0.20206079483032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,4,power_law_1.01,0.21724801063537597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,4,power_law_1.01,0.17548799514770508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.01,0.19098880290985107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.01,0.19226880073547364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.01,0.21391360759735106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.01,0.24791040420532226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.01,0.29626240730285647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.01,0.387993597984314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.01,0.48816637992858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.01,0.6808256149291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.01,0.8546367645263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,128,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,4,balanced,0.06079466640949249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,4,balanced,0.06439466774463654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,4,balanced,0.07047466437021892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,4,balanced,0.08828799923261006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,4,balanced,0.1202826698621114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,4,balanced,0.1817973256111145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,4,balanced,0.16775466998418173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,4,balanced,0.17016534010569254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,4,balanced,0.1691946585973104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,4,balanced,0.16875199476877847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,4,balanced,0.16795200109481812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,4,balanced,0.16977065801620483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,4,balanced,0.1749066710472107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,4,balanced,0.17213332653045654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,4,balanced,0.1794933279355367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,4,balanced,0.1818079948425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,4,balanced,0.1851146618525187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,4,balanced,0.199455996354421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,4,balanced,0.2113866607348124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,4,balanced,0.23576533794403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,4,balanced,0.25588266054789227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,4,balanced,0.30931733051935834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,4,balanced,0.3577386538187663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,4,balanced,0.5211840073267618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,4,balanced,0.6262186765670776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,4,balanced,0.8868107000986735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,4,balanced,1.1512266794840496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,128,balanced,0.03169066707293192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,128,balanced,0.03145066648721695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,128,balanced,0.0315786674618721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,128,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,128,balanced,0.035445332527160645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,128,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,128,balanced,0.034703999757766724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,128,balanced,0.03732266773780187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,128,balanced,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,128,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,128,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,128,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,128,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,128,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,128,balanced,0.03965866565704346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,128,balanced,0.041322665909926094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,128,balanced,0.04162133236726125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,128,balanced,0.04349866509437561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,128,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,128,balanced,0.0476800004641215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,128,balanced,0.04789333542188009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,128,balanced,0.052229334910710655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,128,balanced,0.05613866448402405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,128,balanced,0.06846933563550313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,128,balanced,0.07867733140786488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,128,balanced,0.10378666718800862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,128,balanced,0.12327466408411662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,128,balanced,0.039861333866914116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,8,power_law_1.2,0.06855040192604064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,8,power_law_1.2,0.06820480227470398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,8,power_law_1.2,0.07573760151863099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,8,power_law_1.2,0.08947839736938476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,8,power_law_1.2,0.08280320167541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,8,power_law_1.2,0.0889855980873108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,8,power_law_1.2,0.09440000057220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,8,power_law_1.2,0.09224320054054261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,8,power_law_1.2,0.09651200175285339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,8,power_law_1.2,0.09651200175285339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,8,power_law_1.2,0.09774079918861389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,8,power_law_1.2,0.10136959552764893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,8,power_law_1.2,0.10319360494613647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,8,power_law_1.2,0.10951039791107178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,8,power_law_1.2,0.11153919696807861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,8,power_law_1.2,0.11822719573974609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,8,power_law_1.2,0.12567039728164672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,8,power_law_1.2,0.14261120557785034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.2,0.16417280435562134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.2,0.20325760841369628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.2,0.22464001178741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.2,0.27576959133148193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.2,0.37233920097351075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.2,0.4930751800537109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.2,0.6080383777618408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,64,power_law_1.01,0.05240319967269898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,64,power_law_1.01,0.04325119853019714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,64,power_law_1.01,0.04164479970932007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,64,power_law_1.01,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,64,power_law_1.01,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,64,power_law_1.01,0.04450559914112091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,64,power_law_1.01,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,64,power_law_1.01,0.047116801142692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,64,power_law_1.01,0.047417598962783816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,64,power_law_1.01,0.04733439981937408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,64,power_law_1.01,0.048614400625228885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,64,power_law_1.01,0.05156480073928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,64,power_law_1.01,0.05137280225753784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,64,power_law_1.01,0.055103999376296994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,64,power_law_1.01,0.05953279733657837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,64,power_law_1.01,0.06108800172805786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,64,power_law_1.01,0.06741759777069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,64,power_law_1.01,0.07783039808273315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,64,power_law_1.01,0.08288639783859253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.2,0.844268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,64,power_law_1.01,0.10219520330429077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,64,power_law_1.01,0.11628160476684571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,64,power_law_1.01,0.1458624005317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,64,power_law_1.01,0.17079039812088012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,64,power_law_1.01,0.2311552047729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,64,power_law_1.01,0.2888063907623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,64,power_law_1.01,0.4150656223297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,64,power_law_1.01,0.5250239849090577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,128,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.2,1.0972352027893066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,2,balanced,0.07446399827798207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,2,balanced,0.11498666803042094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,2,balanced,0.16872000694274902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,2,balanced,0.16087999939918518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,2,balanced,0.16381866733233133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,2,balanced,0.15862933794657388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,2,balanced,0.17603733142217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,2,balanced,0.17883199453353882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,2,balanced,0.17256534099578857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,2,balanced,0.16178133090337118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,2,balanced,0.16506666938463846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,2,balanced,0.15659733613332114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,2,balanced,0.15429866313934326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,2,balanced,0.1544533371925354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,2,balanced,0.13672000169754028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,2,balanced,0.14257599910100302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,2,balanced,0.12826666235923767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,2,balanced,0.23274133602778116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,2,balanced,0.19004799922307333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,2,balanced,0.25204267104466754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,2,balanced,0.31360532840092975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,2,balanced,0.4349386692047119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,2,balanced,0.5530986785888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,2,balanced,0.8046826521555582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,2,balanced,1.0560479958852131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,2,balanced,1.555898666381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,2,balanced,2.0456159909566245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,128,balanced,0.04176533222198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,128,balanced,0.04589866598447164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,128,balanced,0.042352000872294106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,128,balanced,0.04428266485532125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,128,balanced,0.050101334849993386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,128,balanced,0.05007466673851013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,128,balanced,0.050197333097457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,128,balanced,0.05409599840641022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,128,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,128,balanced,0.06451733410358429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,128,balanced,0.0665226678053538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,128,balanced,0.08290666838486989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,128,balanced,0.09699733058611552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,128,balanced,0.12370666861534119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,128,balanced,0.14831466476122537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,128,balanced,0.20729066928227743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,128,balanced,0.24778133630752563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,1,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,1,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,1,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,1,balanced,0.06809600194295247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,1,balanced,0.10399466753005981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,1,balanced,0.10226133465766907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,1,balanced,0.10331199566523235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,1,balanced,0.10338667035102844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,1,balanced,0.10705600182215373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,1,balanced,0.10842133561770122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,1,balanced,0.10995733737945557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,1,balanced,0.11329600214958191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,1,balanced,0.1151626706123352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,1,balanced,0.122789333264033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,1,balanced,0.1285760005315145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,1,balanced,0.1309706668059031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,1,balanced,0.1421119968096415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,1,balanced,0.16078399618466696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,1,balanced,0.1800480087598165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,1,balanced,0.2516746719678243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,1,balanced,0.2795040011405945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,1,balanced,0.3978773355484009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,1,balanced,0.4750560124715169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,1,balanced,0.6857706705729166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,1,balanced,0.8548479874928793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,1,balanced,1.2542133331298828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,1,balanced,1.6057333946228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,4,power_law_1.2,0.03089280128479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,4,power_law_1.2,0.029747200012207032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,4,power_law_1.2,0.028364801406860353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,4,power_law_1.2,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,4,power_law_1.2,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,4,power_law_1.2,0.038406398892402646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,4,power_law_1.2,0.05803520083427429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,4,power_law_1.2,0.044787201285362246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,4,power_law_1.2,0.06230400204658508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,4,power_law_1.2,0.05059840083122254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,4,power_law_1.2,0.0561024010181427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,4,power_law_1.2,0.06865919828414917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,4,power_law_1.2,0.07502080202102661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,4,power_law_1.2,0.07264000177383423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,4,power_law_1.2,0.10425599813461303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,4,power_law_1.2,0.08419200181961059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,4,power_law_1.2,0.09831039905548096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,4,power_law_1.2,0.13756799697875977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.2,0.1624127984046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.2,0.2296895980834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.2,0.29049599170684814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.2,0.41863040924072265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.2,0.5298111915588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.2,0.7810431957244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,64,balanced,0.05496533215045929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,64,balanced,0.05429866909980774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,64,balanced,0.0562720000743866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,64,balanced,0.05832533538341522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,64,balanced,0.06248533229033152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,64,balanced,0.07748800019423167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,64,balanced,0.07301866511503856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,64,balanced,0.07717866698900859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,64,balanced,0.07728533446788788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,64,balanced,0.07318399846553802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,64,balanced,0.07245866457621257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,64,balanced,0.0776800016562144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,64,balanced,0.07812266548474629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,64,balanced,0.07724266747633617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,64,balanced,0.08098666866620381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,64,balanced,0.08096000055472057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,64,balanced,0.0836906631787618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,64,balanced,0.09358400106430054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,64,balanced,0.09992532928784688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,64,balanced,0.11127466956774394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,64,balanced,0.12657599647839865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,64,balanced,0.15150400002797446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,16,power_law_1.01,0.05129600167274475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,64,balanced,0.1758293310801188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,64,balanced,0.23774933815002441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,64,balanced,0.2839360038439433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,64,balanced,0.3885440031687419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,64,balanced,0.49664000670115155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,16,power_law_1.01,0.058534401655197146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,16,power_law_1.01,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,16,power_law_1.01,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,16,power_law_1.01,0.0574783980846405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.2,1.0191743850708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,16,power_law_1.01,0.059385597705841064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,16,power_law_1.01,0.05988479852676391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,16,power_law_1.01,0.05909119844436646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,16,power_law_1.01,0.0646399974822998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,16,power_law_1.01,0.06263039708137512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,16,power_law_1.01,0.06471679806709289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,16,power_law_1.01,0.06622080206871032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,16,power_law_1.01,0.067603200674057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,16,power_law_1.01,0.07033600211143494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,16,power_law_1.01,0.07682560086250305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,16,power_law_1.01,0.08316799998283386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,16,power_law_1.01,0.08512639999389648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,16,power_law_1.01,0.09642879962921143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,16,power_law_1.01,0.09742079973220825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,16,power_law_1.01,0.11894400119781494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,16,power_law_1.01,0.13634560108184815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,16,power_law_1.01,0.1737663984298706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,16,power_law_1.01,0.20599040985107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,16,power_law_1.01,0.26871039867401125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,16,power_law_1.01,0.32647678852081297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,16,power_law_1.01,0.4702591896057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,16,power_law_1.01,0.6018303871154785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.2,1.5112704277038573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,4,power_law_1.01,0.02627840042114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,4,power_law_1.01,0.0253248006105423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,4,power_law_1.01,0.023897600173950196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,4,power_law_1.01,0.030086401104927062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,4,power_law_1.01,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,4,power_law_1.01,0.032416000962257385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,4,power_law_1.01,0.043987199664115906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,4,power_law_1.01,0.03770880103111267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,4,power_law_1.01,0.08186240196228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,4,power_law_1.01,0.07980160117149353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,4,power_law_1.01,0.0807424008846283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,4,power_law_1.01,0.0839743971824646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,4,power_law_1.01,0.07749119997024537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,4,power_law_1.01,0.06116480231285095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,4,power_law_1.01,0.08846719861030579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,4,power_law_1.01,0.07390720248222352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,4,power_law_1.01,0.09460480213165283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,4,power_law_1.01,0.11755520105361938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.01,0.14218239784240722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.01,0.20078721046447753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.01,0.25559039115905763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.01,0.35871360301971433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,8,power_law_1.2,0.07978879809379577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,8,power_law_1.2,0.09617279767990113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,8,power_law_1.2,0.09275519847869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,8,power_law_1.2,0.11223679780960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,8,power_law_1.2,0.11909120082855225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,8,power_law_1.2,0.12329599857330323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,8,power_law_1.2,0.12482559680938721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,8,power_law_1.2,0.13159040212631226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,8,power_law_1.2,0.1259071946144104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,8,power_law_1.2,0.12551679611206054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,8,power_law_1.2,0.1288831949234009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,8,power_law_1.2,0.13479679822921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,8,power_law_1.2,0.13433599472045898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,8,power_law_1.2,0.13820799589157104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,8,power_law_1.2,0.1453760027885437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,8,power_law_1.2,0.14881919622421264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.01,0.4685567855834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,8,power_law_1.2,0.15101439952850343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,8,power_law_1.2,0.17716480493545533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,8,power_law_1.2,0.1917248010635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,8,power_law_1.2,0.22703359127044678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,8,power_law_1.2,0.2663680076599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,8,power_law_1.2,0.36457600593566897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,8,power_law_1.2,0.45425920486450194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,8,power_law_1.2,0.6475520133972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.01,0.6773375988006591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,8,power_law_1.2,0.8669695854187012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.2,1.9996671676635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,8,power_law_1.2,1.3515839576721191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,8,power_law_1.2,1.6558399200439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.01,0.8896384239196777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.01,1.307091236114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,2,balanced,0.05509333312511444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,2,balanced,0.06834666430950165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,2,balanced,0.08761599659919739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,8,power_law_1.2,0.09071999788284302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,2,balanced,0.13095466295878092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,2,balanced,0.20229866107304892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,8,power_law_1.2,0.09648640155792236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,2,balanced,0.26528533299763996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,2,balanced,0.2691413362820943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,2,balanced,0.2696853280067444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,2,balanced,0.272490660349528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,2,balanced,0.27160000801086426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,2,balanced,0.2757173379262288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,2,balanced,0.282586673895518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,2,balanced,0.2844480077425639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,2,balanced,0.28961066404978436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,2,balanced,0.3009173274040222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,2,balanced,0.308405339717865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,2,balanced,0.3219573299090068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,2,balanced,0.36903464794158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,2,balanced,0.3818986813227336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,2,8,power_law_1.2,0.11957759857177734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,2,balanced,0.525605320930481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,2,8,power_law_1.2,0.12556159496307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,2,balanced,0.5383626619974772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.01,1.7254528045654296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,2,balanced,0.8442933559417725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,2,balanced,0.8479093710581461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,2,balanced,1.3053279717763264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,2,balanced,1.4243466059366863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,2,8,power_law_1.2,0.12083840370178223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,2,8,power_law_1.2,0.13378560543060303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,2,8,power_law_1.2,0.1375615954399109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,2,8,power_law_1.2,0.1403839945793152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,2,balanced,2.141424020131429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,2,8,power_law_1.2,0.14035199880599974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,2,8,power_law_1.2,0.14103039503097534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,2,balanced,2.6445120175679526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,2,8,power_law_1.2,0.1416383981704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,2,8,power_law_1.2,0.14433280229568482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,2,8,power_law_1.2,0.14958080053329467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,2,8,power_law_1.2,0.1507840037345886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,2,8,power_law_1.2,0.1574463963508606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,2,8,power_law_1.2,0.15859199762344361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,2,8,power_law_1.2,0.17344000339508056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,2,8,power_law_1.2,0.19088000059127808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,2,8,power_law_1.2,0.21877760887145997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,2,8,power_law_1.2,0.2606911897659302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,2,8,power_law_1.2,0.3079040050506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,2,8,power_law_1.2,0.4061823844909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,2,8,power_law_1.2,0.5004223823547364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,2,8,power_law_1.2,0.7805183887481689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,2,8,power_law_1.2,0.948140811920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,2,8,power_law_1.2,1.3477120399475098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,2,8,power_law_1.2,1.8983999252319337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,4,power_law_1.01,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,4,power_law_1.01,0.030419200658798218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,4,power_law_1.01,0.029951998591423036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,4,power_law_1.01,0.032953599095344545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,4,power_law_1.01,0.032876798510551454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,4,power_law_1.01,0.03415040075778961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,4,power_law_1.01,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,4,power_law_1.01,0.035795199871063235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,4,power_law_1.01,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,4,power_law_1.01,0.038073599338531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,4,power_law_1.01,0.03900800049304962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,4,power_law_1.01,0.04030719995498657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,4,power_law_1.01,0.04071680009365082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,4,power_law_1.01,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,4,power_law_1.01,0.051123201847076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,4,power_law_1.01,0.054092800617218016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,4,power_law_1.01,0.06002560257911682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,4,power_law_1.01,0.0699455976486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,4,power_law_1.01,0.08144000172615051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,4,power_law_1.01,0.10980479717254639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,4,power_law_1.01,0.1314560055732727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,4,power_law_1.01,0.1689919948577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,4,power_law_1.01,0.22458879947662352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,4,power_law_1.01,0.2963520050048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,4,power_law_1.01,0.3930943965911865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,4,power_law_1.01,0.5611135959625244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,2,balanced,0.039887999494870506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,2,balanced,0.05307200054327647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,2,balanced,0.07656000057856242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,2,balanced,0.12196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,2,balanced,0.20493866999944052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,2,balanced,0.20959466695785522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,2,balanced,0.20941867431004843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,2,balanced,0.20772800842920938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,2,balanced,0.21027199427286783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,2,balanced,0.21144533157348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,2,balanced,0.21326400836308798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,2,balanced,0.21668267250061035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,2,balanced,0.21648534138997397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,2,balanced,0.21968533595403036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,2,balanced,0.2294293244679769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,2,balanced,0.23054399092992148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,2,balanced,0.2355626622835795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,2,balanced,0.2472106615702311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,2,balanced,0.26389867067337036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,2,balanced,0.30138667424519855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,2,balanced,0.32578667004903156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,2,balanced,0.3983999888102214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,2,balanced,0.4434719880421956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,2,balanced,0.6038613319396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,2,balanced,0.7030080159505209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,2,balanced,0.9955999851226807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,2,balanced,1.2793707052866619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,4,power_law_1.01,0.6719488143920899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,4,power_law_1.2,0.050470399856567386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,4,power_law_1.2,0.05511680245399475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,4,power_law_1.2,0.0639743983745575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,4,power_law_1.2,0.07253760099411011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,4,power_law_1.2,0.08910719752311706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,4,power_law_1.2,0.10510720014572143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,4,power_law_1.2,0.11651840209960937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,4,power_law_1.2,0.11806080341339112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,4,power_law_1.2,0.12020479440689087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,4,power_law_1.2,0.11472640037536622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,4,power_law_1.2,0.11842559576034546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,4,power_law_1.2,0.12088320255279542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,4,power_law_1.2,0.12466559410095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,4,power_law_1.2,0.12702720165252684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,4,power_law_1.2,0.13319040536880494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,4,power_law_1.2,0.13819520473480223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,4,power_law_1.2,0.14569599628448487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,4,power_law_1.2,0.15943679809570313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,4,power_law_1.2,0.16757760047912598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,4,power_law_1.2,0.19303679466247559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,4,power_law_1.2,0.22112638950347902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,4,power_law_1.2,0.28557438850402833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,128,power_law_1.2,0.052243202924728394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,4,power_law_1.2,0.2819200038909912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,128,power_law_1.2,0.052249598503112796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,4,power_law_1.2,0.34999680519104004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,128,power_law_1.2,0.040992000699043275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,4,power_law_1.2,0.4276735782623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,128,power_law_1.2,0.043993601202964784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,128,power_law_1.2,0.04408319890499115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,128,power_law_1.2,0.044563201069831845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,128,power_law_1.2,0.045638400316238406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,128,power_law_1.2,0.045664000511169436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,128,power_law_1.2,0.046086400747299194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,128,power_law_1.2,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,128,power_law_1.2,0.047200000286102294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,128,power_law_1.2,0.04776319861412048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,128,power_law_1.2,0.04897280037403107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,128,power_law_1.2,0.049804800748825075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,128,power_law_1.2,0.05511040091514587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,128,power_law_1.2,0.05559679865837097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,128,power_law_1.2,0.05856639742851257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,128,power_law_1.2,0.06826239824295044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,128,power_law_1.2,0.07208960056304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,128,power_law_1.2,0.08595839738845826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,128,power_law_1.2,0.09975039958953857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,128,power_law_1.2,0.1296895980834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,128,power_law_1.2,0.15913599729537964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,128,power_law_1.2,0.2223680019378662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,128,power_law_1.2,0.28136959075927737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,128,power_law_1.2,0.4157951831817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,128,power_law_1.2,0.546227216720581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,4,power_law_1.2,0.636569595336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,4,power_law_1.2,0.6563327789306641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,8,balanced,0.050586665670077004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,8,balanced,0.05608533322811127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,8,balanced,0.05821333328882853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,8,balanced,0.06924266616503398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,8,balanced,0.0844106674194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,8,balanced,0.09330667058626811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,8,balanced,0.09507733583450317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,8,balanced,0.0906773308912913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,8,balanced,0.09096533060073853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,8,balanced,0.09479467074076335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,8,balanced,0.08904533584912618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,8,balanced,0.09294399619102478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,8,balanced,0.09656000137329102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,8,balanced,0.09298666318257649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,8,balanced,0.09846400221188863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,8,balanced,0.10114666819572449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,8,balanced,0.10083199540774028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,8,balanced,0.11267200112342834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,8,balanced,0.1209386686484019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,8,balanced,0.13476799925168356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,8,balanced,0.1500746707121531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,8,balanced,0.19953600565592447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,8,balanced,0.22804800669352213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,8,balanced,0.3062346577644348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,8,balanced,0.3927573362986247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,8,balanced,0.522437334060669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,8,balanced,0.6840480168660482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,32,4,power_law_1.2,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,32,4,power_law_1.2,0.049420800805091855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,32,4,power_law_1.2,0.047628799080848695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,32,4,power_law_1.2,0.0576960027217865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,32,4,power_law_1.2,0.061875200271606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,32,4,power_law_1.2,0.06892160177230836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,32,4,power_law_1.2,0.08268160223960877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,32,4,power_law_1.2,0.08444799780845642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,32,4,power_law_1.2,0.08459519743919372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,32,4,power_law_1.2,0.08350719809532166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,32,4,power_law_1.2,0.08477439880371093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,32,4,power_law_1.2,0.08577920198440551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,32,4,power_law_1.2,0.0898688018321991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,32,4,power_law_1.2,0.09297919869422913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,32,4,power_law_1.2,0.10166399478912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,32,4,power_law_1.2,0.10625280141830444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,32,4,power_law_1.2,0.11651840209960937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,32,4,power_law_1.2,0.13456640243530274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,32,4,power_law_1.2,0.1515391945838928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,32,4,power_law_1.2,0.17704960107803344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,32,4,power_law_1.2,0.2025279998779297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,32,4,power_law_1.2,0.257804799079895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,32,4,power_law_1.2,0.3158463954925537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,32,4,power_law_1.2,0.45927038192749026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,32,4,power_law_1.2,0.5627903938293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,32,4,power_law_1.2,0.7969791889190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,32,4,power_law_1.2,1.109823989868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,2,power_law_1.01,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,2,power_law_1.01,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,2,power_law_1.01,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,2,power_law_1.01,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,2,power_law_1.01,0.04281600117683411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,2,power_law_1.01,0.052985602617263795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,2,power_law_1.01,0.05377920269966126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,2,power_law_1.01,0.05413119792938233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,2,power_law_1.01,0.05490559935569763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,2,power_law_1.01,0.056364798545837404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,2,power_law_1.01,0.05774719715118408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,2,power_law_1.01,0.061952000856399535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,2,power_law_1.01,0.06384000182151794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,2,power_law_1.01,0.06895359754562377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,2,power_law_1.01,0.07618560194969178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,2,power_law_1.01,0.08069760203361512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,2,power_law_1.01,0.08327680230140685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,2,power_law_1.01,0.09787520170211791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,2,power_law_1.01,0.11555839776992798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,2,power_law_1.01,0.13836159706115722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,2,power_law_1.01,0.187827205657959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,2,power_law_1.01,0.24795520305633545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,2,power_law_1.01,0.28704640865325926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,2,power_law_1.01,0.40929279327392576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,2,power_law_1.01,0.5179135799407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,2,power_law_1.01,0.731987190246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,2,power_law_1.01,0.9424063682556152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,8,power_law_1.01,0.05797119736671448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,8,power_law_1.01,0.06106879711151123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,8,power_law_1.01,0.06320000290870667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,8,power_law_1.01,0.06720640063285828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,8,power_law_1.01,0.07084159851074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,8,power_law_1.01,0.07034239768981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,8,power_law_1.01,0.07580800056457519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,8,power_law_1.01,0.07301759719848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,8,power_law_1.01,0.07535359859466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,8,power_law_1.01,0.07550079822540283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,8,power_law_1.01,0.07682560086250305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,8,power_law_1.01,0.07870079874992371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,8,power_law_1.01,0.08181120157241821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,8,power_law_1.01,0.08379520177841186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,8,power_law_1.01,0.09271680116653443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,8,power_law_1.01,0.09321600198745728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,8,power_law_1.01,0.10483200550079345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,8,power_law_1.01,0.12024960517883301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.01,0.13068799972534179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.01,0.1713088035583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.01,0.1750656008720398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.01,0.2227328062057495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.01,0.2905791997909546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.01,0.3969151973724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.01,0.5165056228637696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.01,0.7049024105072021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.01,0.900921630859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,32,power_law_1.2,0.030169600248336793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,32,power_law_1.2,0.027347201108932497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,32,power_law_1.2,0.027327999472618103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,32,power_law_1.2,0.0293503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,32,power_law_1.2,0.030112001299858093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,32,power_law_1.2,0.031430399417877196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,32,power_law_1.2,0.03214719891548157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,32,power_law_1.2,0.032076799869537355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,64,power_law_1.2,0.030342400074005127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,64,power_law_1.2,0.027852800488471986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,64,power_law_1.2,0.027603200078010558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,64,power_law_1.2,0.029702401161193846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,64,power_law_1.2,0.02961280047893524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,64,power_law_1.2,0.03133440017700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,64,power_law_1.2,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,32,power_law_1.2,0.03290880024433136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,64,power_law_1.2,0.03206399977207184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,32,power_law_1.2,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,32,power_law_1.2,0.03333120048046112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,64,power_law_1.2,0.03221760094165802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,64,power_law_1.2,0.03225600123405457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,64,power_law_1.2,0.033024001121521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,64,power_law_1.2,0.03381119966506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,32,power_law_1.2,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,64,power_law_1.2,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,64,power_law_1.2,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,64,power_law_1.2,0.039603200554847715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,64,power_law_1.2,0.04133760035037994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,64,power_law_1.2,0.04281600117683411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,64,power_law_1.2,0.0472896009683609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,32,power_law_1.2,0.034355199337005614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,64,power_law_1.2,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,64,power_law_1.2,0.05774719715118408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,64,power_law_1.2,0.05912320017814636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,64,power_law_1.2,0.07678719758987426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,32,power_law_1.2,0.035596799850463864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,64,power_law_1.2,0.09365760087966919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,64,power_law_1.2,0.1334015965461731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,64,power_law_1.2,0.16834559440612792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,32,balanced,0.042965332667032875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,32,balanced,0.04252799848715464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,32,balanced,0.042693331837654114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,32,balanced,0.04595200220743815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,32,balanced,0.04757333298524221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,32,balanced,0.050714666644732155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,32,balanced,0.05178666611512502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,32,power_law_1.2,0.04035840034484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,32,balanced,0.05136533578236898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,32,balanced,0.05422399938106537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,32,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,32,power_law_1.2,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,32,balanced,0.054154664278030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,32,balanced,0.0581226646900177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,32,power_law_1.2,0.04439679980278015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,32,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,32,balanced,0.06025599936644236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,32,power_law_1.2,0.04837760031223297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,32,power_law_1.2,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,32,power_law_1.2,0.06653439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,32,balanced,0.06433600187301636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,32,power_law_1.2,0.08278399705886841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,32,balanced,0.06486933430035909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,32,balanced,0.0711839993794759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,32,balanced,0.07640533149242401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,32,power_law_1.2,0.11451519727706909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,32,balanced,0.08011733492215474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,32,balanced,0.0941439966360728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,32,balanced,0.10281067093213399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,32,balanced,0.12993066509564719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,32,balanced,0.15053866306940714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,32,balanced,0.20153599977493286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,32,balanced,0.24437866608301798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,32,balanced,0.3442933162053426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,32,power_law_1.2,0.14886399507522582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,32,balanced,0.4318506717681885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,32,power_law_1.2,0.19650559425354003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,4,power_law_1.2,0.017510400712490083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,4,power_law_1.2,0.016832000017166136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,4,power_law_1.2,0.017113600671291352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,4,power_law_1.2,0.01809920072555542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,4,power_law_1.2,0.018028800189495087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,4,power_law_1.2,0.018240000307559966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,4,power_law_1.2,0.0212351992726326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,4,power_law_1.2,0.018624000251293182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,4,power_law_1.2,0.021657599508762358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,4,power_law_1.2,0.022086399793624877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,4,power_law_1.2,0.022015999257564544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,4,power_law_1.2,0.02311680018901825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,4,power_law_1.2,0.023929600417613984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,4,power_law_1.2,0.030963200330734252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,4,power_law_1.2,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,4,power_law_1.2,0.032287999987602234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,32,power_law_1.2,0.25623040199279784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,4,power_law_1.2,0.03315199911594391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,4,power_law_1.2,0.03800959885120392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,16,power_law_1.2,0.021036800742149354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,16,power_law_1.2,0.025811201333999632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,16,power_law_1.2,0.025235199928283693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,16,power_law_1.2,0.0255295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,16,power_law_1.2,0.024294400215148927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,16,power_law_1.2,0.0240447998046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.2,0.04105600118637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,16,power_law_1.2,0.025446400046348572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,16,power_law_1.2,0.027008000016212463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,16,power_law_1.2,0.02690559923648834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,16,power_law_1.2,0.0274944007396698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,16,power_law_1.2,0.02876800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,16,power_law_1.2,0.030118399858474733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,16,power_law_1.2,0.03115519881248474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,16,power_law_1.2,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,16,power_law_1.2,0.04142079949378967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,16,power_law_1.2,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,16,power_law_1.2,0.05783680081367493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.2,0.05356159806251526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,16,power_law_1.2,0.06358399987220764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,16,power_law_1.2,0.07138559818267823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,16,power_law_1.2,0.09106559753417968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,16,power_law_1.2,0.07013760209083557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,16,power_law_1.2,0.08621439933776856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,16,power_law_1.2,0.09735040068626404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,16,power_law_1.2,0.12450560331344604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,16,power_law_1.2,0.15644799470901488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,16,power_law_1.2,0.22951040267944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,16,power_law_1.2,0.27658240795135497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,64,power_law_1.2,0.2598912000656128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.2,0.05973119735717773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.2,0.07905920147895813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,32,power_law_1.2,0.37897601127624514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.2,0.09624959826469422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,1,balanced,0.05753066639105479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,1,balanced,0.07453333338101704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,1,balanced,0.1027893324693044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,1,balanced,0.1606613298257192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,1,balanced,0.26339733600616455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,1,balanced,0.46296000480651855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,1,balanced,0.4694933493932088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,1,balanced,0.4735039869944255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,1,balanced,0.4807893435160319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,1,balanced,0.4806613524754842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,1,balanced,0.4875146547953288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,1,balanced,0.49371198813120526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,1,balanced,0.501530647277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,1,balanced,0.5079840024312338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,1,balanced,0.52074134349823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,1,balanced,0.5305813153584799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,1,balanced,0.5475253264109293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,1,balanced,0.5943413178126017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,1,balanced,0.6267999807993571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,1,balanced,0.7122933069864908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,1,balanced,0.7930400371551514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,1,balanced,1.0344159603118896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,1,balanced,1.1592533588409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,1,balanced,1.59552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,1,balanced,1.91373872756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,1,balanced,2.7923253377278647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.2,0.1248703956604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,1,balanced,3.535792032877604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,32,power_law_1.2,0.5753983974456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.2,0.15890560150146485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,64,power_law_1.2,0.2990528106689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.2,0.22239999771118163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,1,8,balanced,0.11755733688672383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,1,8,balanced,0.11531200011571248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,1,8,balanced,0.11619200309117635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,1,8,balanced,0.12447999914487202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,1,8,balanced,0.11911466717720032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,1,8,balanced,0.11632532874743144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,1,8,balanced,0.11937066912651062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,1,8,balanced,0.11773332953453064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,1,8,balanced,0.1220693290233612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,1,8,balanced,0.11700266599655151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,1,8,balanced,0.11522133151690166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,1,8,balanced,0.10803199807802837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,1,8,balanced,0.11548800269762675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,1,8,balanced,0.11086400349934895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,1,8,balanced,0.09899200002352397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,1,8,balanced,0.10108799735705058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,1,8,balanced,0.0967733363310496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,1,8,balanced,0.15176533659299216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,1,8,balanced,0.13134400049845377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,1,8,balanced,0.16685867309570312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,1,8,balanced,0.19933332999547324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,1,8,balanced,0.2732853293418884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,1,8,balanced,0.34462400277455646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,1,8,balanced,0.47802666823069256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,1,8,balanced,0.6175786654154459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,1,8,balanced,0.8824213345845541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,1,8,balanced,1.1483573118845622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.2,0.2858880043029785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,8,balanced,0.043696001172065735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,8,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,8,balanced,0.04980266590913137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,8,balanced,0.07321600119272868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,8,balanced,0.0992693305015564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,8,balanced,0.15194666385650635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,8,balanced,0.15315733353296915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,8,balanced,0.15524267156918845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,8,balanced,0.1548853317896525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,8,balanced,0.1565600037574768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,8,balanced,0.15583466490109762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,8,balanced,0.15777066349983215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,8,balanced,0.1613866686820984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,8,balanced,0.16125333309173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,8,balanced,0.17007466157277426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,8,balanced,0.17156267166137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,8,balanced,0.17667200167973837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,8,balanced,0.18997865915298462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,8,balanced,0.20070934295654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,8,balanced,0.22890132665634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,8,balanced,0.2595626711845398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,8,balanced,0.31012266874313354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,8,balanced,0.3560640017191569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,8,balanced,0.4681440194447835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,8,balanced,0.5662613312403361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,8,balanced,0.7894720236460367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,8,balanced,0.9853333632151285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,8,power_law_1.2,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,8,power_law_1.2,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,8,power_law_1.2,0.05990399718284607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,8,power_law_1.2,0.06836479902267456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,8,power_law_1.2,0.07550719976425171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,8,power_law_1.2,0.07886719703674316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,8,power_law_1.2,0.07920640110969543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,8,power_law_1.2,0.07933440208435058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,8,power_law_1.2,0.07601280212402343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,8,power_law_1.2,0.07832319736480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,8,power_law_1.2,0.08078719973564148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,8,power_law_1.2,0.08335360288619995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,8,power_law_1.2,0.0849727988243103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,8,power_law_1.2,0.08773120045661927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,8,power_law_1.2,0.09649279713630676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,8,power_law_1.2,0.10043519735336304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,8,power_law_1.2,0.10497280359268188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,8,power_law_1.2,0.12220159769058228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,8,power_law_1.2,0.12636159658432006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,8,power_law_1.2,0.1487552046775818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,8,power_law_1.2,0.17329920530319215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,8,power_law_1.2,0.22330238819122314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,8,power_law_1.2,0.2663808107376099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,8,power_law_1.2,0.3512511968612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,8,power_law_1.2,0.44327678680419924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,8,power_law_1.2,0.6376575946807861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,8,power_law_1.2,0.8436415672302247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,2,8,power_law_1.2,0.057183998823165896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,2,8,power_law_1.2,0.05740159749984741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,2,8,power_law_1.2,0.05911039710044861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,2,8,power_law_1.2,0.06268159747123718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,2,8,power_law_1.2,0.06273279786109924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,2,8,power_law_1.2,0.06339840292930603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,2,8,power_law_1.2,0.06421759724617004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,2,8,power_law_1.2,0.0651199996471405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,2,8,power_law_1.2,0.07173759937286377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,2,8,power_law_1.2,0.07163519859313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,2,8,power_law_1.2,0.07249280214309692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,2,8,power_law_1.2,0.0844223976135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,2,8,power_law_1.2,0.08766720294952393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,2,8,power_law_1.2,0.08984959721565247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,2,8,power_law_1.2,0.11492480039596557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,2,8,power_law_1.2,0.11013760566711425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,2,8,power_law_1.2,0.12845439910888673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,2,8,power_law_1.2,0.18391040563583375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,2,8,power_law_1.2,0.2096127986907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,2,8,power_law_1.2,0.2944191932678223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,2,8,power_law_1.2,0.38454399108886717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,4,power_law_1.01,0.04136959910392761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,4,power_law_1.01,0.041971200704574586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,4,power_law_1.01,0.04505600035190582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,4,power_law_1.01,0.04524799883365631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,4,power_law_1.01,0.04776960015296936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,4,power_law_1.01,0.0465472012758255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,4,power_law_1.01,0.046463999152183535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,4,power_law_1.01,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,4,power_law_1.01,0.04733439981937408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,4,power_law_1.01,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,4,power_law_1.01,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,4,power_law_1.01,0.04774399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,4,power_law_1.01,0.04844799935817719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,4,power_law_1.01,0.048844799399375916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,4,power_law_1.01,0.055244797468185426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,4,power_law_1.01,0.05799040198326111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,2,8,power_law_1.2,0.5455743789672851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,4,power_law_1.01,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,4,power_law_1.01,0.07032319903373718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.01,0.08023679852485657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.01,0.09550079703330994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,16,4,power_law_1.01,0.05934720039367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,4,balanced,0.051967998345692955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,4,balanced,0.07037866612275441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,4,balanced,0.09539733330408733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,4,balanced,0.1516639987627665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,4,balanced,0.2538986603418986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,4,balanced,0.44705601533253986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,4,balanced,0.643392006556193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.01,0.11411199569702149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,4,balanced,0.6449013153711954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,4,balanced,0.6483680009841919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,4,balanced,0.6489226818084717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,16,4,power_law_1.01,0.06364160180091857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,4,balanced,0.6537866592407227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,4,balanced,0.6633173227310181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,4,balanced,0.6653439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,16,4,power_law_1.01,0.06661120057106018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,4,balanced,0.662063995997111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,16,4,power_law_1.01,0.07848320007324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.01,0.14113919734954833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,16,4,power_law_1.01,0.08611199855804444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,16,4,power_law_1.01,0.08600959777832032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,2,8,power_law_1.2,0.7042751789093018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,16,4,power_law_1.01,0.1012992024421692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,16,4,power_law_1.01,0.10767999887466431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,16,4,power_law_1.01,0.10144000053405762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,16,4,power_law_1.01,0.10529279708862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,4,balanced,0.6736053625742594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,4,balanced,0.6791093349456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,4,balanced,0.6898612976074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,4,balanced,0.7110133171081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,4,balanced,0.7300000190734863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,4,balanced,0.7696426709493002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,4,balanced,0.8082506656646729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,16,4,power_law_1.01,0.10999679565429688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,4,balanced,0.8872799873352051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,16,4,power_law_1.01,0.11013760566711425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,4,balanced,0.9687146345774332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,4,balanced,1.1109920342763264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,4,balanced,1.3099413712819417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,16,4,power_law_1.01,0.1146239995956421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,4,balanced,1.5886774063110352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,16,4,power_law_1.01,0.11739519834518433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,16,4,power_law_1.01,0.12213120460510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,4,balanced,2.1448586781819663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,16,4,power_law_1.01,0.12814079523086547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,16,4,power_law_1.01,0.13345279693603515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,16,4,power_law_1.01,0.15062400102615356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.01,0.168723201751709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.01,0.19866880178451538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.01,0.23403520584106446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.01,0.29964160919189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.01,0.17599999904632568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.01,0.38421120643615725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.01,0.5184127807617187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.01,0.6739583969116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.01,0.9685183525085449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,2,8,power_law_1.2,0.980288028717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.01,1.2133695602416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.01,0.2630399942398071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,2,8,power_law_1.2,1.3746560096740723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.01,0.30409600734710696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.01,0.43144960403442384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.01,0.5730112075805665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,2,8,power_law_1.2,1.9952064514160157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,16,power_law_1.2,0.026131200790405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,16,power_law_1.2,0.045337599515914914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,2,8,power_law_1.2,2.696019172668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,16,power_law_1.2,0.04464640021324158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,16,power_law_1.2,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,16,power_law_1.2,0.0452672004699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,16,power_law_1.2,0.04431999921798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,16,power_law_1.2,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,16,power_law_1.2,0.045977601408958436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,16,power_law_1.2,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,16,power_law_1.2,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,16,power_law_1.2,0.05007359981536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,16,power_law_1.2,0.05002239942550659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,16,power_law_1.2,0.05508480072021484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,16,power_law_1.2,0.06094080209732056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,16,power_law_1.2,0.06551039814949036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,16,power_law_1.2,0.06627200245857238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,16,power_law_1.2,0.10568959712982177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,16,power_law_1.2,0.1299072027206421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,16,power_law_1.2,0.14336639642715454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,16,power_law_1.2,0.18154239654541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,16,power_law_1.2,0.1148095965385437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,16,power_law_1.2,0.14370559453964232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,16,power_law_1.2,0.18824959993362428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,16,power_law_1.2,0.24288640022277833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,16,power_law_1.2,0.3039103984832764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,16,power_law_1.2,0.43168001174926757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,16,power_law_1.2,0.5971327781677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,4,power_law_1.2,0.021267199516296388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,4,power_law_1.2,0.02167679965496063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,4,power_law_1.2,0.02481279969215393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,4,power_law_1.2,0.028038400411605834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,4,power_law_1.2,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,4,power_law_1.2,0.04085119962692261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,4,power_law_1.2,0.04168959856033325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,4,power_law_1.2,0.04229120016098022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,4,power_law_1.2,0.04466559886932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,4,power_law_1.2,0.04766719937324524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,4,power_law_1.2,0.04872319996356964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,4,power_law_1.2,0.05135999917984009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,4,power_law_1.2,0.061312001943588254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,4,power_law_1.2,0.05859839916229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,4,power_law_1.2,0.059571200609207155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,4,power_law_1.2,0.09207040071487427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,4,power_law_1.2,0.09879040122032165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,4,power_law_1.2,0.10165760517120362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,4,power_law_1.2,0.1053887963294983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,4,power_law_1.2,0.12152320146560669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,4,power_law_1.2,0.1314239978790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,4,power_law_1.2,0.1666111946105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,4,power_law_1.2,0.18346240520477294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,4,power_law_1.2,0.2504767894744873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,4,power_law_1.2,0.3078464031219482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,32,4,power_law_1.2,0.01973759979009628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,32,4,power_law_1.2,0.020710399746894835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,32,4,power_law_1.2,0.022368000447750093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,32,4,power_law_1.2,0.02595199942588806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,32,4,power_law_1.2,0.028275200724601747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,32,4,power_law_1.2,0.029606398940086365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,32,4,power_law_1.2,0.03182080090045929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,32,4,power_law_1.2,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,32,4,power_law_1.2,0.0350847989320755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,32,4,power_law_1.2,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,32,4,power_law_1.2,0.0359360009431839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,32,4,power_law_1.2,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,32,4,power_law_1.2,0.03857280015945434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,32,4,power_law_1.2,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,32,4,power_law_1.2,0.043558400869369504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,32,4,power_law_1.2,0.04686720073223114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,32,4,power_law_1.2,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,32,4,power_law_1.2,0.06802560091018676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,32,4,power_law_1.2,0.10374399423599243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,32,4,power_law_1.2,0.1211967945098877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,32,4,power_law_1.2,0.1235200047492981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,32,4,power_law_1.2,0.14470399618148805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,32,4,power_law_1.2,0.16531200408935548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,32,4,power_law_1.2,0.22008960247039794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,32,4,power_law_1.2,0.2741503953933716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,32,4,power_law_1.2,0.3554303884506226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,32,4,power_law_1.2,0.47055997848510744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,4,power_law_1.2,0.41030402183532716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,16,balanced,0.05817066629727682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,16,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,16,balanced,0.06523199876149495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,16,balanced,0.08463999629020691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,16,balanced,0.1109386682510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,16,balanced,0.16089600324630737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,16,balanced,0.16155200203259787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,16,balanced,0.16036267081896463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,16,balanced,0.16354667147000632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,16,balanced,0.16189866264661154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,16,balanced,0.1642453372478485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,16,balanced,0.16669867436091104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,16,balanced,0.1673333247502645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,16,balanced,0.1684266726175944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,16,balanced,0.18480000893274942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,16,balanced,0.18546666701634726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,16,balanced,0.18905067443847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,16,balanced,0.20654400189717612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,16,balanced,0.2137920061747233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,16,balanced,0.22466667493184408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,16,balanced,0.24110400676727295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,16,balanced,0.27410133679707843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,16,balanced,0.30460800727208454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,16,balanced,0.40399467945098877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,16,balanced,0.46586132049560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,16,balanced,0.645413319269816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,16,balanced,0.8367679913838705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,4,power_law_1.2,0.5935935974121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,2,power_law_1.2,0.034329599142074584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,2,power_law_1.2,0.03563520014286041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,2,power_law_1.2,0.04211840033531189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,2,power_law_1.2,0.05198720097541809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,2,power_law_1.2,0.06111360192298889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,2,power_law_1.2,0.08044800162315369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,2,power_law_1.2,0.07951359748840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,2,power_law_1.2,0.08683519959449768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,2,power_law_1.2,0.08936319947242737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,2,power_law_1.2,0.09142400026321411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,2,power_law_1.2,0.09219200015068055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,2,power_law_1.2,0.09505919814109802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,2,power_law_1.2,0.09832320213317872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,2,power_law_1.2,0.1043328046798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,2,power_law_1.2,0.10825599431991577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,2,power_law_1.2,0.11258879899978638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,2,power_law_1.2,0.13461120128631593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,2,power_law_1.2,0.15312000513076782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,2,power_law_1.2,0.15056639909744263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,2,power_law_1.2,0.18529920578002929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,2,power_law_1.2,0.20371201038360595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,2,power_law_1.2,0.279417610168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,2,power_law_1.2,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,2,power_law_1.2,0.039103999733924866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,2,power_law_1.2,0.05177599787712097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,2,power_law_1.2,0.0694208025932312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,2,power_law_1.2,0.07876480221748353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,2,power_law_1.2,0.10414719581604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,2,power_law_1.2,0.10909440517425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,2,power_law_1.2,0.1155392050743103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,2,power_law_1.2,0.11816960573196411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,2,power_law_1.2,0.12050559520721435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,2,power_law_1.2,0.12224639654159546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,2,power_law_1.2,0.12346880435943604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,2,power_law_1.2,0.1257024049758911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,2,power_law_1.2,0.12993919849395752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,2,power_law_1.2,0.1361855983734131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,2,power_law_1.2,0.14223359823226928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,2,power_law_1.2,0.1422271966934204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,2,power_law_1.2,0.1569983959197998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,2,power_law_1.2,0.1677183985710144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,2,power_law_1.2,0.19314559698104858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,2,power_law_1.2,0.2318592071533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,2,power_law_1.2,0.334604811668396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,2,power_law_1.2,0.26626560688018797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,2,power_law_1.2,0.33685760498046874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,2,power_law_1.2,0.4747007846832275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,2,power_law_1.2,0.6038591861724854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,2,power_law_1.2,0.43340158462524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,2,power_law_1.2,0.8419455528259278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,2,power_law_1.2,0.529311990737915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,16,4,balanced,0.02958400050799052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,16,4,balanced,0.031258667508761086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,16,4,balanced,0.031221332649389904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,16,4,balanced,0.03746666759252548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,16,4,balanced,0.04400533437728882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,16,4,balanced,0.04553066690762838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,16,4,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,16,4,balanced,0.04370133578777313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,16,4,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,16,4,balanced,0.044112001856168113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,16,4,balanced,0.04397333165009817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,16,4,balanced,0.04589866598447164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,16,4,balanced,0.045824001232783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,16,4,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,16,4,balanced,0.055546666185061135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,16,4,balanced,0.05425066749254862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,2,power_law_1.2,1.0820544242858887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,16,4,balanced,0.0591893345117569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,16,4,balanced,0.06654400130112965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,16,4,balanced,0.075162669022878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,16,4,balanced,0.0925333301226298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,16,4,balanced,0.10517866412798564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,16,4,balanced,0.13606933752695718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,16,4,balanced,0.16198399662971497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,16,4,balanced,0.22000000874201456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,2,power_law_1.2,0.7519423961639404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,16,4,balanced,0.2690879901250203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,16,4,balanced,0.391487995783488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,16,4,balanced,0.49039467175801593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,2,power_law_1.2,0.9484607696533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,4,balanced,0.03081600119670232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,4,balanced,0.0312266672650973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,4,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,4,balanced,0.03846933444341024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,4,balanced,0.04554666578769684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,4,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,4,balanced,0.04612799982229868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,4,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,4,balanced,0.04770666857560476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,4,balanced,0.047797332207361855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,4,balanced,0.04894933104515076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,4,balanced,0.05006400247414907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,4,balanced,0.05340266724427541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,4,balanced,0.054746667544047035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,4,balanced,0.062047998110453285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,4,balanced,0.06402666866779327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,4,balanced,0.06442133088906606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,4,balanced,0.07500266532103221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,4,balanced,0.08461333314577739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,64,power_law_1.01,0.022278399765491487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,4,balanced,0.1135040024916331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,64,power_law_1.01,0.022099199891090392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,64,power_law_1.01,0.022393600642681123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,64,power_law_1.01,0.0230335995554924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,64,power_law_1.01,0.024780799448490144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,64,power_law_1.01,0.032646399736404416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,64,power_law_1.01,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,4,balanced,0.12485333283742268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,4,balanced,0.16846400499343872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,4,balanced,0.203167994817098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,4,balanced,0.2815679907798767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,4,balanced,0.36353600025177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,4,balanced,0.5217333237330118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,4,balanced,0.6790719827016195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,64,power_law_1.01,0.055667197704315184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,64,power_law_1.01,0.05086719989776611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,64,power_law_1.01,0.0521664023399353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,64,power_law_1.01,0.05235840082168579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,64,power_law_1.01,0.05175679922103882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,64,power_law_1.01,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,64,power_law_1.01,0.05123839974403381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,64,power_law_1.01,0.050496000051498416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,64,power_law_1.01,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,64,power_law_1.01,0.0535103976726532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,64,power_law_1.01,0.05373439788818359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.01,0.06076800227165222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.01,0.08247039914131164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.01,0.09185919761657715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.01,0.11430399417877198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.01,0.14135040044784547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.01,0.19973119497299194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.01,0.26446080207824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.01,0.3648576021194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.01,0.5103807926177979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,16,power_law_1.2,0.14350719451904298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,16,power_law_1.2,0.24975359439849854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,16,power_law_1.2,0.31356799602508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,16,power_law_1.2,0.32819199562072754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,16,power_law_1.2,0.3395711898803711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,16,power_law_1.2,0.3264960050582886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,16,power_law_1.2,0.25575039386749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,16,power_law_1.2,0.3334719896316528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,16,power_law_1.2,0.28792319297790525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,16,power_law_1.2,0.2667840003967285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,16,power_law_1.2,0.2880000114440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,16,power_law_1.2,0.30185599327087403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,16,power_law_1.2,0.3005183935165405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,16,power_law_1.2,0.3035648107528687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,16,power_law_1.2,0.31641600131988523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,16,power_law_1.2,0.30719358921051027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,16,power_law_1.2,0.3389631986618042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,16,power_law_1.2,0.395795202255249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,16,power_law_1.2,0.4370368003845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,16,power_law_1.2,0.5008959770202637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,8,power_law_1.01,0.08008319735527039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,8,power_law_1.01,0.08130559921264649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,8,power_law_1.01,0.07982720136642456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,8,power_law_1.01,0.08300799727439881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,8,power_law_1.01,0.08348159790039063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,8,power_law_1.01,0.08849279880523682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,8,power_law_1.01,0.09879040122032165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,8,power_law_1.01,0.09525759816169739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,16,power_law_1.2,0.5387455940246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,8,power_law_1.01,0.10131839513778687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,8,power_law_1.01,0.10529919862747192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,16,power_law_1.2,0.6912767887115479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,8,power_law_1.01,0.1017408013343811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,8,power_law_1.01,0.10490880012512208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,16,power_law_1.2,0.7476863861083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,8,power_law_1.01,0.11400320529937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,16,power_law_1.2,0.9386176109313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,1,power_law_1.01,0.05628160238265991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,1,power_law_1.01,0.07852799892425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,16,power_law_1.2,1.0011584281921386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,16,power_law_1.2,1.3119232177734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,8,power_law_1.01,0.12124799489974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,8,power_law_1.01,0.12907520532608033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,8,power_law_1.01,0.14532480239868165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,16,power_law_1.2,1.6482559204101563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,8,power_law_1.01,0.1598912000656128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,8,power_law_1.01,0.2153536081314087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,1,power_law_1.01,0.09059200286865235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,1,power_law_1.01,0.13240959644317626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,1,power_law_1.01,0.13519359827041627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,8,power_law_1.01,0.2561919927597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,1,power_law_1.01,0.13708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,8,power_law_1.01,0.3562623977661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,8,power_law_1.01,0.4632256031036377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,1,power_law_1.01,0.14128639698028564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,1,power_law_1.01,0.14226560592651366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,1,power_law_1.01,0.1413632035255432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,1,power_law_1.01,0.1466431975364685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,1,power_law_1.01,0.1473088026046753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,8,power_law_1.01,0.6377471923828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,1,power_law_1.01,0.15418239831924438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,1,power_law_1.01,0.1564288020133972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,1,power_law_1.01,0.16151679754257203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,1,power_law_1.01,0.18414080142974854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,1,power_law_1.01,0.19130879640579224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,1,power_law_1.01,0.19875839948654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,1,power_law_1.01,0.23253118991851807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,1,power_law_1.01,0.26241919994354246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,8,power_law_1.01,0.8707072257995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,1,power_law_1.01,0.32948479652404783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,1,power_law_1.01,0.3903167963027954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,8,power_law_1.01,1.258841609954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,1,power_law_1.01,0.5243008136749268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,1,power_law_1.01,0.6486015796661377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,1,power_law_1.01,0.9140416145324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,8,power_law_1.01,1.6479040145874024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,1,power_law_1.01,1.1451456069946289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,8,power_law_1.01,2.6091007232666015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,1,power_law_1.01,1.6537408828735352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,8,power_law_1.01,3.290617752075195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,1,power_law_1.2,0.1579519987106323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,1,power_law_1.2,0.25564160346984866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,1,power_law_1.2,0.4574272155761719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,1,power_law_1.2,0.6888895988464355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,1,power_law_1.2,0.9131008148193359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,1,power_law_1.2,1.6601408004760743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,1,power_law_1.2,1.77142391204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,1,power_law_1.2,1.8542015075683593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,1,power_law_1.2,1.9443775177001954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,1,power_law_1.2,1.9633920669555665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,1,power_law_1.2,2.016691207885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,1,power_law_1.2,2.093926429748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,1,power_law_1.2,2.124729537963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,1,power_law_1.2,2.128531265258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,1,power_law_1.2,2.2194623947143555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,1,power_law_1.2,2.2915199279785154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,1,power_law_1.2,2.3223743438720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,1,balanced,0.0706879993279775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,1,balanced,0.095360000928243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,1,balanced,0.14849066734313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,1,power_law_1.2,2.53505916595459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,1,balanced,0.24921600023905435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,1,balanced,0.44338667392730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,1,balanced,0.8396053314208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,1,balanced,0.8451840082804362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,1,power_law_1.2,2.801465606689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,1,balanced,0.8493866920471191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,1,balanced,0.8494773705800375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,1,balanced,0.8547946612040201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,1,balanced,0.8641973336537679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,1,power_law_1.2,3.2962112426757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,1,balanced,0.8701386451721191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,1,balanced,0.8780372937520345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,1,balanced,0.888159990310669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,1,balanced,0.8991146882375082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,1,balanced,0.910213311513265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,1,balanced,0.925061305363973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,1,balanced,0.9697973728179932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,1,balanced,1.01145601272583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,1,power_law_1.2,3.0249536514282225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,2,power_law_1.2,0.09670400023460388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,1,balanced,1.089845339457194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,2,power_law_1.2,0.14677759408950805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,1,balanced,1.1563146909077961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,2,power_law_1.2,0.20499839782714843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,1,balanced,1.2945813337961833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,1,balanced,1.4718079566955566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,2,power_law_1.2,0.3103487968444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,1,balanced,1.8858240445454915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,1,power_law_1.2,3.5576702117919923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,2,power_law_1.2,0.4590591907501221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,1,balanced,2.154047966003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,1,power_law_1.01,2.156070327758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,1,power_law_1.2,4.071347045898437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,2,power_law_1.2,0.6108799934387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,1,balanced,3.1544265747070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,2,power_law_1.2,0.6369215965270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,2,power_law_1.2,0.8881407737731933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,1,power_law_1.2,4.924812698364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,1,balanced,3.84549872080485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,2,power_law_1.2,0.8827648162841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,2,power_law_1.2,0.9569855690002441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,2,power_law_1.2,0.9918463706970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,2,power_law_1.2,1.0114368438720702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,1,power_law_1.2,6.548313903808594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,2,power_law_1.2,1.013145637512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,2,power_law_1.2,1.0395968437194825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,2,power_law_1.2,1.077894401550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,2,power_law_1.2,1.0769280433654784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,2,power_law_1.2,1.154470443725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,2,power_law_1.2,1.2240320205688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,2,power_law_1.2,1.3435071945190429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,2,power_law_1.2,1.4918399810791017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,2,power_law_1.2,1.5698368072509765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,2,power_law_1.2,1.8272512435913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,2,power_law_1.2,1.783942413330078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,2,power_law_1.2,2.1868671417236327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,2,power_law_1.2,2.71399040222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,2,power_law_1.2,3.1889215469360352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,1,power_law_1.2,8.256185913085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,2,power_law_1.2,4.369971084594726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,1,power_law_1.2,10.474668884277344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,256,power_law_1.2,0.04748159945011139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,256,power_law_1.2,0.04743039906024933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,256,power_law_1.2,0.048556798696517946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,256,power_law_1.2,0.051744002103805545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,256,power_law_1.2,0.052172797918319705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,256,power_law_1.2,0.055129599571228025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,256,power_law_1.2,0.05648639798164368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,256,power_law_1.2,0.059910398721694944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,256,power_law_1.2,0.060140800476074216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,256,power_law_1.2,0.06458240151405334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,256,power_law_1.2,0.0609279990196228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,256,power_law_1.2,0.06355199813842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,256,power_law_1.2,0.06769919991493226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,256,power_law_1.2,0.06613119840621948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,256,power_law_1.2,0.07904000282287597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,256,power_law_1.2,0.07895680069923401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,256,power_law_1.2,0.08637440204620361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,256,power_law_1.2,0.10179200172424316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,256,power_law_1.2,0.11007360219955445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,256,power_law_1.2,0.14380160570144654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,256,power_law_1.2,0.17609599828720093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,256,power_law_1.2,0.24642560482025147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,256,power_law_1.2,0.3138495922088623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,256,power_law_1.2,0.47854080200195315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,256,power_law_1.2,0.7022016048431396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,256,power_law_1.2,1.0854975700378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,8,1,power_law_1.2,0.034176000952720643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,8,1,power_law_1.2,0.03950720131397247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,256,power_law_1.2,1.4124159812927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,8,1,power_law_1.2,0.05286399722099304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,8,1,power_law_1.2,0.06744319796562195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,8,1,power_law_1.2,0.08030080199241638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,8,1,power_law_1.2,0.11986559629440308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,8,1,power_law_1.2,0.1254464030265808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,8,1,power_law_1.2,0.12739839553833007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,8,1,power_law_1.2,0.13258880376815796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,8,1,power_law_1.2,0.13509759902954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,8,1,power_law_1.2,0.14236799478530884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,8,1,power_law_1.2,0.14815360307693481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,8,1,power_law_1.2,0.15403519868850707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,8,1,power_law_1.2,0.1598207950592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,8,1,power_law_1.2,0.1688447952270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,8,1,power_law_1.2,0.17608959674835206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,8,1,power_law_1.2,0.18777600526809693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,8,1,power_law_1.2,0.2163007974624634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,8,1,power_law_1.2,0.2431999921798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,8,1,power_law_1.2,0.2995327949523926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,8,1,power_law_1.2,0.3425856113433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,8,1,power_law_1.2,0.4528831958770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,8,1,power_law_1.2,0.566374397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,8,1,power_law_1.2,0.7552063941955567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,8,1,power_law_1.2,0.9492223739624024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,8,1,power_law_1.2,1.3499135971069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,8,power_law_1.01,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,8,power_law_1.01,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,8,power_law_1.01,0.030508801341056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,8,power_law_1.01,0.034995201230049136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,8,power_law_1.01,0.03809280097484589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,8,power_law_1.01,0.037747201323509214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,8,power_law_1.01,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,8,power_law_1.01,0.035750401020050046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,8,power_law_1.01,0.03729279935359955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,8,power_law_1.01,0.037887999415397645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,8,power_law_1.01,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,8,power_law_1.01,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,8,power_law_1.01,0.04037120044231415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,8,power_law_1.01,0.04312320053577423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,8,power_law_1.01,0.049055999517440795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,8,power_law_1.01,0.04955520033836365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,8,power_law_1.01,0.053324800729751584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,8,power_law_1.01,0.05739520192146301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,8,power_law_1.01,0.059910398721694944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,8,power_law_1.01,0.07183359861373902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,8,power_law_1.01,0.07388799786567687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,8,power_law_1.01,0.09066879749298096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,8,power_law_1.01,0.10304000377655029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,8,power_law_1.01,0.134169602394104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,8,power_law_1.01,0.16117119789123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,8,power_law_1.01,0.224947190284729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,8,power_law_1.01,0.28624639511108396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,4,power_law_1.2,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,4,power_law_1.2,0.04485760033130646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,4,power_law_1.2,0.05062400102615357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,4,power_law_1.2,0.06156799793243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,4,power_law_1.2,0.06183040142059326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,4,power_law_1.2,0.06409599781036376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,4,power_law_1.2,0.06583679914474487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,4,power_law_1.2,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,4,power_law_1.2,0.0692031979560852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,4,power_law_1.2,0.07187839746475219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,8,1,power_law_1.2,1.8144447326660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,4,power_law_1.2,0.069760000705719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,4,power_law_1.2,0.07556480169296265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,4,power_law_1.2,0.07376639842987061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,4,power_law_1.2,0.08042880296707153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,32,power_law_1.2,0.034720000624656674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,4,power_law_1.2,0.08895999789237977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,32,power_law_1.2,0.03177599906921387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,4,power_law_1.2,0.10042879581451417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,4,power_law_1.2,0.1085312008857727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,4,power_law_1.2,0.13348480463027954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,32,power_law_1.2,0.030771198868751525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,32,power_law_1.2,0.03224959969520569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,32,power_law_1.2,0.03364480137825012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,32,power_law_1.2,0.03466239869594574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,32,power_law_1.2,0.035078400373458864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,32,power_law_1.2,0.03452160060405731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,32,power_law_1.2,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,32,power_law_1.2,0.03542400002479553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,32,power_law_1.2,0.03600000143051148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,4,power_law_1.2,0.1468608021736145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,32,power_law_1.2,0.03702400028705597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,32,power_law_1.2,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,32,power_law_1.2,0.038726401329040525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,32,power_law_1.2,0.04178560078144074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,32,power_law_1.2,0.04284160137176514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,32,power_law_1.2,0.045286399126052854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,32,power_law_1.2,0.04933759868144989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,32,power_law_1.2,0.050246399641036985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,32,power_law_1.2,0.059468799829483034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,4,power_law_1.2,0.18699519634246825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,32,power_law_1.2,0.06688640117645264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,4,power_law_1.2,0.21152639389038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,32,power_law_1.2,0.08376320004463196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,4,power_law_1.2,0.2964735984802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,16,power_law_1.2,0.01934719979763031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,16,power_law_1.2,0.019487999379634857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,16,power_law_1.2,0.01860480010509491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,32,power_law_1.2,0.09315199851989746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,16,power_law_1.2,0.019251200556755065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,16,power_law_1.2,0.01940480023622513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,16,power_law_1.2,0.01973759979009628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,16,power_law_1.2,0.021465599536895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,16,power_law_1.2,0.021184000372886657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,16,power_law_1.2,0.0215488001704216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,16,power_law_1.2,0.022169600427150726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,16,power_law_1.2,0.022655999660491942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,16,power_law_1.2,0.02389120012521744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,16,power_law_1.2,0.023839999735355378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,16,power_law_1.2,0.027142399549484254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,16,power_law_1.2,0.027955201268196107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,16,power_law_1.2,0.029228800535202028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,16,power_law_1.2,0.04270080029964447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,16,power_law_1.2,0.04563199877738953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,16,power_law_1.2,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,4,power_law_1.2,0.4168064117431641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,16,power_law_1.2,0.0482367992401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,32,power_law_1.2,0.1335039973258972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,16,power_law_1.2,0.052876800298690796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,16,power_law_1.2,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,16,power_law_1.2,0.081632000207901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,16,power_law_1.2,0.10224000215530396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,16,power_law_1.2,0.12351360321044921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,16,power_law_1.2,0.17993600368499757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,16,power_law_1.2,0.2169856071472168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,64,power_law_1.01,0.04901120066642761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,64,power_law_1.01,0.03651840090751648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,64,power_law_1.01,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,64,power_law_1.01,0.034169599413871765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,64,power_law_1.01,0.03116160035133362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,64,power_law_1.01,0.032864001393318173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,64,power_law_1.01,0.03240320086479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,64,power_law_1.01,0.032364800572395325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,64,power_law_1.01,0.03240959942340851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,64,power_law_1.01,0.032864001393318173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,64,power_law_1.01,0.03320319950580597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,64,power_law_1.01,0.03407360017299652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,64,power_law_1.01,0.03559040129184723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,32,power_law_1.2,0.16243200302124022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,64,power_law_1.01,0.036883199214935304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,64,power_law_1.01,0.042335999011993405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,64,power_law_1.01,0.043033599853515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,64,power_law_1.01,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,64,power_law_1.01,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,64,power_law_1.01,0.04867840111255646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,64,power_law_1.01,0.05720319747924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,64,power_law_1.01,0.060140800476074216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,64,power_law_1.01,0.06872320175170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,64,power_law_1.01,0.09505280256271362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,64,power_law_1.01,0.11100159883499146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,64,power_law_1.01,0.13352960348129272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,64,power_law_1.01,0.16428159475326537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,64,power_law_1.01,0.21482880115509034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,4,power_law_1.2,0.5428927898406982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,4,power_law_1.2,0.6700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,32,power_law_1.2,0.21774721145629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,2,power_law_1.01,0.15045759677886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,2,power_law_1.01,0.24782719612121581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,2,power_law_1.01,0.4423935890197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,2,power_law_1.01,0.5589632034301758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,2,power_law_1.01,0.9522624015808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,2,power_law_1.01,1.1972800254821778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,2,power_law_1.01,1.318188762664795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,2,power_law_1.01,1.792857551574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,2,power_law_1.01,1.7835071563720704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,4,power_law_1.2,0.1174015998840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,2,power_law_1.01,1.808358383178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,4,power_law_1.2,0.13319040536880494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,4,power_law_1.2,0.1562432050704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,2,power_law_1.01,1.9207679748535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,4,power_law_1.2,0.1928320050239563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,4,power_law_1.2,0.2236095905303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,2,power_law_1.01,1.9231168746948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,4,power_law_1.2,0.2816704034805298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,2,power_law_1.01,1.958220863342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,2,power_law_1.01,1.9972288131713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,2,power_law_1.01,2.0762176513671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,4,power_law_1.2,0.28360960483551023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,4,power_law_1.2,0.30490880012512206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,2,power_law_1.01,2.118329620361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,4,power_law_1.2,0.3052095890045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,4,power_law_1.2,0.31926400661468507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,4,power_law_1.2,0.3094847917556763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,2,power_law_1.01,2.1109376907348634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,4,power_law_1.2,0.3288511991500854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,4,power_law_1.2,0.331987190246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,2,power_law_1.01,2.2671552658081056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,4,power_law_1.2,0.3469183921813965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,4,power_law_1.2,0.35518081188201905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,4,power_law_1.2,0.37900800704956056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,4,power_law_1.2,0.39743359088897706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,4,power_law_1.2,0.43016958236694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,4,power_law_1.2,0.492844820022583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,4,power_law_1.2,0.6188608169555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,2,power_law_1.01,2.396134376525879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,4,power_law_1.2,0.6828735828399658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,4,power_law_1.2,0.8554495811462403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,2,power_law_1.01,2.7080703735351563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,4,power_law_1.2,1.07391357421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,4,power_law_1.2,0.9825599670410157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,2,power_law_1.01,2.743110466003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,2,power_law_1.01,3.2301761627197267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,4,power_law_1.2,1.3416959762573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,2,power_law_1.01,3.1613759994506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,4,power_law_1.2,1.6896127700805663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,2,power_law_1.01,3.776678466796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,4,power_law_1.2,2.3895103454589846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,2,power_law_1.01,4.294342422485352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,4,power_law_1.2,3.3299007415771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,2,power_law_1.01,5.29615364074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,2,power_law_1.01,7.086822509765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,32,power_law_1.2,0.2875967979431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,4,power_law_1.2,1.1820608139038087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,1,power_law_1.01,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,1,power_law_1.01,0.07993599772453308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,1,power_law_1.01,0.10828160047531128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,1,power_law_1.01,0.15493760108947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,1,power_law_1.01,0.23357439041137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,1,power_law_1.01,0.3080895900726318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,1,power_law_1.01,0.364467191696167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,1,power_law_1.01,0.46469759941101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,1,power_law_1.01,0.48083839416503904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,1,power_law_1.01,0.5066624164581299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,1,power_law_1.01,0.5242176055908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,1,power_law_1.01,0.5374144077301025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,32,power_law_1.2,0.10663679838180543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,32,power_law_1.2,0.10542720556259155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,32,power_law_1.2,0.10167039632797241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,32,power_law_1.2,0.0887167990207672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,32,power_law_1.2,0.09629439711570739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,32,power_law_1.2,0.09698560237884521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,32,power_law_1.2,0.10698879957199096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,1,power_law_1.01,0.5528960227966309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,32,power_law_1.2,0.09987840056419373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,1,power_law_1.01,0.5960127830505371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,32,power_law_1.2,0.1083456039428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,1,power_law_1.01,0.6073344230651856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,32,power_law_1.2,0.1098431944847107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,32,power_law_1.2,0.10195200443267823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,1,power_law_1.01,0.6299776077270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,32,power_law_1.2,0.10903040170669556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,1,power_law_1.01,0.6413887977600098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,32,power_law_1.2,0.10520960092544555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,32,power_law_1.2,0.10797439813613892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,1,power_law_1.01,0.7052031993865967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,32,power_law_1.2,0.11304320096969604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,1,power_law_1.01,0.770739221572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,32,power_law_1.2,0.11844480037689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,1,power_law_1.01,0.8996800422668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,32,power_law_1.2,0.13150719404220582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,32,power_law_1.2,0.14417920112609864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,1,power_law_1.01,0.9790271759033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.2,0.16425600051879882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,1,power_law_1.01,1.204806423187256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,1,power_law_1.01,1.5021183967590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,1,power_law_1.01,1.9251583099365235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.2,0.19544960260391236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.2,0.21658880710601808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.2,0.2630975961685181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,1,power_law_1.01,2.215622329711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.2,0.32943999767303467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.2,0.5400256156921387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,1,power_law_1.01,2.9071935653686523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.2,0.6396351814270019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,1,power_law_1.01,3.7450366973876954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.2,1.0833600044250489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,32,balanced,0.06133866806825002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,32,balanced,0.062074666221936546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,32,balanced,0.06273066500822704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,32,balanced,0.07642666498819987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,32,balanced,0.09477333227793376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,32,balanced,0.13541332880655924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,32,balanced,0.18891199429829916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,32,balanced,0.16882665952046713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,32,balanced,0.16673600673675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,32,balanced,0.18173867464065552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,32,balanced,0.16291733582814535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,32,balanced,0.16862932840983072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,32,balanced,0.16664000352223715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,32,balanced,0.16715200742085776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,32,balanced,0.17204799254735312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,32,balanced,0.17247466246287027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,32,balanced,0.18980266650517783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,32,balanced,0.17358932892481485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,32,balanced,0.19812800486882529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,32,balanced,0.20164799690246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,32,balanced,0.2139093279838562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,32,balanced,0.2466986576716105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,32,balanced,0.27082133293151855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,32,balanced,0.33560534318288165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,32,balanced,0.48710934321085614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,32,balanced,0.5762240091959635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,32,balanced,0.7719146410624186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.2,1.2739328384399413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,2,power_law_1.2,0.046003198623657225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,2,power_law_1.2,0.04725759923458099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,2,power_law_1.2,0.04943360090255737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,2,power_law_1.2,0.05663999915122986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,2,power_law_1.2,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,2,power_law_1.2,0.07521920204162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,2,power_law_1.2,0.08514559864997864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,2,power_law_1.2,0.08570880293846131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,2,power_law_1.2,0.08939520120620728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,2,power_law_1.2,0.0911296010017395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,2,power_law_1.2,0.09441919922828675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,2,power_law_1.2,0.09573760032653808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,2,power_law_1.2,0.09879040122032165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,2,power_law_1.2,0.10178560018539429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,2,power_law_1.2,0.10911359786987304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,2,power_law_1.2,0.11547520160675048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,2,power_law_1.2,0.1335744023323059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,2,power_law_1.2,0.1483839988708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,2,power_law_1.2,0.16657279729843139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,2,power_law_1.2,0.195961594581604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,2,power_law_1.2,0.2298111915588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,2,power_law_1.2,0.2972095966339111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,2,power_law_1.2,0.3640383958816528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,2,power_law_1.2,0.4904064178466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,2,power_law_1.2,0.6411647796630859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,2,power_law_1.2,0.8900992393493652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,2,power_law_1.2,1.1681792259216308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,1,balanced,0.05402666827042898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,1,balanced,0.058245331048965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,1,balanced,0.07542933523654938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,1,balanced,0.11009599765141805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,1,balanced,0.16030399998029074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,8,power_law_1.01,0.07964159846305847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,8,power_law_1.01,0.10523519515991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,1,balanced,0.2528853416442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,8,power_law_1.01,0.0998848021030426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,8,power_law_1.01,0.11361919641494751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,8,power_law_1.01,0.12127360105514526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,8,power_law_1.01,0.13312000036239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,1,balanced,0.25521600246429443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,1,balanced,0.25760000944137573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,1,balanced,0.26128532489140827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,1,balanced,0.26338134209314984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,1,balanced,0.2655093272527059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,1,balanced,0.2732693354288737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,1,balanced,0.2787253260612488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,1,balanced,0.2835093339284261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,1,balanced,0.29522132873535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,1,balanced,0.3038613398869832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,1,balanced,0.3294559915860494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,1,balanced,0.3771786689758301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,8,power_law_1.01,0.13470720052719115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,1,balanced,0.408517320950826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,1,balanced,0.48798934618632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,8,power_law_1.01,0.12878079414367677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,1,balanced,0.5783146619796753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,8,power_law_1.01,0.13328640460968016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,1,balanced,0.7302613258361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,8,power_law_1.01,0.1351680040359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,1,power_law_1.2,0.021913599967956544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,8,power_law_1.01,0.13417600393295287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,1,power_law_1.2,0.030457600951194763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,8,power_law_1.01,0.1441856026649475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,1,power_law_1.2,0.04552960097789764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,8,power_law_1.01,0.14425599575042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,1,power_law_1.2,0.05151360034942627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,8,power_law_1.01,0.14755840301513673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,1,power_law_1.2,0.06611199975013733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,8,power_law_1.01,0.16034560203552245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,1,power_law_1.2,0.09376000165939331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,1,balanced,0.8692746957143148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,1,power_law_1.2,0.1021888017654419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,1,balanced,1.244815985361735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,1,power_law_1.2,0.10618879795074462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,1,power_law_1.2,0.11067520380020142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,1,balanced,1.5454079310099285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,1,power_law_1.2,0.11146240234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,1,power_law_1.2,0.10899200439453124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,1,power_law_1.2,0.11504640579223632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,8,power_law_1.01,0.16601599454879762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,8,power_law_1.01,0.1771456003189087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,8,power_law_1.01,0.20618879795074463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,1,balanced,2.232111930847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,1,power_law_1.2,0.1217919945716858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,1,power_law_1.2,0.11281919479370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,8,power_law_1.01,0.2280832052230835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,1,power_law_1.2,0.12305279970169067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,1,balanced,2.906192143758138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,1,power_law_1.2,0.12836480140686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,1,power_law_1.2,0.15312639474868775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,8,power_law_1.01,0.2904383897781372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,1,power_law_1.2,0.17887359857559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,8,power_law_1.01,0.3303679943084717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.2,0.18646399974822997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,8,power_law_1.01,0.4238527774810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.2,0.23884799480438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,8,power_law_1.01,0.49322237968444826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,8,power_law_1.01,0.6315968036651611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,8,power_law_1.01,0.8082304000854492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,8,power_law_1.01,1.153337574005127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.2,0.22341759204864503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.2,0.2812864065170288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,8,power_law_1.01,1.462764835357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.2,0.2958656072616577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.2,0.38565759658813475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.2,0.4636544227600098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.2,0.6363264083862304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,8,power_law_1.2,0.04599680006504059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,8,power_law_1.2,0.061484801769256595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,8,power_law_1.2,0.05539839863777161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,8,power_law_1.2,0.06767359972000123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,8,power_law_1.2,0.07067520022392274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,8,power_law_1.2,0.0782912015914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,8,power_law_1.2,0.08060160279273987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,8,power_law_1.2,0.07089279890060425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,8,power_law_1.2,0.07406079769134521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,8,power_law_1.2,0.07386879920959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,8,power_law_1.2,0.07762560248374939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,8,power_law_1.2,0.08395519852638245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,8,power_law_1.2,0.08458880186080933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,8,power_law_1.2,0.08557440042495727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,8,power_law_1.2,0.0913856029510498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,8,power_law_1.2,0.09562240242958069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,8,power_law_1.2,0.0964735984802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,8,power_law_1.2,0.11347839832305909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,8,power_law_1.2,0.11642240285873413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,8,power_law_1.2,0.13715200424194335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,8,power_law_1.2,0.1460543990135193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,8,power_law_1.2,0.18873599767684937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,8,power_law_1.2,0.22414720058441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,8,power_law_1.2,0.3029184103012085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,8,power_law_1.2,0.3469631910324097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,8,power_law_1.2,0.5339136123657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,8,power_law_1.2,0.6886784076690674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.2,0.8086912155151367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,4,1,balanced,0.05594133337338766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,4,1,balanced,0.05816533168156942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,4,1,balanced,0.0673280010620753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,4,1,balanced,0.0888426701227824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,4,1,balanced,0.12355200449625652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,4,1,balanced,0.11899733543395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,4,1,balanced,0.11974400281906128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,4,1,balanced,0.12157332897186279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,4,1,balanced,0.12066133817036946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,4,1,balanced,0.12249066432317098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,4,1,balanced,0.12576533357302347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,4,1,balanced,0.12823999921480814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,4,1,balanced,0.1313653290271759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,4,1,balanced,0.13738666971524557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,4,1,balanced,0.13985066612561545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,4,1,balanced,0.14660267035166422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,4,1,balanced,0.15684266885121664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,4,1,balanced,0.180074671904246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,4,1,balanced,0.20121600230534872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,4,1,balanced,0.2478826642036438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,4,1,balanced,0.28818132479985553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,4,1,balanced,0.43049601713816327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,4,1,balanced,0.5216319958368937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,4,1,balanced,0.7591306368509928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,4,1,balanced,0.9896907011667887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,4,1,balanced,1.4464586575826008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,8,8,power_law_1.01,0.06829439997673034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,4,1,balanced,1.8876214027404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,8,8,power_law_1.01,0.08064000010490417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,8,8,power_law_1.01,0.0845632016658783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,8,8,power_law_1.01,0.09260159730911255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,8,8,power_law_1.01,0.09841920137405395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,8,8,power_law_1.01,0.09646720290184022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,8,8,power_law_1.01,0.1018496036529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,8,8,power_law_1.01,0.10016640424728393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,8,8,power_law_1.01,0.10213119983673095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,8,8,power_law_1.01,0.10254080295562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,8,8,power_law_1.01,0.1011199951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,8,8,power_law_1.01,0.10387200117111206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,8,8,power_law_1.01,0.10487680435180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,8,8,power_law_1.01,0.10636160373687745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,8,8,power_law_1.01,0.1170240044593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,8,8,power_law_1.01,0.11560319662094116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,8,8,power_law_1.01,0.11935360431671142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,8,8,power_law_1.01,0.13197439908981323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,8,8,power_law_1.01,0.14079999923706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,8,8,power_law_1.01,0.1735551953315735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,8,8,power_law_1.01,0.18882559537887572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,8,8,power_law_1.01,0.23045120239257813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,8,8,power_law_1.01,0.2859839916229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,4,power_law_1.01,0.047712001204490664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,8,8,power_law_1.01,0.3901439905166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,4,power_law_1.01,0.05187199711799621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,8,8,power_law_1.01,0.49850239753723147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,4,power_law_1.01,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,4,power_law_1.01,0.06807039976119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,8,8,power_law_1.01,0.6761472225189209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,4,power_law_1.01,0.08466560244560242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,8,8,power_law_1.01,0.9253952026367187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,4,power_law_1.01,0.09281920194625855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,4,power_law_1.01,0.09921280145645142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,4,power_law_1.01,0.10655360221862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,4,power_law_1.01,0.11019519567489625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,4,power_law_1.01,0.11138559579849243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,4,power_law_1.01,0.10677119493484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,4,power_law_1.01,0.11562240123748779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,4,power_law_1.01,0.11864320039749146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,4,power_law_1.01,0.12937599420547485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,64,power_law_1.2,0.03187200129032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,64,power_law_1.2,0.03345920145511627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,4,power_law_1.01,0.13585280179977416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,64,power_law_1.2,0.027929601073265076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,4,power_law_1.01,0.14404480457305907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,64,power_law_1.2,0.02956799864768982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,64,power_law_1.2,0.03038719892501831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,4,power_law_1.01,0.15010559558868408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,4,power_law_1.01,0.1675968050956726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,4,power_law_1.01,0.1938688039779663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,4,power_law_1.01,0.23917438983917236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,64,power_law_1.2,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,4,power_law_1.01,0.2650496006011963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,64,power_law_1.2,0.030988800525665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,64,power_law_1.2,0.031142398715019226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,4,power_law_1.01,0.3540927886962891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,64,power_law_1.2,0.03115519881248474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,64,power_law_1.2,0.031116798520088196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,4,power_law_1.01,0.44410881996154783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,4,power_law_1.01,0.601087999343872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,4,power_law_1.01,0.7598400115966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,4,power_law_1.01,1.1207807540893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,64,power_law_1.2,0.03235200047492981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,64,power_law_1.2,0.031948798894882204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,64,power_law_1.2,0.032339200377464294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,64,power_law_1.2,0.03407360017299652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,64,power_law_1.2,0.036831998825073244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,64,power_law_1.2,0.038924801349639895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,64,power_law_1.2,0.04142079949378967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,64,power_law_1.2,0.04348799884319306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,64,power_law_1.2,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,64,power_law_1.2,0.054655998945236206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,64,power_law_1.2,0.058348798751831056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,64,power_law_1.2,0.08307200074195861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,64,power_law_1.2,0.08426240086555481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,64,power_law_1.2,0.10948480367660522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,4,power_law_1.01,1.4928000450134278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,64,power_law_1.2,0.15697920322418213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,64,power_law_1.2,0.2237312078475952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,64,power_law_1.2,0.21818881034851073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,64,balanced,0.01923199991385142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,64,balanced,0.019258666783571243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,64,balanced,0.01932266727089882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,64,balanced,0.018954666952292126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,64,balanced,0.020015999674797058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,64,balanced,0.023018665611743927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,64,balanced,0.023077333966890972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,64,balanced,0.03252800057331721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,64,balanced,0.03329599897066752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,64,balanced,0.0332640012105306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,64,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,64,balanced,0.03326933334271113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,64,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,64,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,64,balanced,0.03949866692225138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,64,balanced,0.035599999129772186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,64,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,64,balanced,0.04257600009441376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,64,balanced,0.0452159990866979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,64,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,64,balanced,0.06578133503595988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,64,balanced,0.08060266574223836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,64,balanced,0.09834667046864827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,64,balanced,0.13020267089207968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,64,balanced,0.16662399967511496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,64,balanced,0.23231999079386392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,64,balanced,0.3001493414243062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,8,8,power_law_1.2,0.05134080052375793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,8,8,power_law_1.2,0.06794880032539367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,8,8,power_law_1.2,0.08305919766426087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,8,8,power_law_1.2,0.09381759762763978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,8,8,power_law_1.2,0.10609920024871826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,8,8,power_law_1.2,0.11934080123901367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,8,8,power_law_1.2,0.12318079471588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,8,8,power_law_1.2,0.12407679557800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,8,8,power_law_1.2,0.11727999448776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,8,8,power_law_1.2,0.10858880281448365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,8,8,power_law_1.2,0.12484480142593384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,8,8,power_law_1.2,0.12620799541473388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,8,8,power_law_1.2,0.12866560220718384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,8,8,power_law_1.2,0.12379519939422608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,8,8,power_law_1.2,0.13267199993133544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,8,8,power_law_1.2,0.13849600553512573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,8,8,power_law_1.2,0.1438655972480774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,8,8,power_law_1.2,0.15690239667892455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,8,8,power_law_1.2,0.18286720514297486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,8,8,power_law_1.2,0.21878399848937988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,8,8,power_law_1.2,0.23264000415802003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,8,8,power_law_1.2,0.2916032075881958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,8,8,power_law_1.2,0.3816704034805298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,8,8,power_law_1.2,0.47588481903076174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,8,8,power_law_1.2,0.5977536201477051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,8,8,power_law_1.2,0.8787263870239258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,8,8,power_law_1.2,1.0521344184875487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,16,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,16,balanced,0.05189866820971171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,16,balanced,0.06990399956703186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,16,balanced,0.0993280013402303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,16,balanced,0.1520853340625763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,16,balanced,0.24626133839289346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,16,balanced,0.34879998366038006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,16,balanced,0.34698665142059326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,16,balanced,0.3487519820531209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,16,balanced,0.3497920036315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,16,balanced,0.3527573347091675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,16,balanced,0.3512959877649943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,16,balanced,0.3564266761144002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,16,balanced,0.3599040110905965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,16,balanced,0.36167999108632404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,16,balanced,0.36764268080393475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,16,balanced,0.3678133487701416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,16,balanced,0.37812264760335285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,16,balanced,0.38317867120107013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,16,balanced,0.4022293488184611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,16,balanced,0.4193706512451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,16,balanced,0.45766933759053546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,16,balanced,0.4957866668701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,16,balanced,0.5583626826604208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,16,balanced,0.665231982866923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,16,balanced,0.7858186562856039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,16,balanced,1.0062133471171062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,4,power_law_1.2,0.031744000315666196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,4,power_law_1.2,0.03164800107479095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,4,power_law_1.2,0.033548799157142636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,4,power_law_1.2,0.039289599657058714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,4,power_law_1.2,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,4,power_law_1.2,0.04976640045642853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,4,power_law_1.2,0.0525056004524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,4,power_law_1.2,0.05384320020675659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,4,power_law_1.2,0.05583360195159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,4,power_law_1.2,0.05674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,4,power_law_1.2,0.05488640069961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,4,power_law_1.2,0.057145601511001586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,4,power_law_1.2,0.057920002937316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,4,power_law_1.2,0.06255999803543091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,4,power_law_1.2,0.06765440106391907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,4,power_law_1.2,0.06802560091018676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,4,power_law_1.2,0.06906880140304565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,4,power_law_1.2,0.07912319898605347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,4,power_law_1.2,0.07976319789886474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,4,power_law_1.2,0.089055997133255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,4,power_law_1.2,0.10239360332489014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,4,power_law_1.2,0.1325503945350647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,4,power_law_1.2,0.14878079891204835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,4,power_law_1.2,0.201580810546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,4,power_law_1.2,0.2572927951812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,4,power_law_1.2,0.33253118991851804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,4,power_law_1.2,0.4521279811859131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,1,power_law_1.2,0.018291200697422027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,1,power_law_1.2,0.01988479942083359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,1,power_law_1.2,0.026739200949668883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,1,power_law_1.2,0.03415040075778961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,1,power_law_1.2,0.0451200008392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,1,power_law_1.2,0.05689600110054016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,1,power_law_1.2,0.07210239768028259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,1,power_law_1.2,0.07663360238075256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,1,power_law_1.2,0.07798399925231933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,1,power_law_1.2,0.07956479787826538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,1,power_law_1.2,0.08584319949150085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,1,power_law_1.2,0.08928639888763427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,1,power_law_1.2,0.09264000058174134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,1,power_law_1.2,0.10340479612350464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,1,power_law_1.2,0.09844480156898498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,1,power_law_1.2,0.1039423942565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,1,power_law_1.2,0.11434880495071412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,1,power_law_1.2,0.11431039571762085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,1,power_law_1.2,0.12615679502487182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,1,power_law_1.2,0.13449599742889404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,1,power_law_1.2,0.15066879987716675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,4,power_law_1.2,0.01722240000963211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,1,power_law_1.2,0.2935744047164917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,4,power_law_1.2,0.018086400628089905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,1,power_law_1.2,0.3580096006393433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,4,power_law_1.2,0.01738879978656769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,1,power_law_1.2,0.2942784070968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,4,power_law_1.2,0.017900800704956053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,4,power_law_1.2,0.018080000579357148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,1,power_law_1.2,0.35152640342712405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,32,power_law_1.2,0.15688320398330688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,4,power_law_1.2,0.019571200013160706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,4,power_law_1.2,0.02056960016489029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,1,power_law_1.2,0.44475522041320803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,4,power_law_1.2,0.022700800001621245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,4,power_law_1.2,0.025593599677085875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,32,power_law_1.2,0.15459200143814086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,1,power_law_1.2,0.5485824108123779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,4,power_law_1.2,0.02643199861049652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,4,power_law_1.2,0.02258560061454773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,4,power_law_1.2,0.024031999707221984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,4,power_law_1.2,0.02383359968662262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,4,power_law_1.2,0.027539199590682982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,4,power_law_1.2,0.02995840013027191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,4,power_law_1.2,0.03111039996147156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,4,power_law_1.2,0.03294720053672791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,4,power_law_1.2,0.037273600697517395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,4,power_law_1.2,0.051820802688598636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,4,power_law_1.2,0.06149759888648987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,32,power_law_1.2,0.1509376049041748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,4,power_law_1.2,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,32,power_law_1.2,0.15637760162353515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,4,power_law_1.2,0.06993280053138733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,4,power_law_1.2,0.07838720083236694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,32,power_law_1.2,0.1575103998184204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,32,power_law_1.2,0.1559231996536255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,32,power_law_1.2,0.1577855944633484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,32,power_law_1.2,0.15591039657592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,32,power_law_1.2,0.14523520469665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,4,power_law_1.2,0.10497280359268188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,32,power_law_1.2,0.1606528043746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,32,power_law_1.2,0.1601855993270874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,32,power_law_1.2,0.16184959411621094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,32,power_law_1.2,0.16567039489746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,32,power_law_1.2,0.17063679695129394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,32,power_law_1.2,0.17511680126190185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,32,power_law_1.2,0.17716480493545533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,32,power_law_1.2,0.19007999897003175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,32,power_law_1.2,0.21288959980010985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,32,power_law_1.2,0.2401792049407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,32,power_law_1.2,0.3022592067718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,32,power_law_1.2,0.3591936111450195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,4,power_law_1.2,0.03548159897327423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,4,power_law_1.2,0.030214399099349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,4,power_law_1.2,0.03175680041313171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,4,power_law_1.2,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,4,power_law_1.2,0.035180801153182985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,4,power_law_1.2,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,4,power_law_1.2,0.12705279588699342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,4,power_law_1.2,0.038553598523139956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,4,power_law_1.2,0.03818239867687225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,4,power_law_1.2,0.03888640105724335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,32,power_law_1.2,0.4794112205505371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,4,power_law_1.2,0.03937920033931732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,4,power_law_1.2,0.04229759871959686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,32,power_law_1.2,0.5748799800872803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,32,power_law_1.2,0.8522432327270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,32,power_law_1.2,1.052614402770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,4,power_law_1.2,0.04366720020771027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,4,power_law_1.2,0.045977601408958436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,4,power_law_1.2,0.04636160135269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,4,power_law_1.2,0.052262401580810545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,4,power_law_1.2,0.05383679866790771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,4,power_law_1.2,0.05796480178833008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,4,power_law_1.2,0.06469759941101075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,4,power_law_1.2,0.07566720247268677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,4,power_law_1.2,0.09520639777183533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,4,power_law_1.2,0.11239680051803588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,4,power_law_1.2,0.14625279903411864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,32,power_law_1.2,1.7982336044311524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,4,power_law_1.2,0.17959680557250976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,32,power_law_1.2,2.1842624664306642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,4,power_law_1.2,0.17505279779434205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,4,power_law_1.2,0.24630401134490967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,4,power_law_1.2,0.3302527904510498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,8,power_law_1.2,0.03361920118331909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,8,power_law_1.2,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,8,power_law_1.2,0.04252159893512726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,8,power_law_1.2,0.055801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,8,power_law_1.2,0.05203199982643127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,8,power_law_1.2,0.05548800230026245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,4,power_law_1.2,0.49045119285583494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,8,power_law_1.2,0.06570240259170532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,8,power_law_1.2,0.0655232012271881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,8,power_law_1.2,0.06402559876441956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,8,power_law_1.2,0.0886847972869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,8,power_law_1.2,0.0819263994693756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,8,power_law_1.2,0.07668480277061462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,8,power_law_1.2,0.11445120573043824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,8,power_law_1.2,0.1148800015449524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,8,power_law_1.2,0.1296064019203186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,8,power_law_1.2,0.11478400230407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,8,power_law_1.2,0.1144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,8,power_law_1.2,0.12174719572067261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.2,0.13075200319290162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.2,0.1401087999343872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.2,0.15132800340652466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.2,0.20149118900299073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.2,0.23249919414520265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.2,0.3163840055465698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.2,0.4068160057067871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.2,0.5771391868591309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.2,0.7574592113494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,4,power_law_1.2,0.22760961055755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,2,power_law_1.2,0.023398399353027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,2,power_law_1.2,0.028262400627136232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,2,power_law_1.2,0.04445439875125885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,2,power_law_1.2,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,2,power_law_1.2,0.06234880089759827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,2,power_law_1.2,0.08339840173721313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,2,power_law_1.2,0.08776320219039917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,2,power_law_1.2,0.09362559914588928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,2,power_law_1.2,0.09656959772109985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,2,power_law_1.2,0.10185600519180298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,2,power_law_1.2,0.10249600410461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,2,power_law_1.2,0.10796159505844116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,2,power_law_1.2,0.11328639984130859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,2,power_law_1.2,0.10937600135803223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,2,power_law_1.2,0.11519360542297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,2,power_law_1.2,0.121670401096344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,2,power_law_1.2,0.12753920555114745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,2,power_law_1.2,0.13640960454940795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.2,0.15564800500869752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.2,0.1988927960395813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.2,0.2375488042831421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.2,0.2744191884994507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.2,0.3216383934020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.2,0.4066304206848145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,4,power_law_1.2,0.6581888198852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.2,0.4821760177612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.2,0.6820799827575683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,16,power_law_1.01,0.06887680292129517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,16,power_law_1.01,0.08080639839172363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,16,power_law_1.01,0.0790336012840271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,16,power_law_1.01,0.07776640057563781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,16,power_law_1.01,0.08065279722213745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,16,power_law_1.01,0.0716480016708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,16,power_law_1.01,0.07419520020484924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,16,power_law_1.01,0.07677440047264099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,16,power_law_1.01,0.07477120161056519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,16,power_law_1.01,0.07503359913825988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,16,power_law_1.01,0.07565439939498901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,16,power_law_1.01,0.077292799949646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,16,power_law_1.01,0.07670400142669678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,16,power_law_1.01,0.07736960053443909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,16,power_law_1.01,0.08407040238380432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,16,power_law_1.01,0.08447359800338745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,16,power_law_1.01,0.08869760036468506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,16,power_law_1.01,0.09898880124092102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,16,power_law_1.01,0.10431360006332398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,16,power_law_1.01,0.12083840370178223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,16,power_law_1.01,0.1394752025604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,16,power_law_1.01,0.17176320552825927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,16,power_law_1.01,0.2019968032836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,16,power_law_1.01,0.2649791955947876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,16,power_law_1.01,0.3446719884872437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,16,power_law_1.01,0.4678463935852051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,16,power_law_1.01,0.6134975910186767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.2,0.8625727653503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,16,power_law_1.2,0.042847999930381776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,16,power_law_1.2,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,16,power_law_1.2,0.039059200882911684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,16,power_law_1.2,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,16,power_law_1.2,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,16,power_law_1.2,0.0430976003408432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,16,power_law_1.2,0.04335359930992126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,16,power_law_1.2,0.04455040097236633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,16,power_law_1.2,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,16,power_law_1.2,0.04418559968471527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,16,power_law_1.2,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,16,power_law_1.2,0.05495679974555969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,16,power_law_1.2,0.05580800175666809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,16,power_law_1.2,0.057606399059295654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,16,power_law_1.2,0.06245120167732239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,16,power_law_1.2,0.06332799792289734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,16,power_law_1.2,0.07249280214309692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,16,power_law_1.2,0.08135679960250855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,16,power_law_1.2,0.09521920084953309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,16,power_law_1.2,0.12028800249099732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,16,power_law_1.2,0.1418944001197815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,16,power_law_1.2,0.19935359954833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,16,power_law_1.2,0.28677120208740237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,64,power_law_1.2,0.12013440132141114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,64,power_law_1.2,0.07288960218429566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,64,power_law_1.2,0.07044479846954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,64,power_law_1.2,0.07470719814300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,64,power_law_1.2,0.0766592025756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,64,power_law_1.2,0.06975359916687011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,64,power_law_1.2,0.07459840178489685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,64,power_law_1.2,0.07920640110969543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,4,balanced,0.07032533486684163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,4,balanced,0.0718453327814738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,64,power_law_1.2,0.07609599828720093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,4,balanced,0.07668800155321757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,4,balanced,0.09262399872144063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,64,power_law_1.2,0.07802879810333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,64,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,4,balanced,0.10897599657376607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,64,power_law_1.2,0.07982079982757569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,4,balanced,0.14250133434931436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,64,power_law_1.2,0.07942399978637696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,64,power_law_1.2,0.08224639892578126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,16,power_law_1.2,0.4313024044036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,64,power_law_1.2,0.09100160002708435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,64,power_law_1.2,0.09225599765777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,64,power_law_1.2,0.09505280256271362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,64,power_law_1.2,0.10924160480499268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,64,power_law_1.2,0.12191359996795655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,4,balanced,0.1418880025545756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,4,balanced,0.14338666200637817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,4,balanced,0.14652799566586813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,4,balanced,0.14422399799029031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,4,balanced,0.14666133125623068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,4,balanced,0.150325338045756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,4,balanced,0.15056533614794412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,4,balanced,0.15414399902025858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,4,balanced,0.16030399998029074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,4,balanced,0.16075199842453003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,64,power_law_1.2,0.14339840412139893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,4,balanced,0.16532267133394876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,4,balanced,0.1816800038019816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,4,balanced,0.19056000312169394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,4,balanced,0.22324266036351523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,4,balanced,0.25271467367808026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,4,balanced,0.31569067637125653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,4,balanced,0.3802666664123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,64,power_law_1.2,0.16830719709396363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,64,power_law_1.2,0.2085184097290039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,64,power_law_1.2,0.2547904014587402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,64,power_law_1.2,0.35459840297698975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,4,balanced,0.5450933376948038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,64,power_law_1.2,0.4527103900909424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,4,balanced,0.6680800120035807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,64,power_law_1.2,0.6074560165405274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,4,balanced,0.9578346411387125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,4,balanced,1.2571093241373699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,64,power_law_1.2,0.8054207801818848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,16,power_law_1.2,0.5320576190948486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,1,power_law_1.01,0.05533440113067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,1,power_law_1.01,0.05885440111160278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,1,power_law_1.01,0.073471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,1,power_law_1.01,0.10428800582885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,1,power_law_1.01,0.13303680419921876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,1,power_law_1.01,0.1657663941383362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,1,power_law_1.01,0.21519360542297364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,1,power_law_1.01,0.2181567907333374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,1,power_law_1.01,0.22798080444335939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,1,power_law_1.01,0.22986879348754882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,1,power_law_1.01,0.2377023935317993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,1,power_law_1.01,0.24634881019592286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,2,power_law_1.2,0.02024320065975189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,1,power_law_1.01,0.2550719976425171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,1,power_law_1.01,0.26425600051879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,1,power_law_1.01,0.2701375961303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,1,power_law_1.01,0.27879040241241454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,1,power_law_1.01,0.30312960147857665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,1,power_law_1.01,0.35114240646362305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,1,power_law_1.01,0.38604800701141356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,1,power_law_1.01,0.47233920097351073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,1,power_law_1.01,0.5355264186859131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,1,power_law_1.01,0.6894464015960693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,1,power_law_1.01,0.8457728385925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,2,power_law_1.2,0.024633599817752837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,1,power_law_1.01,1.1337087631225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,2,power_law_1.2,0.033228799700737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,2,power_law_1.2,0.04523519873619079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,1,power_law_1.01,1.4436415672302245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,2,power_law_1.2,0.051283198595047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,1,power_law_1.01,1.9857088088989259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,2,power_law_1.2,0.06391040086746216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,2,power_law_1.2,0.07639039754867553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,1,power_law_1.01,2.5195968627929686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,2,power_law_1.2,0.07888640165328979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,2,power_law_1.2,0.0838591992855072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,2,power_law_1.2,0.08696320056915283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,2,power_law_1.2,0.09100800156593322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,2,power_law_1.2,0.0893567979335785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,2,power_law_1.2,0.0923200011253357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,2,power_law_1.2,0.09764479994773864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,2,power_law_1.2,0.09934719800949096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,2,power_law_1.2,0.10025600194931031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,2,power_law_1.2,0.1064255952835083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,2,power_law_1.2,0.11610879898071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,2,power_law_1.2,0.1210752010345459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,2,power_law_1.2,0.24186880588531495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,2,power_law_1.2,0.2747391939163208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,2,power_law_1.2,0.20696959495544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,2,power_law_1.2,0.2326143980026245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,2,power_law_1.2,0.27457919120788576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,2,power_law_1.2,0.32344961166381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,2,power_law_1.2,0.43343358039855956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,2,power_law_1.2,0.5322303771972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,16,power_law_1.2,0.7996223926544189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,1,power_law_1.2,0.02054399996995926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,1,power_law_1.2,0.02290560007095337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,1,power_law_1.2,0.029824000597000123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,1,power_law_1.2,0.043296000361442565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,1,power_law_1.2,0.06005120277404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,1,power_law_1.2,0.07687039971351624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,1,power_law_1.2,0.09549440145492553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,1,power_law_1.2,0.1046720027923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,1,power_law_1.2,0.10638079643249512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,1,power_law_1.2,0.10929919481277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,1,power_law_1.2,0.11100800037384033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,1,power_law_1.2,0.11560319662094116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,1,power_law_1.2,0.12008960247039795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,1,power_law_1.2,0.12340480089187622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,1,power_law_1.2,0.1257024049758911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,1,power_law_1.2,0.1333184003829956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,1,power_law_1.2,0.14145920276641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,1,power_law_1.2,0.16241919994354248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,1,power_law_1.2,0.17555199861526488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,4,power_law_1.2,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,1,power_law_1.2,0.21222400665283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,4,power_law_1.2,0.03983359932899475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,1,power_law_1.2,0.24849278926849366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,4,power_law_1.2,0.047276800870895384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,1,power_law_1.2,0.327891206741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,1,power_law_1.2,0.33983359336853025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,1,power_law_1.2,0.4592639923095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,1,power_law_1.2,0.5683328151702881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,1,power_law_1.2,0.7960576057434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,4,power_law_1.2,0.05998079776763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,4,power_law_1.2,0.058297598361968996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,1,power_law_1.2,1.0314751625061036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,4,power_law_1.2,0.0778432011604309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,4,power_law_1.2,0.07932800054550171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,4,power_law_1.2,0.08194559812545776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,4,power_law_1.2,0.08455039858818054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,4,power_law_1.2,0.08170239925384522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,4,power_law_1.2,0.08446720242500305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,4,power_law_1.2,0.08929280042648316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,4,power_law_1.2,0.09296000003814697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,4,power_law_1.2,0.09546239972114563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,4,power_law_1.2,0.10450559854507446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,4,power_law_1.2,0.10572160482406616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,4,power_law_1.2,0.1123263955116272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,4,power_law_1.2,0.12584960460662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,4,power_law_1.2,0.14309120178222656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,4,power_law_1.2,0.1750272035598755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,4,power_law_1.2,0.19857280254364013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,4,power_law_1.2,0.26513280868530276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,16,power_law_1.2,1.044422435760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,1,balanced,0.08628799517949422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,1,balanced,0.13593066732088724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,1,balanced,0.23561600844065347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,1,balanced,0.23944532871246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,1,balanced,0.24222399791081747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,1,balanced,0.2422879934310913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,1,balanced,0.2441813349723816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,1,balanced,0.24798399209976196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,1,balanced,0.2490826646486918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,1,balanced,0.25037866830825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,1,balanced,0.25410666068394977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,1,balanced,0.2663840055465698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,1,balanced,0.2702453335126241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,1,balanced,0.27796266476313275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,1,balanced,0.2868160009384155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,1,balanced,0.2919466694196065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,1,balanced,0.30869332949320477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,1,balanced,0.38443199793497723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,1,balanced,0.3882933457692464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,1,balanced,0.5442613363265991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,1,balanced,0.6046186685562134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,1,balanced,0.8567039966583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,4,power_law_1.2,0.35015039443969725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,1,balanced,1.0658559799194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,1,balanced,1.5754559834798176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,1,balanced,2.0831626256306968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,1,balanced,3.052805264790853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,1,balanced,4.043600082397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,4,power_law_1.2,0.4802879810333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,4,power_law_1.2,0.5591807842254639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,4,power_law_1.2,0.8541376113891601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,4,balanced,0.03369600077470144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,4,balanced,0.029498666524887085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,4,balanced,0.028175999720891316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,4,balanced,0.031258667508761086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,4,balanced,0.02942399928967158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,4,balanced,0.03201066702604294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,4,balanced,0.031328000128269196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,4,balanced,0.031178665657838184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,4,balanced,0.031712000568707786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,4,balanced,0.03160533308982849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,4,balanced,0.0315786674618721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,4,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,4,balanced,0.031888000667095184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,4,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,4,balanced,0.04153066625197729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,4,balanced,0.0397173340121905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,4,balanced,0.039621333281199135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,4,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,4,balanced,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,4,balanced,0.04986133178075155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,4,balanced,0.05314666529496511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,4,balanced,0.07050666709740956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,4,balanced,0.08265066643555959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,4,balanced,0.10925333698590596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,4,balanced,0.1243946651617686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,4,balanced,0.16538133223851523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,4,balanced,0.19919466972351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,32,balanced,0.06311466793219249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,32,balanced,0.06180266539255778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,32,balanced,0.06277333199977875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,32,balanced,0.07468266785144806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,32,balanced,0.10094400246938069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,32,balanced,0.14223466316858926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,32,balanced,0.13505066434542337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,32,balanced,0.1316213309764862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,32,balanced,0.1279093325138092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,32,balanced,0.13216533263524374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,32,balanced,0.12682132919629416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,32,balanced,0.12639466921488443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,32,balanced,0.12914133071899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,32,balanced,0.12782399853070578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,32,balanced,0.13269866506258646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,32,balanced,0.13381333152453104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,32,balanced,0.13470932841300964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,32,balanced,0.14804266889890036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,32,balanced,0.14152000347773233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,32,balanced,0.16085867087046304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,32,balanced,0.16505066553751627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,32,balanced,0.18869866927464804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,32,balanced,0.21428799629211426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,32,balanced,0.3091573317845662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,32,balanced,0.350544015566508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,32,balanced,0.4837760130564372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,32,balanced,0.6172853310902914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,4,power_law_1.2,1.1037504196166992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,8,power_law_1.2,0.026528000831604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,8,power_law_1.2,0.03091199994087219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,8,power_law_1.2,0.038553598523139956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,8,power_law_1.2,0.040166398882865904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,8,power_law_1.2,0.036524799466133115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,4,power_law_1.2,0.08517119884490967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,8,power_law_1.2,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,4,power_law_1.2,0.10261119604110717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,8,power_law_1.2,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,8,power_law_1.2,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,4,power_law_1.2,0.11255040168762206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,8,power_law_1.2,0.0717311978340149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,4,power_law_1.2,0.14999040365219116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,8,power_law_1.2,0.06967679858207702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,8,power_law_1.2,0.06759679913520814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,4,power_law_1.2,0.15813119411468507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,8,power_law_1.2,0.08552320003509521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,4,power_law_1.2,0.17786240577697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,8,power_law_1.2,0.08602880239486695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,8,power_law_1.2,0.08521599769592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,4,power_law_1.2,0.20936319828033448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,8,power_law_1.2,0.09463040232658386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,4,power_law_1.2,0.1898687958717346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,8,power_law_1.2,0.10167039632797241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,8,power_law_1.2,0.08739200234413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,8,power_law_1.2,0.09146239757537841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.2,0.09557759761810303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.2,0.10740480422973633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.2,0.11690880060195923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.2,0.15073920488357545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,4,power_law_1.2,0.19894399642944335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.2,0.1803007960319519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,4,power_law_1.2,0.21205759048461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,4,power_law_1.2,0.19452159404754638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,4,power_law_1.2,0.1980288028717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,4,power_law_1.2,0.20891520977020264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,4,power_law_1.2,0.2142335891723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,4,power_law_1.2,0.21721599102020264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,4,power_law_1.2,0.22101759910583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,4,power_law_1.2,0.22723839282989503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,4,power_law_1.2,0.25552639961242674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.2,0.27857279777526855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.2,0.31349759101867675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.2,0.24791040420532226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.2,0.3496959924697876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.2,0.45389437675476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.2,0.5344575881958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.2,0.6812992095947266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.2,0.9036735534667969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.2,1.311577606201172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.2,1.6540544509887696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.2,0.3097856044769287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,32,balanced,0.05843733251094818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,32,balanced,0.06015466650327047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,32,balanced,0.05977599819501241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,32,balanced,0.07239466905593872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,32,balanced,0.08630399902661641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,32,balanced,0.12115200360616048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,32,balanced,0.11542399724324544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,32,balanced,0.12204800049463908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,32,balanced,0.1160426636536916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,32,balanced,0.11869866649309795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,32,balanced,0.12204266587893169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,32,balanced,0.11667199929555257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,32,balanced,0.11524266997973125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,32,balanced,0.11940266688664754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,32,balanced,0.12345066666603088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,32,balanced,0.12417067090670268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,32,balanced,0.12187199791272481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,32,balanced,0.12785599629084268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,32,balanced,0.13409599661827087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,32,balanced,0.15288000305493674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,32,balanced,0.1562399963537852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,32,balanced,0.1785279909769694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,32,balanced,0.20336000124613443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,32,balanced,0.28777599334716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,32,balanced,0.32899200916290283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,32,balanced,0.45453333854675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,32,balanced,0.5788480043411255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,1,power_law_1.01,0.03431679904460907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,1,power_law_1.01,0.03614720106124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,1,power_law_1.01,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,1,power_law_1.01,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,1,power_law_1.01,0.06117759943008423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,1,power_law_1.01,0.07820159792900086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,1,power_law_1.01,0.08154240250587463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.2,0.49132800102233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,1,power_law_1.01,0.08352640271186829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,1,power_law_1.01,0.08423039913177491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,1,power_law_1.01,0.08602880239486695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,1,power_law_1.01,0.09100800156593322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,1,power_law_1.01,0.0934719979763031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,1,power_law_1.01,0.09833599925041199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,1,power_law_1.01,0.10663679838180543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,1,power_law_1.01,0.11400959491729737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,1,power_law_1.01,0.12045439481735229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,1,power_law_1.01,0.12666239738464355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,1,power_law_1.01,0.1505280017852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,1,power_law_1.01,0.1752768039703369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,1,power_law_1.01,0.22026240825653076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,1,power_law_1.01,0.2765568017959595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,1,power_law_1.01,0.3759360074996948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,1,power_law_1.01,0.4578559875488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,1,power_law_1.01,0.6246592044830322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,1,power_law_1.01,0.8024448394775391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,64,balanced,0.046037331223487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,64,balanced,0.04598933458328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,64,balanced,0.04799466828505198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,64,balanced,0.05306133131186167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,64,balanced,0.05398933092753092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,64,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,64,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,64,balanced,0.05603733162085215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,64,balanced,0.056847999493281044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,64,balanced,0.058133333921432495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,64,balanced,0.0581279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,64,balanced,0.060677334666252136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,64,balanced,0.062037333846092224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,1,power_law_1.01,1.1640000343322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,64,balanced,0.06200533111890157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,64,balanced,0.0680266668399175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,64,balanced,0.0683786670366923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,64,balanced,0.07239999870459239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,64,balanced,0.08886933326721191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,1,power_law_1.01,1.502400016784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,64,balanced,0.08866133292516072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,64,balanced,0.11739200353622437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,64,balanced,0.13201066851615906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,64,balanced,0.18910932540893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,64,balanced,0.18865066766738892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,64,balanced,0.24461867411931357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,64,balanced,0.2874773343404134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,64,balanced,0.38811735312143963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,64,balanced,0.46724267800649005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.2,0.5739200115203857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,2,64,power_law_1.2,0.03498879969120026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,2,64,power_law_1.2,0.03475840091705322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,2,64,power_law_1.2,0.03079040050506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,2,64,power_law_1.2,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,2,64,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,2,64,power_law_1.2,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,2,64,power_law_1.2,0.03336319923400879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,2,64,power_law_1.2,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,2,64,power_law_1.2,0.033881598711013795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,2,64,power_law_1.2,0.03487359881401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,2,64,power_law_1.2,0.03521279990673065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,2,64,power_law_1.2,0.035897600650787356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,2,64,power_law_1.2,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,2,64,power_law_1.2,0.040243199467658995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,2,64,power_law_1.2,0.04590719938278198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,2,64,power_law_1.2,0.04872959852218628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,2,64,power_law_1.2,0.04754559993743897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,2,64,power_law_1.2,0.05904639959335327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,2,64,power_law_1.2,0.06364799737930298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,2,64,power_law_1.2,0.08672639727592468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,2,64,power_law_1.2,0.09307519793510437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,2,64,power_law_1.2,0.12648320198059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,2,64,power_law_1.2,0.1327679991722107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,2,64,power_law_1.2,0.18250240087509156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,2,64,power_law_1.2,0.22689919471740722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,32,balanced,0.029738667110602062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,32,balanced,0.026159999271233875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,32,balanced,0.02717333287000656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,32,balanced,0.029680001238981884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,32,balanced,0.03142933299144109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,32,balanced,0.033258666594823204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,32,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,32,balanced,0.03332799921433131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,32,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,32,balanced,0.03329066683848699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,32,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,32,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,32,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,32,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,32,balanced,0.037946666280428566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,32,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,32,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,32,balanced,0.04176533222198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,32,balanced,0.04005866746107737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,32,balanced,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,32,balanced,0.046682665745417275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,32,balanced,0.053770666321118675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,32,balanced,0.05607999861240387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,32,balanced,0.08270399769147237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,32,balanced,0.10930666327476501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,32,balanced,0.14660799503326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,2,16,power_law_1.01,0.08354560136795045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,2,16,power_law_1.01,0.08949120044708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,2,16,power_law_1.01,0.08718079924583436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,2,16,power_law_1.01,0.0956928014755249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,32,balanced,0.17594132820765176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,2,16,power_law_1.01,0.09582080245018006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,2,16,power_law_1.01,0.09515519738197327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,2,16,power_law_1.01,0.0940671980381012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,2,16,power_law_1.01,0.09607040286064147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,2,16,power_law_1.01,0.09616640210151672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,2,16,power_law_1.01,0.09861119985580444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,2,16,power_law_1.01,0.10081280469894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,2,16,power_law_1.01,0.10381439924240113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,2,16,power_law_1.01,0.10464639663696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,2,16,power_law_1.01,0.10947840213775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,2,16,power_law_1.01,0.11728639602661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,2,16,power_law_1.01,0.118995201587677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,2,16,power_law_1.01,0.128985595703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,2,16,power_law_1.01,0.15081599950790406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,2,16,power_law_1.01,0.16584960222244263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,2,16,power_law_1.01,0.21390719413757325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,2,16,power_law_1.01,0.23654398918151856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,2,16,power_law_1.01,0.30147199630737304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,2,16,power_law_1.01,0.346995210647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,2,16,power_law_1.01,0.45438079833984374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,2,16,power_law_1.01,0.5489855766296386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,2,16,power_law_1.01,0.8249471664428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,2,16,power_law_1.01,0.9923199653625489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,2,64,power_law_1.2,0.3209536075592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,2,64,power_law_1.2,0.4954239845275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,1,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,1,power_law_1.01,0.036582401394844054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,1,power_law_1.01,0.03779839873313904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,1,power_law_1.01,0.04805119931697845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,1,power_law_1.01,0.04854399859905243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,1,power_law_1.01,0.048876801133155824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,1,power_law_1.01,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,1,power_law_1.01,0.04940800070762634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,1,power_law_1.01,0.05017600059509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,1,power_law_1.01,0.05120639801025391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,1,power_law_1.01,0.05167359709739685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,1,power_law_1.01,0.05381760001182556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,1,power_law_1.01,0.05741440057754517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,1,power_law_1.01,0.05948160290718078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,1,power_law_1.01,0.0726144015789032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,1,power_law_1.01,0.07869439721107482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,1,power_law_1.01,0.08039680123329163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,1,power_law_1.01,0.09067519903182983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.01,0.11288959980010986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,256,power_law_1.01,0.05975040197372437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,256,power_law_1.01,0.06324480175971985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,256,power_law_1.01,0.058873599767684935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,256,power_law_1.01,0.065830397605896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,256,power_law_1.01,0.06293759942054748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,1,power_law_1.2,0.049728000164031984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,256,power_law_1.01,0.06517120003700257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,1,power_law_1.2,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,256,power_law_1.01,0.06502400040626526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,1,power_law_1.2,0.053401601314544675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,256,power_law_1.01,0.06549760103225707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,1,power_law_1.2,0.06307200193405152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,256,power_law_1.01,0.06636800169944763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,1,power_law_1.2,0.07418879866600037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,256,power_law_1.01,0.06854400038719177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,1,power_law_1.2,0.086080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,1,power_law_1.2,0.10444799661636353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,1,power_law_1.2,0.10669440031051636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,256,power_law_1.01,0.06624000072479248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,1,power_law_1.2,0.11020799875259399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,256,power_law_1.01,0.0667967975139618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.01,0.14190720319747924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,1,power_law_1.2,0.11281280517578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,256,power_law_1.01,0.06814720034599304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,1,power_law_1.2,0.11637760400772094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,256,power_law_1.01,0.06813439726829529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,1,power_law_1.2,0.12261120080947877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,256,power_law_1.01,0.07214720249176025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,1,power_law_1.2,0.1278720021247864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,1,power_law_1.2,0.12624640464782716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,1,power_law_1.2,0.1325055956840515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,1,power_law_1.2,0.13698559999465942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,1,power_law_1.2,0.15919359922409057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,1,power_law_1.2,0.18379520177841185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.01,0.1678976058959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,1,power_law_1.2,0.2068864107131958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,1,power_law_1.2,0.2520064115524292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,1,power_law_1.2,0.2982912063598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,1,power_law_1.2,0.3769023895263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,256,power_law_1.01,0.0695743978023529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,256,power_law_1.01,0.07044479846954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,256,power_law_1.01,0.08750720024108886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.01,0.08683519959449768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.01,0.11230080127716065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.01,0.12795519828796387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,1,power_law_1.2,0.4719679832458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.01,0.16670080423355102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,1,power_law_1.2,0.6328703880310058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,1,power_law_1.2,0.8067071914672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,1,power_law_1.2,1.1384767532348632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,1,power_law_1.2,1.4759103775024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.01,0.19958399534225463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.01,0.2715775966644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.01,0.3399807929992676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.01,0.47309441566467286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.01,0.6435391902923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.01,0.23248639106750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,1,power_law_1.01,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,1,power_law_1.01,0.10076160430908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,1,power_law_1.01,0.1516543984413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,1,power_law_1.01,0.2279616117477417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,8,power_law_1.2,0.08561279773712158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,8,power_law_1.2,0.11316479444503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,8,power_law_1.2,0.11489280462265014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,8,power_law_1.2,0.12338559627532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,8,power_law_1.2,0.14412800073623658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,1,power_law_1.01,0.29913599491119386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,8,power_law_1.2,0.14659199714660645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,1,power_law_1.01,0.4620607852935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,8,power_law_1.2,0.1579967975616455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,1,power_law_1.01,0.4957376003265381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,8,power_law_1.2,0.15741440057754516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,1,power_law_1.01,0.5091392040252686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,1,power_law_1.01,0.5236671924591064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,1,power_law_1.01,0.5250944137573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.01,0.2988607883453369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,1,power_law_1.01,0.5452991962432862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,1,power_law_1.01,0.5704768180847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,1,power_law_1.01,0.5698048114776612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,1,power_law_1.01,0.5779007911682129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,8,power_law_1.2,0.16001919507980347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,1,power_law_1.01,0.6086976051330566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,8,power_law_1.2,0.1552448034286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,1,power_law_1.01,0.626854419708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,8,power_law_1.2,0.1572160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,1,power_law_1.01,0.6552576065063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,1,power_law_1.01,0.7382016181945801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,1,power_law_1.01,0.7500736236572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,1,power_law_1.01,0.8732288360595704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,8,power_law_1.2,0.15635199546813966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,1,power_law_1.01,0.8989184379577637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,8,power_law_1.2,0.16740479469299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,8,power_law_1.2,0.1677248001098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,8,power_law_1.2,0.17726080417633056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,8,power_law_1.2,0.17737599611282348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,8,power_law_1.2,0.1879744052886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,1,power_law_1.01,1.0954815864562988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,1,power_law_1.01,1.290015983581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.01,0.44896640777587893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,1,power_law_1.01,1.6531328201293944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,8,power_law_1.2,0.2067199945449829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,1,power_law_1.01,2.3730752944946287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,8,power_law_1.2,0.21974399089813232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,1,power_law_1.01,3.0934335708618166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,8,power_law_1.2,0.2572160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.01,0.5768383979797364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,8,power_law_1.2,0.2945535898208618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,1,power_law_1.01,3.828665542602539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,8,power_law_1.2,0.38747520446777345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,8,power_law_1.2,0.4735551834106445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,8,power_law_1.2,0.6830080032348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,8,power_law_1.2,0.8583680152893066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,8,power_law_1.2,1.2484543800354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,8,power_law_1.2,1.6184959411621094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.01,0.871571159362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.01,1.1444928169250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,4,1,power_law_1.01,0.03991680145263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,4,1,power_law_1.01,0.0456959992647171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,4,1,power_law_1.01,0.04968959987163544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,4,1,power_law_1.01,0.06792320013046264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,4,1,power_law_1.01,0.06858879923820496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,4,1,power_law_1.01,0.06928640007972717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,4,1,power_law_1.01,0.0700543999671936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,4,1,power_law_1.01,0.07221119999885559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,4,1,power_law_1.01,0.07178239822387696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,4,1,power_law_1.01,0.07234560251235962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,4,1,power_law_1.01,0.07445759773254394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,4,1,power_law_1.01,0.07672960162162781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,4,1,power_law_1.01,0.0804032027721405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,4,1,power_law_1.01,0.0833728015422821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,4,1,power_law_1.01,0.09233279824256897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,4,1,power_law_1.01,0.09620479941368103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,4,1,power_law_1.01,0.10080000162124633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,4,1,power_law_1.01,0.11750400066375732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.01,0.133241605758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.01,0.16510720252990724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.01,0.20071680545806886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,16,power_law_1.2,0.028326401114463808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,16,power_law_1.2,0.049593600630760196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,16,power_law_1.2,0.04683519899845123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.01,0.26425600051879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,16,power_law_1.2,0.049721598625183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,16,power_law_1.2,0.046828800439834596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,16,power_law_1.2,0.04584319889545441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,16,power_law_1.2,0.05488640069961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,16,power_law_1.2,0.05413119792938233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,16,power_law_1.2,0.0707584023475647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,16,power_law_1.2,0.06817280054092408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,16,power_law_1.2,0.07061120271682739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,16,power_law_1.2,0.09473919868469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,16,power_law_1.2,0.09729920029640197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,16,power_law_1.2,0.09657599925994872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,16,power_law_1.2,0.10588159561157226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,16,power_law_1.2,0.10202879905700683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.01,0.32328319549560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,16,power_law_1.2,0.09799039959907532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,16,power_law_1.2,0.09798399806022644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.2,0.10714880228042603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.2,0.1279360055923462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.2,0.1329408049583435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.01,0.44542717933654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.2,0.1630911946296692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.2,0.20215680599212646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.2,0.2867327928543091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.2,0.36591999530792235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.2,0.5068992137908935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.2,0.6690624237060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.01,0.5658688068389892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,2,power_law_1.01,0.033081600069999696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,2,power_law_1.01,0.03080959916114807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,2,power_law_1.01,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,2,power_law_1.01,0.03728640079498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,2,power_law_1.01,0.038950398564338684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,2,power_law_1.01,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,2,power_law_1.01,0.04417920112609863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,2,power_law_1.01,0.04420480132102966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,2,power_law_1.01,0.04888319969177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,2,power_law_1.01,0.049235200881958006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,2,power_law_1.01,0.05252479910850525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,2,power_law_1.01,0.056454402208328244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,2,power_law_1.01,0.05852800011634827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,2,power_law_1.01,0.06185600161552429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,2,power_law_1.01,0.06965759992599488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,2,power_law_1.01,0.07497599720954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,2,power_law_1.01,0.07431679964065552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,2,power_law_1.01,0.08727040290832519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,2,power_law_1.01,0.10030080080032348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,2,power_law_1.01,0.12265599966049194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,2,power_law_1.01,0.1498751997947693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,2,power_law_1.01,0.19740159511566163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,2,power_law_1.01,0.24257919788360596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,2,power_law_1.01,0.3439296007156372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,2,power_law_1.01,0.4515520095825195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,2,power_law_1.01,0.6566720008850098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,2,power_law_1.01,0.8905407905578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.01,0.8064767837524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.01,1.0754816055297851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,16,4,power_law_1.2,0.030988800525665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,16,4,power_law_1.2,0.031564798951148984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,16,4,power_law_1.2,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,16,4,power_law_1.2,0.03237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,16,4,power_law_1.2,0.032262399792671204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,16,4,power_law_1.2,0.03402239978313446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,16,4,power_law_1.2,0.035385599732398985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,16,4,power_law_1.2,0.0353408008813858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,16,4,power_law_1.2,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,16,4,power_law_1.2,0.03814400136470795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,16,4,power_law_1.2,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,16,4,power_law_1.2,0.039468801021575926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,16,4,power_law_1.2,0.04051840007305145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,16,4,power_law_1.2,0.044972801208496095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,16,4,power_law_1.2,0.04876160025596619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,16,4,power_law_1.2,0.05297279953956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,16,4,power_law_1.2,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,16,4,power_law_1.2,0.06967039704322815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,16,4,power_law_1.2,0.08305919766426087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,8,power_law_1.2,0.06505600214004517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,8,power_law_1.2,0.0818943977355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,8,power_law_1.2,0.08089600205421447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,8,power_law_1.2,0.0967423975467682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,16,4,power_law_1.2,0.10602240562438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,8,power_law_1.2,0.10920319557189942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,8,power_law_1.2,0.09797120094299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,8,power_law_1.2,0.11139839887619019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,8,power_law_1.2,0.1144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,8,power_law_1.2,0.11331839561462402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,8,power_law_1.2,0.11742719411849975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,8,power_law_1.2,0.1194815993309021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,8,power_law_1.2,0.11967359781265259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,8,power_law_1.2,0.12344959974288941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,16,4,power_law_1.2,0.14032000303268433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,8,power_law_1.2,0.12480000257492066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,8,power_law_1.2,0.13583999872207642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,8,power_law_1.2,0.1398911952972412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,8,power_law_1.2,0.1497215986251831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,8,power_law_1.2,0.16706559658050538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.2,0.18946559429168702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.2,0.22227199077606202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.2,0.25174400806427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.2,0.32303359508514407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.2,0.39267840385437014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.2,0.47687039375305174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.2,0.6675648212432861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.2,0.9398528099060058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.2,1.2619968414306642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,16,4,power_law_1.2,0.17292159795761108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,8,power_law_1.01,0.15008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,8,power_law_1.01,0.24063360691070557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,8,power_law_1.01,0.4395264148712158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,8,power_law_1.01,0.36890239715576173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,8,power_law_1.01,0.4698239803314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,8,power_law_1.01,0.5752448081970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,16,4,power_law_1.2,0.2294015884399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,8,power_law_1.01,0.5206592082977295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,8,power_law_1.01,0.637446403503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,8,power_law_1.01,0.5707903861999511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,8,power_law_1.01,0.5770624160766602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,8,power_law_1.01,0.6049471855163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,8,power_law_1.01,0.5509439945220947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,8,power_law_1.01,0.5854400157928467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,8,power_law_1.01,0.5551680088043213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,8,power_law_1.01,0.5739776134490967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,8,power_law_1.01,0.5360191822052002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,8,power_law_1.01,0.6054848194122314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,8,power_law_1.01,0.6509312152862549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,8,power_law_1.01,0.6699520111083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,8,power_law_1.01,0.8374336242675782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,8,power_law_1.01,0.8
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,16,4,power_law_1.2,0.33973119258880613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,8,power_law_1.01,0.9392895698547363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,8,power_law_1.01,1.148755168914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,8,power_law_1.01,1.3681535720825195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,8,power_law_1.01,1.3496640205383301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,2,power_law_1.01,0.09887359738349914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,2,power_law_1.01,0.11299840211868287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,8,power_law_1.01,1.7540864944458008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,2,power_law_1.01,0.13164160251617432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,2,power_law_1.01,0.15289599895477296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,8,power_law_1.01,2.468563270568848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,2,power_law_1.01,0.1763584017753601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,2,power_law_1.01,0.1931712031364441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,2,power_law_1.01,0.2104192018508911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,2,power_law_1.01,0.223801589012146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,2,power_law_1.01,0.23288319110870362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,2,power_law_1.01,0.2357248067855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,2,power_law_1.01,0.24151039123535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,2,power_law_1.01,0.2461184024810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,2,power_law_1.01,0.25407359600067136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,2,power_law_1.01,0.26357760429382326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,2,power_law_1.01,0.2660032033920288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,2,power_law_1.01,0.27182080745697024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,2,power_law_1.01,0.28460800647735596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,2,power_law_1.01,0.3196671962738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,2,power_law_1.01,0.3571968078613281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,2,power_law_1.01,0.44456958770751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,2,power_law_1.01,0.5243391990661621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,16,4,power_law_1.2,0.36462080478668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,2,power_law_1.01,0.6623487949371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,2,power_law_1.01,0.8053248405456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,2,power_law_1.01,1.1353023529052735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,2,power_law_1.01,1.403071975708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,2,power_law_1.01,2.1388416290283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,2,power_law_1.01,2.6726591110229494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,16,16,power_law_1.2,0.04816640019416809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,16,16,power_law_1.2,0.062105602025985716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,16,16,power_law_1.2,0.059680002927780154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,16,4,power_law_1.2,0.5985983848571778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,16,16,power_law_1.2,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,16,16,power_law_1.2,0.05609599947929382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,16,16,power_law_1.2,0.0573311984539032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,16,16,power_law_1.2,0.057920002937316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,16,16,power_law_1.2,0.059001600742340087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,16,16,power_law_1.2,0.05877760052680969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,16,16,power_law_1.2,0.05944960117340088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,16,16,power_law_1.2,0.060249602794647215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,16,16,power_law_1.2,0.06472960114479065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,16,16,power_law_1.2,0.06727679967880248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,16,16,power_law_1.2,0.07045120000839233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,16,16,power_law_1.2,0.07875840067863464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,16,16,power_law_1.2,0.08263040184974671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,16,16,power_law_1.2,0.09265919923782348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,16,16,power_law_1.2,0.1173632025718689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,16,16,power_law_1.2,0.12183680534362792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,16,16,power_law_1.2,0.15013760328292847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,16,16,power_law_1.2,0.17593599557876588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,16,16,power_law_1.2,0.2357952117919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,16,16,power_law_1.2,0.29420158863067625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,16,16,power_law_1.2,0.41276159286499026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,16,16,power_law_1.2,0.5420159816741943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,16,16,power_law_1.2,0.7727231979370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,16,16,power_law_1.2,1.0231743812561036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,16,4,power_law_1.2,0.7373184204101563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,16,power_law_1.01,0.07285119891166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,16,power_law_1.01,0.05475199818611145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,16,power_law_1.01,0.05064319968223572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,16,power_law_1.01,0.05529599785804749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,16,power_law_1.01,0.050195199251174924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,16,power_law_1.01,0.05429120063781738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,16,power_law_1.01,0.056518399715423585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,16,power_law_1.01,0.053388798236846925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,16,power_law_1.01,0.054764801263809205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,16,power_law_1.01,0.05443840026855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,16,power_law_1.01,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,16,power_law_1.01,0.05674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,16,power_law_1.01,0.058208000659942624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,16,power_law_1.01,0.0603007972240448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,16,power_law_1.01,0.06575359702110291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,16,power_law_1.01,0.06598399877548218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,16,power_law_1.01,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,16,power_law_1.01,0.07149440050125122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,16,power_law_1.01,0.07726719975471497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,16,power_law_1.01,0.08772479891777038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,16,power_law_1.01,0.08910719752311706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,16,power_law_1.01,0.10691839456558228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,16,power_law_1.01,0.135481595993042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,16,power_law_1.01,0.16357760429382323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,16,power_law_1.01,0.19651199579238893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,16,power_law_1.01,0.26257920265197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,16,power_law_1.01,0.2998464107513428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,8,power_law_1.2,0.06435199975967407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,8,power_law_1.2,0.06760960221290588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,128,power_law_1.2,0.01852799952030182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,8,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,128,power_law_1.2,0.02083200067281723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,8,power_law_1.2,0.07740160226821899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,8,power_law_1.2,0.08143360018730164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,8,power_law_1.2,0.09150720238685608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,8,power_law_1.2,0.09199360013008118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,8,power_law_1.2,0.09484800100326538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,128,power_law_1.2,0.019411200284957887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,128,power_law_1.2,0.018982400000095368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,128,power_law_1.2,0.022752000391483305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,128,power_law_1.2,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,128,power_law_1.2,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,8,power_law_1.2,0.09395840167999267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,128,power_law_1.2,0.03294079899787903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,8,power_law_1.2,0.09668480157852173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,8,power_law_1.2,0.09662079811096191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,128,power_law_1.2,0.033011201024055484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,8,power_law_1.2,0.09611520171165466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,128,power_law_1.2,0.03345920145511627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,8,power_law_1.2,0.10156799554824829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,128,power_law_1.2,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,128,power_law_1.2,0.033024001121521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,128,power_law_1.2,0.032691198587417605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,128,power_law_1.2,0.03266560137271881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,128,power_law_1.2,0.03708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,128,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,128,power_law_1.2,0.03628160059452057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,128,power_law_1.2,0.03888640105724335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.2,0.042905598878860474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,8,power_law_1.2,0.10313600301742554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.2,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.2,0.06043519973754883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.2,0.0738752007484436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.2,0.08942080140113831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,8,power_law_1.2,0.10720640420913696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,8,power_law_1.2,0.11000319719314575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,8,power_law_1.2,0.11975680589675904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,8,power_law_1.2,0.1313279986381531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,8,power_law_1.2,0.14606720209121704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,8,power_law_1.2,0.17328640222549438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,8,power_law_1.2,0.22069759368896485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.2,0.12535040378570556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,8,power_law_1.2,0.31816959381103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,8,power_law_1.2,0.3602112054824829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,8,power_law_1.2,0.5644927978515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,8,power_law_1.2,0.702400016784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.2,0.1565824031829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,4,power_law_1.2,0.021542400121688843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,4,power_law_1.2,0.022752000391483305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,4,power_law_1.2,0.024691200256347655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,4,power_law_1.2,0.02911359965801239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,4,power_law_1.2,0.031206399202346802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,4,power_law_1.2,0.03457280099391937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,4,power_law_1.2,0.03754239976406097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,4,power_law_1.2,0.04033919870853424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,4,power_law_1.2,0.040428799390792844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,4,power_law_1.2,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,4,power_law_1.2,0.04500479996204376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.2,0.2235647916793823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,4,power_law_1.2,0.045491200685501096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,4,power_law_1.2,0.04711039960384369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,4,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,4,power_law_1.2,0.05861120223999024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,4,power_law_1.2,0.061945599317550656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,4,power_law_1.2,0.06935039758682252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,4,power_law_1.2,0.09639040231704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.2,0.12131839990615845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.2,0.1272320032119751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.2,0.14008959531784057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.2,0.16363519430160522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.2,0.18877439498901366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.2,0.23779840469360353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.2,0.300383996963501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.2,0.4132351875305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,8,power_law_1.2,1.008236789703369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.2,0.5243135929107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.2,0.290009593963623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,2,2,power_law_1.2,0.05978239774703979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,2,2,power_law_1.2,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,2,2,power_law_1.2,0.07877119779586791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,2,2,power_law_1.2,0.09684479832649232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,2,2,power_law_1.2,0.09736319780349731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,2,2,power_law_1.2,0.09810559749603272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,2,2,power_law_1.2,0.09844480156898498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,2,2,power_law_1.2,0.09926400184631348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,2,2,power_law_1.2,0.09868159890174866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,2,2,power_law_1.2,0.09987199902534485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,2,2,power_law_1.2,0.10091520547866821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,2,2,power_law_1.2,0.11909120082855225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,2,2,power_law_1.2,0.12139519453048705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,2,2,power_law_1.2,0.12634240388870238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,2,2,power_law_1.2,0.13415679931640626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,2,2,power_law_1.2,0.13420799970626832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,2,2,power_law_1.2,0.1500991940498352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,2,2,power_law_1.2,0.18205440044403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,64,balanced,0.04818666477998098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,64,balanced,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,64,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,64,balanced,0.052042668064435325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,64,balanced,0.05226666728655497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,64,balanced,0.06630399823188782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,64,balanced,0.06565866867701213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,64,balanced,0.06628266473611195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,2,2,power_law_1.2,0.2008960008621216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,64,balanced,0.07307200133800507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,64,balanced,0.07111466427644093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,64,balanced,0.07276799778143565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,2,2,power_law_1.2,0.2714495897293091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,8,power_law_1.2,1.6479488372802735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,64,balanced,0.08067733546098073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,64,balanced,0.0801386684179306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,64,balanced,0.07532266775767009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,2,2,power_law_1.2,0.33803520202636717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,64,balanced,0.08802133798599243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,64,balanced,0.09293867150942485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,64,balanced,0.09213866790135701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,64,balanced,0.10525332887967427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,64,balanced,0.10941333572069804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,64,balanced,0.12463466326395671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,64,balanced,0.13808533549308777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,64,balanced,0.1713386575380961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,64,balanced,0.19738133748372397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,2,2,power_law_1.2,0.4382336139678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,64,balanced,0.249071995417277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,64,balanced,0.29787200689315796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,64,balanced,0.400490681330363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,64,balanced,0.5209226608276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,2,2,power_law_1.2,0.5898496150970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,2,power_law_1.01,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,2,power_law_1.01,0.04750719964504242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,2,power_law_1.01,0.049779200553894044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,2,power_law_1.01,0.058387202024459836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,2,power_law_1.01,0.0680895984172821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,2,power_law_1.01,0.07708799839019775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,2,power_law_1.01,0.08878719806671143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,2,power_law_1.01,0.08824319839477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,8,power_law_1.2,0.15729279518127443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,8,power_law_1.2,0.17305599451065062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,8,power_law_1.2,0.24055678844451905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,8,power_law_1.2,0.26012799739837644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,8,power_law_1.2,0.22215039730072023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,2,power_law_1.01,0.08924800157546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,2,power_law_1.01,0.09106559753417968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,2,power_law_1.01,0.09395840167999267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,2,power_law_1.01,0.09442560076713562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,2,power_law_1.01,0.09771519899368286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,2,power_law_1.01,0.10124800205230713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,2,power_law_1.01,0.10743039846420288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,2,power_law_1.01,0.11020799875259399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,8,power_law_1.2,0.2854079961776733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,2,power_law_1.01,0.12154239416122437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,8,power_law_1.2,0.27861120700836184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,2,power_law_1.01,0.14513280391693115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,8,power_law_1.2,0.27733120918273924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,8,power_law_1.2,0.28318080902099607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,8,power_law_1.2,0.3061568021774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,8,power_law_1.2,0.3101311922073364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,8,power_law_1.2,0.31663999557495115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,2,power_law_1.01,0.16381440162658692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,2,power_law_1.01,0.18900480270385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,8,power_law_1.2,0.317305588722229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,8,power_law_1.2,0.334553599357605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,8,power_law_1.2,0.341811203956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,8,power_law_1.2,0.358188796043396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,8,power_law_1.2,0.3945024013519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,8,power_law_1.2,0.4608128070831299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,2,power_law_1.01,0.22058238983154296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,8,power_law_1.2,0.47175040245056155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,2,power_law_1.01,0.2757951974868774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,8,power_law_1.2,0.535865592956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,2,power_law_1.01,0.35556480884552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,2,power_law_1.01,0.48288640975952146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,2,2,power_law_1.2,1.0295552253723144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,2,power_law_1.01,0.6181183815002441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,8,power_law_1.2,0.6890431880950928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,8,power_law_1.2,0.8152447700500488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,8,power_law_1.2,0.8624320030212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,8,power_law_1.2,0.9913727760314941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,8,power_law_1.2,1.3540096282958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,2,power_law_1.01,0.8403327941894532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,2,power_law_1.01,1.117734432220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,8,power_law_1.2,1.878291130065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,8,power_law_1.2,2.3451839447021485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,16,power_law_1.2,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,2,2,power_law_1.2,1.2923199653625488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,16,power_law_1.2,0.03722879886627197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,16,power_law_1.2,0.040582400560379026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,16,power_law_1.2,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,16,power_law_1.2,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,16,power_law_1.2,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,16,power_law_1.2,0.05793280005455017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,16,power_law_1.2,0.05539199709892273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,16,power_law_1.2,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,16,power_law_1.2,0.08309760093688964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,16,power_law_1.2,0.08277119994163513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,16,power_law_1.2,0.08364160060882568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,16,power_law_1.2,0.0906112015247345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,16,power_law_1.2,0.08843520283699036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,16,power_law_1.2,0.0806335985660553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,16,power_law_1.2,0.08027520179748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,16,power_law_1.2,0.07515519857406616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,16,power_law_1.2,0.08145279884338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.2,0.09102079868316651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.2,0.1021888017654419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.2,0.1195904016494751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.2,0.14718719720840454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.2,0.18411519527435302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.2,0.2553024053573608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.2,0.3362368106842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.2,0.48051838874816893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.2,0.6140736103057861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,2,2,power_law_1.2,1.5581631660461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,2,2,power_law_1.2,2.6225791931152345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,4,balanced,0.04204266766707102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,4,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,4,balanced,0.04691733419895172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,4,balanced,0.054042667150497437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,4,balanced,0.06422933439413707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,4,balanced,0.07659199833869934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,4,balanced,0.0782773345708847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,4,balanced,0.07967466612656911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,4,balanced,0.07826666533946991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,4,balanced,0.08040000001589458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,4,balanced,0.07926933467388153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,4,balanced,0.08077866832415263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,4,balanced,0.08239999910195668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,4,balanced,0.08527466654777527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,4,balanced,0.08859200278917949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,4,balanced,0.09302933017412822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,4,balanced,0.09573866923650105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,4,balanced,0.10724266370137532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,4,balanced,0.11365866661071777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,4,balanced,0.1495786706606547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,4,balanced,0.1673706571261088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,4,balanced,0.23298132419586182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,4,balanced,0.2577173312505086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,4,balanced,0.37467201550801593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,4,balanced,0.46000532309214276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,4,balanced,0.6240959962209066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,4,balanced,0.7976693312327067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,64,power_law_1.01,0.058982402086257935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,64,power_law_1.01,0.055193597078323366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,64,power_law_1.01,0.05423359870910645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,64,power_law_1.01,0.05663359761238098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,64,power_law_1.01,0.05783680081367493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,64,power_law_1.01,0.06049280166625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,64,power_law_1.01,0.06062080264091492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,64,power_law_1.01,0.06474239826202392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,64,power_law_1.01,0.06147199869155884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,64,power_law_1.01,0.06505600214004517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,64,power_law_1.01,0.0682752013206482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,64,power_law_1.01,0.0692351996898651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,64,power_law_1.01,0.07036160230636597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,64,power_law_1.01,0.0743552029132843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,2,power_law_1.2,0.04418559968471527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,64,power_law_1.01,0.0896448016166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,2,power_law_1.2,0.046060800552368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,2,power_law_1.2,0.04755200147628784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,64,power_law_1.01,0.09230080246925354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,2,power_law_1.2,0.053299200534820554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,64,power_law_1.01,0.09129599928855896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,2,power_law_1.2,0.06120960116386413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,64,power_law_1.01,0.10393600463867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,2,power_law_1.2,0.06586880087852479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,64,power_law_1.01,0.11628799438476563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,2,power_law_1.2,0.07326080203056336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,2,power_law_1.2,0.0760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,2,power_law_1.2,0.0788096010684967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,2,power_law_1.2,0.08041599988937378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,2,power_law_1.2,0.08320000171661376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,2,power_law_1.2,0.08800640106201171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,2,power_law_1.2,0.0904640018939972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,2,power_law_1.2,0.09887359738349914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,2,power_law_1.2,0.10813440084457397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,64,power_law_1.01,0.13594239950180054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,2,power_law_1.2,0.1134719967842102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,64,power_law_1.01,0.15623040199279786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,2,power_law_1.2,0.11704959869384765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,2,power_law_1.2,0.1347839951515198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,64,power_law_1.01,0.20206079483032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,2,power_law_1.2,0.1491328001022339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,64,power_law_1.01,0.23990399837493898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,64,power_law_1.01,0.3190144062042236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,64,power_law_1.01,0.3985984086990356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,64,power_law_1.01,0.6022208213806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,2,power_law_1.2,0.18804479837417604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,64,power_law_1.01,0.8016703605651856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,32,power_law_1.2,0.05826560258865356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,32,power_law_1.2,0.05726720094680786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,2,power_law_1.2,0.22900478839874266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,32,power_law_1.2,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,2,power_law_1.2,0.2940671920776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,32,power_law_1.2,0.05681279897689819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,2,power_law_1.2,0.3670592069625854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,32,power_law_1.2,0.05550720095634461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,2,power_law_1.2,0.5132480144500733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,32,power_law_1.2,0.05553920269012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,32,power_law_1.2,0.056729602813720706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,32,power_law_1.2,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,32,power_law_1.2,0.05923839807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,32,power_law_1.2,0.05911679863929749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,32,power_law_1.2,0.05796480178833008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,32,power_law_1.2,0.060185599327087405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,32,power_law_1.2,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,2,power_law_1.2,0.6615808010101318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,2,power_law_1.2,0.9326720237731934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,2,power_law_1.2,1.2774463653564454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,32,power_law_1.2,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,32,power_law_1.2,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,32,power_law_1.2,0.0669376015663147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,32,power_law_1.2,0.070278400182724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,32,power_law_1.2,0.0778432011604309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,32,power_law_1.2,0.08289920091629029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,32,power_law_1.2,0.1013375997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,32,power_law_1.2,0.1134335994720459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,32,power_law_1.2,0.14448000192642213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,32,power_law_1.2,0.1728127956390381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,32,power_law_1.2,0.231712007522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,32,power_law_1.2,0.3025791883468628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,32,power_law_1.2,0.4355648040771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,32,power_law_1.2,0.5382400035858155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,1,power_law_1.2,0.11505919694900513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,1,power_law_1.2,0.15297919511795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,1,power_law_1.2,0.21668479442596436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,1,power_law_1.2,0.29497599601745605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,1,power_law_1.2,0.38360960483551027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,1,power_law_1.2,0.4946176052093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,1,power_law_1.2,0.6815552234649658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,1,power_law_1.2,0.6998847961425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,1,power_law_1.2,0.7177216053009033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,1,power_law_1.2,0.7303296089172363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,1,power_law_1.2,0.7666175842285157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,1,power_law_1.2,0.7705920219421387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,1,power_law_1.2,0.784000015258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,1,power_law_1.2,0.8031744003295899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,1,power_law_1.2,0.8293312072753907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,1,power_law_1.2,0.8472319602966308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,1,power_law_1.2,0.05241600275039673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,1,power_law_1.2,0.8835776329040528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,1,power_law_1.2,0.05631359815597534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,1,power_law_1.2,0.06181120276451111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,1,power_law_1.2,0.9707455635070801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,1,power_law_1.2,0.08547840118408204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,1,power_law_1.2,1.036128044128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,1,power_law_1.2,0.10659840106964111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,1,power_law_1.2,0.12584320306777955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,1,power_law_1.2,1.154355239868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,1,power_law_1.2,0.16145919561386107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,1,power_law_1.2,1.2826623916625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,1,power_law_1.2,0.16589440107345582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,1,power_law_1.2,0.17360639572143555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,1,power_law_1.2,0.178713595867157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,1,power_law_1.2,0.18840960264205933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,1,power_law_1.2,0.19950720071792602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,1,power_law_1.2,0.2126784086227417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,1,power_law_1.2,0.22581760883331298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,1,power_law_1.2,1.5321215629577636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,1,power_law_1.2,0.2445823907852173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,1,power_law_1.2,0.2600575923919678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,1,power_law_1.2,1.8060735702514648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,1,power_law_1.2,0.25591039657592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,1,power_law_1.2,0.30436480045318604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,1,power_law_1.2,2.2884864807128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,1,power_law_1.2,2.815519905090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,1,power_law_1.2,0.3423487901687622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,1,power_law_1.2,0.4318079948425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,1,power_law_1.2,3.812255859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,1,power_law_1.2,0.5065983772277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,1,power_law_1.2,0.6705728054046631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,1,power_law_1.2,0.8374464035034179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,1,power_law_1.2,4.805875015258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,1,power_law_1.2,1.1429759979248046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,1,power_law_1.2,1.4425536155700684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,1,power_law_1.2,2.0531328201293944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,1,power_law_1.2,2.6796159744262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,2,balanced,0.06414933502674103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,2,balanced,0.08541867136955261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,2,balanced,0.12199466427167256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,2,balanced,0.19869865973790488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,2,balanced,0.3315626581509908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,2,balanced,0.46003735065460205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,2,balanced,0.4628373384475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,2,balanced,0.46306665738423664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,2,balanced,0.46963731447855633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,2,balanced,0.46876800060272217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,2,balanced,0.47485331694285077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,2,balanced,0.4807093143463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,2,balanced,0.484442671140035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,2,balanced,0.49504534403483075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,2,balanced,0.5054346720377604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,2,balanced,0.5109279950459799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,2,balanced,0.5261600017547607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,2,balanced,0.5644533236821493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,2,balanced,0.592847983042399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,2,balanced,0.7911039988199869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,2,balanced,0.7345759868621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,2,balanced,0.9725813070933024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,2,balanced,1.0237173239390056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,2,balanced,1.5027893384297688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,2,balanced,1.6594187418619792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,2,balanced,2.62500794728597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,2,balanced,2.9676478703816733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,2,balanced,0.07026133437951405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,2,balanced,0.09563199679056804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,2,balanced,0.1471573313077291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,2,balanced,0.24658133586247763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,2,balanced,0.44676800568898517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,2,balanced,0.841818650563558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,2,balanced,1.2409439881642659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,2,balanced,1.2293226718902588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,2,balanced,1.229482650756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,2,balanced,1.2301759719848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,2,balanced,1.2397493521372478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,2,balanced,1.245802640914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,2,balanced,1.251967986424764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,2,balanced,1.2574773629506428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,2,balanced,1.2617653210957844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,2,balanced,1.2774933179219563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,2,balanced,1.291221300760905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,2,balanced,1.3126346270243328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,2,balanced,1.3403466542561848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,2,balanced,1.4005653063456218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,2,balanced,1.4572052955627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,2,balanced,1.5493812561035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,2,balanced,1.6380586624145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,2,balanced,1.8168212572733562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,2,balanced,2.1039786338806152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,2,balanced,2.42905600865682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,2,balanced,3.3278398513793945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,32,power_law_1.2,0.0434112012386322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,32,power_law_1.2,0.035385599732398985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,32,power_law_1.2,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,32,power_law_1.2,0.03511039912700653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,32,power_law_1.2,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,32,power_law_1.2,0.03592959940433502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,32,power_law_1.2,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,32,power_law_1.2,0.036800000071525577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,32,power_law_1.2,0.03697920143604279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,32,power_law_1.2,0.03761920034885406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,32,power_law_1.2,0.03738240003585815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,32,power_law_1.2,0.0389055997133255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,32,power_law_1.2,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,32,power_law_1.2,0.040454399585723874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,32,power_law_1.2,0.04441600143909454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,32,power_law_1.2,0.04707840085029602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,32,power_law_1.2,0.04902400076389313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,32,power_law_1.2,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,32,power_law_1.2,0.05857279896736145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,32,power_law_1.2,0.06778240203857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,32,power_law_1.2,0.07219840288162231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,32,power_law_1.2,0.08673279881477355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,32,power_law_1.2,0.10506880283355713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,32,power_law_1.2,0.14608000516891478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,32,power_law_1.2,0.1909824013710022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,32,power_law_1.2,0.3159424066543579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,32,power_law_1.2,0.3320255994796753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,32,2,balanced,0.021642667551835377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,32,2,balanced,0.02186666677395503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,32,2,balanced,0.023749334116776783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,32,2,balanced,0.028069332242012024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,32,2,balanced,0.03716800113519033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,32,2,balanced,0.044165333112080894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,32,2,balanced,0.05051200091838837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,32,2,balanced,0.05236266553401947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,32,2,balanced,0.05656533439954122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,32,2,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,32,2,balanced,0.06027733286221822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,32,2,balanced,0.06284800171852112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,32,2,balanced,0.06516266862551372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,32,2,balanced,0.06683200101057689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,32,2,balanced,0.07026133437951405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,32,2,balanced,0.07292800148328145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,32,2,balanced,0.07588266829649608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,32,2,balanced,0.08915199836095174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,32,2,balanced,0.10166399677594502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,32,2,balanced,0.14039466778437296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,32,2,balanced,0.15358933806419373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,32,2,balanced,0.2283253272374471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,32,2,balanced,0.26197866598765057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,32,2,balanced,0.30453334252039593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,32,2,balanced,0.34807467460632324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,32,2,balanced,0.5530826648076376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,32,2,balanced,0.650383989016215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,128,power_law_1.01,0.050886398553848265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,128,power_law_1.01,0.05122560262680054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,128,power_law_1.01,0.051283198595047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,128,power_law_1.01,0.052908802032470705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,8,8,power_law_1.2,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,128,power_law_1.01,0.05314559936523437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,8,8,power_law_1.2,0.0761023998260498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,8,8,power_law_1.2,0.07378559708595275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,8,8,power_law_1.2,0.08475520014762879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,8,8,power_law_1.2,0.08684800267219543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,128,power_law_1.01,0.053990399837493895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,8,8,power_law_1.2,0.08996480107307434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,128,power_law_1.01,0.055622398853302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,8,8,power_law_1.2,0.09822720289230347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,128,power_law_1.01,0.05434240102767944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,8,8,power_law_1.2,0.09397119879722596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,8,8,power_law_1.2,0.10017919540405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,8,8,power_law_1.2,0.09666560292243957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,8,8,power_law_1.2,0.09735040068626404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,8,8,power_law_1.2,0.0981760025024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,128,power_law_1.01,0.055718398094177245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,128,power_law_1.01,0.05493119955062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,128,power_law_1.01,0.05534719824790955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,128,power_law_1.01,0.05666559934616089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,128,power_law_1.01,0.057708799839019775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,128,power_law_1.01,0.05740159749984741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,128,power_law_1.01,0.061791998147964475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,8,8,power_law_1.2,0.10309760570526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,128,power_law_1.01,0.06180480122566223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,128,power_law_1.01,0.06389120221138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,128,power_law_1.01,0.07178879976272583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,128,power_law_1.01,0.07787520289421082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,128,power_law_1.01,0.09106559753417968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,8,8,power_law_1.2,0.10439679622650147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,128,power_law_1.01,0.10066560506820679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,8,8,power_law_1.2,0.11114879846572875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,128,power_law_1.01,0.12525440454483033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,8,8,power_law_1.2,0.11228159666061402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,128,power_law_1.01,0.15221760272979737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,8,8,power_law_1.2,0.11813119649887086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,128,power_law_1.01,0.20492799282073976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,8,8,power_law_1.2,0.13574399948120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,128,power_law_1.01,0.2502016067504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,8,8,power_law_1.2,0.14568320512771607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,128,power_law_1.01,0.35105280876159667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,8,8,power_law_1.2,0.17856639623641968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,128,power_law_1.01,0.4610176086425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,8,8,power_law_1.2,0.22298240661621094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,8,8,power_law_1.2,0.28780159950256345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,8,8,power_law_1.2,0.3612735986709595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,8,8,power_law_1.2,0.5482111930847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,8,8,power_law_1.2,0.6781184196472168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,8,8,power_law_1.2,1.0360447883605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,8,8,power_law_1.2,1.4153023719787599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,4,power_law_1.2,0.028403198719024657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,4,power_law_1.2,0.029651200771331786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,4,power_law_1.2,0.02945919930934906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,4,power_law_1.2,0.030623999238014222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,4,power_law_1.2,0.03036159873008728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,4,power_law_1.2,0.03201279938220978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,4,power_law_1.2,0.032691198587417605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,4,power_law_1.2,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,4,power_law_1.2,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,4,power_law_1.2,0.03443840146064758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,4,power_law_1.2,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,4,power_law_1.2,0.036556801199913024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,4,power_law_1.2,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,4,power_law_1.2,0.039078399538993835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,4,power_law_1.2,0.046675199270248414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,4,power_law_1.2,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,4,power_law_1.2,0.05003520250320435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,4,power_law_1.2,0.07203199863433837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,4,power_law_1.2,0.08271999955177307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,4,power_law_1.2,0.058880001306533813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,4,power_law_1.2,0.08892800211906433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,4,power_law_1.2,0.10120960474014282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,4,power_law_1.2,0.10680960416793824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,4,power_law_1.2,0.06432639956474304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,4,power_law_1.2,0.11174399852752685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,4,power_law_1.2,0.12607359886169434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,4,power_law_1.2,0.1288831949234009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,4,power_law_1.2,0.12984960079193114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,4,power_law_1.2,0.13063679933547973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,4,power_law_1.2,0.12967679500579835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,4,power_law_1.2,0.1363136053085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,4,power_law_1.2,0.1403839945793152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,4,power_law_1.2,0.07731199860572815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,4,power_law_1.2,0.1436095952987671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,4,power_law_1.2,0.14941439628601075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,4,power_law_1.2,0.15407999753952026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,4,power_law_1.2,0.16343679428100585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,4,power_law_1.2,0.18101119995117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,4,power_law_1.2,0.20523519515991212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,4,power_law_1.2,0.25471999645233157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,4,power_law_1.2,0.31110401153564454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,4,power_law_1.2,0.4021120071411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,4,power_law_1.2,0.49237761497497556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,4,power_law_1.2,0.09747200012207032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,4,power_law_1.2,0.7566527843475341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,4,power_law_1.2,0.9480959892272949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,4,power_law_1.2,0.12710399627685548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,4,power_law_1.2,1.2981504440307616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,4,power_law_1.2,1.9693376541137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,4,power_law_1.2,0.1591871976852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,4,power_law_1.01,0.020985600352287293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,4,power_law_1.01,0.020428800582885744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,4,power_law_1.01,0.020172800123691558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,4,power_law_1.01,0.02035840004682541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,4,power_law_1.01,0.020147199928760528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,4,power_law_1.01,0.02064639925956726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,4,power_law_1.01,0.02452480047941208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,4,power_law_1.01,0.02369920015335083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,4,power_law_1.01,0.04880639910697937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,4,power_law_1.01,0.04796159863471985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,4,power_law_1.01,0.049344000220298764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,4,power_law_1.01,0.04170239865779877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,4,power_law_1.2,0.2135551929473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,4,power_law_1.01,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,4,power_law_1.01,0.03686400055885315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,4,power_law_1.01,0.04357120096683502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,4,power_law_1.01,0.03880319893360138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,4,power_law_1.01,0.04254719913005829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,4,power_law_1.01,0.05485439896583557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.01,0.0599295973777771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.01,0.08268160223960877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.01,0.10195839405059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.01,0.13642239570617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,4,power_law_1.2,0.2988352060317993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.01,0.1646016001701355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.01,0.23647360801696776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,32,power_law_1.2,0.0514240026473999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,32,power_law_1.2,0.04835839867591858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,32,power_law_1.2,0.047116801142692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,32,power_law_1.2,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,32,power_law_1.2,0.05108479857444763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,32,power_law_1.2,0.052902400493621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,32,power_law_1.2,0.0535103976726532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,32,power_law_1.2,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,32,power_law_1.2,0.05371519923210144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,32,power_law_1.2,0.05537279844284058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,32,power_law_1.2,0.05562880039215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,32,power_law_1.2,0.05947520136833191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,32,power_law_1.2,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.01,0.3008320093154907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,32,power_law_1.2,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,32,power_law_1.2,0.06378239989280701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,32,power_law_1.2,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,32,power_law_1.2,0.07289599776268005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,32,power_law_1.2,0.08243839740753174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,32,power_law_1.2,0.09008640050888062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,32,power_law_1.2,0.11136640310287475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,32,power_law_1.2,0.13230719566345214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,32,power_law_1.2,0.16632319688796998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,32,power_law_1.2,0.22410879135131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,32,power_law_1.2,0.3053056001663208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,32,power_law_1.2,0.3726464033126831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,32,power_law_1.2,0.5626239776611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,32,power_law_1.2,0.7939263820648194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,4,power_law_1.2,0.3918272018432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,64,power_law_1.01,0.08010240197181702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,64,power_law_1.01,0.0765824019908905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,64,power_law_1.01,0.0754047989845276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.01,0.43639039993286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,64,power_law_1.01,0.07946879863739013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,64,power_law_1.01,0.07871360182762147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,64,power_law_1.01,0.07628160119056701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,64,power_law_1.01,0.07909119725227357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,64,power_law_1.01,0.0807039976119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,64,power_law_1.01,0.0804095983505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,64,power_law_1.01,0.08398079872131348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,64,power_law_1.01,0.0839743971824646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,64,power_law_1.01,0.08588799834251404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,64,power_law_1.01,0.08486400246620178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,64,power_law_1.01,0.08843520283699036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,64,power_law_1.01,0.09659519791603088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,64,power_law_1.01,0.09459199905395507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,64,power_law_1.01,0.10543999671936036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,64,power_law_1.01,0.11210240125656128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,64,power_law_1.01,0.12250880002975464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,64,power_law_1.01,0.15242880582809448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,64,power_law_1.01,0.1820736050605774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,64,power_law_1.01,0.23228800296783447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,64,power_law_1.01,0.28132479190826415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,64,power_law_1.01,0.39166080951690674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,64,power_law_1.01,0.5000959873199463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,64,power_law_1.01,0.6942527770996094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,64,power_law_1.01,0.921343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.01,0.5730559825897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,4,power_law_1.2,0.5535871982574463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,4,8,power_law_1.01,0.04227840006351471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,4,8,power_law_1.01,0.049344000220298764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,4,8,power_law_1.01,0.04810880124568939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,4,8,power_law_1.01,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,4,8,power_law_1.01,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,4,8,power_law_1.01,0.05389440059661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,4,8,power_law_1.01,0.05539839863777161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,4,8,power_law_1.01,0.047891199588775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,4,8,power_law_1.01,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,4,8,power_law_1.01,0.049107199907302855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,4,8,power_law_1.01,0.05084800124168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,4,8,power_law_1.01,0.05279359817504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,4,8,power_law_1.01,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,4,8,power_law_1.01,0.05943040251731872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,4,8,power_law_1.01,0.0679744005203247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,4,8,power_law_1.01,0.0741375982761383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,4,8,power_law_1.01,0.07212799787521362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,4,8,power_law_1.01,0.08642560243606567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,4,8,power_law_1.01,0.1086527943611145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,4,8,power_law_1.01,0.12510080337524415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,4,8,power_law_1.01,0.15164159536361693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,4,8,power_law_1.01,0.20606720447540283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,4,8,power_law_1.01,0.2536448001861572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,4,8,power_law_1.01,0.3187583923339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,1,1,power_law_1.01,0.08702719807624817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,1,1,power_law_1.01,0.12357120513916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,1,1,power_law_1.01,0.14882559776306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,1,1,power_law_1.01,0.23931519985198973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,1,1,power_law_1.01,0.2377471923828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,1,1,power_law_1.01,0.24712319374084474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,1,1,power_law_1.01,0.24725120067596434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,1,1,power_law_1.01,0.25340800285339354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,1,1,power_law_1.01,0.2579583883285522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,1,1,power_law_1.01,0.2569407939910889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,1,1,power_law_1.01,0.2598720073699951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,1,1,power_law_1.01,0.2766207933425903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,1,1,power_law_1.01,0.28614399433135984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,1,1,power_law_1.01,0.29806718826293943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,1,1,power_law_1.01,0.3214656114578247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,1,1,power_law_1.01,0.3394752025604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,1,1,power_law_1.01,0.3454848051071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,1,1,power_law_1.01,0.40250239372253416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,1,1,power_law_1.01,0.46679039001464845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,4,8,power_law_1.01,0.48419837951660155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,1,1,power_law_1.01,0.5743552207946777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,1,1,power_law_1.01,0.6968704223632812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,1,1,power_law_1.01,0.9457216262817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,1,1,power_law_1.01,1.1769984245300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,4,8,power_law_1.01,0.6627967834472657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,1,1,power_law_1.01,1.68035831451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,1,1,power_law_1.01,2.1968704223632813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,32,power_law_1.2,0.02834559977054596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,32,power_law_1.2,0.030297601222991945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,32,power_law_1.2,0.02683520019054413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,32,power_law_1.2,0.030976000428199767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,32,power_law_1.2,0.0297791987657547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,32,power_law_1.2,0.029721599817276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,32,power_law_1.2,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,32,power_law_1.2,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,32,power_law_1.2,0.04074240028858185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,32,power_law_1.2,0.03945600092411041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,32,power_law_1.2,0.04357759952545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,32,power_law_1.2,0.04803839921951294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,32,power_law_1.2,0.055308800935745236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,32,power_law_1.2,0.08064000010490417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,32,power_law_1.2,0.08144000172615051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,32,power_law_1.2,0.08554880023002624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,4,8,power_law_1.01,0.8579456329345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,32,power_law_1.2,0.09273599982261657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,32,power_law_1.2,0.10716160535812377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.2,0.12298879623413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.2,0.1574399948120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.2,0.10469119548797608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.2,0.1441215991973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.2,0.1742784023284912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.2,0.23605759143829347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.2,0.30767359733581545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.2,0.45168638229370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.2,0.5990719795227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,1,1,power_law_1.01,3.153152084350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,2,power_law_1.2,0.11624319553375244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,2,power_law_1.2,0.1329408049583435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,2,power_law_1.2,0.16729600429534913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,2,power_law_1.2,0.1668352007865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,2,power_law_1.2,0.19100799560546874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,2,power_law_1.2,0.18944640159606935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,2,power_law_1.2,0.18532480001449586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,2,power_law_1.2,0.18451199531555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,2,power_law_1.2,0.18210560083389282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,2,power_law_1.2,0.1849727988243103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,2,power_law_1.2,0.17943040132522584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,2,power_law_1.2,0.2093503952026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,2,power_law_1.2,0.1815808057785034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,2,power_law_1.2,0.19322880506515502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,2,power_law_1.2,0.22716801166534423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,2,power_law_1.2,0.2375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,2,power_law_1.2,0.2511552095413208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,2,power_law_1.2,0.34150400161743166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,1,1,power_law_1.01,4.160787200927734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.2,0.3707583904266357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.2,0.4395135879516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,2,2,power_law_1.2,0.10384000539779663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.2,0.5485055923461915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,2,2,power_law_1.2,0.12517759799957276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,2,2,power_law_1.2,0.16865919828414916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,2,2,power_law_1.2,0.19639040231704713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,2,2,power_law_1.2,0.23180160522460938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,2,2,power_law_1.2,0.34371840953826904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,2,2,power_law_1.2,0.3584831953048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,2,2,power_law_1.2,0.3643199920654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.2,0.8416383743286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,2,2,power_law_1.2,0.3535871982574463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,2,2,power_law_1.2,0.3863039970397949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,2,2,power_law_1.2,0.3806080102920532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,2,2,power_law_1.2,0.3903615951538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,2,2,power_law_1.2,0.38796799182891845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,2,2,power_law_1.2,0.39750399589538576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,8,1,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,2,2,power_law_1.2,0.4015679836273193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,8,1,balanced,0.060090666015942894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,8,1,balanced,0.08819199601809184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,8,1,balanced,0.09222933650016785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.2,0.9518591880798339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,8,1,balanced,0.09309333562850952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,8,1,balanced,0.09372799595197041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,8,1,balanced,0.09589333335558574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,8,1,balanced,0.09699199597040813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,8,1,balanced,0.09611200292905171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,8,1,balanced,0.09700799981753032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,8,1,balanced,0.09771733482678731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,8,1,balanced,0.10121599833170573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,8,1,balanced,0.10170132915178935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,8,1,balanced,0.10388267040252686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,8,1,balanced,0.11063999931017558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,8,1,balanced,0.11504000425338745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,2,2,power_law_1.2,0.4261184215545654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,2,2,power_law_1.2,0.42335357666015627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,2,2,power_law_1.2,0.45706877708435056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.2,0.5117568016052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.2,0.5628096103668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,8,1,balanced,0.11527466773986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.2,0.6681151866912842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.2,0.8516672134399415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,8,balanced,0.052527998884518944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,8,balanced,0.05450133482615153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,8,balanced,0.06247466802597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,8,balanced,0.0688213308652242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.2,0.9598079681396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,8,balanced,0.09297600388526917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,8,balanced,0.1360160013039907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,8,balanced,0.1357973317305247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,8,balanced,0.1267519990603129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,8,balanced,0.13355200489362082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,8,balanced,0.12777066230773926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,8,balanced,0.12437867124875386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,8,balanced,0.1276800036430359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,8,balanced,0.12921067078908285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,8,balanced,0.12657066186269125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,8,balanced,0.13339199622472128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,8,balanced,0.13356266419092813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,8,balanced,0.13691733280817667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,8,balanced,0.14350933829943338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,8,balanced,0.15198399623235068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,8,balanced,0.17009600003560385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,8,balanced,0.1837973395983378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.2,1.329535961151123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,8,balanced,0.21925334135691324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,8,1,balanced,0.1497173309326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,8,balanced,0.251093327999115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,8,1,balanced,0.16058666507403055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,8,balanced,0.3601280053456624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,8,balanced,0.42626134554545086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.2,1.681817626953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,8,balanced,0.6007680098215739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,8,balanced,0.7664960225423177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,8,1,balanced,0.21523733933766684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,8,1,balanced,0.26708267132441205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,8,1,balanced,0.37830932935078937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,8,1,balanced,0.47882131735483807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,8,1,balanced,0.6781919797261556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,8,1,balanced,0.883466641108195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.2,2.2548608779907227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,8,1,balanced,1.2776532967885335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,8,1,balanced,1.6774826049804688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.2,1.3079680442810058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.2,2.9402816772460936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.2,1.6877824783325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.2,2.6978687286376952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.2,4.087539291381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,4,balanced,0.04228266576925913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,4,balanced,0.04154666761557261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,4,balanced,0.05589333176612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,4,balanced,0.06025599936644236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,4,balanced,0.06234133243560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,4,balanced,0.062208001812299095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,4,balanced,0.0625546673933665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,4,balanced,0.06437866886456807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,4,balanced,0.06320000191529591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,4,balanced,0.06433066725730896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,4,balanced,0.06409599880377452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,128,power_law_1.2,0.0302592009305954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,128,power_law_1.2,0.03043839931488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,128,power_law_1.2,0.030924800038337707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,128,power_law_1.2,0.03281280100345611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,128,power_law_1.2,0.03409920036792755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,4,balanced,0.07421333094437917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,4,balanced,0.07434666653474171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,4,balanced,0.07459199925263722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,128,power_law_1.2,0.03588480055332184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,4,balanced,0.07753600180149078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,4,balanced,0.07894933223724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,4,balanced,0.08683733145395915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,4,balanced,0.09359467029571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,4,balanced,0.0997973382472992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,4,balanced,0.13557333747545877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,4,balanced,0.14691733320554098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,4,balanced,0.2034346659978231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,4,balanced,0.25040533145268756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,4,balanced,0.35250667730967206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,4,balanced,0.450602650642395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,4,balanced,0.6616853475570679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,4,balanced,0.8726399739583334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,128,power_law_1.2,0.035776001214981076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,128,power_law_1.2,0.03653759956359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,128,power_law_1.2,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,128,power_law_1.2,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,128,power_law_1.2,0.041631999611854556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,128,power_law_1.2,0.04381439983844757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,128,power_law_1.2,0.04095999896526337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,128,power_law_1.2,0.045228800177574156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,128,power_law_1.2,0.05203199982643127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,128,power_law_1.2,0.05373439788818359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,128,power_law_1.2,0.06025599837303162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,128,power_law_1.2,0.06757760047912598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,128,power_law_1.2,0.08076159954071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,128,power_law_1.2,0.11291520595550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,128,power_law_1.2,0.14146560430526733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,128,power_law_1.2,0.20181760787963868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,128,power_law_1.2,0.25621120929718016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,128,power_law_1.2,0.43182082176208497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,2,power_law_1.2,0.13032959699630736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,128,power_law_1.2,0.5785408020019531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,2,power_law_1.2,0.14801280498504638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,2,power_law_1.2,0.1998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,2,power_law_1.2,0.26122241020202636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,2,power_law_1.2,0.3217983961105347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,2,power_law_1.2,0.49996161460876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,2,power_law_1.2,0.5211967945098877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,2,power_law_1.2,0.5327167987823487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,2,power_law_1.2,0.5434239864349365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,2,power_law_1.2,0.5464704036712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,2,power_law_1.2,0.575712013244629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,2,power_law_1.2,0.5900095939636231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,2,power_law_1.2,0.6084928035736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,2,power_law_1.2,0.6419072151184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,2,power_law_1.2,0.6564288139343262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,2,power_law_1.2,0.6774464130401612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,2,power_law_1.2,0.7045248031616211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,2,power_law_1.2,0.7884543895721435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,2,power_law_1.2,0.8752703666687012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,2,power_law_1.2,1.0646976470947265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,2,power_law_1.2,1.1895487785339356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,2,power_law_1.2,1.4417344093322755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,2,power_law_1.2,1.82238712310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,2,power_law_1.2,2.3189056396484373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,2,power_law_1.2,2.9761600494384766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,2,power_law_1.2,4.058361434936524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,2,power_law_1.2,5.228179168701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,128,power_law_1.2,1.0032511711120606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,16,balanced,0.05187733471393585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,16,balanced,0.07457600037256877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,16,balanced,0.10040533542633057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,16,balanced,0.15589333573977152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,16,balanced,0.2217386762301127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,16,balanced,0.2561013301213582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,16,balanced,0.2558880050977071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,16,balanced,0.25616000096003216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,16,balanced,0.2593066692352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,16,balanced,0.2560693422953288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,16,balanced,0.25889066855112713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,16,balanced,0.2622186740239461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,16,balanced,0.26261333624521893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,16,balanced,0.2646080056826274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,16,balanced,0.26945600907007855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,16,balanced,0.2714933355649312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,16,balanced,0.2770559986432393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,16,balanced,0.3069919943809509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,16,balanced,0.3025760054588318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,16,balanced,0.3789920012156169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,16,balanced,0.35693331559499103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,16,balanced,0.5667413473129272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,16,balanced,0.42736534277598065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,16,balanced,0.6690613428751627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,16,balanced,0.6263146797815958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,16,balanced,0.9618399937947592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,16,balanced,0.9715627034505209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,128,power_law_1.2,1.294713592529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,16,power_law_1.2,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,16,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,16,power_law_1.2,0.03640959858894348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,16,power_law_1.2,0.036051198840141296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,16,power_law_1.2,0.036748799681663516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,16,power_law_1.2,0.03765760064125061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,16,power_law_1.2,0.03877759873867035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,16,power_law_1.2,0.040191999077796935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,16,power_law_1.2,0.04196479916572571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,16,power_law_1.2,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,16,power_law_1.2,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,16,power_law_1.2,0.049004799127578734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,16,power_law_1.2,0.05160319805145264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,16,power_law_1.2,0.050457602739334105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,16,power_law_1.2,0.05854079723358154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,16,power_law_1.2,0.0632960021495819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,16,power_law_1.2,0.066048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,16,power_law_1.2,0.07405440211296081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,16,power_law_1.2,0.09143679738044738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,16,power_law_1.2,0.10957440137863159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,16,power_law_1.2,0.1302847981452942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,16,power_law_1.2,0.16506240367889405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,8,balanced,0.03350933392842611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,8,balanced,0.033402666449546814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,8,balanced,0.033615998923778534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,8,balanced,0.03583466758330663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,8,balanced,0.03941866755485535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,8,balanced,0.03967999915281931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,8,balanced,0.039861333866914116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,8,balanced,0.041573333243529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,8,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,8,balanced,0.04015466570854187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,8,balanced,0.03977066775163015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,8,balanced,0.04177066683769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,8,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,8,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,8,balanced,0.04860266546408335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,8,balanced,0.04997866849104563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,8,balanced,0.0498986691236496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,16,power_law_1.2,0.21123199462890624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,16,power_law_1.2,0.2747391939163208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,8,balanced,0.05176533261934916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,8,balanced,0.05813866853713989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,8,balanced,0.06442666550477345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,8,balanced,0.07254933317502339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,8,balanced,0.09491200248400371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,8,balanced,0.10730133454004924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,8,balanced,0.1411786675453186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,8,balanced,0.1718239982922872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,8,balanced,0.2390186587969462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,8,balanced,0.3188053369522095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,16,power_law_1.2,0.3865279912948608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,8,balanced,0.029674666623274486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,8,balanced,0.030053332448005676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,8,balanced,0.029482667644818623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,16,power_law_1.2,0.5691520214080811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,8,balanced,0.032773333291212715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,8,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,8,balanced,0.0373333344856898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,8,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,8,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,8,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,8,balanced,0.03591466695070267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,8,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,8,balanced,0.03734400123357773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,8,balanced,0.038160001238187156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,8,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,8,balanced,0.04595733185609182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,8,balanced,0.04377600053946177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,8,balanced,0.04590400060017904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,8,balanced,0.04784533381462097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,8,balanced,0.0540533314148585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,8,balanced,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,8,balanced,0.06650133430957794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,8,balanced,0.08719467123349507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,8,balanced,0.10693333546320598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,8,balanced,0.14108799894650778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,8,balanced,0.17764800786972046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,8,balanced,0.24425599972407022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,8,balanced,0.3012373248736064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,8,power_law_1.01,0.05059199929237366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,8,power_law_1.01,0.07249919772148132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,8,power_law_1.01,0.06921600103378296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,8,power_law_1.01,0.08889600038528442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,8,power_law_1.01,0.07900159955024719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,8,power_law_1.01,0.09499520063400269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,8,power_law_1.01,0.10683519840240478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,8,power_law_1.01,0.10363520383834839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,8,power_law_1.01,0.09987840056419373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,8,power_law_1.01,0.13722879886627198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,8,power_law_1.01,0.13431040048599244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,8,power_law_1.01,0.12698240280151368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,8,power_law_1.01,0.19088000059127808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,8,power_law_1.01,0.1991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,8,power_law_1.01,0.19811840057373048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,8,power_law_1.01,0.21063039302825928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,8,power_law_1.01,0.20193281173706054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,8,power_law_1.01,0.1919360041618347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.01,0.20709118843078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.01,0.20332798957824708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.01,0.21750400066375733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,16,power_law_1.2,0.5976319789886475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.01,0.2784895896911621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.01,0.3345279932022095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.01,0.4556159973144531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.01,0.6041535854339599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.01,0.8389120101928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.01,1.151910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,32,balanced,0.059605335195859276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,32,balanced,0.06028266747792562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,32,balanced,0.060229331254959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,32,balanced,0.06801066795984904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,32,balanced,0.07739733159542084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,32,balanced,0.09232532978057861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,32,balanced,0.10455999771753947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,32,balanced,0.10520000259081523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,32,balanced,0.1074720025062561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,32,balanced,0.105295995871226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,32,balanced,0.10750400026639302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,32,balanced,0.10966933767000835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,32,balanced,0.11061333616574605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,32,balanced,0.11335999766985576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,32,balanced,0.12224533160527547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,32,balanced,0.12299199899037679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,32,balanced,0.12980266412099203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,32,balanced,0.1368000010649363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,32,balanced,0.1448319951693217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,32,balanced,0.1622933348019918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,32,balanced,0.18006932735443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,32,balanced,0.21845867236455283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,32,balanced,0.25244800249735516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,32,balanced,0.3167733351389567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,32,balanced,0.40753599007924396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,32,balanced,0.5475200017293295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,32,balanced,0.7128213246663412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,32,power_law_1.01,0.092985600233078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,32,power_law_1.01,0.08906880021095276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,32,power_law_1.01,0.08577280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,32,power_law_1.01,0.0886847972869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,32,power_law_1.01,0.08477439880371093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,32,power_law_1.01,0.08190720081329346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,32,power_law_1.01,0.0822272002696991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,32,power_law_1.01,0.08638079762458802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,32,power_law_1.01,0.08131840229034423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,32,power_law_1.01,0.08518400192260742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,32,power_law_1.01,0.08727040290832519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,32,power_law_1.01,0.08695679903030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,32,power_law_1.01,0.0881600022315979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,32,power_law_1.01,0.08686720132827759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,32,power_law_1.01,0.09267200231552124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,2,power_law_1.01,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,32,power_law_1.01,0.09427199959754944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,2,power_law_1.01,0.058822399377822875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,32,power_law_1.01,0.09653120040893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,32,power_law_1.01,0.10729600191116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,32,power_law_1.01,0.11940480470657348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,32,power_law_1.01,0.13612799644470214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,32,power_law_1.01,0.15265920162200927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,32,power_law_1.01,0.1966271996498108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,2,power_law_1.01,0.07018240094184876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,2,power_law_1.01,0.0861631989479065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,2,power_law_1.01,0.10255359411239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,2,power_law_1.01,0.12519680261611937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,2,power_law_1.01,0.14606080055236817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,32,power_law_1.01,0.22907519340515137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,2,power_law_1.01,0.15440640449523926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,32,power_law_1.01,0.31411840915679934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,2,power_law_1.01,0.16032639741897584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,2,power_law_1.01,0.1585088014602661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,2,power_law_1.01,0.16153600215911865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,2,power_law_1.01,0.164409601688385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,2,power_law_1.01,0.1701632022857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,2,power_law_1.01,0.17344640493392943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,2,power_law_1.01,0.18577280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,2,power_law_1.01,0.18984320163726806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,32,power_law_1.01,0.39830400943756106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,2,power_law_1.01,0.19560320377349855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,32,power_law_1.01,0.5201344013214111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,32,power_law_1.01,0.7142079830169678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,2,power_law_1.01,0.21648640632629396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,2,power_law_1.01,0.23093760013580322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,2,power_law_1.01,0.2624128103256226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,2,power_law_1.01,0.2954240083694458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,2,power_law_1.01,0.3659840106964111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,2,power_law_1.01,0.43727359771728513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,2,power_law_1.01,0.5572800159454345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,2,power_law_1.01,0.6886975765228271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,2,power_law_1.01,1.0002304077148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,2,power_law_1.01,1.1916095733642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,16,power_law_1.01,0.018662400543689728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,16,power_law_1.01,0.018297599256038667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,16,power_law_1.01,0.018329599499702455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,16,power_law_1.01,0.018303999304771425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,16,power_law_1.01,0.020479999482631683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,16,power_law_1.01,0.02200320065021515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,16,power_law_1.01,0.024883200228214265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,16,power_law_1.01,0.024953599274158477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,16,power_law_1.01,0.035123199224472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,16,power_law_1.01,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,16,power_law_1.01,0.03500800132751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,16,power_law_1.01,0.03802880048751831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,16,power_law_1.01,0.03708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,16,power_law_1.01,0.037215998768806456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,16,power_law_1.01,0.036883199214935304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,16,power_law_1.01,0.03720319867134094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,16,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,16,power_law_1.01,0.04035199880599975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,16,power_law_1.01,0.04190720021724701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,16,power_law_1.01,0.047737601399421695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,16,power_law_1.01,0.05102720260620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,16,power_law_1.01,0.06372479796409607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,16,power_law_1.01,0.07793920040130616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,16,power_law_1.01,0.10574079751968384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,16,power_law_1.01,0.12578560113906861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,16,power_law_1.01,0.18485759496688842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,16,power_law_1.01,0.22965118885040284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,2,16,power_law_1.2,0.04170880019664765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,2,16,power_law_1.2,0.03925119936466217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,2,16,power_law_1.2,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,2,16,power_law_1.2,0.03959679901599884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,2,16,power_law_1.2,0.0398719996213913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,2,16,power_law_1.2,0.0418368011713028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,2,16,power_law_1.2,0.04264959990978241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,2,16,power_law_1.2,0.04245760142803192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,2,16,power_law_1.2,0.042387199401855466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,2,16,power_law_1.2,0.0425024002790451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,2,16,power_law_1.2,0.0434112012386322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,2,16,power_law_1.2,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,2,16,power_law_1.2,0.04398080110549927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,2,16,power_law_1.2,0.04591360092163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,2,16,power_law_1.2,0.048793599009513855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,2,16,power_law_1.2,0.05018240213394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,2,16,power_law_1.2,0.052857601642608644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,2,16,power_law_1.2,0.058143997192382814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,2,16,power_law_1.2,0.06273919939994813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,2,16,power_law_1.2,0.06845440268516541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,2,16,power_law_1.2,0.08193280100822449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,2,16,power_law_1.2,0.10442880392074586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,2,16,power_law_1.2,0.1189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,2,16,power_law_1.2,0.1626688003540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,2,16,power_law_1.2,0.19676159620285033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,1,2,balanced,0.03577066709597906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,1,2,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,1,2,balanced,0.06278933087984721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,1,2,balanced,0.10345066587130229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,1,2,balanced,0.16491732994715372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,1,2,balanced,0.21547200282414755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,1,2,balanced,0.214026669661204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,1,2,balanced,0.21449599663416544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,1,2,balanced,0.21586666504542032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,1,2,balanced,0.217631995677948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,1,2,balanced,0.21760533253351846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,1,2,balanced,0.21994133790334067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,1,2,balanced,0.21970132986704508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,1,2,balanced,0.22182399034500122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,1,2,balanced,0.22881599267323813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,1,2,balanced,0.22779200474421182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,1,2,balanced,0.23286932706832886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,1,2,balanced,0.255237340927124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,1,2,balanced,0.24849599599838257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,1,2,balanced,0.30161066850026447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,1,2,balanced,0.2790186603864034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,1,2,balanced,0.4381493330001831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,1,2,balanced,0.35334400335947674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,1,2,balanced,0.5471413135528564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,1,2,balanced,0.5410079956054688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,2,16,power_law_1.2,0.3010175943374634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,1,2,balanced,0.8533439636230469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,1,2,balanced,0.809829314549764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,2,16,power_law_1.2,0.37649919986724856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,1,balanced,0.187226672967275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,1,balanced,0.3213546673456828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,1,balanced,0.5961013237635294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,1,balanced,1.1329546769460042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,1,balanced,2.178640047709147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,1,balanced,3.2021280924479165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,1,balanced,3.217850685119629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,1,balanced,3.2173598607381186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,1,balanced,3.234805425008138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,1,balanced,3.245973269144694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,1,balanced,3.2684265772501626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,1,balanced,3.2747252782185874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,2,balanced,0.04837333162625631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,2,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,1,balanced,3.2936054865519204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,2,balanced,0.07226133346557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,1,balanced,3.3175414403279624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,2,balanced,0.10626666744550069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,2,balanced,0.1593226691087087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,2,balanced,0.2592159907023112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,1,balanced,3.3543574015299478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,2,balanced,0.2608906626701355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,2,balanced,0.2629759907722473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,1,balanced,3.3770081202189126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,2,balanced,0.2639146645863851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,2,balanced,0.2695573369661967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,2,balanced,0.2735893328984578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,2,balanced,0.27672000726064044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,2,balanced,0.27982934315999347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,2,balanced,0.28942400217056274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,2,balanced,0.29611732562383014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,2,balanced,0.30266666412353516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,2,balanced,0.31363733609517414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,2,balanced,0.3442666530609131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,2,balanced,0.36513598759969074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,2,balanced,0.4127039909362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,2,balanced,0.4604853391647339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,1,balanced,3.4386399586995444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,2,balanced,0.5932426850001017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,2,balanced,0.6940426826477051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,2,balanced,0.9240960280100504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,1,balanced,3.741685231526693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,2,balanced,1.1241172949473064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,2,balanced,1.6305120786031086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,1,balanced,3.6043306986490884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,2,balanced,2.0439039866129556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,1,balanced,4.46994145711263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,1,balanced,3.93501345316569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,1,balanced,5.854447682698567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,1,balanced,4.70468807220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,1,balanced,8.102314631144205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,1,balanced,7.440186818440755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,1,balanced,13.224650065104166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,1,balanced,11.17520014444987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,64,power_law_1.01,0.06549760103225707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,64,power_law_1.01,0.06182399988174438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,64,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,64,power_law_1.01,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,64,power_law_1.01,0.04855040013790131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,64,power_law_1.01,0.05103999972343445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,64,power_law_1.01,0.05103359818458557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,64,power_law_1.01,0.051641601324081424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,64,power_law_1.01,0.05299199819564819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,64,power_law_1.01,0.05386880040168762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,64,power_law_1.01,0.05386880040168762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,64,power_law_1.01,0.05555840134620667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,64,power_law_1.01,0.05822719931602478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,64,power_law_1.01,0.060915201902389526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,64,power_law_1.01,0.06510720252990723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,64,power_law_1.01,0.06986240148544312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,64,power_law_1.01,0.07237759828567505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,64,power_law_1.01,0.08544639945030212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,64,power_law_1.01,0.08789119720458985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,64,power_law_1.01,0.10499199628829955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,64,power_law_1.01,0.11820160150527954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,64,power_law_1.01,0.1536960005760193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,64,power_law_1.01,0.17671040296554566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,64,power_law_1.01,0.2184704065322876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,64,power_law_1.01,0.28682239055633546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,64,power_law_1.01,0.375980806350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,64,power_law_1.01,0.49898881912231446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,4,power_law_1.01,0.04845440089702606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,4,power_law_1.01,0.06479359865188598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,4,power_law_1.01,0.0756991982460022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,4,power_law_1.01,0.09423999786376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,4,power_law_1.01,0.10025600194931031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,4,power_law_1.01,0.10689280033111573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,4,power_law_1.01,0.12872320413589478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,4,power_law_1.01,0.12807040214538573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,4,power_law_1.01,0.12993919849395752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,4,power_law_1.01,0.12454400062561036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,4,power_law_1.01,0.13605120182037353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,4,power_law_1.01,0.14466559886932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,4,power_law_1.01,0.14633599519729615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,4,power_law_1.01,0.15374079942703248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,2,power_law_1.2,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,4,power_law_1.01,0.15592960119247437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,2,power_law_1.2,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,2,power_law_1.2,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,2,power_law_1.2,0.07057279944419861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,2,power_law_1.2,0.08451840281486511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,2,power_law_1.2,0.11131520271301269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,2,power_law_1.2,0.11975680589675904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,2,power_law_1.2,0.1180351972579956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,2,power_law_1.2,0.12537599802017213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,4,power_law_1.01,0.1704192042350769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,2,power_law_1.2,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,4,power_law_1.01,0.17140480279922485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,2,power_law_1.2,0.13191039562225343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,4,power_law_1.01,0.1842560052871704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,2,power_law_1.2,0.13651200532913207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,2,power_law_1.2,0.13976320028305053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,2,power_law_1.2,0.14718719720840454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,2,power_law_1.2,0.15734399557113649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,2,power_law_1.2,0.16465280055999756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,2,power_law_1.2,0.17015039920806885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,2,power_law_1.2,0.19441920518875122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,2,power_law_1.2,0.21091840267181397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,2,power_law_1.2,0.2693183898925781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,4,power_law_1.01,0.2008064031600952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,2,power_law_1.2,0.30030078887939454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,2,power_law_1.2,0.4056896209716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,4,power_law_1.01,0.2357759952545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,2,power_law_1.2,0.5025023937225341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,2,power_law_1.2,0.6746304035186768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,2,power_law_1.2,0.8148927688598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,4,power_law_1.01,0.26776320934295655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,4,power_law_1.01,0.3135103940963745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,2688,22,512,1,128,power_law_1.01,0.07088639736175537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,4,power_law_1.01,0.4036736011505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,4,power_law_1.01,0.5092991828918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,2688,22,512,1,128,power_law_1.01,0.07008640170097351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,4,power_law_1.01,0.649721622467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,2688,22,512,1,128,power_law_1.01,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,4,power_law_1.01,0.8663680076599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,2688,22,512,1,128,power_law_1.01,0.07767040133476258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,4,power_law_1.01,1.1677056312561036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,2688,22,512,1,128,power_law_1.01,0.07529600262641907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,2688,22,512,1,128,power_law_1.01,0.0782912015914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,2688,22,512,1,128,power_law_1.01,0.08546559810638428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,2688,22,512,1,128,power_law_1.01,0.0850048005580902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,2688,22,512,1,128,power_law_1.01,0.08379520177841186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,2688,22,512,1,128,power_law_1.01,0.08568320274353028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,2688,22,512,1,128,power_law_1.01,0.08540160059928895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,2688,22,512,1,128,power_law_1.01,0.09155200123786926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,2688,22,512,1,128,power_law_1.01,0.0911679983139038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,2688,22,512,1,128,power_law_1.01,0.09285119771957398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,2688,22,512,1,128,power_law_1.01,0.1009600043296814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,2688,22,512,1,128,power_law_1.01,0.10185600519180298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,2688,22,512,1,128,power_law_1.01,0.10960639715194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,2688,22,512,1,128,power_law_1.01,0.12641919851303102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,2688,22,512,1,128,power_law_1.01,0.1517184019088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,2688,22,512,1,128,power_law_1.01,0.19088640213012695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,2688,22,512,1,128,power_law_1.01,0.21612160205841063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,2,power_law_1.2,1.231430435180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,2688,22,512,1,128,power_law_1.01,0.3020096063613892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,2688,22,512,1,128,power_law_1.01,0.34226560592651367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,2688,22,512,1,128,power_law_1.01,0.45915517807006834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,2688,22,512,1,128,power_law_1.01,0.6243135929107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,2688,22,512,1,128,power_law_1.01,0.9057087898254395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,2688,22,512,1,128,power_law_1.01,1.1413951873779298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,8,power_law_1.2,0.13690880537033082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,8,power_law_1.2,0.20600318908691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,8,power_law_1.2,0.20508160591125488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,8,power_law_1.2,0.24854400157928466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,8,power_law_1.2,0.28647680282592775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,8,power_law_1.2,0.2971839904785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,8,power_law_1.2,0.267903995513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,8,power_law_1.2,0.2596800088882446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,8,power_law_1.2,0.2886143922805786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,8,power_law_1.2,0.29815680980682374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,8,power_law_1.2,0.29142398834228517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,8,power_law_1.2,0.29582719802856444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,8,power_law_1.2,0.3059776067733765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,8,power_law_1.2,0.3119935989379883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,8,power_law_1.2,0.3105664014816284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,8,power_law_1.2,0.31468799114227297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,8,power_law_1.2,0.32247679233551024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,8,power_law_1.2,0.3421056032180786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.2,0.37832319736480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.2,0.3971519947052002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.2,0.49881601333618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.2,0.5839935779571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.2,0.6803135871887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.2,0.9295999526977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,8,power_law_1.2,0.029267200827598573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.2,1.0529727935791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,8,power_law_1.2,0.029766398668289184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,8,power_law_1.2,0.02736000120639801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,8,power_law_1.2,0.029292801022529603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.2,1.5957056045532227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,8,power_law_1.2,0.029760000109672547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,8,power_law_1.2,0.02971520125865936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,8,power_law_1.2,0.03022719919681549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,8,power_law_1.2,0.030041599273681642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.2,1.8136959075927734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,8,power_law_1.2,0.030950400233268737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,8,power_law_1.2,0.03076480031013489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,8,power_law_1.2,0.03178240060806274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,8,power_law_1.2,0.03187200129032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,8,power_law_1.2,0.03225600123405457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,8,power_law_1.2,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,8,power_law_1.2,0.03837440013885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,8,power_law_1.2,0.03919999897480011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,8,power_law_1.2,0.04190720021724701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,8,power_law_1.2,0.049497601389884946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,8,power_law_1.2,0.05297279953956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,8,power_law_1.2,0.06335359811782837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,8,power_law_1.2,0.07216640114784241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,8,power_law_1.2,0.0997376024723053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,2,power_law_1.2,1.670457649230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,16,balanced,0.021290667355060577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,16,balanced,0.022143999735514324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,16,balanced,0.023120000958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,16,balanced,0.023045333723227184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,16,balanced,0.023141334454218548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,16,balanced,0.02517866591612498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,16,balanced,0.02792533238728841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,16,balanced,0.02811199923356374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,16,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,16,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,16,balanced,0.03540800015131632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,16,balanced,0.05624533196290334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,16,balanced,0.05620799958705902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,16,balanced,0.0568800022204717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,16,balanced,0.058373332023620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,16,balanced,0.06170133252938589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,16,balanced,0.06033066908518473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,16,balanced,0.06774933139483134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,16,balanced,0.07667199770609538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,16,balanced,0.09000533819198608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,16,balanced,0.10504532853762309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,16,balanced,0.13320533434549967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,16,balanced,0.16475199659665427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,16,balanced,0.20774932702382407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,16,balanced,0.2605653405189514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,16,balanced,0.3877439896265666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,16,balanced,0.48959465821584064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,8,power_law_1.2,0.11077120304107665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,8,power_law_1.2,0.14154239892959594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,8,power_law_1.2,0.1863360047340393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,8,power_law_1.2,0.2544384002685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,1,power_law_1.2,0.15501439571380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,1,power_law_1.2,0.2547647953033447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,1,power_law_1.2,0.43873281478881837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,1,power_law_1.2,0.8400959968566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,1,power_law_1.2,1.3200127601623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,1,power_law_1.2,1.8230144500732421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,1,power_law_1.2,2.2176895141601562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,1,power_law_1.2,3.013574409484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,1,power_law_1.2,3.196076774597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,1,power_law_1.2,3.312051010131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,1,power_law_1.2,3.4672000885009764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,1,power_law_1.2,3.650668716430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,1,power_law_1.2,3.669913482666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,1,power_law_1.2,3.9153217315673827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,1,power_law_1.2,4.0469825744628904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,1,power_law_1.2,4.204755020141602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,1,power_law_1.2,4.351532745361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,1,power_law_1.2,4.765241622924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,8,power_law_1.2,0.3320512056350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,1,power_law_1.2,5.042099380493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,1,power_law_1.2,5.691507339477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,1,power_law_1.2,5.365702438354492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,1,power_law_1.2,6.046393585205078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,1,power_law_1.2,7.255999755859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,1,power_law_1.2,8.88028793334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,1,power_law_1.2,8.762297821044921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,1,power_law_1.2,10.469062042236327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,1,power_law_1.2,14.504998779296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,128,balanced,0.01915733392039935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,128,balanced,0.021029333273569744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,128,balanced,0.020576000213623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,128,balanced,0.021317332983016968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,128,balanced,0.026026666164398193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,128,balanced,0.03781333317359289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,128,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,128,balanced,0.03533333291610082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,128,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,128,balanced,0.03521066655715307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,128,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,128,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,128,balanced,0.03331733246644338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,128,balanced,0.03349333256483078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,128,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,128,balanced,0.03939199944337209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,128,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,128,balanced,0.038058665891488395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,128,balanced,0.04761599997679392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,128,balanced,0.05193600058555603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,128,balanced,0.06346666812896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,128,balanced,0.08449600140253703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,128,balanced,0.10513066252072652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,128,balanced,0.14317867159843445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,128,balanced,0.1824000080426534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,128,balanced,0.2603413263956706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,128,balanced,0.3393919865290324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,2,balanced,0.055104002356529236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,2,balanced,0.06837333242098491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,2,balanced,0.09442666172981262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,2,balanced,0.1495573321978251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,2,balanced,0.24748265743255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,2,balanced,0.44412267208099365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,2,balanced,0.641648014386495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,2,balanced,0.6421333154042562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,2,balanced,0.6435093482335409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,2,balanced,0.6474666595458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,2,balanced,0.6481226682662964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,2,balanced,0.6547146638234457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,2,balanced,0.6612000068028768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,2,balanced,0.6684587001800537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,2,balanced,0.6736213366190592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,2,balanced,0.6809066931406657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,2,balanced,0.6956213315327963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,2,balanced,0.7224693298339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,2,balanced,0.7436213493347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,2,balanced,0.803114652633667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,2,balanced,0.8321546713511149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,2,balanced,0.9387679894765218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,2,balanced,1.014954646428426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,2,balanced,1.1983786424001057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,2,balanced,1.471951961517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,2,balanced,1.8988426526387532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,2,balanced,2.433114687601725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,4,balanced,0.06392000118891399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,4,balanced,0.06325866778691609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,4,balanced,0.07284266750017802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,4,balanced,0.07885866860548656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,4,balanced,0.07685866455237071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,4,balanced,0.0769760012626648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,4,balanced,0.07874133189519246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,4,balanced,0.07252799967924754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,4,balanced,0.07458666463692983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,4,balanced,0.0729013333717982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,4,balanced,0.07482666770617168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,4,balanced,0.06966933111349742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,4,balanced,0.06866666674613953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,4,balanced,0.07494399944941203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,4,balanced,0.07678399980068207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,4,balanced,0.07377066711584727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,4,balanced,0.07714666426181793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,4,balanced,0.09172266721725464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,4,balanced,0.09101333220799764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,4,balanced,0.1150986651579539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,4,balanced,0.12661866346995035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,4,balanced,0.16428800423940024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,4,balanced,0.20099733273188272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,4,balanced,0.27346134185791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,4,balanced,0.3362666765848796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,4,balanced,0.48006399472554523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,4,balanced,0.6148159901301066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,4,power_law_1.2,0.018323199450969697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,4,power_law_1.2,0.018278400599956512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,4,power_law_1.2,0.01823360025882721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,4,power_law_1.2,0.018374399840831758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,4,power_law_1.2,0.01834239959716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,4,power_law_1.2,0.01910399943590164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,4,power_law_1.2,0.021881599724292756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,4,power_law_1.2,0.022547200322151184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,8,power_law_1.2,0.06719359755516052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,8,power_law_1.2,0.07505919933319091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,8,power_law_1.2,0.08149759769439698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,8,power_law_1.2,0.0950655996799469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,8,power_law_1.2,0.10533119440078735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,4,power_law_1.2,0.045542401075363156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,4,power_law_1.2,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,128,power_law_1.2,0.07019519805908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,128,power_law_1.2,0.06823679804801941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,128,power_law_1.2,0.06832640171051026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,128,power_law_1.2,0.07247359752655029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,128,power_law_1.2,0.07086079716682434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,128,power_law_1.2,0.07656319737434387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,4,power_law_1.2,0.04643200039863586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,4,power_law_1.2,0.033318400382995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,4,power_law_1.2,0.03395200073719025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,4,power_law_1.2,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,4,power_law_1.2,0.03521920144557953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,8,power_law_1.2,0.11145600080490112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,4,power_law_1.2,0.03687039911746979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,128,power_law_1.2,0.07386879920959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,128,power_law_1.2,0.07697920203208923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,128,power_law_1.2,0.07333760261535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,128,power_law_1.2,0.0745855987071991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,128,power_law_1.2,0.07566720247268677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,128,power_law_1.2,0.0802944004535675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,128,power_law_1.2,0.07867519855499268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,4,power_law_1.2,0.040678399801254275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,128,power_law_1.2,0.0802944004535675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,8,power_law_1.2,0.10814080238342286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,128,power_law_1.2,0.08407040238380432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,128,power_law_1.2,0.0866815984249115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,8,power_law_1.2,0.10785919427871704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,4,power_law_1.2,0.04380159974098206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,8,power_law_1.2,0.10965119600296021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,8,power_law_1.2,0.1110975980758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,8,power_law_1.2,0.11250560283660889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,8,power_law_1.2,0.11598720550537109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,8,power_law_1.2,0.11745920181274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.2,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,8,power_law_1.2,0.12417919635772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,128,power_law_1.2,0.09255040287971497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,8,power_law_1.2,0.1303231954574585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,128,power_law_1.2,0.1027135968208313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,8,power_law_1.2,0.1356927990913391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,128,power_law_1.2,0.10837759971618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,8,power_law_1.2,0.1430848002433777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.2,0.053235197067260744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,128,power_law_1.2,0.12647680044174195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,128,power_law_1.2,0.14261120557785034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,8,power_law_1.2,0.17032320499420167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,128,power_law_1.2,0.19248640537261963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,8,power_law_1.2,0.19242240190505983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,128,power_law_1.2,0.2337536096572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,8,power_law_1.2,0.23557119369506835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,128,power_law_1.2,0.33016960620880126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,8,power_law_1.2,0.2796992063522339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,128,power_law_1.2,0.4294591903686523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,8,power_law_1.2,0.3591232061386108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.2,0.061408001184463504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,128,power_law_1.2,0.6233856201171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,8,power_law_1.2,0.4162816047668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,8,power_law_1.2,0.5909183979034424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,8,power_law_1.2,0.7106368064880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,128,power_law_1.2,0.9111616134643554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,8,power_law_1.2,0.941004753112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,8,power_law_1.2,1.159347152709961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.2,0.08627840280532836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.2,0.09908480048179627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.2,0.1342144012451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.2,0.17134079933166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,8,balanced,0.029685333371162415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,8,balanced,0.029898665845394135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,8,balanced,0.029706666866938274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,8,balanced,0.031888000667095184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,8,balanced,0.036789332826932274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,8,balanced,0.03986666599909464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,8,balanced,0.037946666280428566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,8,balanced,0.0645546664794286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,8,balanced,0.03990400085846583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,8,balanced,0.06621866424878438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,8,balanced,0.07698133091131847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,8,balanced,0.09513599673906963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,8,balanced,0.12707733114560446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,8,balanced,0.18348799149195352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,8,balanced,0.18430399894714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,8,balanced,0.1829813321431478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,8,balanced,0.18614933888117471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,8,balanced,0.04019733270009359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,8,balanced,0.18452266852060953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,8,balanced,0.03995199998219808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,8,balanced,0.1865653395652771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,8,balanced,0.18965333700180054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,8,balanced,0.19035200277964273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,8,balanced,0.19316800435384116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,8,balanced,0.20363734165827432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,8,balanced,0.20336000124613443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,8,balanced,0.2075786590576172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,8,balanced,0.04027733455101649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,8,balanced,0.03994133323431015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,8,balanced,0.0421066681543986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,8,balanced,0.04214933514595032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,8,balanced,0.04417600234349569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,8,balanced,0.0517546683549881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,8,balanced,0.04972266654173533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,8,balanced,0.0566293348868688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,8,balanced,0.053472002347310386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,8,balanced,0.07302933434645335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,8,balanced,0.22900799910227457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,8,balanced,0.07428266604741414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,8,balanced,0.09353599945704143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,8,balanced,0.10296000043551128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,8,balanced,0.1402506629625956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,8,balanced,0.23426133394241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,8,balanced,0.16620799899101257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,8,balanced,0.25886933008829754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,8,balanced,0.22508267561594644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,8,balanced,0.2791093389193217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,8,balanced,0.267194668451945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,8,balanced,0.32526934146881104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,8,balanced,0.37225067615509033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,4,balanced,0.09147199988365173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,4,balanced,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,4,balanced,0.10117333134015401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,4,balanced,0.11652266979217529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,4,balanced,0.13262400031089783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,4,balanced,0.17456533511479697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,4,balanced,0.19987199703852335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,4,balanced,0.19921066363652548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,4,balanced,0.20195200045903525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,4,balanced,0.1995946764945984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,4,balanced,0.20356265703837076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,4,balanced,0.20705066124598184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,8,balanced,0.5128639936447144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,4,balanced,0.20785599946975708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,4,balanced,0.21207465728123984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,4,balanced,0.22132800022761026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,4,balanced,0.22166933616002402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,4,balanced,0.22866666316986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,4,balanced,0.24766933917999268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,4,balanced,0.2616213361422221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,8,balanced,0.6053173144658407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,4,balanced,0.2977439959843953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.2,0.24437119960784912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,4,balanced,0.33296533425649005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,8,balanced,0.8561973571777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,8,balanced,1.1139787038167317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,4,balanced,0.40889068444569904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,4,balanced,0.4907840092976888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,4,balanced,0.6437439918518066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,4,balanced,0.8645333449045817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,4,balanced,1.1738080183664958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,4,balanced,1.5571786562601726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,2,128,power_law_1.01,0.042931199073791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,2,128,power_law_1.01,0.05077760219573975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,2,128,power_law_1.01,0.04044159948825836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,2,128,power_law_1.01,0.04206080138683319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,2,128,power_law_1.01,0.04151679873466492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,2,128,power_law_1.01,0.043647998571395875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,2,128,power_law_1.01,0.044915199279785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,2,128,power_law_1.01,0.04496000111103058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,2,128,power_law_1.01,0.044998401403427125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,2,128,power_law_1.01,0.045664000511169436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,2,128,power_law_1.01,0.05016319751739502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,2,128,power_law_1.01,0.05311359763145447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,2,128,power_law_1.01,0.054028797149658206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,2,128,power_law_1.01,0.05704960227012634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,2,128,power_law_1.01,0.059724801778793336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,2,128,power_law_1.01,0.06090880036354065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,2,128,power_law_1.01,0.055251199007034305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,2,128,power_law_1.01,0.06332160234451294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,2,128,power_law_1.01,0.06559360027313232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,2,128,power_law_1.01,0.07818880081176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,2,128,power_law_1.01,0.0886784017086029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,2,128,power_law_1.01,0.10927360057830811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,2,128,power_law_1.01,0.12761600017547609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,2,128,power_law_1.01,0.167084801197052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,2,128,power_law_1.01,0.21030399799346924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,2,32,power_law_1.2,0.06063359975814819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,2,128,power_law_1.01,0.2891455888748169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,2,32,power_law_1.2,0.05803520083427429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,2,128,power_law_1.01,0.3713023900985718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,2,32,power_law_1.2,0.057132798433303836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,2,32,power_law_1.2,0.05287039875984192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,2,32,power_law_1.2,0.055769598484039305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,2,32,power_law_1.2,0.05904639959335327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,2,32,power_law_1.2,0.06156799793243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,2,32,power_law_1.2,0.06223359704017639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,2,32,power_law_1.2,0.06373119950294495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,2,32,power_law_1.2,0.06677119731903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,2,32,power_law_1.2,0.06704000234603882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,2,32,power_law_1.2,0.07246720194816589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,2,32,power_law_1.2,0.07169920206069946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,2,32,power_law_1.2,0.07662720084190369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,2,32,power_law_1.2,0.08533120155334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,2,32,power_law_1.2,0.08855040073394775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,2,32,power_law_1.2,0.10003839731216431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,2,32,power_law_1.2,0.10859520435333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,2,32,power_law_1.2,0.11559040546417236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,2,32,power_law_1.2,0.15296000242233276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,2,32,power_law_1.2,0.1667456030845642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,2,32,power_law_1.2,0.21144320964813232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,2,32,power_law_1.2,0.281439995765686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.2,0.304915189743042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,2,32,power_law_1.2,0.3902208089828491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,2,32,power_law_1.2,0.4988096237182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,2,32,power_law_1.2,0.7227263927459717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,2,32,power_law_1.2,0.8874048233032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,32,power_law_1.01,0.05980799794197082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,32,power_law_1.01,0.06180480122566223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,32,power_law_1.01,0.05538560152053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,32,power_law_1.01,0.060844802856445314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,32,power_law_1.01,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,32,power_law_1.01,0.06342399716377259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,1,power_law_1.2,0.07278079986572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,32,power_law_1.01,0.062438398599624634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,32,power_law_1.01,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,32,power_law_1.01,0.06490240097045899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,32,power_law_1.01,0.06546559929847717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,32,power_law_1.01,0.06812800168991089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,32,power_law_1.01,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,32,power_law_1.01,0.07651839852333069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,1,power_law_1.2,0.09609599709510804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,32,power_law_1.01,0.07388160228729249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,32,power_law_1.01,0.08257279992103576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,32,power_law_1.01,0.08538879752159119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,32,power_law_1.01,0.08809599876403809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,32,power_law_1.01,0.10539519786834717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,32,power_law_1.01,0.11898239850997924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,32,power_law_1.01,0.1454975962638855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,1,power_law_1.2,0.13229440450668334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,1,power_law_1.2,0.20743680000305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,1,power_law_1.2,0.28740479946136477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,1,power_law_1.2,0.3625792026519775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,1,power_law_1.2,0.5428991794586182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,32,power_law_1.01,0.17943040132522584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,1,power_law_1.2,0.5690432071685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,32,power_law_1.01,0.2296191930770874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,1,power_law_1.2,0.5890495777130127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,32,power_law_1.01,0.23283839225769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,1,power_law_1.2,0.5940927982330322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,32,power_law_1.01,0.315116810798645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,1,power_law_1.2,0.630841588973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,32,power_law_1.01,0.38472959995269773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,1,power_law_1.2,0.6513984203338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,1,power_law_1.2,0.6688127994537354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,1,power_law_1.2,0.6918975830078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,1,power_law_1.2,0.7160704135894775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,1,power_law_1.2,0.7254528045654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,1,power_law_1.2,0.7761983871459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,32,power_law_1.01,0.5349247932434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,1,power_law_1.2,0.8816639900207519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,32,power_law_1.01,0.6835968017578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,1,power_law_1.2,0.9111424446105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,1,power_law_1.2,1.0469120025634766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,1,power_law_1.2,1.107423973083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,1,power_law_1.2,1.32359037399292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,1,power_law_1.2,1.5668543815612792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,1,power_law_1.2,2.0229183197021485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,1,power_law_1.2,2.418988800048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,1,power_law_1.2,3.2880321502685548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,1,power_law_1.2,4.200991821289063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,32,4,power_law_1.2,0.0457152009010315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,32,4,power_law_1.2,0.04981119930744171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,32,4,power_law_1.2,0.05257599949836731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,32,4,power_law_1.2,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,32,4,power_law_1.2,0.06529279947280883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,32,4,power_law_1.2,0.0686464011669159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,32,4,power_law_1.2,0.07827839851379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,32,4,power_law_1.2,0.07923200130462646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,32,4,power_law_1.2,0.08017920255661011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,32,4,power_law_1.2,0.07841280102729797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,32,4,power_law_1.2,0.08704000115394592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,32,4,power_law_1.2,0.09187840223312378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,32,4,power_law_1.2,0.0948415994644165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,32,4,power_law_1.2,0.10300159454345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,32,4,power_law_1.2,0.11373440027236939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,32,4,power_law_1.2,0.11499520540237426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,32,4,power_law_1.2,0.12233599424362182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,32,4,power_law_1.2,0.14190720319747924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,32,4,power_law_1.2,0.1701696038246155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,32,4,power_law_1.2,0.21076478958129882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,32,4,power_law_1.2,0.23459200859069823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,32,4,power_law_1.2,0.3297152042388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,32,4,power_law_1.2,0.3993151903152466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,32,4,power_law_1.2,0.5694143772125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,32,4,power_law_1.2,0.6817599773406983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,32,4,power_law_1.2,1.0208191871643066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,32,4,power_law_1.2,1.4259072303771974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,4,power_law_1.2,0.0884223997592926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,4,power_law_1.2,0.1251456022262573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,4,power_law_1.2,0.1356608033180237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,4,power_law_1.2,0.20120320320129395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,4,power_law_1.2,0.23239679336547853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,4,power_law_1.2,0.2277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,4,power_law_1.2,0.3261375904083252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,4,power_law_1.2,0.3343616008758545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,4,power_law_1.2,0.34898560047149657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,4,power_law_1.2,0.3196799993515015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,4,power_law_1.2,0.34012799263000487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,4,power_law_1.2,0.3539007902145386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,4,power_law_1.2,0.34576001167297366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,4,power_law_1.2,0.3568959951400757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,4,power_law_1.2,0.3863424062728882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,4,power_law_1.2,0.40120320320129393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,4,power_law_1.2,0.4075136184692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,4,power_law_1.2,0.45841279029846194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,4,power_law_1.2,0.47192959785461425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,4,power_law_1.2,0.5327424049377442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,4,power_law_1.2,0.5460224151611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,4,power_law_1.2,0.6532991886138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,4,power_law_1.2,0.7880127906799317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,4,power_law_1.2,0.9770048141479493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,4,power_law_1.2,1.1035584449768066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,4,power_law_1.2,1.568627166748047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,4,power_law_1.2,1.911404800415039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,4,1,power_law_1.2,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,4,1,power_law_1.2,0.07386239767074584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,4,1,power_law_1.2,0.10090880393981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,4,1,power_law_1.2,0.12973439693450928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,4,1,power_law_1.2,0.1633471965789795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,4,1,power_law_1.2,0.2617023944854736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,4,1,power_law_1.2,0.28358399868011475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,4,1,power_law_1.2,0.2975872039794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,4,1,power_law_1.2,0.30467839241027833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,4,1,power_law_1.2,0.309881591796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,4,1,power_law_1.2,0.32101759910583494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,4,1,power_law_1.2,0.33540480136871337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,4,1,power_law_1.2,0.35233919620513915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,4,1,power_law_1.2,0.35062398910522463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,4,balanced,0.03997866561015447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,4,balanced,0.03990933299064636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,4,balanced,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,4,balanced,0.05078400174776713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,4,balanced,0.06305066744486491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,4,balanced,0.08761066198348999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,4,balanced,0.08866666754086812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,4,balanced,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,4,balanced,0.08483733733495076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,4,balanced,0.08532266815503438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,4,balanced,0.08715200424194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,4,balanced,0.08860266208648682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,4,balanced,0.09030933181444804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,4,balanced,0.08959999680519104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,4,balanced,0.09543466567993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,4,balanced,0.095360000928243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,4,balanced,0.10454400380452473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,4,1,power_law_1.2,0.36375041007995607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,4,balanced,0.1132586697737376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,4,1,power_law_1.2,0.382259202003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,4,1,power_law_1.2,0.4119872093200684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,4,1,power_law_1.2,0.47920641899108884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,4,1,power_law_1.2,0.5057536125183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,4,1,power_law_1.2,0.6216703891754151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,4,balanced,0.11830400427182515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,4,1,power_law_1.2,0.7662464141845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,4,balanced,0.13554666439692178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,4,balanced,0.15148267149925232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,4,balanced,0.19131199518839517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,4,1,power_law_1.2,0.9771072387695312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,4,balanced,0.21252266565958658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,4,balanced,0.29611732562383014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,4,1,power_law_1.2,1.1001215934753419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,4,1,power_law_1.2,1.4482624053955078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,4,balanced,0.3612906535466512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,4,balanced,0.5145653486251831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,4,balanced,0.6509013175964355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,4,1,power_law_1.2,1.8790912628173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,16,power_law_1.2,0.12350080013275147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,16,power_law_1.2,0.17882879972457885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,16,power_law_1.2,0.16355839967727662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,16,power_law_1.2,0.17433600425720214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,16,power_law_1.2,0.17694720029830932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,16,power_law_1.2,0.15266560316085814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,16,power_law_1.2,0.1559615969657898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,16,power_law_1.2,0.1628543972969055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,16,power_law_1.2,0.17546240091323853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,4,1,power_law_1.2,2.5691455841064452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,16,power_law_1.2,0.1693887948989868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,16,power_law_1.2,0.17975679636001587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,16,power_law_1.2,0.16976640224456788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,16,power_law_1.2,0.17687679529190065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,16,power_law_1.2,0.17663359642028809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,16,power_law_1.2,0.19883519411087036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,16,power_law_1.2,0.20281600952148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,16,power_law_1.2,0.21289598941802979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,16,power_law_1.2,0.22561280727386473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,16,power_law_1.2,0.2358975887298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,16,power_law_1.2,0.2835776090621948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,16,power_law_1.2,0.3237567901611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,16,power_law_1.2,0.40919041633605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,16,power_law_1.2,0.5168767929077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,16,power_law_1.2,0.7545216083526611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,16,power_law_1.2,0.9122624397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,16,power_law_1.2,1.3653056144714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,16,power_law_1.2,2.0169855117797852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,4,1,power_law_1.2,3.2245121002197266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,1,power_law_1.01,0.10101759433746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,1,power_law_1.01,0.15262720584869385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,1,power_law_1.01,0.25652480125427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,1,power_law_1.01,0.4094655990600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,1,power_law_1.01,0.5266240119934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,1,power_law_1.01,0.8672320365905761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,1,power_law_1.01,0.9348480224609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,1,power_law_1.01,0.9450943946838379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,1,power_law_1.01,0.9672703742980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,1,power_law_1.01,1.0290559768676757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,1,power_law_1.01,1.032691192626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,1,power_law_1.01,1.0461376190185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,1,power_law_1.01,1.0686335563659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,1,power_law_1.01,1.098310375213623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,1,power_law_1.01,1.1400511741638184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,1,power_law_1.01,1.1691712379455566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,1,power_law_1.01,1.2281984329223632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,1,power_law_1.01,1.3667712211608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,1,power_law_1.01,1.3979071617126464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,1,power_law_1.01,1.5984319686889648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,1,power_law_1.01,1.6183040618896485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,1,power_law_1.01,1.928563117980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,1,power_law_1.01,2.2650880813598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,1,power_law_1.01,2.8844736099243162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,1,power_law_1.01,4.061196899414062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,1,power_law_1.01,5.118246459960938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,1,power_law_1.01,6.0924736022949215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,8,power_law_1.01,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,8,power_law_1.01,0.052134400606155394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,8,power_law_1.01,0.05804799795150757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,8,power_law_1.01,0.05811840295791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,8,power_law_1.01,0.06503679752349853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,8,power_law_1.01,0.06851199865341187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,8,power_law_1.01,0.06858879923820496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,8,power_law_1.01,0.06933760046958923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,8,power_law_1.01,0.07055360078811646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,8,power_law_1.01,0.07134079933166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,8,power_law_1.01,0.07304959893226623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,8,power_law_1.01,0.07588480114936828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,8,power_law_1.01,0.07819520235061646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,8,power_law_1.01,0.08496639728546143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,8,power_law_1.01,0.09144319891929627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,8,power_law_1.01,0.09652479887008666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,8,power_law_1.01,0.10568959712982177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,8,power_law_1.01,0.12943999767303466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,8,power_law_1.01,0.13467520475387573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,8,power_law_1.01,0.17096960544586182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,8,power_law_1.01,0.19827840328216553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,8,power_law_1.01,0.2694783926010132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,8,power_law_1.01,0.3375744104385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,8,power_law_1.01,0.4669055938720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,8,power_law_1.01,0.6180223941802978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,8,power_law_1.01,0.8201024055480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,8,power_law_1.01,1.114956760406494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,1,power_law_1.01,0.09776639938354492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,1,power_law_1.01,0.13181439638137818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,1,power_law_1.01,0.1883455991744995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,1,power_law_1.01,0.30186240673065184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,1,power_law_1.01,0.40728960037231443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,1,power_law_1.01,0.6567232131958007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,1,power_law_1.01,0.7112192153930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,1,power_law_1.01,0.7282495975494385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,1,power_law_1.01,0.7448192119598389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,1,power_law_1.01,0.7489727973937989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,1,power_law_1.01,0.7963007926940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,1,power_law_1.01,0.8111359596252441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,1,power_law_1.01,0.8334976196289062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,1,power_law_1.01,0.8521792411804199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,1,power_law_1.01,0.86627197265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,1,power_law_1.01,0.8883008003234864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,1,power_law_1.01,0.944486427307129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,1,power_law_1.01,1.0362624168395995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,1,power_law_1.01,1.142195224761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,1,power_law_1.01,1.3488320350646972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,1,power_law_1.01,1.3953023910522462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,1,power_law_1.01,1.71396484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,1,power_law_1.01,1.9956031799316407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,1,power_law_1.01,2.5963712692260743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,64,power_law_1.2,0.0840511977672577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,64,power_law_1.2,0.0827072024345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,64,power_law_1.2,0.08130559921264649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,1,power_law_1.01,3.485753631591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,64,power_law_1.2,0.08565120100975036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,64,power_law_1.2,0.07744640111923218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,64,power_law_1.2,0.08524799942970276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,64,power_law_1.2,0.08477439880371093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,64,power_law_1.2,0.08931840062141419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,1,power_law_1.01,4.71923828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,64,power_law_1.2,0.0865664005279541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,64,power_law_1.2,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,64,power_law_1.2,0.09208959937095643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,64,power_law_1.2,0.08911359906196595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,64,power_law_1.2,0.08979200124740601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,1,power_law_1.01,5.749785614013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,64,power_law_1.2,0.0867904007434845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,64,power_law_1.2,0.09544960260391236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,64,power_law_1.2,0.09692800045013428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,64,power_law_1.2,0.09850239753723145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,64,power_law_1.2,0.1071936011314392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,64,power_law_1.2,0.11646720170974731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,64,power_law_1.2,0.14439040422439575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,64,power_law_1.2,0.16694400310516358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,64,power_law_1.2,0.21283199787139892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,64,power_law_1.2,0.24376959800720216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,64,power_law_1.2,0.35948159694671633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,64,power_law_1.2,0.4215231895446777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,64,power_law_1.2,0.649241590499878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,64,power_law_1.2,0.8325311660766601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,16,4,balanced,0.08178133269151051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,16,4,balanced,0.08462400237719218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,16,4,balanced,0.08798933029174805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,16,4,balanced,0.09963200489679973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,16,4,balanced,0.11522666613260905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,16,4,balanced,0.14458666245142618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,16,4,balanced,0.16856000820795694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,16,4,balanced,0.16916799545288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,16,4,balanced,0.17154133319854736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,16,4,balanced,0.1694506605466207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,16,4,balanced,0.17232000827789307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,16,4,balanced,0.1767573356628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,16,4,balanced,0.17658666769663492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,16,4,balanced,0.1827146609624227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,16,4,balanced,0.19003732999165854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,16,4,balanced,0.18997865915298462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,16,4,balanced,0.19761067628860474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,16,4,balanced,0.21572266022364298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,16,4,balanced,0.22986133893330893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,16,4,balanced,0.2643893361091614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,32,balanced,0.04824000100294749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,16,4,balanced,0.30195732911427814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,32,balanced,0.031504000226656594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,32,balanced,0.02958400050799052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,16,4,balanced,0.37941332658131915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,32,balanced,0.031541332602500916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,32,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,16,4,balanced,0.45977067947387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,16,4,balanced,0.6160159905751547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,16,4,balanced,0.8218879699707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,16,4,balanced,1.1316320101420085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,16,4,balanced,1.4888906478881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,32,balanced,0.03385066737731298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,32,balanced,0.03375466664632162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,32,balanced,0.03188266605138779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,32,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,32,balanced,0.03370666752258936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,32,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,32,balanced,0.033941333492596946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,32,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,32,balanced,0.034703999757766724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,32,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,32,balanced,0.03772266705830892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,32,balanced,0.03822933385769526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,32,balanced,0.04182399809360504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,32,balanced,0.039674667020638786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,32,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,32,balanced,0.04274666806062063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,32,balanced,0.05273066461086273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,32,balanced,0.0536053329706192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,32,balanced,0.06846933563550313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,32,balanced,0.07102400064468384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,32,balanced,0.09402666489283244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,32,balanced,0.10047466556231181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,2,balanced,0.061039999127388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,2,balanced,0.0640533318122228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,2,balanced,0.07099199791749318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,2,balanced,0.08910933136940002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,2,balanced,0.11846933762232463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,2,balanced,0.11948800086975098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,2,balanced,0.12142933408419292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,2,balanced,0.12168533603350322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,2,balanced,0.12224533160527547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,2,balanced,0.12293333808581035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,2,balanced,0.12308800220489502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,2,balanced,0.12845866878827414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,2,balanced,0.12920000155766806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,2,balanced,0.1317813297112783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,2,balanced,0.1344000001748403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,2,balanced,0.1376426617304484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,2,balanced,0.1423786679903666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,2,balanced,0.1562879979610443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,2,balanced,0.169813334941864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,2,balanced,0.1982133388519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,2,balanced,0.22925333182017008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,2,balanced,0.3309440016746521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,2,balanced,0.39443735281626385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,2,balanced,0.5619519948959351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,2,balanced,0.7241333325703939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,2,balanced,1.0561599731445312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,2,balanced,1.3899413744608562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,1,power_law_1.01,0.01754239946603775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,1,power_law_1.01,0.018272000551223754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,1,power_law_1.01,0.020659199357032774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,1,power_law_1.01,0.028198400139808656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,1,power_law_1.01,0.033606401085853575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,1,power_law_1.01,0.04589439928531647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,1,power_law_1.01,0.05330560207366943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,1,power_law_1.01,0.05697280168533325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,1,power_law_1.01,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,1,power_law_1.01,0.0607807993888855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,1,power_law_1.01,0.06183680295944214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,1,power_law_1.01,0.06254079937934875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,1,power_law_1.01,0.0655232012271881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,1,power_law_1.01,0.06769279837608337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,1,power_law_1.01,0.06849920153617858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,1,power_law_1.01,0.07051519751548767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,1,power_law_1.01,0.0741312026977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,1,power_law_1.01,0.08391680121421814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,1,power_law_1.01,0.09440640211105347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,1,power_law_1.01,0.11898239850997924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,1,power_law_1.01,0.13963520526885986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,1,power_law_1.01,0.18534400463104247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,1,power_law_1.01,0.19906560182571412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,1,power_law_1.01,0.2669055938720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,1,power_law_1.01,0.349126410484314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,1,power_law_1.01,0.4723008155822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,1,power_law_1.01,0.6118015766143798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,128,balanced,0.05426666637261709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,128,balanced,0.05596800148487091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,128,balanced,0.05584533512592316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,128,balanced,0.05875733494758606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,128,balanced,0.0565280020236969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,128,balanced,0.06404800216356914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,128,balanced,0.06534400085608165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,128,balanced,0.06639466683069865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,128,balanced,0.06623999774456024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,128,balanced,0.06821333368619283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,128,balanced,0.0702400008837382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,128,balanced,0.06832000116507213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,128,balanced,0.07252799967924754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,128,balanced,0.07670933504899342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,128,balanced,0.08060266574223836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,128,balanced,0.07870399951934814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,128,balanced,0.08563199639320374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,128,balanced,0.0986400047938029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,128,balanced,0.1116426686445872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,128,balanced,0.13733866810798645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,128,balanced,0.1558133363723755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,128,balanced,0.2010186711947123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,128,balanced,0.24117332696914673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,128,balanced,0.32520532608032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,128,balanced,0.41227734088897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,128,balanced,0.574293335278829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,128,balanced,0.7393013636271158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,8,16,power_law_1.2,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,8,16,power_law_1.2,0.05618559718132019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,8,16,power_law_1.2,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,8,16,power_law_1.2,0.05902720093727112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,8,16,power_law_1.2,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,8,16,power_law_1.2,0.05736960172653198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,8,16,power_law_1.2,0.059628802537918094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,8,16,power_law_1.2,0.06170880198478699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,8,16,power_law_1.2,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,8,16,power_law_1.2,0.061868798732757566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,8,16,power_law_1.2,0.06480640172958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,8,16,power_law_1.2,0.06830719709396363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,8,16,power_law_1.2,0.07093759775161743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,8,16,power_law_1.2,0.0722495973110199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,8,16,power_law_1.2,0.07993599772453308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,8,16,power_law_1.2,0.08350080251693726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,8,16,power_law_1.2,0.08537600040435792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,8,16,power_law_1.2,0.10214400291442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,8,16,power_law_1.2,0.09868159890174866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,8,16,power_law_1.2,0.13246079683303832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,8,16,power_law_1.2,0.1389631986618042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,8,16,power_law_1.2,0.19098880290985107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,8,16,power_law_1.2,0.2192768096923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,8,16,power_law_1.2,0.3086911916732788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,8,16,power_law_1.2,0.401964807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,8,16,power_law_1.2,0.5722623825073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,8,16,power_law_1.2,0.7013823986053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,2,power_law_1.01,0.020351999998092653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,2,power_law_1.01,0.02276480048894882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,2,power_law_1.01,0.029356798529624938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,2,power_law_1.01,0.03744640052318573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,2,power_law_1.01,0.044736000895500186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,2,power_law_1.01,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,2,power_law_1.01,0.05834239721298218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,2,power_law_1.01,0.06207360029220581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,2,power_law_1.01,0.0652671992778778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,2,power_law_1.01,0.06446080207824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,2,power_law_1.01,0.06977279782295227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,2,power_law_1.01,0.07391999959945679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,2,power_law_1.01,0.0729856014251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,2,power_law_1.01,0.07593600153923034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,16,8,balanced,0.0440586656332016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,16,8,balanced,0.044341335693995156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,2,power_law_1.01,0.08035200238227844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,16,8,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,16,8,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,2,power_law_1.01,0.08602880239486695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,16,8,balanced,0.05990933378537496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,16,8,balanced,0.08014399806658427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,2,power_law_1.01,0.08862079977989197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,16,8,balanced,0.08216000099976857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,16,8,balanced,0.08270933230717976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,2,power_law_1.01,0.16076159477233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,2,power_law_1.01,0.18839679956436156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,2,power_law_1.01,0.1721984028816223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,16,8,balanced,0.08267733454704285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,16,8,balanced,0.08508267005284627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,2,power_law_1.01,0.1876479983329773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,16,8,balanced,0.08388266960779826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,16,8,balanced,0.08522133032480876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,16,8,balanced,0.08692800005276997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,16,8,balanced,0.0892639954884847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,16,8,balanced,0.09282666444778442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,16,8,balanced,0.09568533301353455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,16,8,balanced,0.10216533144315083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,16,8,balanced,0.1111199955145518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,16,8,balanced,0.11657599608103435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,16,8,balanced,0.13170666495958963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,16,8,balanced,0.1462613344192505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,16,8,balanced,0.18635733922322592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,16,8,balanced,0.21912533044815063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,2,power_law_1.01,0.20631680488586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,16,8,balanced,0.2879839936892192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,2,power_law_1.01,0.25011839866638186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,16,8,balanced,0.35683735211690265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,2,power_law_1.01,0.3177088022232056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,2,power_law_1.01,0.4036223888397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,2,power_law_1.01,0.5429567813873291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,2,power_law_1.01,0.6574783802032471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,16,8,balanced,0.4941386779149373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,16,8,balanced,0.6281866629918417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,2,64,power_law_1.2,0.021356800198554994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,2,64,power_law_1.2,0.021516799926757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,2,64,power_law_1.2,0.021465599536895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,2,64,power_law_1.2,0.022073599696159362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,2,64,power_law_1.2,0.023999999463558196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,2,64,power_law_1.2,0.03091840147972107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,2,64,power_law_1.2,0.04979200065135956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,2,64,power_law_1.2,0.05050879716873169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,2,64,power_law_1.2,0.05103999972343445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,2,64,power_law_1.2,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,2,64,power_law_1.2,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,2,64,power_law_1.2,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,2,64,power_law_1.2,0.050732797384262084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,2,64,power_law_1.2,0.04984959959983826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,2,64,power_law_1.2,0.048691201210021975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,2,64,power_law_1.2,0.04860160052776337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,2,64,power_law_1.2,0.04906879961490631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,2,64,power_law_1.2,0.05256959795951843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,2,64,power_law_1.2,0.058374398946762086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,2,64,power_law_1.2,0.07519360184669495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,2,64,power_law_1.2,0.08626559972763062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,2,64,power_law_1.2,0.11545599699020385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,2,64,power_law_1.2,0.1374079942703247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,2,64,power_law_1.2,0.19769599437713622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,2,64,power_law_1.2,0.24491519927978517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,2,64,power_law_1.2,0.36392319202423096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,2,64,power_law_1.2,0.5199808120727539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,8,power_law_1.01,0.05570560097694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,8,power_law_1.01,0.05735679864883423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,8,power_law_1.01,0.05541120171546936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,8,power_law_1.01,0.0591871976852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,8,power_law_1.01,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,8,power_law_1.01,0.05943040251731872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,8,power_law_1.01,0.06378880143165588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,8,power_law_1.01,0.06689280271530151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,8,power_law_1.01,0.06870399713516236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,8,power_law_1.01,0.07145599722862243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,8,power_law_1.01,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,8,power_law_1.01,0.07227519750595093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,8,power_law_1.01,0.07654399871826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,8,power_law_1.01,0.08076159954071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,8,power_law_1.01,0.08945919871330262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,8,power_law_1.01,0.09502080082893372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,8,power_law_1.01,0.11198079586029053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,8,power_law_1.01,0.14431359767913818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,8,power_law_1.01,0.15546879768371583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,8,power_law_1.01,0.22227199077606202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,8,power_law_1.01,0.2583359956741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,8,power_law_1.01,0.3473920106887817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,8,power_law_1.01,0.4337344169616699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,8,power_law_1.01,0.6092671871185302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,8,power_law_1.01,0.8385600090026856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,2,power_law_1.2,0.038726401329040525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,2,power_law_1.2,0.04005120098590851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,2,power_law_1.2,0.039980798959732056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,2,power_law_1.2,0.045561599731445315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,2,power_law_1.2,0.045459198951721194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,2,power_law_1.2,0.046515199542045596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,2,power_law_1.2,0.04629760086536407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,2,power_law_1.2,0.04747520089149475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,2,power_law_1.2,0.04601599872112274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,2,power_law_1.2,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,2,power_law_1.2,0.04652160108089447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,2,power_law_1.2,0.04854399859905243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,2,power_law_1.2,0.049055999517440795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,2,power_law_1.2,0.0494592010974884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,2,power_law_1.2,0.0553600013256073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,2,power_law_1.2,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,2,power_law_1.2,0.06021760106086731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,2,power_law_1.2,0.07080960273742676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.2,0.08377599716186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.2,0.09673600196838379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,1,balanced,0.022954667607943218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,1,balanced,0.02514133354028066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.2,0.10983680486679077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,1,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,1,balanced,0.050954664746920265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,1,balanced,0.07720533510049184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,1,balanced,0.13136000434557596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,1,balanced,0.134442667166392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,1,balanced,0.13782399892807007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,1,balanced,0.1423786679903666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,1,balanced,0.1466506620248159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,1,balanced,0.1537546714146932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,1,balanced,0.1604639987150828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,1,balanced,0.16971200704574585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,1,balanced,0.1845653255780538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,1,balanced,0.17074666420618692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,1,balanced,0.1800959904988607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,1,balanced,0.1960053245226542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,1,balanced,0.19670399030049643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,1,balanced,0.21659199396769205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,1,balanced,0.2267413338025411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,1,balanced,0.24681599934895834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,1,balanced,0.48676268259684247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,1,balanced,0.5539146661758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,1,balanced,0.45761064688364667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,1,balanced,0.5035146474838257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,1,balanced,0.8374293645222982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,1,balanced,0.9323573112487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.2,0.1627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,2,32,power_law_1.01,0.09267839789390564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,2,32,power_law_1.01,0.1031872034072876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,2,32,power_law_1.01,0.0982208013534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,2,power_law_1.01,0.051507198810577394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,2,32,power_law_1.01,0.10290559530258178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,2,power_law_1.01,0.07271040081977845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,2,power_law_1.01,0.08566399812698364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,2,32,power_law_1.01,0.10357760190963745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,2,power_law_1.01,0.11477760076522828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,2,32,power_law_1.01,0.10528639554977418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,2,power_law_1.01,0.1389631986618042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,2,power_law_1.01,0.17030400037765503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,2,32,power_law_1.01,0.10302720069885254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,2,power_law_1.01,0.21021440029144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,2,power_law_1.01,0.21971840858459474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,2,power_law_1.01,0.2283263921737671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,2,power_law_1.01,0.23496320247650146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,2,power_law_1.01,0.23164799213409423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,2,power_law_1.01,0.2400063991546631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,2,power_law_1.01,0.24449920654296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,2,32,power_law_1.01,0.09946240186691284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,2,power_law_1.01,0.24983038902282714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,2,power_law_1.01,0.2686784029006958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,2,32,power_law_1.01,0.09966719746589661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,2,power_law_1.01,0.2827840089797974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,2,32,power_law_1.01,0.09895039796829223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,2,power_law_1.01,0.30037760734558105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,2,power_law_1.01,0.33113598823547363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,2,power_law_1.01,0.36469759941101076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,2,power_law_1.01,0.4237696170806885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,2,power_law_1.01,0.4990015983581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,2,power_law_1.01,0.6236544132232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,2,32,power_law_1.01,0.10051200389862061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,2,power_law_1.01,0.7496191978454589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,8,power_law_1.01,1.3830016136169434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,2,32,power_law_1.01,0.10537600517272949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,2,power_law_1.01,1.0365951538085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,2,32,power_law_1.01,0.10586880445480347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,2,32,power_law_1.01,0.10746879577636718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,2,power_law_1.01,1.2526464462280273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,2,32,power_law_1.01,0.11776000261306763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,2,power_law_1.01,1.7665151596069335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,2,power_law_1.01,2.306732749938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.2,0.18094079494476317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,2,32,power_law_1.01,0.12094719409942627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.2,0.24649600982666015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,2,32,power_law_1.01,0.12470400333404541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,2,32,power_law_1.01,0.13598079681396485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,2,32,power_law_1.01,0.14932479858398437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,2,32,power_law_1.01,0.17345279455184937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,2,32,power_law_1.01,0.2022144079208374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,2,32,power_law_1.01,0.250547194480896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,2,32,power_law_1.01,0.2920383930206299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,2,32,power_law_1.01,0.42096638679504395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,2,32,power_law_1.01,0.5623744010925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,2,32,power_law_1.01,0.7328832149505615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,2,32,power_law_1.01,0.9756863594055176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.2,0.34767999649047854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,8,power_law_1.01,1.734342384338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,2,4,power_law_1.01,0.11384960412979125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,2,4,power_law_1.01,0.10790400505065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,2,4,power_law_1.01,0.11956479549407958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,2,4,power_law_1.01,0.1192255973815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,2,4,power_law_1.01,0.12015360593795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,2,4,power_law_1.01,0.12324479818344117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,2,4,power_law_1.01,0.13593599796295167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,2,4,power_law_1.01,0.13055360317230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,2,4,power_law_1.01,0.13303040266036986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,2,4,power_law_1.01,0.14653439521789552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,2,4,power_law_1.01,0.14113279581069946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,2,4,power_law_1.01,0.15100159645080566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,2,4,power_law_1.01,0.14843519926071166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,2,4,power_law_1.01,0.1867967963218689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,2,4,power_law_1.01,0.19761919975280762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,2,4,power_law_1.01,0.21225600242614745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,2,4,power_law_1.01,0.20988800525665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,2,4,power_law_1.01,0.23364479541778566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,2,4,power_law_1.01,0.3106816053390503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,2,4,power_law_1.01,0.4583168029785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,2,4,power_law_1.01,0.48636798858642577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,2,4,power_law_1.01,0.6456704139709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,4,16,power_law_1.2,0.01852799952030182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,4,16,power_law_1.2,0.019443200528621675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,4,16,power_law_1.2,0.019385600090026857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,4,16,power_law_1.2,0.020032000541687012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,4,16,power_law_1.2,0.021408000588417055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,2,4,power_law_1.01,0.9308032035827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,4,16,power_law_1.2,0.022969600558280946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,4,16,power_law_1.2,0.02529279887676239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,4,16,power_law_1.2,0.023955200612545014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,4,16,power_law_1.2,0.028601598739624024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,4,16,power_law_1.2,0.02805759906768799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,4,16,power_law_1.2,0.02800000011920929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,4,16,power_law_1.2,0.04136320054531097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,4,16,power_law_1.2,0.04102399945259094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,4,16,power_law_1.2,0.04131200015544891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,4,16,power_law_1.2,0.043731200695037845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,4,16,power_law_1.2,0.04405120015144348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,4,16,power_law_1.2,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,4,16,power_law_1.2,0.043673598766326906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,4,16,power_law_1.2,0.04696959853172302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,4,16,power_law_1.2,0.05257599949836731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,2,4,power_law_1.01,1.249772834777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,4,16,power_law_1.2,0.059545600414276124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.2,0.4401599884033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,4,16,power_law_1.2,0.07349759936332703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,4,16,power_law_1.2,0.08899840116500854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,4,16,power_law_1.2,0.11161600351333618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,4,16,power_law_1.2,0.13779200315475465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,4,16,power_law_1.2,0.18807679414749146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,4,16,power_law_1.2,0.23904640674591066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,2,4,power_law_1.01,1.481158447265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,2,4,power_law_1.01,2.1988479614257814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.2,0.6461952209472657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,2,4,power_law_1.01,3.128646469116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,8,balanced,0.024495999018351238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,8,balanced,0.02102400114138921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,8,balanced,0.023024000227451324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,8,balanced,0.023711999257405598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,8,balanced,0.02720533311367035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,8,balanced,0.036864000062147774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,8,balanced,0.039488000174363456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,8,balanced,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,8,balanced,0.04046933352947235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,8,balanced,0.04274666806062063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,8,balanced,0.04610133171081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,8,balanced,0.04970666766166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,8,balanced,0.055439998706181846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,8,balanced,0.05394133428732554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,8,balanced,0.0558186670144399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,8,balanced,0.06020799775918325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,8,balanced,0.06645866731802623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,8,balanced,0.07858133316040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,8,balanced,0.1048906644185384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,8,balanced,0.14057599504788718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,8,balanced,0.13081600268681845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,8,balanced,0.15627732872962952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,8,balanced,0.18383999665578207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,8,balanced,0.23053866624832153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,8,balanced,0.28165332476298016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,8,balanced,0.4258933464686076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,8,balanced,0.5237386624018351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,4,power_law_1.2,0.0456959992647171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,4,power_law_1.2,0.049728000164031984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,4,power_law_1.2,0.052095997333526614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,4,power_law_1.2,0.058764797449111936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,4,power_law_1.2,0.07079039812088013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,4,power_law_1.2,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,4,power_law_1.2,0.0840511977672577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,4,power_law_1.2,0.09375360012054443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,4,power_law_1.2,0.0880832016468048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,4,power_law_1.2,0.09020159840583801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,4,power_law_1.2,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,4,power_law_1.2,0.09505280256271362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,4,power_law_1.2,0.0975488007068634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,4,power_law_1.2,0.10351359844207764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,4,power_law_1.2,0.11143679618835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,4,power_law_1.2,0.11663360595703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,4,power_law_1.2,0.12332799434661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,4,power_law_1.2,0.15239039659500123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,4,power_law_1.2,0.15822720527648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,4,power_law_1.2,0.19650559425354003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,4,power_law_1.2,0.21122560501098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,4,power_law_1.2,0.30353920459747313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,4,power_law_1.2,0.35696001052856446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,4,power_law_1.2,0.4839104175567627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,4,power_law_1.2,0.6582079887390136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,4,power_law_1.2,0.8688832283020019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,4,power_law_1.2,1.1846336364746093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,32,power_law_1.01,0.01940480023622513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,32,power_law_1.01,0.01903360038995743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,32,power_law_1.01,0.018918399512767792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,32,power_law_1.01,0.01982080042362213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,32,power_law_1.01,0.023865599930286408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,32,power_law_1.01,0.03020159900188446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,16,balanced,0.041637333730856575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,16,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,16,balanced,0.04349866509437561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,16,balanced,0.04721599817276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,16,balanced,0.049653331438700356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,16,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,16,balanced,0.0537066658337911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,16,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,16,balanced,0.05215999980767568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,16,balanced,0.05407466491063436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,16,balanced,0.05522133409976959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,16,balanced,0.054085334142049156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,16,balanced,0.05585066477457682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,16,balanced,0.05823466678460439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,16,balanced,0.06006933252016703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,16,balanced,0.062352001667022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,16,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,16,balanced,0.06716266771157582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,16,balanced,0.07218133409818013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,16,balanced,0.09267200032869975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,16,balanced,0.10388799508412679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,16,balanced,0.13159466783205667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,16,balanced,0.14453333616256714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,16,balanced,0.1934986710548401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,16,balanced,0.22277865807215372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,16,balanced,0.3063146670659383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,16,balanced,0.376911997795105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,1,power_law_1.01,0.05440000295639038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,1,power_law_1.01,0.058297598361968996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,1,power_law_1.01,0.06595199704170226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,1,power_law_1.01,0.09701759815216064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,1,power_law_1.01,0.12190719842910766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,1,power_law_1.01,0.1467136025428772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,1,power_law_1.01,0.18071039915084838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,1,power_law_1.01,0.19285119771957399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,1,power_law_1.01,0.19900799989700318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,1,power_law_1.01,0.20096640586853026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,1,power_law_1.01,0.2103679895401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,1,power_law_1.01,0.2220992088317871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,1,power_law_1.01,0.23338239192962645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,1,power_law_1.01,0.2511744022369385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,1,power_law_1.01,0.2720128059387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,1,power_law_1.01,0.28536961078643797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,1,power_law_1.01,0.275161600112915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,1,power_law_1.01,0.33477120399475097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,1,power_law_1.01,0.371942400932312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,1,power_law_1.01,0.46521601676940916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,1,power_law_1.01,0.5544256210327149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,1,power_law_1.01,0.7341440200805665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,1,power_law_1.01,0.9200448036193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,1,power_law_1.01,1.262764835357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,1,power_law_1.01,1.5995200157165528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,1,power_law_1.01,2.274425506591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,1,power_law_1.01,2.972761535644531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,8,16,balanced,0.05462933580080668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,8,16,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,8,16,balanced,0.05595199763774872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,8,16,balanced,0.06036800146102905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,8,16,balanced,0.06737066805362701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,8,16,balanced,0.08321066697438557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,8,16,balanced,0.09296533465385437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,8,16,balanced,0.0926080048084259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,8,16,balanced,0.09473066528638203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,8,16,balanced,0.09339732925097148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,8,16,balanced,0.09517866373062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,8,16,balanced,0.09847467144330342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,8,16,balanced,0.09790399670600891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,8,16,balanced,0.09901866316795349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,8,16,balanced,0.1112000048160553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,8,16,balanced,0.11078932881355286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,8,16,balanced,0.12371733784675598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,8,16,balanced,0.13523733615875244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,32,2,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,32,2,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,32,2,balanced,0.052005335688591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,32,2,balanced,0.06610666712125142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,8,16,balanced,0.13514666755994162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,32,2,balanced,0.0943946639696757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,8,16,balanced,0.15638933579126993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,8,16,balanced,0.1753066579500834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,8,16,balanced,0.21407999595006308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,8,16,balanced,0.25204267104466754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,8,16,balanced,0.3310239911079407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,8,16,balanced,0.4243786732355754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,8,16,balanced,0.5946613152821859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,32,2,balanced,0.11354133486747742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,8,16,balanced,0.7870559692382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,32,2,balanced,0.11688533425331116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,32,2,balanced,0.10624000430107117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,32,2,balanced,0.1067626674969991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,32,2,balanced,0.10900800426801045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,32,2,balanced,0.11011200149854024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,32,2,balanced,0.1128480037053426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,32,2,balanced,0.1156160036722819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,32,2,balanced,0.118559996287028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,32,2,balanced,0.12313066919644673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,32,2,balanced,0.12735999623934427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,32,2,balanced,0.13405866424242655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,32,2,balanced,0.15821866194407144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,32,2,balanced,0.16806934277216592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,32,2,balanced,0.2360639969507853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,32,2,balanced,0.24380266666412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,32,2,balanced,0.36902932325998944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,32,2,balanced,0.40934399763743085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,32,2,balanced,0.5954613288243612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,32,2,balanced,0.7231733004252116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,32,2,balanced,1.083456039428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,32,2,balanced,1.3408800760904949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,2,power_law_1.01,0.07190399765968322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,2,power_law_1.01,0.09608319997787476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,2,power_law_1.01,0.14703359603881835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,2,power_law_1.01,0.17530879974365235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,2,power_law_1.01,0.2659008026123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,2,power_law_1.01,0.34766080379486086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,2,power_law_1.01,0.39888639450073243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,2,power_law_1.01,0.49050240516662597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,2,power_law_1.01,0.48665599822998046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,2,power_law_1.01,0.4848320007324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,2,power_law_1.01,0.5179264068603515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,2,power_law_1.01,0.5237567901611329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,2,power_law_1.01,0.528492784500122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,2,power_law_1.01,0.5566912174224854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,2,power_law_1.01,0.5717631816864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,2,power_law_1.01,0.557689619064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,2,power_law_1.01,0.6114048004150391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,2,power_law_1.01,0.6671232223510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,2,power_law_1.01,0.7236671924591065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,2,power_law_1.01,0.819974422454834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,2,power_law_1.01,0.8297408103942872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,2,power_law_1.01,1.0016639709472657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,2,power_law_1.01,1.164352035522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,2,power_law_1.01,1.4333632469177247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,64,power_law_1.2,0.03880319893360138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,64,power_law_1.2,0.038406398892402646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,2,power_law_1.01,1.6121856689453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,2,power_law_1.01,2.1756927490234377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,64,power_law_1.2,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,64,power_law_1.2,0.039129599928855896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,2,power_law_1.01,2.880780792236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,64,power_law_1.2,0.04023039937019348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,64,power_law_1.2,0.04207360148429871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,64,power_law_1.2,0.04251520037651062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,64,power_law_1.2,0.042694398760795595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,64,power_law_1.2,0.04336639940738678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,64,power_law_1.2,0.04408960044384003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,64,power_law_1.2,0.05000960230827332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,64,power_law_1.2,0.05137280225753784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,64,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,64,power_law_1.2,0.04739840030670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,64,power_law_1.2,0.05187839865684509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,64,power_law_1.2,0.052985602617263795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,64,power_law_1.2,0.05546240210533142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,64,power_law_1.2,0.06407039761543273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,64,power_law_1.2,0.06746879816055298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,64,power_law_1.2,0.08212479948997498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,64,power_law_1.2,0.09022719860076904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,64,power_law_1.2,0.11715199947357177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,64,power_law_1.2,0.12880640029907225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,64,power_law_1.2,0.18575359582901002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,64,power_law_1.2,0.22429440021514893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,64,power_law_1.2,0.32943360805511473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,64,power_law_1.2,0.41431679725646975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,8,power_law_1.2,0.04870400130748749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,8,power_law_1.2,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,8,power_law_1.2,0.06159359812736511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,2,power_law_1.2,0.02134400010108948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,8,power_law_1.2,0.07933440208435058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,8,power_law_1.2,0.08640639781951905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,8,power_law_1.2,0.08362240195274354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,8,power_law_1.2,0.09010559916496277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,8,power_law_1.2,0.08883200287818908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,2,power_law_1.2,0.026054400205612182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,2,power_law_1.2,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,8,power_law_1.2,0.09105920195579528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,2,power_law_1.2,0.043628799915313723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,8,power_law_1.2,0.09292160272598267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,2,power_law_1.2,0.04791040122509003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,8,power_law_1.2,0.09155840277671815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,2,power_law_1.2,0.063264000415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,8,power_law_1.2,0.08979840278625488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,2,power_law_1.2,0.06611199975013733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,8,power_law_1.2,0.09685760140419006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,2,power_law_1.2,0.06883199810981751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,8,power_law_1.2,0.09751679897308349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,8,power_law_1.2,0.10696959495544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,8,power_law_1.2,0.10990719795227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,8,power_law_1.2,0.12037119865417481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,8,power_law_1.2,0.1324031949043274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,8,power_law_1.2,0.13999999761581422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,8,power_law_1.2,0.17543679475784302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,2,power_law_1.2,0.07171840071678162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,2,power_law_1.2,0.07181439995765686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,2,power_law_1.2,0.07427840232849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,2,power_law_1.2,0.07434880137443542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,2,power_law_1.2,0.07500159740447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,2,power_law_1.2,0.0779263973236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,2,power_law_1.2,0.08111360073089599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,2,power_law_1.2,0.0855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,2,power_law_1.2,0.09028480052947999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,8,power_law_1.2,0.1872063994407654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,2,power_law_1.2,0.16785279512405396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.2,0.19338879585266114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,8,power_law_1.2,0.23322880268096924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.2,0.14904320240020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.2,0.16073600053787232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,8,power_law_1.2,0.29386239051818847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.2,0.192576003074646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,8,power_law_1.2,0.38446080684661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,8,power_law_1.2,0.4976191997528076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.2,0.21935999393463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,8,power_law_1.2,0.7426047801971436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,8,power_law_1.2,0.920473575592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.2,0.29420158863067625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.2,0.3473855972290039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.2,0.507596778869629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,1,2,power_law_1.01,0.1850816011428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,1,2,power_law_1.01,0.21110401153564454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,1,2,power_law_1.01,0.22758400440216064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,1,2,power_law_1.01,0.3556992053985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,1,2,power_law_1.01,0.38576641082763674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,1,2,power_law_1.01,0.35545599460601807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,1,2,power_law_1.01,0.39950079917907716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,1,2,power_law_1.01,0.41294078826904296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,1,2,power_law_1.01,0.41809282302856443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,1,2,power_law_1.01,0.43761281967163085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,1,2,power_law_1.01,0.43242878913879396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,1,2,power_law_1.01,0.4776832103729248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,1,2,power_law_1.01,0.44501757621765137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,1,2,power_law_1.01,0.49407358169555665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,1,2,power_law_1.01,0.5337279796600342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,1,2,power_law_1.01,0.5519999980926513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,1,2,power_law_1.01,0.5467648029327392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,1,2,power_law_1.01,0.6636159896850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,1,2,power_law_1.01,0.7900479793548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,1,2,power_law_1.01,1.0354175567626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,1,2,power_law_1.01,1.0642751693725585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,1,2,power_law_1.01,1.7508415222167968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,1,2,power_law_1.01,1.9901119232177735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,1,2,power_law_1.01,2.9435775756835936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,1,2,power_law_1.01,4.526220703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,64,power_law_1.2,0.10399999618530273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,64,power_law_1.2,0.10280959606170655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,64,power_law_1.2,0.10210560560226441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,64,power_law_1.2,0.1034559965133667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,64,power_law_1.2,0.10604159832000733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,64,power_law_1.2,0.09992319941520691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,64,power_law_1.2,0.1043328046798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,1,8,balanced,0.029487999776999157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,1,8,balanced,0.02942933390537898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,64,power_law_1.2,0.10822399854660034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,1,8,balanced,0.030986666679382324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,1,8,balanced,0.035301332672437034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,64,power_law_1.2,0.10040960311889649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,64,power_law_1.2,0.10620160102844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,64,power_law_1.2,0.10680320262908935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,64,power_law_1.2,0.11183359622955322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,64,power_law_1.2,0.11287039518356323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,64,power_law_1.2,0.11584639549255371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,64,power_law_1.2,0.12178560495376586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,1,8,balanced,0.04960533479849497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,4,balanced,0.055871998270352684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,4,balanced,0.07428800066312154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,4,balanced,0.09913600484530131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,4,balanced,0.14933866262435913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,4,balanced,0.2476266622543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,4,balanced,0.44652799765268963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,4,balanced,0.6416853268941244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,4,balanced,0.6434826850891113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,4,balanced,0.6424213250478109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,4,balanced,0.6396533250808716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,1,8,balanced,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,64,power_law_1.2,0.12570879459381104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,64,power_law_1.2,0.14283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,64,power_law_1.2,0.16663039922714235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,64,power_law_1.2,0.1793984055519104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,64,power_law_1.2,0.24816639423370362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,4,balanced,0.6484906673431396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,64,power_law_1.2,0.22927360534667968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,1,8,balanced,0.05681600173314413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,1,8,balanced,0.05867200096448263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,1,8,balanced,0.05861866474151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,1,8,balanced,0.05783466498057047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,1,8,balanced,0.059903999169667564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,1,8,balanced,0.059631998340288796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,1,8,balanced,0.061946665247281395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,1,8,balanced,0.061674664417902626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,1,8,balanced,0.0653546651204427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,1,8,balanced,0.06817600131034851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,1,8,balanced,0.06855466465155284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,1,8,balanced,0.07467733323574066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,1,8,balanced,0.07332266867160797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,1,8,balanced,0.08685333530108134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,1,8,balanced,0.0885813335577647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,1,8,balanced,0.10950932900110881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,64,power_law_1.2,0.3429120063781738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,1,8,balanced,0.11762666702270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,1,8,balanced,0.1712053418159485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,64,power_law_1.2,0.34172160625457765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,64,power_law_1.2,0.45619840621948243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,64,power_law_1.2,0.5574912071228028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,4,balanced,0.6537173191706339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,64,power_law_1.2,0.806060791015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,1,8,balanced,0.18036266167958578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,1,8,balanced,0.2453599969546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,64,power_law_1.2,0.9116479873657226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,1,8,balanced,0.2941439946492513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,4,balanced,0.6541066567103068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,4,balanced,0.6517333189646403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,4,balanced,0.6627466678619385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,4,balanced,0.6680586338043213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,4,balanced,0.6748853524525961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,4,balanced,0.692202647527059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,4,balanced,0.7053706645965576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,4,balanced,0.7451466719309489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,4,balanced,0.7711679935455322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,4,balanced,0.8312373161315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,4,balanced,0.8944053649902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,4,balanced,0.999621311823527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,4,balanced,1.179807980855306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,4,balanced,1.3923039436340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,4,balanced,1.8531306584676106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.2,0.65065598487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,1,2,power_law_1.01,6.226796722412109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,1,2,power_law_1.01,8.251872253417968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,1,power_law_1.2,0.020729599893093108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,1,power_law_1.2,0.024396799504756927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,1,power_law_1.2,0.033657601475715636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,1,power_law_1.2,0.04389120042324066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,1,power_law_1.2,0.05861120223999024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,1,power_law_1.2,0.07839360237121581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,1,power_law_1.2,0.08604800105094909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,1,power_law_1.2,0.08906880021095276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,1,power_law_1.2,0.09561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,1,power_law_1.2,0.09963520169258118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,1,power_law_1.2,0.10204160213470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,1,power_law_1.2,0.10517760515213012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,1,power_law_1.2,0.11202559471130372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,16,balanced,0.04650666813055674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,1,power_law_1.2,0.10951039791107178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,16,balanced,0.046495998899141945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,1,power_law_1.2,0.11809920072555542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,16,balanced,0.05134400228659312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,1,power_law_1.2,0.11829760074615478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,1,power_law_1.2,0.1308608055114746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,1,power_law_1.2,0.15495680570602416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.2,0.18061439990997313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.2,0.23983359336853027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.2,0.22903680801391602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.2,0.29796481132507324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,16,balanced,0.07503466804822286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,16,balanced,0.09980799754460652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,16,balanced,0.15371200442314148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,16,balanced,0.2038080096244812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,16,balanced,0.20011732975641885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,16,balanced,0.2018293341000875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,16,balanced,0.20205867290496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,16,balanced,0.20746666193008423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,16,balanced,0.20945066213607788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,16,balanced,0.2129866679509481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,16,balanced,0.21066667636235556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,16,balanced,0.2188426653544108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.2,0.3560767889022827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.2,0.44112000465393064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.2,0.5394559860229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,16,balanced,0.22053333123524985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,16,balanced,0.2274186611175537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,16,balanced,0.2402986685434977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,16,balanced,0.2505066593488057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,16,balanced,0.27695999542872113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,16,balanced,0.29914132754007977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,16,balanced,0.34887464841206867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,16,balanced,0.3959093491236369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,16,balanced,0.48465601603190106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,16,balanced,0.6058773199717203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,16,balanced,0.783397356669108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,16,balanced,1.013493299484253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.2,0.732313585281372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.2,0.9251328468322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,8,1,balanced,0.0205226664741834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,8,1,balanced,0.020037333170572918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,8,1,balanced,0.02146133283774058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,8,1,balanced,0.025477332373460133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,8,1,balanced,0.0339626669883728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,8,1,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,8,1,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,8,1,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,8,1,balanced,0.04234666625658671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,8,1,balanced,0.04401599864164988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,8,1,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,8,1,balanced,0.048954665660858154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,8,1,balanced,0.051914667089780174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,8,1,balanced,0.05681600173314413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,8,1,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,8,1,balanced,0.058117335041364036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,8,1,balanced,0.06428266565004985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,8,1,balanced,0.07448533177375793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,8,1,balanced,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,8,1,balanced,0.1179039975007375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,8,1,balanced,0.14447466532389322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,8,1,balanced,0.14697600404421488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,8,1,balanced,0.15507200360298157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,8,1,balanced,0.246671994527181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,8,1,balanced,0.26532800992329914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,8,1,balanced,0.3815573453903198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,8,1,balanced,0.4928213357925415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,64,power_law_1.01,0.02961280047893524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,64,power_law_1.01,0.02885119915008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,64,power_law_1.01,0.028326401114463808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,64,power_law_1.01,0.03048959970474243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,64,power_law_1.01,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,64,power_law_1.01,0.03166080117225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,64,power_law_1.01,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,64,power_law_1.01,0.03265919983386993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,64,power_law_1.01,0.033062401413917544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,64,power_law_1.01,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,64,power_law_1.01,0.034220799803733826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,64,power_law_1.01,0.03546879887580871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,64,power_law_1.01,0.036345601081848145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,64,power_law_1.01,0.037536001205444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,64,power_law_1.01,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,64,power_law_1.01,0.04359039962291718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,64,power_law_1.01,0.04351359903812409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,64,power_law_1.01,0.0487744003534317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,64,power_law_1.01,0.05125120282173157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,64,power_law_1.01,0.061299198865890504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,64,power_law_1.01,0.06728960275650024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,64,power_law_1.01,0.09693440198898315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,64,power_law_1.01,0.1260543942451477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,64,power_law_1.01,0.17047040462493895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,64,power_law_1.01,0.22416000366210936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,64,power_law_1.01,0.272326397895813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,64,power_law_1.01,0.3710783958435059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,32,power_law_1.01,0.10017919540405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,32,power_law_1.01,0.0946943998336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,32,power_law_1.01,0.09178879857063293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,32,power_law_1.01,0.09493119716644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,32,power_law_1.01,0.08815360069274902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,32,power_law_1.01,0.08339200019836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,32,power_law_1.01,0.08977919816970825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,32,power_law_1.01,0.09467520117759705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,32,power_law_1.01,0.0923904001712799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,32,power_law_1.01,0.09427840113639832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,32,power_law_1.01,0.09470720291137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,32,power_law_1.01,0.09544320106506347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,4,power_law_1.2,0.05915520191192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,32,power_law_1.01,0.09515519738197327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,4,power_law_1.2,0.07645440101623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,2,balanced,0.04506133496761322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,2,balanced,0.04603200157483419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,2,balanced,0.04799999793370565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,2,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,2,balanced,0.07231999933719635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,2,balanced,0.09789866209030151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,2,balanced,0.09898133079210918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,2,balanced,0.09921600421269734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,2,balanced,0.10147733489672343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,2,balanced,0.10248000423113506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,2,balanced,0.10309867064158122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,2,balanced,0.10714667042096455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,2,balanced,0.10837866862614949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,2,balanced,0.11012267072995503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,2,balanced,0.117658664782842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,32,power_law_1.01,0.09685119986534119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,2,balanced,0.12138133247693379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,4,power_law_1.2,0.07544320225715637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,2,balanced,0.1283626655737559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,2,balanced,0.14729066689809164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,4,power_law_1.2,0.10654079914093018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,2,balanced,0.16031466921170553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,2,balanced,0.19323732455571493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,4,power_law_1.2,0.10960639715194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,2,balanced,0.21353065967559814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,2,balanced,0.3009653290112813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,4,power_law_1.2,0.13493119478225707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,2,balanced,0.3471519947052002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,4,power_law_1.2,0.16648319959640503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,4,power_law_1.2,0.1629439949989319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,32,power_law_1.01,0.1037824034690857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,4,power_law_1.2,0.1601472020149231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,32,power_law_1.01,0.09846400022506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,4,power_law_1.2,0.16042879819869996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,32,power_law_1.01,0.1046720027923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,4,power_law_1.2,0.17046400308609008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,4,power_law_1.2,0.1700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,2,balanced,0.504319985707601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,4,power_law_1.2,0.1678272008895874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,2,balanced,0.6270720163981119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,4,power_law_1.2,0.1812608003616333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,2,balanced,0.9190560181935629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,4,power_law_1.2,0.19411840438842773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,4,power_law_1.2,0.20138239860534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,2,balanced,1.2075093587239583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,4,power_law_1.2,0.20117759704589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,32,power_law_1.01,0.11650559902191163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,4,power_law_1.2,0.2324608087539673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.01,0.12700159549713136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,4,power_law_1.2,0.23958399295806884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.01,0.1446720004081726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,4,power_law_1.2,0.28808960914611814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,4,power_law_1.2,0.3173311948776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,4,power_law_1.2,0.373036789894104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,4,power_law_1.2,0.462662410736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.01,0.16864000558853148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.01,0.21901440620422363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.01,0.24308478832244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.01,0.3263295888900757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,4,power_law_1.2,0.5977536201477051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.01,0.4257343769073486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,4,power_law_1.2,0.7114048004150391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.01,0.5943935871124267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,4,power_law_1.2,1.0088000297546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.01,0.7613696098327637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,4,power_law_1.2,1.3153471946716309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,8,power_law_1.2,0.08565760254859925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,8,power_law_1.2,0.1244096040725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,8,power_law_1.2,0.12124160528182984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,8,power_law_1.2,0.16597119569778443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,8,power_law_1.2,0.20087039470672607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,8,power_law_1.2,0.15944960117340087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,8,power_law_1.2,0.18815360069274903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,8,power_law_1.2,0.19071359634399415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,8,power_law_1.2,0.19199999570846557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,8,power_law_1.2,0.1942080020904541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,8,power_law_1.2,0.1869503974914551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,8,power_law_1.2,0.19359999895095825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,8,power_law_1.2,0.21050240993499755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,8,power_law_1.2,0.2048896074295044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,8,power_law_1.2,0.22325758934020995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,8,power_law_1.2,0.22351999282836915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,8,power_law_1.2,0.23797121047973632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,8,power_law_1.2,0.2680191993713379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,8,power_law_1.2,0.2788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,8,power_law_1.2,0.3276416063308716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,8,power_law_1.2,0.34223361015319825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,8,power_law_1.2,0.4292031764984131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,8,power_law_1.2,0.4780223846435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,8,power_law_1.2,0.6013440132141114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,8,power_law_1.2,0.7739840030670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,8,power_law_1.2,1.0507328033447265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,8,power_law_1.2,1.2936960220336915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,8,power_law_1.2,0.06867200136184692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,8,power_law_1.2,0.08579840064048767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,8,power_law_1.2,0.08087679743766785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,8,power_law_1.2,0.09434880018234253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,8,power_law_1.2,0.1007040023803711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,8,power_law_1.2,0.0898688018321991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,8,power_law_1.2,0.10632959604263306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,8,power_law_1.2,0.10613759756088256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,8,power_law_1.2,0.1039423942565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,8,power_law_1.2,0.1045375943183899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,8,power_law_1.2,0.10298880338668823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,8,power_law_1.2,0.10944639444351197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,8,power_law_1.2,0.11067520380020142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,8,power_law_1.2,0.1107200026512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,8,power_law_1.2,0.12144639492034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,8,power_law_1.2,0.12462719678878784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,8,power_law_1.2,0.13009920120239257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,8,power_law_1.2,0.14355839490890504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,8,power_law_1.2,0.1613759994506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,8,power_law_1.2,0.18609280586242677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,8,power_law_1.2,0.2237312078475952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,8,power_law_1.2,0.3246592044830322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,8,power_law_1.2,0.40137600898742676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,8,power_law_1.2,0.5550015926361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,8,power_law_1.2,0.7577407836914063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,8,power_law_1.2,1.1461440086364747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,8,power_law_1.2,1.4294848442077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,2,power_law_1.01,0.048153600096702574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,2,power_law_1.01,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,2,power_law_1.01,0.07390080094337463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,2,power_law_1.01,0.09455999732017517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,2,power_law_1.01,0.09856640100479126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,2,power_law_1.01,0.13697919845581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,2,power_law_1.01,0.1430400013923645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,2,power_law_1.01,0.14940799474716188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,2,power_law_1.01,0.1527168035507202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,2,power_law_1.01,0.1525056004524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,2,power_law_1.01,0.15738879442214965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,2,power_law_1.01,0.1625856041908264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,2,power_law_1.01,0.1647744059562683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,2,power_law_1.01,0.1705471992492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,2,power_law_1.01,0.18307199478149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,2,power_law_1.01,0.1893952012062073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,2,power_law_1.01,0.20195200443267822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,2,power_law_1.01,0.22174079418182374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.01,0.25541760921478274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.01,0.2978431940078735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.01,0.34760959148406984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.01,0.44839677810668943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.01,0.537011194229126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.01,0.7469759941101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.01,0.9692031860351562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.01,1.3835391998291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.01,1.6802623748779297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,32,power_law_1.2,0.0808896005153656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,32,power_law_1.2,0.08013439774513245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,32,power_law_1.2,0.07366399765014649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,32,power_law_1.2,0.08071680068969726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,32,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,32,power_law_1.2,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,32,power_law_1.2,0.07761279940605163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,32,power_law_1.2,0.07976959943771363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,32,power_law_1.2,0.08532480001449586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,32,power_law_1.2,0.08315520286560059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,32,power_law_1.2,0.08449919819831848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,32,power_law_1.2,0.08540160059928895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,32,power_law_1.2,0.08556159734725952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,32,power_law_1.2,0.08755840063095092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,32,power_law_1.2,0.09594240188598632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,32,power_law_1.2,0.09756159782409668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,32,power_law_1.2,0.1020799994468689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,32,power_law_1.2,0.11964800357818603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,32,power_law_1.2,0.1331264019012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,32,power_law_1.2,0.15422719717025757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,32,power_law_1.2,0.17909760475158693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,32,power_law_1.2,0.22960638999938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,32,power_law_1.2,0.29322879314422606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,32,power_law_1.2,0.39763200283050537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,32,power_law_1.2,0.5274816036224366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,32,power_law_1.2,0.7531904220581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,32,power_law_1.2,0.9746687889099122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,1,power_law_1.01,0.020934399962425233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,1,power_law_1.01,0.02958720028400421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,1,power_law_1.01,0.043756800889968875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,1,power_law_1.01,0.05472639799118042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,1,power_law_1.01,0.07061759829521179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,1,power_law_1.01,0.09587200284004212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,1,power_law_1.01,0.10234880447387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,1,power_law_1.01,0.105075204372406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,1,power_law_1.01,0.11111040115356445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,1,power_law_1.01,0.11608959436416626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,1,power_law_1.01,0.1122048020362854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,1,power_law_1.01,0.11460479497909545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,1,power_law_1.01,0.121343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,1,power_law_1.01,0.11596800088882446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,1,power_law_1.01,0.12205439805984497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,1,power_law_1.01,0.12684799432754518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,1,power_law_1.01,0.14821120500564575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,1,power_law_1.01,0.17495039701461793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.01,0.17895679473876952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.01,0.23313279151916505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.01,0.2203903913497925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.01,0.28178560733795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.01,0.30166399478912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.01,0.373036789894104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.01,0.46082558631896975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.01,0.6319104194641113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.01,0.7999360084533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,1,32,balanced,0.05942399799823761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,1,32,balanced,0.0587360014518102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,1,32,balanced,0.060005332032839455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,1,32,balanced,0.07186666627724965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,1,32,balanced,0.08481599887212117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,1,32,balanced,0.09295466542243958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,1,32,balanced,0.09428266684214275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,1,32,balanced,0.09294399619102478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,1,32,balanced,0.09479467074076335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,1,32,balanced,0.09404266873995464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,1,32,balanced,0.09550399581591289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,1,32,balanced,0.09691733121871948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,1,32,balanced,0.09773332873980205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,1,32,balanced,0.09930666287740071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,1,32,balanced,0.10779199997584026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,1,32,balanced,0.10824533303578694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,1,32,balanced,0.11036266883214314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,1,32,balanced,0.11760000387827556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,1,32,balanced,0.12121066451072693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,1,32,balanced,0.13246933619181314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,1,32,balanced,0.14617066582043967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,1,32,balanced,0.19322667519251505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,1,32,balanced,0.21407999595006308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,1,32,balanced,0.28036266565322876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,1,32,balanced,0.349237322807312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,1,32,balanced,0.4477599859237671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,1,32,balanced,0.5921866496404012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,8,balanced,0.02346666653951009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,8,balanced,0.023215999205907185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,8,balanced,0.025173333783944447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,8,balanced,0.02510933329661687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,8,balanced,0.027290667096773785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,8,balanced,0.03134933362404505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,8,balanced,0.03126399964094162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,8,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,8,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,8,balanced,0.03503466645876566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,8,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,8,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,8,balanced,0.04567466676235199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,8,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,8,balanced,0.07196799914042155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,8,1,balanced,0.03857066730658213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,8,1,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,8,1,balanced,0.06278400123119354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,8,1,balanced,0.06497066716353099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,8,1,balanced,0.0666186660528183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,8,1,balanced,0.06693333387374878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,8,1,balanced,0.06689066688219707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,8,balanced,0.07425599793593089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,8,balanced,0.07890666524569194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,8,balanced,0.08713600039482117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,8,balanced,0.09175466497739156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,8,balanced,0.1053013304869334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,8,balanced,0.12004799644152324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,8,balanced,0.1575040022532145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,8,balanced,0.18037867546081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,8,balanced,0.23147199551264444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,8,balanced,0.29363733530044556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,8,1,balanced,0.06830933193365733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,8,1,balanced,0.06834133466084798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,8,1,balanced,0.06839466591676076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,8,1,balanced,0.06946133573849995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,8,1,balanced,0.07150400181611379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,8,1,balanced,0.0747680018345515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,8,1,balanced,0.08057599763075511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,8,1,balanced,0.08559466401735942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,8,1,balanced,0.08763200044631958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,8,1,balanced,0.09282132983207703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,8,1,balanced,0.11352533102035522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,8,1,balanced,0.12210133671760559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,8,1,balanced,0.1586186687151591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,8,1,balanced,0.18316266934076944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,8,1,balanced,0.24516799052556357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,8,balanced,0.42950932184855145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,8,1,balanced,0.297541340192159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,8,balanced,0.5429066816965739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,8,1,balanced,0.4238933324813843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,8,1,balanced,0.5511786540349325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,8,1,balanced,0.7863732973734537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,8,1,balanced,1.0265226364135742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,4,balanced,0.058933332562446594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,4,balanced,0.06282133360703786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,4,balanced,0.06666133304437001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,4,balanced,0.07916800181070964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,4,balanced,0.0976693332195282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,4,balanced,0.14628799756368002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,4,balanced,0.14591466387112936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,4,balanced,0.14286399881045023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,4,balanced,0.14789332946141562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,4,balanced,0.14387733737627664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,4,balanced,0.1420906682809194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,4,balanced,0.1406773328781128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,4,balanced,0.1458346645037333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,4,balanced,0.14737600088119507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,4,balanced,0.15310399731000265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,4,balanced,0.15215999881426492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,4,balanced,0.1567573348681132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,4,balanced,0.1702880064646403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,4,balanced,0.18110400438308716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,4,balanced,0.2060746749242147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,4,balanced,0.22755199670791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,4,balanced,0.2761920094490051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,4,balanced,0.32576533158620197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,4,balanced,0.47649065653483075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,4,balanced,0.5798720121383667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,2,power_law_1.01,0.016633599996566772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,4,balanced,0.8163626988728842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,2,power_law_1.01,0.01727360039949417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,2,power_law_1.01,0.018137599527835845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,4,balanced,1.0609333515167236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,2,power_law_1.01,0.020160000026226043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,2,power_law_1.01,0.020876799523830415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,2,power_law_1.01,0.023948800563812257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,2,power_law_1.01,0.024665600061416625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,2,power_law_1.01,0.024748800694942473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,2,power_law_1.01,0.026707199215888978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,2,power_law_1.01,0.026796799898147584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,2,power_law_1.01,0.028217598795890808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,2,power_law_1.01,0.027475199103355406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,2,power_law_1.01,0.02805759906768799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,2,power_law_1.01,0.031046399474143983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,2,power_law_1.01,0.034867200255393985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,2,power_law_1.01,0.035334399342536925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,2,power_law_1.01,0.037196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,2,power_law_1.01,0.05902079939842224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,2,power_law_1.01,0.06220800280570984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,2,power_law_1.01,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,2,power_law_1.01,0.07170559763908387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,2,power_law_1.01,0.08234879970550538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,2,power_law_1.01,0.09899520277976989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,2,power_law_1.01,0.1235967993736267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,2,power_law_1.01,0.15264639854431153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,2,power_law_1.01,0.20430080890655516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,2,power_law_1.01,0.2641151905059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,1,2,power_law_1.01,0.05061759948730469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,1,2,power_law_1.01,0.061849600076675414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,1,2,power_law_1.01,0.06865280270576476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,1,2,power_law_1.01,0.07735679745674133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,1,2,power_law_1.01,0.08540160059928895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,1,2,power_law_1.01,0.10228480100631714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,1,2,power_law_1.01,0.10531840324401856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,1,2,power_law_1.01,0.10635520219802856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,1,2,power_law_1.01,0.10883200168609619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,1,2,power_law_1.01,0.11121920347213746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,1,2,power_law_1.01,0.11111040115356445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,1,2,power_law_1.01,0.11546239852905274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,1,2,power_law_1.01,0.11564799547195434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,1,2,power_law_1.01,0.11935360431671142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,1,2,power_law_1.01,0.12008320093154908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,1,2,power_law_1.01,0.12155519723892212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,1,2,power_law_1.01,0.125164794921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,1,2,power_law_1.01,0.13742079734802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,1,2,power_law_1.01,0.15235840082168578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,1,2,power_law_1.01,0.18101119995117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,1,2,power_law_1.01,0.21116158962249756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,1,2,power_law_1.01,0.27601919174194334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,1,2,power_law_1.01,0.3356031894683838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,8,balanced,0.034671999514102936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,8,balanced,0.03550933301448822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,1,2,power_law_1.01,0.5070208072662353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,8,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,8,balanced,0.047781333327293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,8,balanced,0.06713066498438518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,8,balanced,0.06855999926726024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,1,2,power_law_1.01,0.6344128131866456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,8,balanced,0.06807999809583028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,8,balanced,0.06947200000286102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,8,balanced,0.07048533360163371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,8,balanced,0.0691786656777064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,1,2,power_law_1.01,0.95797119140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,8,balanced,0.07019733389218648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,8,balanced,0.07017066578070323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,8,balanced,0.07333866755167644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,8,balanced,0.0745119998852412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,8,balanced,0.078015998005867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,1,2,power_law_1.01,1.209164810180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,8,balanced,0.08158933122952779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,8,balanced,0.08303999900817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,8,balanced,0.08833600083986919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,8,balanced,0.1025279959042867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,8,balanced,0.1132319966952006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,8,balanced,0.11724266409873962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,8,balanced,0.14600533246994019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,8,balanced,0.16673066218694052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,8,balanced,0.21975467602411905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,8,balanced,0.26868265867233276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,8,balanced,0.3632426659266154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,8,balanced,0.46566931406656903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,4,1,power_law_1.2,0.08240000009536744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,4,1,power_law_1.2,0.09238399863243103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,4,1,power_law_1.2,0.11375999450683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,4,1,power_law_1.2,0.15410560369491577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,4,1,power_law_1.2,0.1928063988685608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,4,1,power_law_1.2,0.23647360801696776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,4,1,power_law_1.2,0.3213567972183228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,4,1,power_law_1.2,0.34175360202789307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,4,1,power_law_1.2,0.34840960502624513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,4,1,power_law_1.2,0.3603071928024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,4,1,power_law_1.2,0.3734976053237915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,4,1,power_law_1.2,0.38893439769744875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,4,1,power_law_1.2,0.40071678161621094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,4,1,power_law_1.2,0.40076799392700196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,4,1,power_law_1.2,0.415231990814209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,4,1,power_law_1.2,0.42835202217102053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,4,1,power_law_1.2,0.4559487819671631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,4,1,power_law_1.2,0.516102409362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.2,0.5767744064331055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.2,0.6772096157073975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.2,0.749721622467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.2,0.9410623550415039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.2,1.133843231201172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.2,1.5329279899597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.2,1.9716863632202148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.2,2.732204818725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,1,balanced,0.029274667302767437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,1,balanced,0.038986665507157646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,1,balanced,0.052255998055140175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,1,balanced,0.07386666536331177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,1,balanced,0.12057066957155864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,1,balanced,0.12583466370900473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.2,3.642259216308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,1,balanced,0.1304586629072825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,1,balanced,0.1362986663977305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,1,balanced,0.14310399691263834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,1,balanced,0.14822399616241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,1,balanced,0.1628106633822123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,1,balanced,0.14668266971906027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,1,balanced,0.15334399541219076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,1,balanced,0.1650826632976532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,1,balanced,0.15963733196258545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,1,balanced,0.16061333815256754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,1,balanced,0.17627199490865073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,1,balanced,0.19036267201105753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,1,balanced,0.19723200798034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,1,balanced,0.41551466782887775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,1,balanced,0.47832000255584717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,1,balanced,0.33664532502492267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,1,balanced,0.3445440133412679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,1,balanced,0.6138399839401245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,8,32,power_law_1.01,0.07164160013198853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,1,balanced,0.6028159856796265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,8,32,power_law_1.01,0.07528319954872131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,1,balanced,0.8683626651763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,8,32,power_law_1.01,0.07146880030632019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,1,balanced,1.1384159723917644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,8,32,power_law_1.01,0.07155839800834655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,8,32,power_law_1.01,0.07158399820327759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,8,32,power_law_1.01,0.0726207971572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,8,32,power_law_1.01,0.06932479739189149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,8,32,power_law_1.01,0.07392640113830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,8,32,power_law_1.01,0.07123839855194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,8,32,power_law_1.01,0.07236480116844177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,8,32,power_law_1.01,0.07642239928245545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,8,32,power_law_1.01,0.07459840178489685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,8,32,power_law_1.01,0.07710719704627991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,8,32,power_law_1.01,0.078847998380661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,8,32,power_law_1.01,0.08250880241394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,8,32,power_law_1.01,0.0834496021270752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,8,32,power_law_1.01,0.09148160219192505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,8,32,power_law_1.01,0.10227839946746826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,8,32,power_law_1.01,0.11039999723434449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,8,32,power_law_1.01,0.13216639757156373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,8,32,power_law_1.01,0.15315200090408326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,8,32,power_law_1.01,0.19430400133132936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,8,32,power_law_1.01,0.23546879291534423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,8,32,power_law_1.01,0.31867520809173583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,8,32,power_law_1.01,0.40522241592407227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,8,32,power_law_1.01,0.5710015773773194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,8,32,power_law_1.01,0.752627182006836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,16,balanced,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,16,balanced,0.052042668064435325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,16,balanced,0.05221333106358846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,16,balanced,0.0561653325955073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,16,balanced,0.060266668597857155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,16,balanced,0.08050133287906647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,16,balanced,0.08072533210118611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,16,balanced,0.07976533472537994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,16,balanced,0.08178133269151051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,16,balanced,0.08078933258851369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,16,balanced,0.08216533561547597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,16,balanced,0.08447466293970744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,16,balanced,0.08433600266774495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,16,balanced,0.08618666728337605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,32,4,balanced,0.044079999128977455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,32,4,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,32,4,balanced,0.046469335754712425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,32,4,balanced,0.053216000398000084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,32,4,balanced,0.0580320010582606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,32,4,balanced,0.07456533114115398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,32,4,balanced,0.07599466542402904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,32,4,balanced,0.07456533114115398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,16,balanced,0.0997279981772105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,16,balanced,0.09919466574986775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,16,balanced,0.10332799951235454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,16,balanced,0.11795733372370402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,16,balanced,0.11788800358772278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,16,balanced,0.13691199819246927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,16,balanced,0.15169599652290344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,16,balanced,0.18329066038131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,16,balanced,0.21386132637659708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,16,balanced,0.2876426577568054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,32,4,balanced,0.07422933479150136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,32,4,balanced,0.07460266848405202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,16,balanced,0.3552159865697225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,32,4,balanced,0.07438399891058604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,32,4,balanced,0.078015998005867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,32,4,balanced,0.08086933195590973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,16,balanced,0.506389339764913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,32,4,balanced,0.08243733147780101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,32,4,balanced,0.09341866771380107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,32,4,balanced,0.09641599655151367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,16,balanced,0.6670880317687988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,32,4,balanced,0.10593066612879436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,32,4,balanced,0.12346133589744568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,32,4,balanced,0.13472533226013184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,32,4,balanced,0.16211199760437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,32,4,balanced,0.18657066424687704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,32,4,balanced,0.2571199933687846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,32,4,balanced,0.302837332089742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,32,4,balanced,0.43980268637339276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,32,4,balanced,0.5550453265508016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,32,4,balanced,0.8063733577728271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,32,4,balanced,1.0667306582132976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,16,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,16,balanced,0.04380266865094503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,16,balanced,0.043605332573254905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,16,balanced,0.04897066454092661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,16,balanced,0.050026665131251015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,16,balanced,0.05789866546789805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,16,balanced,0.07250133156776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,16,balanced,0.07376533250013988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,16,balanced,0.07445866862932841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,16,balanced,0.07524266839027405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,16,balanced,0.075914666056633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,16,balanced,0.0776693324247996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,16,balanced,0.07858666777610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,16,balanced,0.08070933322111766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,16,balanced,0.08513599634170532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,16,balanced,0.08674133817354839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,16,balanced,0.09098133444786072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,16,balanced,0.09990933537483215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,16,balanced,0.10642133156458537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,16,balanced,0.12174933155377705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,16,balanced,0.13424533605575562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,16,balanced,0.16262400150299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,16,balanced,0.20368534326553345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,16,balanced,0.26289065678914386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,16,balanced,0.3258240024248759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,16,balanced,0.4594293435414632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,16,balanced,0.5750453472137451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,4,power_law_1.01,0.031193599104881287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,4,power_law_1.01,0.03133440017700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,4,power_law_1.01,0.030752000212669373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,4,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,4,power_law_1.01,0.03208959996700287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,4,power_law_1.01,0.03411200046539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,4,power_law_1.01,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,4,power_law_1.01,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,4,power_law_1.01,0.03699199855327606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,4,power_law_1.01,0.03859840035438537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,4,balanced,0.05029866596062978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,4,power_law_1.01,0.03701759874820709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,4,balanced,0.06817066669464111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,4,power_law_1.01,0.039654400944709775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,4,power_law_1.01,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,4,power_law_1.01,0.044633600115776065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,4,power_law_1.01,0.05196800231933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,4,power_law_1.01,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,4,balanced,0.09325866897900899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,4,balanced,0.15039466818173727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,4,balanced,0.24527466297149658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,4,balanced,0.440778652826945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,4,balanced,0.4434880018234253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,4,power_law_1.01,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,4,power_law_1.01,0.06897280216217042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,4,balanced,0.4452373186747233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,4,power_law_1.01,0.08325120210647582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,4,balanced,0.4455360174179077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,4,balanced,0.44574932257334393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,4,balanced,0.45081067085266113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,4,balanced,0.45133864879608154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,4,power_law_1.01,0.10252159833908081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,4,balanced,0.4529759883880615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,4,balanced,0.45850133895874023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,4,balanced,0.46275198459625244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,4,power_law_1.01,0.12062720060348511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,4,balanced,0.46698665618896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,4,balanced,0.4729439814885457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,4,balanced,0.48977065086364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,4,balanced,0.5083093245824178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,4,balanced,0.5414506594340006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,4,balanced,0.564026673634847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,4,balanced,0.6106986602147421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,4,balanced,0.7076693375905355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,4,balanced,0.8031040032704672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,4,balanced,0.8873226642608643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,4,balanced,1.3286186854044597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,4,balanced,1.4293120702107747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,4,power_law_1.01,0.14561920166015624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,4,power_law_1.01,0.1804159998893738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,4,power_law_1.01,0.23627519607543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,4,power_law_1.01,0.3689984083175659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,4,power_law_1.01,0.45631999969482423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,4,power_law_1.01,0.6096384048461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,1,power_law_1.2,0.04615679979324341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,1,power_law_1.2,0.04544639885425568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,1,power_law_1.2,0.04609920084476471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,1,power_law_1.2,0.04785279929637909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,1,power_law_1.2,0.04870400130748749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,1,power_law_1.2,0.06077439785003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,1,power_law_1.2,0.07457280158996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,1,power_law_1.2,0.06680960059165955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,1,power_law_1.2,0.08788480162620545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,1,power_law_1.2,0.07985919713973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,1,power_law_1.2,0.10478719472885131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,1,power_law_1.2,0.1201024055480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,1,power_law_1.2,0.1358464002609253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,1,power_law_1.2,0.10103039741516114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,1,power_law_1.2,0.1427135944366455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,1,power_law_1.2,0.12410880327224731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,1,power_law_1.2,0.16263680458068847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,1,2,power_law_1.2,0.14810240268707275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,1,2,power_law_1.2,0.18677760362625123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,1,2,power_law_1.2,0.2577023983001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,1,2,power_law_1.2,0.34167680740356443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,1,2,power_law_1.2,0.3801664113998413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,1,power_law_1.2,0.2113408088684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,1,2,power_law_1.2,0.5750656127929688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,1,2,power_law_1.2,0.6061439990997315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,1,2,power_law_1.2,0.6021567821502686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,1,2,power_law_1.2,0.6259391784667969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,1,2,power_law_1.2,0.6364672183990479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.2,0.2732863903045654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,1,2,power_law_1.2,0.6537471771240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,1,2,power_law_1.2,0.6742015838623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,1,2,power_law_1.2,0.6623871803283692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,1,2,power_law_1.2,0.6517312049865722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,2,16,power_law_1.2,0.0927232027053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,2,16,power_law_1.2,0.14108799695968627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,2,16,power_law_1.2,0.13246079683303832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,2,16,power_law_1.2,0.14342399835586547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,2,16,power_law_1.2,0.1453503966331482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,2,16,power_law_1.2,0.10597120523452759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,1,2,power_law_1.2,0.691648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,2,16,power_law_1.2,0.1398144006729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,1,2,power_law_1.2,0.7203968048095704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,2,16,power_law_1.2,0.1291584014892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,1,2,power_law_1.2,0.732428789138794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,1,2,power_law_1.2,0.7647552013397216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.2,0.3880320072174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.2,0.8431360244750976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.2,0.943295955657959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.2,1.1057087898254394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.2,1.3426624298095704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.2,1.5824319839477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,2,16,power_law_1.2,0.13778560161590575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.2,2.0200639724731446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,2,16,power_law_1.2,0.13912960290908813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,2,16,power_law_1.2,0.14816639423370362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,2,16,power_law_1.2,0.15367679595947265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,2,16,power_law_1.2,0.15306880474090576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,2,16,power_law_1.2,0.14307199716567992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,2,16,power_law_1.2,0.1510655999183655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,2,16,power_law_1.2,0.15475200414657592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,2,16,power_law_1.2,0.17088639736175537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.2,0.5199359893798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,2,16,power_law_1.2,0.19678080081939697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,2,16,power_law_1.2,0.1978495955467224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,32,8,balanced,0.02958933264017105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,32,8,balanced,0.027615999182065327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,2,16,power_law_1.2,0.24677760601043702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,32,8,balanced,0.029279999434947968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,32,8,balanced,0.031632001201311745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,32,8,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,32,8,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,32,8,balanced,0.033717334270477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,32,8,balanced,0.033514666060606636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,32,8,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,32,8,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,32,8,balanced,0.03562133262554804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,32,8,balanced,0.03571200122435888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,32,8,balanced,0.03562133262554804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,32,8,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,32,8,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,32,8,balanced,0.0397173340121905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,32,8,balanced,0.043765331308046974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.2,2.550815963745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,32,8,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,32,8,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,32,8,balanced,0.05815466741720835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,32,8,balanced,0.06224533418814341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,32,8,balanced,0.08259200056393941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,32,8,balanced,0.09770666559537251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,32,8,balanced,0.13085333506266275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,32,8,balanced,0.15416000286738077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,32,8,balanced,0.2206559975941976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,32,8,balanced,0.330512007077535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.2,3.6204288482666014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,2,16,power_law_1.2,0.266483211517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,2,16,power_law_1.2,0.35479040145874025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,2,16,power_law_1.2,0.38973441123962405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,2,16,power_law_1.2,0.5297664165496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.2,4.2100990295410154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,2,16,power_law_1.2,0.5956223964691162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,2,16,power_law_1.2,0.8669631958007813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,2,16,power_law_1.2,1.0423232078552247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.2,0.7421631813049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.2,0.989247989654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.2,1.4562687873840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,4,power_law_1.01,0.099481600522995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,4,power_law_1.01,0.13879679441452025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,4,power_law_1.01,0.14030079841613768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,4,power_law_1.01,0.18622080087661744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,2,power_law_1.2,0.025721600651741026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,4,power_law_1.01,0.17951359748840331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,2,power_law_1.2,0.037990400195121767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,4,power_law_1.01,0.2558527946472168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,4,power_law_1.01,0.2790656089782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,2,power_law_1.2,0.04847359955310822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,2,power_law_1.2,0.06286720037460328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,4,power_law_1.01,0.28286080360412597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,2,power_law_1.2,0.07640320062637329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,2,power_law_1.2,0.09311360120773315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,2,power_law_1.2,0.1310528039932251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,2,power_law_1.2,0.13354239463806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,2,power_law_1.2,0.1358464002609253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,2,power_law_1.2,0.13752319812774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,4,power_law_1.01,0.30354559421539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,2,power_law_1.2,0.1435647964477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,4,power_law_1.01,0.28660480976104735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,2,power_law_1.2,0.14574079513549804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,2,power_law_1.2,0.15063040256500243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,4,power_law_1.01,0.29949440956115725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,2,power_law_1.2,0.1585536003112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,4,power_law_1.01,0.310534405708313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,4,power_law_1.01,0.3240384101867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,4,power_law_1.01,0.32454400062561034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,4,power_law_1.01,0.3388927936553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,4,power_law_1.01,0.35004799365997313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,2,power_law_1.2,0.16135679483413695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,4,power_law_1.01,0.3662847995758057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,2,power_law_1.2,0.16138240098953247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,4,power_law_1.01,0.4240575790405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,2,power_law_1.2,0.17459839582443237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,2,power_law_1.2,0.19721599817276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,2,power_law_1.2,0.20760319232940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,2,power_law_1.2,0.4386752128601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,2,power_law_1.2,0.4955776214599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,4,power_law_1.01,0.41989760398864745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,2,power_law_1.2,0.35018880367279054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,4,power_law_1.01,0.49324798583984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,2,power_law_1.2,0.37587199211120603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,4,power_law_1.01,0.5635519981384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,4,power_law_1.01,0.678495979309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,4,power_law_1.01,0.7750400066375732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,4,power_law_1.01,1.017024040222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,2,power_law_1.2,0.44771838188171387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,4,power_law_1.01,1.3833151817321778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,2,power_law_1.2,0.5127423763275146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,2,power_law_1.2,0.6802752017974854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,4,power_law_1.01,1.7407936096191405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,2,power_law_1.2,0.871174430847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,4,power_law_1.01,2.1524480819702148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.2,1.9172479629516601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.2,2.871001625061035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,1,power_law_1.01,0.04991360008716583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,1,power_law_1.01,0.052179199457168576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,1,power_law_1.01,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,1,power_law_1.01,0.07134079933166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,1,power_law_1.01,0.08730880022048951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,1,power_law_1.01,0.10231679677963257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,1,power_law_1.01,0.11786880493164062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,1,power_law_1.01,0.12410880327224731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,1,power_law_1.01,0.1283136010169983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,1,power_law_1.01,0.13111679553985595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,1,power_law_1.01,0.1333631992340088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,1,power_law_1.01,0.14060159921646118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,1,power_law_1.01,0.14247679710388184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,1,power_law_1.01,0.14909440279006958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,1,power_law_1.01,0.15337599515914918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,1,power_law_1.01,0.1576640009880066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,1,power_law_1.01,0.17868800163269044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,1,power_law_1.01,0.21288321018218995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,1,power_law_1.01,0.2319103956222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,1,power_law_1.01,0.2934079885482788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,1,power_law_1.01,0.3317951917648315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,1,power_law_1.01,0.43865599632263186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,1,power_law_1.01,0.5442368030548096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,1,power_law_1.01,0.7425983905792236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,1,power_law_1.01,0.9702079772949219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,1,power_law_1.01,1.3599231719970704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,1,power_law_1.01,1.7425792694091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.2,3.8001472473144533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,8,32,power_law_1.2,0.06807039976119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,8,32,power_law_1.2,0.06917759776115417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,8,32,power_law_1.2,0.06645119786262513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,8,32,power_law_1.2,0.06651520133018493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,8,32,power_law_1.2,0.0668287992477417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,8,32,power_law_1.2,0.06814720034599304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,8,32,power_law_1.2,0.06673920154571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,8,32,power_law_1.2,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,8,32,power_law_1.2,0.069651198387146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,8,32,power_law_1.2,0.07056639790534973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,8,32,power_law_1.2,0.07217280268669128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,8,32,power_law_1.2,0.07173759937286377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,8,32,power_law_1.2,0.07406719923019409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,8,32,power_law_1.2,0.07781760096549988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,8,32,power_law_1.2,0.08206719756126404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,8,32,power_law_1.2,0.08566399812698364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,8,32,power_law_1.2,0.0920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,8,32,power_law_1.2,0.10645120143890381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,8,32,power_law_1.2,0.12038400173187255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,8,32,power_law_1.2,0.15066239833831788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,8,32,power_law_1.2,0.17820160388946532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,8,32,power_law_1.2,0.2385727882385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,8,32,power_law_1.2,0.27406721115112304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,8,32,power_law_1.2,0.4005631923675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,8,32,power_law_1.2,0.5103487968444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,8,32,power_law_1.2,0.7565311908721923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,8,32,power_law_1.2,0.9451775550842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,2,power_law_1.01,0.0401856005191803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,2,power_law_1.01,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,2,power_law_1.01,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,2,power_law_1.01,0.04812160134315491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,2,power_law_1.01,0.047686401009559634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,2,power_law_1.01,0.04808320105075836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,2,power_law_1.01,0.04522239863872528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,2,power_law_1.01,0.04599039852619171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,2,power_law_1.01,0.04687359929084778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,2,power_law_1.01,0.04752640128135681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,2,power_law_1.01,0.04692479968070984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,2,power_law_1.01,0.04833920001983642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,2,power_law_1.01,0.048825600743293764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,2,power_law_1.01,0.05082880258560181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,2,power_law_1.01,0.056569600105285646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,2,power_law_1.01,0.05864959955215454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,2,power_law_1.01,0.06227200031280518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,2,power_law_1.01,0.06602240204811097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.01,0.07887359857559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.01,0.09300479888916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,16,4,power_law_1.01,0.031174400448799135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,16,4,power_law_1.01,0.029151999950408937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,16,4,power_law_1.01,0.030156800150871278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,16,4,power_law_1.01,0.030976000428199767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.01,0.1124608039855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,16,4,power_law_1.01,0.03164800107479095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,16,4,power_law_1.01,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,16,4,power_law_1.01,0.033292800188064575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,16,4,power_law_1.01,0.03383040130138397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,16,4,power_law_1.01,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,16,4,power_law_1.01,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.01,0.14600319862365724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,16,4,power_law_1.01,0.03589119911193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,16,4,power_law_1.01,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,16,4,power_law_1.01,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,8,2,balanced,0.1036906639734904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,8,2,balanced,0.11157866319020589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,8,2,balanced,0.12380799651145935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,8,2,balanced,0.14518933494885763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,8,2,balanced,0.18416533867518106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,8,2,balanced,0.24875199794769287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,8,2,balanced,0.24888000885645548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,8,2,balanced,0.24892799059549967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,8,2,balanced,0.2524533271789551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,8,2,balanced,0.25250667333602905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,16,4,power_law_1.01,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,16,4,power_law_1.01,0.046515199542045596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,16,4,power_law_1.01,0.048870399594306946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,16,4,power_law_1.01,0.05111039876937866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,16,4,power_law_1.01,0.0562175989151001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,8,2,balanced,0.25703465938568115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,8,2,balanced,0.26368000109990436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.01,0.17867519855499267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,16,4,power_law_1.01,0.062118399143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,16,4,power_law_1.01,0.07858560085296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,8,2,balanced,0.26608532667160034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,8,2,balanced,0.2735946575800578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,8,2,balanced,0.2784266670544942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,8,2,balanced,0.2832640012105306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,16,4,power_law_1.01,0.09409279823303222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,8,2,balanced,0.2926186720530192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,8,2,balanced,0.3264480034510295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,8,2,balanced,0.35150400797526044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,8,2,balanced,0.4175093173980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,8,2,balanced,0.4766133228937785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,8,2,balanced,0.5829866727193197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,8,2,balanced,0.6962933540344238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,8,2,balanced,1.0290186405181885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,8,2,balanced,1.2598453362782795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,8,2,balanced,1.8258026440938313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,16,4,power_law_1.01,0.1273792028427124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,8,2,balanced,2.408661365509033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.01,0.2242432117462158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,16,4,power_law_1.01,0.15521279573440552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.01,0.3049024105072021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,16,4,power_law_1.01,0.22421760559082032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,4,balanced,0.04574400186538696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,4,balanced,0.0476800004641215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,4,balanced,0.0633653352657954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,4,balanced,0.0906773308912913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,4,balanced,0.13431466619173685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,4,balanced,0.2177706758181254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,4,balanced,0.22046399116516113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,4,balanced,0.22076799472173056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,4,balanced,0.22180267175038657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,4,balanced,0.22266666094462076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,4,balanced,0.22408533096313477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,4,balanced,0.22675732771555582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,4,balanced,0.22761066754659018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,4,balanced,0.23193599780400595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,4,balanced,0.23612799247105917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,4,balanced,0.24152533213297525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,4,balanced,0.2442773381868998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,4,balanced,0.2585226694742839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,4,balanced,0.2672906716664632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,4,balanced,0.29366934299468994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,4,balanced,0.32894933223724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,4,balanced,0.3710613250732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,4,balanced,0.41447468598683673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,4,balanced,0.5307413339614868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,4,balanced,0.6242986520131429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,4,balanced,0.8659306367238363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,4,balanced,1.0497386455535889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,16,4,power_law_1.01,0.2781824111938477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.01,0.47068161964416505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,16,2,power_law_1.01,0.08710399866104127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,16,2,power_law_1.01,0.09849600195884704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,16,2,power_law_1.01,0.11720960140228272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,16,2,power_law_1.01,0.1385983943939209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,16,2,power_law_1.01,0.14869120121002197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,16,2,power_law_1.01,0.16193920373916626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,16,2,power_law_1.01,0.18845440149307252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,16,2,power_law_1.01,0.19173120260238646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,16,2,power_law_1.01,0.19331200122833253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,16,2,power_law_1.01,0.19734400510787964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,16,2,power_law_1.01,0.19887360334396362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,16,2,power_law_1.01,0.20686719417572022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,16,2,power_law_1.01,0.21066880226135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,16,2,power_law_1.01,0.2197567939758301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,16,2,power_law_1.01,0.22106239795684815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,16,2,power_law_1.01,0.2248703956604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,16,2,power_law_1.01,0.23775360584259034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,16,2,power_law_1.01,0.27333760261535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,16,2,power_law_1.01,0.31648640632629393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,16,2,power_law_1.01,0.3932287931442261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,16,2,power_law_1.01,0.4840127944946289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,16,2,power_law_1.01,0.5907072067260742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,16,2,power_law_1.01,0.7508287906646729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,16,2,power_law_1.01,1.0720895767211913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,16,2,power_law_1.01,1.3261759757995606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,16,2,power_law_1.01,2.0337663650512696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,16,2,power_law_1.01,2.543129539489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,16,4,power_law_1.01,0.3900543928146362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.01,0.5740032196044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,16,4,power_law_1.01,0.503001594543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,1,power_law_1.01,0.05086719989776611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,1,power_law_1.01,0.05354239940643311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,1,power_law_1.01,0.056505602598190305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,1,power_law_1.01,0.06755840182304382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,1,power_law_1.01,0.06930559873580933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,1,power_law_1.01,0.06853119730949402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,1,power_law_1.01,0.07011200189590454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,1,power_law_1.01,0.07038080096244811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,1,power_law_1.01,0.07048320174217224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,1,power_law_1.01,0.0706496000289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,1,power_law_1.01,0.07052159905433655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,1,power_law_1.01,0.07146239876747132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,1,power_law_1.01,0.0741312026977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,1,power_law_1.01,0.07278720140457154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,1,power_law_1.01,0.08134400248527526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,1,power_law_1.01,0.0831488013267517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,1,power_law_1.01,0.08755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,1,power_law_1.01,0.1
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,256,balanced,0.05620799958705902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,256,balanced,0.06069866816202799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,256,balanced,0.05595199763774872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,1,power_law_1.01,0.12494080066680908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,1,power_law_1.01,0.14229120016098024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,256,balanced,0.06001600126425425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,256,balanced,0.05794133245944977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,256,balanced,0.06001066664854685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,256,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,256,balanced,0.06031466523806254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,256,balanced,0.060362666845321655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,256,balanced,0.06069866816202799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,256,balanced,0.06042666733264923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,256,balanced,0.06412266691525777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,256,balanced,0.06401599943637848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,256,balanced,0.06419200201829274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,256,balanced,0.0702400008837382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,256,balanced,0.07025599976380666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,256,balanced,0.07433066765467326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,256,balanced,0.0827466646830241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,256,balanced,0.08904000123341878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,256,balanced,0.10790399710337321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,256,balanced,0.1220266620318095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,256,balanced,0.16476800044377646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,256,balanced,0.15355733036994934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,256,balanced,0.1975253423055013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,256,balanced,0.2262293299039205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,256,balanced,0.29685332377751666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,256,balanced,0.35554667313893634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,1,power_law_1.01,0.15979520082473755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,1,power_law_1.01,0.22088320255279542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,1,power_law_1.01,0.2694080114364624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,1,power_law_1.01,0.3732032060623169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,1,power_law_1.01,0.48059520721435545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,8,power_law_1.01,0.03299199938774109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,8,power_law_1.01,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,8,power_law_1.01,0.03049600124359131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,8,power_law_1.01,0.0324864000082016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,8,power_law_1.01,0.032876798510551454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,8,power_law_1.01,0.03820799887180328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,8,power_law_1.01,0.06771199703216553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,8,power_law_1.01,0.06776319742202759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,8,power_law_1.01,0.06873599886894226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,8,power_law_1.01,0.06525440216064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,8,power_law_1.01,0.05839359760284424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,8,power_law_1.01,0.0936959981918335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,8,power_law_1.01,0.08741120100021363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,8,power_law_1.01,0.07228800058364868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,8,power_law_1.01,0.10280959606170655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,8,power_law_1.01,0.08278399705886841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,8,power_law_1.01,0.0969983994960785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,8,power_law_1.01,0.1373311996459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,1,power_law_1.01,0.6801087856292725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.01,0.16279679536819458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.01,0.22992639541625975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.01,0.291430401802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.01,0.417632007598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.01,0.5279935836791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.01,0.7822976112365723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,2688,1856,6,128,2,128,power_law_1.01,0.0315200001001358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,2688,1856,6,128,2,128,power_law_1.01,0.03133440017700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,2688,1856,6,128,2,128,power_law_1.01,0.031033599376678468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,2688,1856,6,128,2,128,power_law_1.01,0.03383679986000061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,2688,1856,6,128,2,128,power_law_1.01,0.03386879861354828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,2688,1856,6,128,2,128,power_law_1.01,0.03519999980926514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,2688,1856,6,128,2,128,power_law_1.01,0.035641598701477054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,2688,1856,6,128,2,128,power_law_1.01,0.03583360016345978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,2688,1856,6,128,2,128,power_law_1.01,0.03594239950180054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,2688,1856,6,128,2,128,power_law_1.01,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,2688,1856,6,128,2,128,power_law_1.01,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,2688,1856,6,128,2,128,power_law_1.01,0.036831998825073244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,2688,1856,6,128,2,128,power_law_1.01,0.03770880103111267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,2688,1856,6,128,2,128,power_law_1.01,0.038335999846458434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,2688,1856,6,128,2,128,power_law_1.01,0.04077439904212952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,2688,1856,6,128,2,128,power_law_1.01,0.041606399416923526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,2688,1856,6,128,2,128,power_law_1.01,0.04320639967918396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,2688,1856,6,128,2,128,power_law_1.01,0.04405120015144348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,2688,1856,6,128,2,128,power_law_1.01,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,2688,1856,6,128,2,128,power_law_1.01,0.05030400156974792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,2688,1856,6,128,2,128,power_law_1.01,0.05322239995002746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,2688,1856,6,128,2,128,power_law_1.01,0.06361600160598754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,2688,1856,6,128,2,128,power_law_1.01,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,2688,1856,6,128,2,128,power_law_1.01,0.11176960468292237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,2688,1856,6,128,2,128,power_law_1.01,0.14080640077590942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,2688,1856,6,128,2,128,power_law_1.01,0.19867520332336425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,2688,1856,6,128,2,128,power_law_1.01,0.2407167911529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.01,1.0203968048095704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,1,power_law_1.01,0.8853759765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.01,1.5119615554809571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,2,power_law_1.2,0.02223999947309494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,2,power_law_1.2,0.027622398734092713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,2,power_law_1.2,0.04257279932498932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,2,power_law_1.2,0.05873280167579651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,2,power_law_1.2,0.07779840230941773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,2,power_law_1.2,0.08992639780044556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,2,power_law_1.2,0.11207679510116578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,2,power_law_1.2,0.11731840372085571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,2,power_law_1.2,0.12277120351791382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,2,power_law_1.2,0.12322560548782349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,2,power_law_1.2,0.1226688027381897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,2,power_law_1.2,0.13407360315322875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,2,power_law_1.2,0.13634560108184815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,2,power_law_1.2,0.13540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,2,power_law_1.2,0.14152319431304933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,2,power_law_1.2,0.1459328055381775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,2,power_law_1.2,0.1688639998435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,4,8,power_law_1.01,0.016467200219631196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,4,8,power_law_1.01,0.018137599527835845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,4,8,power_law_1.01,0.017824000120162962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,4,8,power_law_1.01,0.0172992005944252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,4,8,power_law_1.01,0.01764480024576187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,4,8,power_law_1.01,0.018361599743366243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,4,8,power_law_1.01,0.019334399700164796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,2,power_law_1.2,0.19497599601745605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,4,8,power_law_1.01,0.01895039975643158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,4,8,power_law_1.01,0.019475199282169342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.2,0.20446720123291015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,4,8,power_law_1.01,0.0204352006316185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,4,8,power_law_1.01,0.02102400064468384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.2,0.25931520462036134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,4,8,power_law_1.01,0.02133760005235672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,4,8,power_law_1.01,0.022169600427150726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,4,8,power_law_1.01,0.02444159984588623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,4,8,power_law_1.01,0.025536000728607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,4,8,power_law_1.01,0.025824001431465148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,4,8,power_law_1.01,0.03779839873313904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,4,8,power_law_1.01,0.03925119936466217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,4,8,power_law_1.01,0.03894400000572205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.2,0.2549184083938599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.2,0.31637120246887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.2,0.37613439559936523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.2,0.477126407623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,2,balanced,0.1011199951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,2,balanced,0.10297600428263347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.2,0.5779839992523194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,2,balanced,0.11345600088437398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,2,balanced,0.13392000397046408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,4,8,power_law_1.01,0.04305280148983002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.2,0.7489408016204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.2,0.9559552192687988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,2,balanced,0.16749866803487143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,4,8,power_law_1.01,0.046163201332092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,4,8,power_law_1.01,0.05790079832077026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,4,8,power_law_1.01,0.06728960275650024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,4,8,power_law_1.01,0.09042559862136841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,4,8,power_law_1.01,0.10705280303955078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,4,8,power_law_1.01,0.14804480075836182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,4,8,power_law_1.01,0.1901311993598938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,2,balanced,0.2418986757596334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,2,balanced,0.24154667059580484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,2,balanced,0.24315200249354044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,2,balanced,0.24939733743667603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,2,balanced,0.24593067169189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,2,balanced,0.24858667453130087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,2,balanced,0.25484800338745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,2,balanced,0.252405325571696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,2,balanced,0.2589600086212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,1,balanced,0.019509332875410717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,1,balanced,0.021402666966120403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,1,balanced,0.023402666052182514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,1,balanced,0.03126933425664902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,1,balanced,0.048021331429481506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,1,balanced,0.048885335524876915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,1,balanced,0.050757333636283875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,1,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,1,balanced,0.05397866666316986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,1,balanced,0.056159997979799904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,1,balanced,0.05982399980227152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,1,balanced,0.0581279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,1,balanced,0.05982933441797892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,1,balanced,0.06402666866779327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,1,balanced,0.06267733375231425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,1,balanced,0.06215466558933258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,1,balanced,0.06871999800205231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,1,balanced,0.07390399773915608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,1,balanced,0.07863999903202057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,1,balanced,0.13539200027783713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,1,balanced,0.1571946640809377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,1,balanced,0.14326399564743042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,1,balanced,0.14621866742769876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,1,balanced,0.2576106588045756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,1,balanced,0.25575466950734455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,1,balanced,0.3641973336537679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,1,balanced,0.47669867674509686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.01,2.001171112060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,16,power_law_1.01,0.04472320079803467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,16,power_law_1.01,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,16,power_law_1.01,0.035174399614334106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,16,power_law_1.01,0.03697279989719391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,16,power_law_1.01,0.036339199542999266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,16,power_law_1.01,0.03764480054378509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,16,power_law_1.01,0.037651199102401736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,16,power_law_1.01,0.038406398892402646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,16,power_law_1.01,0.03982079923152924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,16,power_law_1.01,0.04165120124816894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,16,power_law_1.01,0.04482559859752655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,16,power_law_1.01,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,16,power_law_1.01,0.05008640289306641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,16,power_law_1.01,0.04737919867038727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,16,power_law_1.01,0.05937920212745666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,16,power_law_1.01,0.06041600108146668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,16,power_law_1.01,0.06244480013847351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,16,power_law_1.01,0.06824319958686828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,16,power_law_1.01,0.08513919711112976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,16,power_law_1.01,0.10276479721069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,16,power_law_1.01,0.11573120355606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,16,power_law_1.01,0.1475391983985901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,16,power_law_1.01,0.17653759717941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,16,power_law_1.01,0.2474368095397949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,16,power_law_1.01,0.32718720436096194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,16,power_law_1.01,0.4601920127868652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,16,power_law_1.01,0.5973631858825683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,16,balanced,0.04199466605981191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,16,balanced,0.04158399999141693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,16,balanced,0.04299733539422353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,16,balanced,0.04753066599369049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,16,balanced,0.04770666857560476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,16,balanced,0.05429866909980774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,16,balanced,0.0552106648683548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,16,balanced,0.05624533196290334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,16,balanced,0.056559999783833824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,16,balanced,0.058970664938290916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,16,balanced,0.057536001006762184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,16,balanced,0.05815466741720835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,16,balanced,0.06193066636721293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,16,balanced,0.062074666221936546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,16,balanced,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,16,balanced,0.0690880020459493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,16,balanced,0.0747519979874293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,16,balanced,0.08858666817347209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,16,balanced,0.0995840032895406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,16,balanced,0.11930666367212932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,16,balanced,0.13327999909718832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,16,balanced,0.17640533049901327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,16,balanced,0.2068906625111898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,16,balanced,0.280730664730072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,16,balanced,0.3582666714986165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,16,balanced,0.502128005027771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,16,balanced,0.63865065574646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,2,4,balanced,0.07709866762161255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,2,4,balanced,0.08469333251317342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,32,power_law_1.01,0.11362559795379638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,32,power_law_1.01,0.11141760349273681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,32,power_law_1.01,0.10597759485244751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,4,power_law_1.01,0.04668160080909729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,32,power_law_1.01,0.11342079639434814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,4,power_law_1.01,0.04991999864578247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,32,power_law_1.01,0.11244800090789794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,4,power_law_1.01,0.05260159969329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,32,power_law_1.01,0.10559359788894654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,4,power_law_1.01,0.06120319962501526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,32,power_law_1.01,0.1098688006401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,32,power_law_1.01,0.11791360378265381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,4,power_law_1.01,0.06567680239677429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,32,power_law_1.01,0.10738559961318969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,4,power_law_1.01,0.07482879757881164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,32,power_law_1.01,0.11290240287780762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,4,power_law_1.01,0.07883520126342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,32,power_law_1.01,0.1155776023864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,32,power_law_1.01,0.11518720388412476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,4,power_law_1.01,0.08288639783859253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,32,power_law_1.01,0.11594239473342896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,4,power_law_1.01,0.08766080141067505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,32,power_law_1.01,0.11900800466537476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,4,power_law_1.01,0.08695039749145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,32,power_law_1.01,0.12244479656219483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,4,power_law_1.01,0.08855040073394775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,32,power_law_1.01,0.12405760288238525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,4,power_law_1.01,0.09448959827423095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,32,power_law_1.01,0.12611199617385865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,32,power_law_1.01,0.14116480350494384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,32,power_law_1.01,0.15407999753952026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,32,power_law_1.01,0.19248640537261963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,32,power_law_1.01,0.21127040386199952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,32,power_law_1.01,0.26577279567718504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,4,power_law_1.01,0.09721599817276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,4,power_law_1.01,0.10266239643096924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,4,power_law_1.01,0.11096960306167603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,4,power_law_1.01,0.11534080505371094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,4,power_law_1.01,0.12373119592666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,4,power_law_1.01,0.1399359941482544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,16,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,16,balanced,0.04996799925963084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,4,power_law_1.01,0.16008319854736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,32,power_law_1.01,0.3300863981246948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,4,power_law_1.01,0.19135359525680543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,32,power_law_1.01,0.46173439025878904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,4,power_law_1.01,0.22668800354003907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,32,power_law_1.01,0.5728447914123536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,4,power_law_1.01,0.29475839138031007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,4,power_law_1.01,0.366048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,4,power_law_1.01,0.49526400566101075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,16,balanced,0.05381333331267039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,16,balanced,0.0643039991458257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,16,balanced,0.07419733206431071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,16,balanced,0.10248532891273499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,4,power_law_1.01,0.659660816192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,4,power_law_1.01,0.900153636932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,32,power_law_1.01,0.7946559906005859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,4,power_law_1.01,1.1870847702026368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,32,power_law_1.01,1.0774911880493163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,16,balanced,0.1023466686407725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,16,balanced,0.10104533036549886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,16,balanced,0.09936533371607463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,16,balanced,0.09476266304651897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,16,balanced,0.09830400347709656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,16,balanced,0.09512533744176228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,16,balanced,0.09705066680908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,16,balanced,0.097680002450943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,16,balanced,0.10591999689737956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,16,balanced,0.10276800394058228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,16,balanced,0.10599999626477559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,1,power_law_1.01,0.07473919987678528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,16,balanced,0.11347732941309611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,16,balanced,0.12177067001660664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,1,power_law_1.01,0.11240320205688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,16,balanced,0.1256533364454905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,16,balanced,0.1346506675084432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,1,power_law_1.01,0.13358720541000366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,16,balanced,0.15703466534614563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,2,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,2,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,2,balanced,0.04991999765237173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,2,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,2,balanced,0.05338666836420695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,2,balanced,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,2,balanced,0.052522664268811546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,2,balanced,0.053861334919929504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,2,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,2,balanced,0.05382933219273885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,2,balanced,0.05409066875775655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,2,balanced,0.053173333406448364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,2,balanced,0.05302399893601736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,2,balanced,0.05585066477457682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,2,balanced,0.061797335743904114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,2,balanced,0.06151466568311056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,2,balanced,0.06452266871929169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,2,balanced,0.07620266576608022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,16,balanced,0.17876799901326498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,16,balanced,0.25499733289082843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,16,balanced,0.29917333523432416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,16,balanced,0.4122399886449178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,16,balanced,0.5204853216807047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,2,balanced,0.08102933565775554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,2,balanced,0.09081600109736125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,2,balanced,0.10098666946093242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,2,balanced,0.13659200072288513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,2,balanced,0.15636799732844034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,2,balanced,0.21241066853205362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,2,balanced,0.26740799347559613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,2,balanced,0.38332800070444745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,2,balanced,0.4822933276494344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,1,power_law_1.01,0.20190079212188722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,1,power_law_1.01,0.2048703908920288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,1,power_law_1.01,0.20242559909820557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,1,power_law_1.01,0.21022720336914064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,1,power_law_1.01,0.2074431896209717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,1,power_law_1.01,0.21359360218048096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,1,power_law_1.01,0.21879680156707765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,1,power_law_1.01,0.22160000801086427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,1,power_law_1.01,0.23046400547027587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,1,power_law_1.01,0.23927679061889648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,1,power_law_1.01,0.2359231948852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,1,power_law_1.01,0.2503551959991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,1,power_law_1.01,0.26377599239349364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,1,power_law_1.01,0.27739520072937013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,1,power_law_1.01,0.3355456113815308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,1,power_law_1.01,0.35983359813690186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,1,power_law_1.01,0.44382081031799314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,1,power_law_1.01,0.5246208190917969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,1,power_law_1.01,0.7006336212158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,1,power_law_1.01,0.8799936294555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,1,power_law_1.01,1.2248703956604003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,4,power_law_1.2,0.028998398780822755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,4,power_law_1.2,0.047353601455688475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,4,power_law_1.2,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,4,power_law_1.2,0.07715200185775757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,4,power_law_1.2,0.08419839739799499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,4,power_law_1.2,0.1038591980934143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,4,power_law_1.2,0.11784319877624512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,4,power_law_1.2,0.12753280401229858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,1,power_law_1.01,1.5434559822082519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,4,power_law_1.2,0.12725119590759276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,4,power_law_1.2,0.12451200485229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,4,power_law_1.2,0.12421120405197143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,4,power_law_1.2,0.1332352042198181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,4,power_law_1.2,0.13538559675216674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,4,power_law_1.2,0.13564800024032592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,4,power_law_1.2,0.14305280447006224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,4,power_law_1.2,0.14922879934310912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,4,power_law_1.2,0.16341760158538818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,4,power_law_1.2,0.17488640546798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.2,0.18936959505081177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.2,0.22410879135131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.2,0.28319358825683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.2,0.31318399906158445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.2,0.36805760860443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.2,0.4727168083190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.2,0.5506175994873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.2,0.793177604675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.2,1.0067456245422364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,1,power_law_1.01,2.1946624755859374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,1,power_law_1.01,2.925049591064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,128,power_law_1.2,0.034483200311660765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,128,power_law_1.2,0.034668800234794614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,128,power_law_1.2,0.03335680067539215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,128,power_law_1.2,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,128,power_law_1.2,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,128,power_law_1.2,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,128,power_law_1.2,0.03813120126724243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,128,power_law_1.2,0.03844479918479919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,128,power_law_1.2,0.038700801134109494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,128,power_law_1.2,0.03889279961585999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,128,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,128,power_law_1.2,0.03994880020618439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,128,power_law_1.2,0.0406464010477066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,128,power_law_1.2,0.04186240136623383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,128,power_law_1.2,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,128,power_law_1.2,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,128,power_law_1.2,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,128,power_law_1.2,0.05196160078048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,128,power_law_1.2,0.05822719931602478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,128,power_law_1.2,0.07090560197830201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,128,power_law_1.2,0.08364160060882568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,128,power_law_1.2,0.12157440185546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,128,power_law_1.2,0.15232640504837036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,128,power_law_1.2,0.20817279815673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,16,16,balanced,0.021456000705560047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,16,16,balanced,0.021344001094500225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,16,16,balanced,0.02107200026512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,16,16,balanced,0.022976001103719074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,16,16,balanced,0.025263999899228413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,16,16,balanced,0.024874667326609295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,16,16,balanced,0.025461333493391674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,16,16,balanced,0.025199999411900837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,16,16,balanced,0.0313226655125618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,16,16,balanced,0.031397332747777305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,16,16,balanced,0.03145066648721695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,16,16,balanced,0.05421333511670431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,16,16,balanced,0.0545653353134791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,16,16,balanced,0.05483733117580414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,16,16,balanced,0.0539626677831014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,16,16,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,16,16,balanced,0.054527997970581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,16,16,balanced,0.062368000547091164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,16,16,balanced,0.07010666529337566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,16,16,balanced,0.08074133098125458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,16,16,balanced,0.09379200140635173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,16,16,balanced,0.1172160009543101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,16,16,balanced,0.1463093360265096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,16,16,balanced,0.19325333833694458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,16,16,balanced,0.23851199944814047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,16,16,balanced,0.35070399443308514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,16,16,balanced,0.4474453528722127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,128,power_law_1.2,0.2754175901412964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,128,power_law_1.2,0.38645761013031005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,1,balanced,0.07257066667079926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,1,balanced,0.09725866715113322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,1,balanced,0.1487626632054647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,1,balanced,0.2516160011291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,1,balanced,0.45338133970896405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,1,balanced,0.841808001200358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,1,balanced,1.2400480111440022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,1,balanced,1.2382559776306152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,1,balanced,1.2428106466929119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,1,balanced,1.2531147003173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,1,balanced,1.254863977432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,1,balanced,1.2677333354949951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,16,balanced,0.04186666508515676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,16,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,16,balanced,0.04347200194994608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,16,balanced,0.04772266745567322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,16,balanced,0.048437332113583885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,16,balanced,0.05299200117588043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,16,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,16,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,16,balanced,0.05422399938106537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,16,balanced,0.055946667989095054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,16,balanced,0.05593599875768026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,16,balanced,0.056287998954455055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,16,balanced,0.058320000767707825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,16,balanced,0.05905599892139435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,16,balanced,0.06423466900984447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,16,balanced,0.06591466565926869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,16,balanced,0.06825600067774455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,16,balanced,0.07642666498819987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,1,balanced,1.2749226888020833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,16,balanced,0.08475733796755473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,16,balanced,0.11529067158699036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,1,balanced,1.2858400344848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,16,balanced,0.1301866670449575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,16,balanced,0.17452265818913779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,1,balanced,1.2994186878204346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,16,balanced,0.1959999998410543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,16,balanced,0.2728266716003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,1,balanced,1.3151146570841472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,16,balanced,0.3285280068715413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,16,balanced,0.45872000853220624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,1,balanced,1.319264014561971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,16,balanced,0.5804320176442465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,1,balanced,1.3699733416239421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,1,balanced,1.4093119303385417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,1,balanced,1.499274730682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,1,balanced,1.5707200368245442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,1,balanced,1.7500106493632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,1,balanced,1.8643892606099446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,1,balanced,2.2272799809773765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,1,balanced,2.662853399912516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,1,balanced,3.295381228129069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,1,balanced,4.481184005737305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,128,power_law_1.2,0.5175295829772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,1,1,balanced,0.055567999680837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,1,1,balanced,0.07632533212502797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,1,1,balanced,0.1193386713663737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,1,1,balanced,0.20709333817164102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,1,1,balanced,0.3760426839192708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,1,1,balanced,0.379807988802592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,1,1,balanced,0.3841386636098226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,1,1,balanced,0.38421865304311115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,1,1,balanced,0.3890399932861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,1,1,balanced,0.38947733243306476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,1,1,balanced,0.3948640028635661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,1,1,balanced,0.3973493178685506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,1,1,balanced,0.4042239983876546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,1,1,balanced,0.4096533457438151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,1,1,balanced,0.41651201248168945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,1,1,balanced,0.42603735129038495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,1,1,balanced,0.4442186752955119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,1,1,balanced,0.47179198265075684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,1,1,balanced,0.5136853456497192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,1,1,balanced,0.569488008817037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,1,1,balanced,0.6304213205973307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,1,1,balanced,0.8340960343678793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,1,1,balanced,0.9033386707305908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,1,1,balanced,1.319648027420044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,1,1,balanced,1.5617280006408691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,1,1,balanced,2.2429653803507485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,1,1,balanced,2.944629351298014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,2,power_law_1.01,0.0709119975566864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,2,power_law_1.01,0.09781119823455811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,2,power_law_1.01,0.12288000583648681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,2,power_law_1.01,0.1865280032157898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,2,power_law_1.01,0.23374719619750978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,2,power_law_1.01,0.2670655965805054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,2,power_law_1.01,0.35729920864105225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,2,power_law_1.01,0.36481919288635256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,2,power_law_1.01,0.3754944086074829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,2,power_law_1.01,0.3701375961303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,2,power_law_1.01,0.4008512020111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,2,power_law_1.01,0.3966592073440552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,2,power_law_1.01,0.4357759952545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,2,power_law_1.01,0.4251264095306396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,2,power_law_1.01,0.4510015964508057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,2,power_law_1.01,0.46141438484191893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,2,power_law_1.01,0.489350414276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,2,power_law_1.01,0.5416384220123291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,2,power_law_1.01,0.5627264022827149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,2,power_law_1.01,0.6500864028930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,2,power_law_1.01,0.6918655872344971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,2,power_law_1.01,0.848691177368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,2,power_law_1.01,0.9687295913696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,2,power_law_1.01,1.2776703834533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,2,power_law_1.01,1.5587455749511718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,2,power_law_1.01,2.1611135482788084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,2,power_law_1.01,2.700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,1,32,balanced,0.063360000650088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,1,32,balanced,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,1,32,balanced,0.06438399851322174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,1,32,balanced,0.07028266787528992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,1,32,balanced,0.08761066198348999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,1,32,balanced,0.12826666235923767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,1,32,balanced,0.15812266866366068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,1,32,balanced,0.157642662525177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,1,32,balanced,0.1639840006828308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,1,32,balanced,0.16511999567349753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,1,32,balanced,0.16609600186347961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,1,32,balanced,0.16217600305875143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,1,32,balanced,0.17561600605646768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,1,32,balanced,0.17820799350738525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,1,32,balanced,0.1769919991493225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,1,32,balanced,0.17969600359598795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,1,32,balanced,0.18361065785090128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,1,32,balanced,0.18341867129007974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,1,32,balanced,0.1994826594988505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,1,32,balanced,0.22197333971659342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,1,32,balanced,0.24114133914311728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,1,32,balanced,0.2870826721191406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,1,32,balanced,0.3281973401705424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,1,32,balanced,0.41650664806365967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,1,32,balanced,0.5624373356501261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,1,32,balanced,0.7153440316518148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,1,32,balanced,0.9351733525594076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,32,balanced,0.03809066613515218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,32,balanced,0.03625066578388214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,32,balanced,0.03195200115442276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,32,balanced,0.035802667339642845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,32,balanced,0.05651199817657471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,32,balanced,0.05852266649405161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,32,balanced,0.057087997595469155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,32,balanced,0.05774400134881338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,32,balanced,0.05925333499908447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,32,balanced,0.05894933144251505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,32,balanced,0.058677335580190025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,32,balanced,0.06020266811052958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,32,balanced,0.059215997656186424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,32,balanced,0.06124266485373179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,32,balanced,0.06503466765085857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,32,balanced,0.0643039991458257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,32,balanced,0.06610666712125142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,32,balanced,0.06653866668542226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,32,balanced,0.0728053351243337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,32,balanced,0.07467199862003326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,32,balanced,0.07621333499749501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,32,power_law_1.01,0.052083200216293334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,32,power_law_1.01,0.05055999755859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,32,power_law_1.01,0.049292799830436704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,32,power_law_1.01,0.05198720097541809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,32,balanced,0.08655466636021932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,4,balanced,0.04710933566093445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,32,balanced,0.09136533737182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,32,balanced,0.10708799958229065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,32,balanced,0.13239999612172446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,32,balanced,0.16552000244458517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,32,balanced,0.1962613264719645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,32,power_law_1.01,0.055238401889801024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,4,balanced,0.049045334259668984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,32,power_law_1.01,0.056377601623535153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,32,power_law_1.01,0.05775359869003296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,32,power_law_1.01,0.05572479963302612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,32,power_law_1.01,0.058406400680541995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,4,balanced,0.06871466835339864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,16,power_law_1.2,0.05648639798164368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,4,balanced,0.09991466999053955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,4,balanced,0.15018666783968607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,4,balanced,0.2499786615371704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,16,power_law_1.2,0.06862720251083373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,4,balanced,0.2510453263918559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,4,balanced,0.25339200099309284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,16,power_law_1.2,0.07649279832839966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,16,power_law_1.2,0.0824895977973938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,32,power_law_1.01,0.059443199634552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,32,power_law_1.01,0.06076800227165222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,16,power_law_1.2,0.08399999737739564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,32,power_law_1.01,0.06110079884529114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,16,power_law_1.2,0.08403840065002441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,4,balanced,0.2568426728248596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,16,power_law_1.2,0.08206080198287964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,16,power_law_1.2,0.0809216022491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,16,power_law_1.2,0.08046079874038696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,16,power_law_1.2,0.08330879807472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,32,power_law_1.01,0.06480640172958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,4,balanced,0.25518399477005005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,16,power_law_1.2,0.08472959995269776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,4,balanced,0.2578773299853007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,4,balanced,0.2625760038693746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,16,power_law_1.2,0.08583679795265198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,4,balanced,0.26947200298309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,16,power_law_1.2,0.08749439716339111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,4,balanced,0.26761066913604736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,4,balanced,0.27591999371846515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,16,power_law_1.2,0.08900480270385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,16,power_law_1.2,0.09932159781455993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,16,power_law_1.2,0.10241919755935669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,32,power_law_1.01,0.07024000287055969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,16,power_law_1.2,0.10756479501724243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,32,power_law_1.01,0.07863039970397949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,16,power_law_1.2,0.12478079795837402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,4,balanced,0.2797279953956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,16,power_law_1.2,0.14005759954452515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,16,power_law_1.2,0.1687999963760376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,16,power_law_1.2,0.19531519412994386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,32,power_law_1.01,0.0744704008102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,32,power_law_1.01,0.08526080250740051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,4,balanced,0.2850933273633321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,32,power_law_1.01,0.09502080082893372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,4,balanced,0.3052746653556824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,32,power_law_1.01,0.10897279977798462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,32,power_law_1.01,0.12592639923095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,32,power_law_1.01,0.14657919406890868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,16,power_law_1.2,0.25718400478363035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,32,power_law_1.01,0.18474240303039552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,16,power_law_1.2,0.2855871915817261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,32,power_law_1.01,0.24428160190582277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,4,balanced,0.32202666997909546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,32,power_law_1.01,0.3414079904556274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,4,balanced,0.3554240067799886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,4,balanced,0.3983946641286214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,32,power_law_1.01,0.47579522132873536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,4,balanced,0.46587200959523517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,4,balanced,0.5411146481831869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,32,power_law_1.01,0.6025152206420898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,16,power_law_1.2,0.3864128112792969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,16,power_law_1.2,0.48691840171813966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,4,balanced,0.690117359161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,16,power_law_1.2,0.6935743808746337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,32,power_law_1.01,0.898528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,4,balanced,0.8194719950358073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,4,balanced,1.1692266464233398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,16,power_law_1.2,0.8844351768493652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,4,balanced,1.4481706619262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,1,power_law_1.01,0.03389439880847931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,1,power_law_1.01,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,1,power_law_1.01,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,1,power_law_1.01,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,1,power_law_1.01,0.03852800130844116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,1,power_law_1.01,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,1,power_law_1.01,0.03999359905719757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,1,power_law_1.01,0.041017600893974306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,1,power_law_1.01,0.04199680089950562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,1,power_law_1.01,0.04202240109443665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,1,power_law_1.01,0.042335999011993405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,1,power_law_1.01,0.0441536009311676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,1,power_law_1.01,0.045587199926376346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,1,power_law_1.01,0.047167998552322385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,1,power_law_1.01,0.05446400046348572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,1,power_law_1.01,0.055955201387405396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,1,power_law_1.01,0.05857279896736145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,1,power_law_1.01,0.06674559712409973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.01,0.07555199861526489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.01,0.0955456018447876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.01,0.10952960252761841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.01,0.14747519493103028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.01,0.17637120485305785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.01,0.24082560539245607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.01,0.3122879981994629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,2,8,power_law_1.01,0.07093120217323304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,2,8,power_law_1.01,0.09623039960861206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,2,8,power_law_1.01,0.14620800018310548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,2,8,power_law_1.01,0.1313472032546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,2,8,power_law_1.01,0.15679359436035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,2,8,power_law_1.01,0.18302719593048095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,2,8,power_law_1.01,0.17065600156784058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,2,8,power_law_1.01,0.19531519412994386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,2,8,power_law_1.01,0.17424000501632692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,2,8,power_law_1.01,0.18487679958343506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,2,8,power_law_1.01,0.17847039699554443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,2,8,power_law_1.01,0.1762112021446228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,2,8,power_law_1.01,0.17587200403213502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,2,8,power_law_1.01,0.19070719480514525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,2,8,power_law_1.01,0.20869119167327882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,2,8,power_law_1.01,0.19600000381469726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,2,8,power_law_1.01,0.20904319286346434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,2,8,power_law_1.01,0.23593599796295167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,2,8,power_law_1.01,0.25678720474243166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,2,8,power_law_1.01,0.30782079696655273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,2,8,power_law_1.01,0.34454400539398194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,2,8,power_law_1.01,0.4387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,2,8,power_law_1.01,0.5184895992279053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,2,8,power_law_1.01,0.676416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,2,8,power_law_1.01,0.7713664054870606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,2,8,power_law_1.01,1.1043392181396485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.01,0.4469183921813965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,2,8,power_law_1.01,1.3564736366271972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.01,0.5836991786956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,2,power_law_1.01,0.0641215980052948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,2,power_law_1.01,0.06962559819221496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,2,power_law_1.01,0.08077440261840821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,2,power_law_1.01,0.10073599815368653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,2,power_law_1.01,0.12041599750518799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,2,power_law_1.01,0.13356800079345704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,2,power_law_1.01,0.160697603225708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,2,power_law_1.01,0.168121600151062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,2,power_law_1.01,0.16862720251083374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,2,power_law_1.01,0.17297279834747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,2,power_law_1.01,0.17562880516052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,2,power_law_1.01,0.1826367974281311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,2,power_law_1.01,0.18896000385284423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,2,power_law_1.01,0.19704960584640502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,2,power_law_1.01,0.20112640857696534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,2,power_law_1.01,0.2008960008621216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,2,power_law_1.01,0.21836159229278565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,2,power_law_1.01,0.23795199394226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.01,0.26796159744262693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.01,0.32145280838012696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.01,0.3702336072921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.01,0.4770815849304199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.01,0.5766335964202881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.01,0.8282112121582031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.01,0.9971648216247558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.01,1.4233344078063965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.01,1.9121408462524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,1,32,balanced,0.06694933275381725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,1,32,balanced,0.06672533353169759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,1,32,balanced,0.06714666883150737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,1,32,balanced,0.0763626645008723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,1,32,balanced,0.08999466896057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,1,32,balanced,0.1316426694393158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,1,32,balanced,0.15034133195877075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,1,32,balanced,0.15069333712259927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,1,32,balanced,0.15381866693496704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,1,32,balanced,0.15169599652290344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,1,32,balanced,0.15342400471369425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,16,8,power_law_1.01,0.0389055997133255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,1,32,balanced,0.1560373306274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,16,8,power_law_1.01,0.039776000380516055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,16,8,power_law_1.01,0.03924480080604553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,16,8,power_law_1.01,0.040268799662590025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,16,8,power_law_1.01,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,16,8,power_law_1.01,0.04062080085277557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,16,8,power_law_1.01,0.04056960046291351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,16,8,power_law_1.01,0.041356799006462094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,16,8,power_law_1.01,0.039987200498580934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,16,8,power_law_1.01,0.039680001139640805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,16,8,power_law_1.01,0.0395904004573822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,16,8,power_law_1.01,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,16,8,power_law_1.01,0.04415999948978424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,16,8,power_law_1.01,0.04337919950485229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,16,8,power_law_1.01,0.04850560128688812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,1,32,balanced,0.15612266461054483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,1,32,balanced,0.15794133146603903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,16,8,power_law_1.01,0.049081599712371825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,1,32,balanced,0.16882665952046713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,1,32,balanced,0.16969066858291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,1,32,balanced,0.17844265699386597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,1,32,balanced,0.18710400660832724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,16,8,power_law_1.01,0.051481598615646364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,1,32,balanced,0.19339734315872192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,16,8,power_law_1.01,0.058796799182891844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,8,power_law_1.2,0.033011201024055484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,8,power_law_1.2,0.03294079899787903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.01,0.06746240258216858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,8,power_law_1.2,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,8,power_law_1.2,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,8,power_law_1.2,0.03611519932746887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.01,0.08233600258827209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,8,power_law_1.2,0.036396801471710205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,8,power_law_1.2,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,8,power_law_1.2,0.03711360096931458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,1,32,balanced,0.20963199933369955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,1,32,balanced,0.22850666443506876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,1,32,balanced,0.26637866099675495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,1,32,balanced,0.3009386658668518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,1,32,balanced,0.3681066830952962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,1,32,balanced,0.47997868061065674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,1,32,balanced,0.6236213445663452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.01,0.09865599870681763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,1,32,balanced,0.7921600341796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,8,power_law_1.2,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,8,power_law_1.2,0.040064001083374025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,8,power_law_1.2,0.04067200124263763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.01,0.12849279642105102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.01,0.14247039556503296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,8,power_law_1.2,0.04394879937171936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,8,power_law_1.2,0.044838398694992065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,8,power_law_1.2,0.05073919892311096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,8,power_law_1.2,0.05989760160446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,8,power_law_1.2,0.06199679970741272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,8,power_law_1.2,0.06986240148544312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,8,power_law_1.2,0.0945088028907776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.01,0.21834878921508788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,8,power_law_1.2,0.11523840427398682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,8,power_law_1.2,0.1564352035522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,8,power_law_1.2,0.198470401763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.01,0.2780735969543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,8,power_law_1.2,0.29312639236450194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,8,power_law_1.2,0.34053759574890136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.01,0.40371198654174806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,8,power_law_1.2,0.5851903915405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.01,0.5209407806396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,4,power_law_1.01,0.04185599982738495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,8,power_law_1.2,0.7481664180755615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,4,power_law_1.01,0.0381632000207901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,4,power_law_1.01,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,4,power_law_1.01,0.04549759924411774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,4,power_law_1.01,0.04712960124015808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,4,power_law_1.01,0.05136640071868896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,4,power_law_1.01,0.052108800411224364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,4,power_law_1.01,0.053913599252700804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,4,power_law_1.01,0.05284479856491089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,4,power_law_1.01,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,4,power_law_1.01,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,4,power_law_1.01,0.058182400465011594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,4,power_law_1.01,0.05901439785957337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,4,power_law_1.01,0.061913597583770755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,4,power_law_1.01,0.06843519806861878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,4,power_law_1.01,0.07174400091171265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,4,power_law_1.01,0.07619199752807618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,4,power_law_1.01,0.08699520230293274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,4,power_law_1.01,0.09436799883842469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,4,power_law_1.01,0.1045375943183899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,4,power_law_1.01,0.12660479545593262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,4,power_law_1.01,0.16808320283889772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,4,power_law_1.01,0.20044159889221191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,4,power_law_1.01,0.2712127923965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,4,power_law_1.01,0.34440319538116454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,4,power_law_1.01,0.551852798461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,4,power_law_1.01,0.6506239891052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,8,power_law_1.2,1.144825553894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,8,power_law_1.2,1.4055295944213868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,2,8,balanced,0.045642669002215065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,2,8,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,2,8,balanced,0.04788800080617269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,2,8,balanced,0.05418666700522105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,2,8,balanced,0.05826666454474131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,2,8,balanced,0.07881066699822743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,2,8,balanced,0.07905599971612294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,2,8,balanced,0.08064533273379008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,2,8,balanced,0.08097066481908162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,2,8,balanced,0.08136533200740814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,2,8,balanced,0.0811466674009959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,2,8,balanced,0.08261333405971527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,2,8,balanced,0.08659733335177104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,2,8,balanced,0.0862559974193573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,2,8,balanced,0.09252799550692241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,2,8,balanced,0.09291199843088786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,2,8,balanced,0.09506666660308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,2,8,balanced,0.1051573355992635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,2,8,balanced,0.11108266313870747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,2,8,balanced,0.12581866979599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,2,8,balanced,0.14831466476122537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,2,8,balanced,0.1756053368250529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,2,8,balanced,0.19957866271336874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,2,8,balanced,0.26910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,2,8,balanced,0.3255733251571655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,2,8,balanced,0.4457813501358032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,2,8,balanced,0.5728319883346558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,8,1,balanced,0.03209066639343897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,8,1,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,8,1,balanced,0.05745066702365875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,8,1,balanced,0.05797333518664042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,8,1,balanced,0.061664000153541565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,8,1,balanced,0.06185600161552429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,8,1,balanced,0.061290666460990906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,8,1,balanced,0.06273066500822704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,8,1,balanced,0.06417599817117055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,8,1,balanced,0.06334400177001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,8,1,balanced,0.06411733229955037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,8,1,balanced,0.0664160003264745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,8,1,balanced,0.06870399912198384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,8,1,balanced,0.069733331600825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,8,1,balanced,0.08099199831485748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,8,1,balanced,0.0830506682395935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,8,1,balanced,0.08667199810345967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,8,1,balanced,0.10357333223025005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,8,1,balanced,0.10893332958221436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,8,1,balanced,0.1402400036652883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,8,1,balanced,0.17255467176437378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,8,1,balanced,0.22307199239730835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,8,1,balanced,0.26845333973566693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,8,1,balanced,0.381168007850647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,8,1,balanced,0.48818667729695636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,8,1,balanced,0.6988053321838379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,8,1,balanced,0.9096799691518148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,16,balanced,0.03978666663169861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,16,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,16,balanced,0.039813332259655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,16,balanced,0.04366933306058248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,16,balanced,0.04363200068473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,16,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,16,balanced,0.05153066913286845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,16,balanced,0.051829333106676735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,16,balanced,0.05186666548252106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,16,balanced,0.05247466762860616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,16,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,16,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,16,balanced,0.055999999245007835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,16,balanced,0.05421333511670431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,16,balanced,0.061887999375661217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,16,balanced,0.060309335589408875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,16,balanced,0.06410133341948192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,16,balanced,0.07028799752394359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,16,balanced,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,16,balanced,0.086709330479304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,16,balanced,0.10308800141016643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,16,balanced,0.13799466689427695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,16,balanced,0.16525333126386008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,16,balanced,0.223306675752004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,16,balanced,0.26800533135732013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,16,balanced,0.36770133177439374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,16,balanced,0.47624532381693524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,2,power_law_1.01,0.058841598033905027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,2,power_law_1.01,0.06490880250930786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,2,power_law_1.01,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,2,power_law_1.01,0.0874176025390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,2,power_law_1.01,0.09710080027580262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,2,power_law_1.01,0.1245695948600769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,2,power_law_1.01,0.12609280347824098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,2,power_law_1.01,0.13379839658737183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,2,power_law_1.01,0.13340799808502196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,2,power_law_1.01,0.13761279582977295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,2,power_law_1.01,0.13939199447631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,2,power_law_1.01,0.1417407989501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,2,power_law_1.01,0.14428160190582276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,2,power_law_1.01,0.14911999702453613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,2,power_law_1.01,0.15734399557113649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,2,power_law_1.01,0.1600767970085144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,2,power_law_1.01,0.17414400577545167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,2,power_law_1.01,0.2048192024230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,2,power_law_1.01,0.22447359561920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,2,power_law_1.01,0.27532799243927003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,2,power_law_1.01,0.3229248046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,2,power_law_1.01,0.43216638565063475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,2,power_law_1.01,0.5669248104095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,2,power_law_1.01,0.7646656036376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,2,power_law_1.01,0.9971776008605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,2,power_law_1.01,1.3913472175598145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,2,power_law_1.01,1.9118463516235351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,2,2,power_law_1.2,0.09874560236930847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,2,2,power_law_1.2,0.11770880222320557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,2,2,power_law_1.2,0.11723519563674926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,2,2,power_law_1.2,0.17180800437927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,2,2,power_law_1.2,0.16155519485473632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,2,2,power_law_1.2,0.1626688003540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,2,2,power_law_1.2,0.1686336040496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,2,2,power_law_1.2,0.1646656036376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,2,2,power_law_1.2,0.1629439949989319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,2,2,power_law_1.2,0.16431360244750975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,2,2,power_law_1.2,0.15400960445404052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,2,2,power_law_1.2,0.17985279560089112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,2,2,power_law_1.2,0.18435200452804565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,2,2,power_law_1.2,0.1754240036010742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,2,2,power_law_1.2,0.20288639068603515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,2,2,power_law_1.2,0.2114880084991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,2,2,power_law_1.2,0.2192512035369873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,2,2,power_law_1.2,0.2617664098739624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.2,0.3220223903656006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.2,0.36243839263916017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.2,0.4959551811218262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.2,0.6085375785827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.2,0.8510527610778809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,1,balanced,0.07497066756089528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,1,balanced,0.10725866754849751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,1,balanced,0.16064533591270447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,1,balanced,0.15479999780654907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,1,balanced,0.15916799505551657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,1,balanced,0.1683573325475057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,1,balanced,0.15918399890263876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,1,balanced,0.1646880010763804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,1,balanced,0.15268799662590027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,1,balanced,0.1502240002155304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,1,balanced,0.16034666697184244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,1,balanced,0.14607466260592142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,1,balanced,0.14512000481287637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,1,balanced,0.14272532860438028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,1,balanced,0.1406613290309906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,1,balanced,0.13589333494504294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,1,balanced,0.1390666663646698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,1,balanced,0.24216532707214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,1,balanced,0.2018400033315023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,1,balanced,0.2680266698201497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,1,balanced,0.3311573266983032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,1,balanced,0.46010132630666095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,1,balanced,0.58842666943868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,1,balanced,0.8534933725992838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,1,balanced,1.1245919863382976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,1,balanced,1.6525386174519856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,1,balanced,2.1953226725260415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.2,1.1827967643737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.2,1.549830436706543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,4,power_law_1.01,0.037887999415397645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,4,power_law_1.01,0.03735679984092712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,4,power_law_1.01,0.039878401160240176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,4,power_law_1.01,0.04240640103816986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,4,power_law_1.01,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,4,power_law_1.01,0.044352000951766966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,4,power_law_1.01,0.04438399970531463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,4,power_law_1.01,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,4,power_law_1.01,0.04562560021877289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,4,power_law_1.01,0.04684160053730011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,4,power_law_1.01,0.04663679897785187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,4,power_law_1.01,0.05848960280418396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,1,power_law_1.01,0.07697920203208923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,4,power_law_1.01,0.0592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,1,power_law_1.01,0.08538240194320679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,4,power_law_1.01,0.06277120113372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,1,power_law_1.01,0.10650240182876587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,1,power_law_1.01,0.1475711941719055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,1,power_law_1.01,0.18477439880371094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,1,power_law_1.01,0.22602880001068115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,1,power_law_1.01,0.29570560455322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,1,power_law_1.01,0.30198400020599364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,1,power_law_1.01,0.3083967924118042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,1,power_law_1.01,0.3158463954925537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,1,power_law_1.01,0.3276927947998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,1,power_law_1.01,0.3357952117919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,1,power_law_1.01,0.3507776021957397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,1,power_law_1.01,0.3553663969039917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,4,power_law_1.01,0.07667199969291687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,1,power_law_1.01,0.36570239067077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,4,power_law_1.01,0.0782975971698761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,1,power_law_1.01,0.37808001041412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,1,power_law_1.01,0.3971519947052002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,4,power_law_1.01,0.09218559861183166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,1,power_law_1.01,0.44800639152526855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,4,power_law_1.01,0.10771199464797973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,1,power_law_1.01,0.5062143802642822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,1,power_law_1.01,0.615392017364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,1,power_law_1.01,0.663865613937378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,1,power_law_1.01,0.8536576271057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,1,power_law_1.01,1.037286376953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,4,power_law_1.01,0.12565759420394898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,4,power_law_1.01,0.16405119895935058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,1,power_law_1.01,1.3931584358215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,1,power_law_1.01,1.789311981201172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,1,power_law_1.01,2.513510322570801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,1,power_law_1.01,3.290156936645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,4,power_law_1.01,0.21240320205688476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,2,power_law_1.01,0.049881601333618165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,2,power_law_1.01,0.05444480180740356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,2,power_law_1.01,0.06734079718589783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,2,power_law_1.01,0.08158079981803894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,2,power_law_1.01,0.09438080191612244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,2,power_law_1.01,0.11108479499816895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,2,power_law_1.01,0.13114240169525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,2,power_law_1.01,0.13801599740982057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,2,power_law_1.01,0.14250880479812622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,2,power_law_1.01,0.14984960556030275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,2,power_law_1.01,0.15212160348892212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.2,2.256883239746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,2,power_law_1.01,0.1577023983001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,2,power_law_1.01,0.16050560474395753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,2,power_law_1.01,0.17311359643936158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,2,power_law_1.01,0.18764159679412842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,2,power_law_1.01,0.19414399862289428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,2,power_law_1.01,0.20107519626617432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,2,power_law_1.01,0.22819199562072753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,2,power_law_1.01,0.25622398853302003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,4,power_law_1.01,0.33955841064453124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,2,power_law_1.01,0.31094400882720946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,2,power_law_1.01,0.3659456014633179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,2,power_law_1.01,0.4852543830871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,2,power_law_1.01,0.5835840225219726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,2,power_law_1.01,0.7998655796051025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,2,power_law_1.01,1.0127936363220216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,2,power_law_1.01,1.3758015632629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,2,power_law_1.01,1.7952512741088866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,4,power_law_1.01,0.43882241249084475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,4,power_law_1.01,0.6115776062011719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.2,2.600268745422363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,4,power_law_1.01,0.7069439888000488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,32,8,power_law_1.01,0.04472959935665131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,32,8,power_law_1.01,0.04895359873771667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,32,8,power_law_1.01,0.050367999076843264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,32,8,power_law_1.01,0.051820802688598636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,32,8,power_law_1.01,0.055244797468185426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,32,8,power_law_1.01,0.056678402423858645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,32,8,power_law_1.01,0.05719040036201477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,32,8,power_law_1.01,0.05882880091667175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,32,8,power_law_1.01,0.05880320072174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,32,8,power_law_1.01,0.06080639958381653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,32,8,power_law_1.01,0.06268159747123718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,32,8,power_law_1.01,0.0640064001083374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,32,8,power_law_1.01,0.06609920263290406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,32,8,power_law_1.01,0.0717631995677948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,32,8,power_law_1.01,0.0784704029560089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,32,8,power_law_1.01,0.0812608003616333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,32,8,power_law_1.01,0.0832256019115448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,32,8,power_law_1.01,0.09552000164985656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,32,8,power_law_1.01,0.10656640529632569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,32,8,power_law_1.01,0.14088959693908693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,32,8,power_law_1.01,0.16588799953460692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,32,8,power_law_1.01,0.21818881034851073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,32,8,power_law_1.01,0.2684351921081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,32,8,power_law_1.01,0.3867840051651001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,32,8,power_law_1.01,0.49006080627441406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,32,8,power_law_1.01,0.721401596069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,32,8,power_law_1.01,0.8628607749938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,4,power_law_1.01,1.189030361175537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,2,balanced,0.09562666217486064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,2,balanced,0.12266666690508525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,2,balanced,0.16665066281954447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,2,balanced,0.26523200670878094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,2,balanced,0.43969066937764484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,2,balanced,0.6028159856796265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,2,balanced,0.6169066826502482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,2,balanced,0.6228106816609701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,2,balanced,0.6219199895858765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,2,balanced,0.627349336942037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,2,balanced,0.6366719802220663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,2,balanced,0.6415573358535767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,2,balanced,0.6470400094985962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,2,balanced,0.6591093142827352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,2,balanced,0.6696586608886719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,2,balanced,0.6808906396230062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,2,balanced,0.7040320237477621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,2,balanced,0.7557226816813151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,2,balanced,0.8002879619598389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,2,balanced,1.18012801806132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,2,balanced,1.0280053615570068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,2,balanced,1.4528533617655437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,2,balanced,1.567893346150716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,2,balanced,2.3231147130330405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,2,balanced,2.598192056020101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,2,balanced,3.9134238560994468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,2,balanced,4.759482701619466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,4,power_law_1.01,0.04986880123615265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,4,power_law_1.01,0.05777919888496399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,4,power_law_1.01,0.06096640229225159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,4,power_law_1.01,0.06675840020179749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,4,power_law_1.01,0.06892799735069274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,4,power_law_1.01,0.08420479893684388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,4,power_law_1.01,0.08567039966583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,4,power_law_1.01,0.08919680118560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,4,power_law_1.01,0.09297919869422913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,4,power_law_1.01,0.09502080082893372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,4,power_law_1.01,0.09508479833602905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,4,power_law_1.01,0.09889919757843017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,4,power_law_1.01,0.1027519941329956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,4,power_law_1.01,0.10549119710922242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,4,power_law_1.01,0.11173759698867798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,4,power_law_1.01,0.11800960302352906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,4,power_law_1.01,0.12499200105667115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,4,power_law_1.01,0.14142719507217408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,4,power_law_1.01,0.16225279569625856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,4,power_law_1.01,0.2048640012741089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,4,power_law_1.01,0.2485759973526001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,4,power_law_1.01,0.3072959899902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,4,power_law_1.01,0.4112448215484619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,4,power_law_1.01,0.5774655818939209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,4,power_law_1.01,0.6910848140716552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,4,power_law_1.01,1.0240127563476562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,4,power_law_1.01,1.3198016166687012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,4,power_law_1.01,1.7030399322509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,8,power_law_1.2,0.05050240159034729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,8,power_law_1.2,0.0536191999912262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,8,power_law_1.2,0.07962239980697632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,8,power_law_1.2,0.0813696026802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,8,power_law_1.2,0.08330879807472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,8,power_law_1.2,0.08806399703025818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,8,power_law_1.2,0.10202879905700683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,8,power_law_1.2,0.10121599435806275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,8,power_law_1.2,0.0923200011253357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,8,power_law_1.2,0.14143359661102295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,8,power_law_1.2,0.12255359888076782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,8,power_law_1.2,0.11923199892044067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,8,power_law_1.2,0.19471999406814575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,8,power_law_1.2,0.19958399534225463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,8,power_law_1.2,0.20887041091918945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,8,power_law_1.2,0.20277760028839112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,8,power_law_1.2,0.18896000385284423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,8,power_law_1.2,0.18888959884643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.2,0.1968384027481079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.2,0.2110975980758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.2,0.24853758811950682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.2,0.32343039512634275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.2,0.37096319198608396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.2,0.5074687957763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.2,0.6426047801971435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.2,0.959660816192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.2,1.2250816345214843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,4,64,power_law_1.01,0.020563200116157532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,4,64,power_law_1.01,0.019865599274635316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,4,64,power_law_1.01,0.019897599518299103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,4,64,power_law_1.01,0.020467199385166168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,4,64,power_law_1.01,0.021996800601482392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,4,64,power_law_1.01,0.02682879865169525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,4,64,power_law_1.01,0.047577598690986635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,4,64,power_law_1.01,0.04922240078449249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,4,64,power_law_1.01,0.04111360013484955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,4,64,power_law_1.01,0.04191359877586365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,4,64,power_law_1.01,0.041715198755264284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,4,64,power_law_1.01,0.04291839897632599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,4,64,power_law_1.01,0.04156799912452698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,4,64,power_law_1.01,0.0421887993812561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,4,64,power_law_1.01,0.04252159893512726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,4,64,power_law_1.01,0.041433599591255185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,4,64,power_law_1.01,0.04474239945411682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,4,64,power_law_1.01,0.04671359956264496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,4,64,power_law_1.01,0.05181440114974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,4,64,power_law_1.01,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,4,64,power_law_1.01,0.07145599722862243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,4,64,power_law_1.01,0.09720320105552674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,4,64,power_law_1.01,0.11523200273513794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,4,64,power_law_1.01,0.14852479696273804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,4,64,power_law_1.01,0.20142080783843994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,4,64,power_law_1.01,0.28942720890045165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,8,power_law_1.2,0.039673599600791934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,8,power_law_1.2,0.04245119988918304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,4,64,power_law_1.01,0.3590912103652954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,8,power_law_1.2,0.04258559942245484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,8,power_law_1.2,0.04577920138835907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,8,power_law_1.2,0.04945279955863953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,8,power_law_1.2,0.049568000435829165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,8,power_law_1.2,0.04983679950237274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,8,power_law_1.2,0.051052802801132204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,8,power_law_1.2,0.052339202165603636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,8,power_law_1.2,0.051923197507858274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,8,power_law_1.2,0.05389440059661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,8,power_law_1.2,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,8,power_law_1.2,0.057126402854919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,8,power_law_1.2,0.060819202661514284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,8,power_law_1.2,0.06743040084838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,8,power_law_1.2,0.06844159960746765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,8,power_law_1.2,0.07209600210189819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,8,power_law_1.2,0.08339200019836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,8,power_law_1.2,0.0870464026927948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,8,power_law_1.2,0.10806399583816528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,8,power_law_1.2,0.12359039783477783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,8,power_law_1.2,0.17142399549484252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,8,power_law_1.2,0.225164794921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,8,power_law_1.2,0.2935296058654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,8,power_law_1.2,0.37085440158843996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,8,power_law_1.2,0.5396863937377929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,8,power_law_1.2,0.690828800201416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,2,power_law_1.01,0.0654591977596283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,2,power_law_1.01,0.08013439774513245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,2,power_law_1.01,0.09060479998588562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,2,power_law_1.01,0.11588480472564697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,2,power_law_1.01,0.14447360038757323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,2,power_law_1.01,0.15841280221939086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,2,power_law_1.01,0.20173439979553223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,2,power_law_1.01,0.20069758892059325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,2,power_law_1.01,0.19820799827575683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,2,power_law_1.01,0.2121920108795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,2,power_law_1.01,0.2163520097732544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,2,power_law_1.01,0.219596791267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,2,power_law_1.01,0.22531840801239014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,2,power_law_1.01,0.22843520641326903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,2,power_law_1.01,0.24096639156341554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,2,power_law_1.01,0.24349439144134521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,2,power_law_1.01,0.2661952018737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,2,power_law_1.01,0.29205119609832764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,2,power_law_1.01,0.3137984037399292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,2,power_law_1.01,0.373907208442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,2,power_law_1.01,0.4155519962310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,2,power_law_1.01,0.5098495960235596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,2,power_law_1.01,0.623635196685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,2,power_law_1.01,0.8093119621276855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,2,power_law_1.01,1.0262016296386718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,2,power_law_1.01,1.4206975936889648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,2,power_law_1.01,1.8218368530273437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,4,balanced,0.044010668992996216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,4,balanced,0.04378666480382284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,4,balanced,0.048341333866119385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,4,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,4,balanced,0.08205333352088928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,4,balanced,0.093231995900472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,4,balanced,0.09327999750773112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,4,balanced,0.09491200248400371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,4,balanced,0.09538666407267253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,4,balanced,0.09703999757766724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,4,balanced,0.09594666957855225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,4,balanced,0.09980266292889912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,4,balanced,0.10409067074457805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,4,balanced,0.1046453317006429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,4,balanced,0.11338667074839275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,4,balanced,0.11642666657765706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,4,balanced,0.12117866675059001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,4,balanced,0.13865066568056741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,4,balanced,0.14829333623250326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,4,balanced,0.18931732575098673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,4,balanced,0.21157866716384888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,4,balanced,0.2943306763966878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,4,balanced,0.33007999261220294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,4,balanced,0.4836426575978597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,4,balanced,0.5962666670481364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,4,balanced,0.8389066855112711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,4,balanced,1.0813120206197102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,2,2,balanced,0.03828266759713491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,2,2,balanced,0.05725333094596863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,2,2,balanced,0.08396266897519429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,2,2,balanced,0.08624000350634257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,2,2,balanced,0.08827199538548787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,2,2,balanced,0.08814932902654012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,2,2,balanced,0.08733333150545756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,2,2,balanced,0.08961600065231323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,2,2,balanced,0.09104532996813457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,2,2,balanced,0.0918933351834615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,2,2,balanced,0.0934879978497823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,2,2,balanced,0.09502399961153667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,2,2,balanced,0.09673600395520528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,2,2,balanced,0.09757866462071736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,2,2,balanced,0.11327466368675232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,2,2,balanced,0.1164479951063792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,2,2,balanced,0.11587199568748474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,2,2,balanced,0.1399999956289927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,2,2,balanced,0.1435466706752777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,2,2,balanced,0.18102399508158365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,2,2,balanced,0.21011199553807577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,2,2,balanced,0.28437866767247516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,2,2,balanced,0.3473600149154663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,2,2,balanced,0.49347201983133954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,2,2,balanced,0.615503986676534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,2,2,balanced,0.8969226678212484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,2,2,balanced,1.1624746322631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,2048,768,8,128,1,16,power_law_1.2,0.04169600009918213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,2048,768,8,128,1,16,power_law_1.2,0.03656960129737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,2048,768,8,128,1,16,power_law_1.2,0.03466239869594574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,2048,768,8,128,1,16,power_law_1.2,0.035667198896408084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,2048,768,8,128,1,16,power_law_1.2,0.035571199655532834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,2048,768,8,128,1,16,power_law_1.2,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,2048,768,8,128,1,16,power_law_1.2,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,2048,768,8,128,1,16,power_law_1.2,0.03687680065631867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,2048,768,8,128,1,16,power_law_1.2,0.03736959993839264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,2048,768,8,128,1,16,power_law_1.2,0.03765760064125061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,2048,768,8,128,1,16,power_law_1.2,0.04015359878540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,2048,768,8,128,1,16,power_law_1.2,0.039654400944709775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,2048,768,8,128,1,16,power_law_1.2,0.039852800965309146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,2048,768,8,128,1,16,power_law_1.2,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,2048,768,8,128,1,16,power_law_1.2,0.04705280065536499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,2048,768,8,128,1,16,power_law_1.2,0.047200000286102294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,2048,768,8,128,1,16,power_law_1.2,0.04989440143108368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,2048,768,8,128,1,16,power_law_1.2,0.05716480016708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,2048,768,8,128,1,16,power_law_1.2,0.059487998485565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,2048,768,8,128,1,16,power_law_1.2,0.06865919828414917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,2048,768,8,128,1,16,power_law_1.2,0.07731840014457703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,2048,768,8,128,1,16,power_law_1.2,0.10051200389862061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,2048,768,8,128,1,16,power_law_1.2,0.12037760019302368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,2048,768,8,128,1,16,power_law_1.2,0.17077759504318238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,2048,768,8,128,1,16,power_law_1.2,0.2078336000442505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,2048,768,8,128,1,16,power_law_1.2,0.29090559482574463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,16,balanced,0.023584000766277313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,16,balanced,0.023157333334287006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,16,balanced,0.021407999098300934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,16,balanced,0.025205334027608235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,16,balanced,0.027802666028340656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,16,balanced,0.028922667105992634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,16,balanced,0.029311999678611755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,16,balanced,0.031354665756225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,16,balanced,0.02932800104220708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,16,balanced,0.02977066735426585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,16,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,16,balanced,0.033946665624777474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,16,balanced,0.033717334270477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,16,1,power_law_1.2,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,16,balanced,0.03509866694609324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,16,1,power_law_1.2,0.03349120020866394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,16,balanced,0.0356480007370313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,16,1,power_law_1.2,0.041971200704574586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,16,1,power_law_1.2,0.054611200094223024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,16,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,16,balanced,0.037845333417256675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,16,balanced,0.03801066676775614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,16,balanced,0.04200000067551931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,16,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,16,balanced,0.04819199939568838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,16,balanced,0.05235733091831207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,16,balanced,0.058490668733914696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,16,balanced,0.07674133280913036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,16,1,power_law_1.2,0.056543999910354616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,16,balanced,0.09403733412424724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,16,balanced,0.11544000109036763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,16,balanced,0.1367573340733846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,16,1,power_law_1.2,0.05675519704818725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,16,1,power_law_1.2,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,16,1,power_law_1.2,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,16,1,power_law_1.2,0.06037759780883789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,16,1,power_law_1.2,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,16,1,power_law_1.2,0.060761600732803345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,16,1,power_law_1.2,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,16,1,power_law_1.2,0.06613759994506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,16,1,power_law_1.2,0.07390720248222352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,16,1,power_law_1.2,0.08325120210647582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,16,1,power_law_1.2,0.08603519797325135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,16,1,power_law_1.2,0.08199679851531982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,2048,768,8,128,1,16,power_law_1.2,0.4007743835449219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,16,1,power_law_1.2,0.0963648021221161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,16,1,power_law_1.2,0.10214400291442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,16,1,power_law_1.2,0.1236672043800354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,16,1,power_law_1.2,0.14773759841918946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,16,1,power_law_1.2,0.18651520013809203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,16,1,power_law_1.2,0.22885758876800538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,16,1,power_law_1.2,0.3038464069366455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,16,1,power_law_1.2,0.3796799898147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,1,power_law_1.01,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,1,power_law_1.01,0.03902080059051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,1,power_law_1.01,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,1,power_law_1.01,0.05988479852676391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,1,power_law_1.01,0.07259520292282104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,1,power_law_1.01,0.09145600199699402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,1,power_law_1.01,0.09591040015220642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,1,power_law_1.01,0.09783040285110474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,1,power_law_1.01,0.10273280143737792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,1,power_law_1.01,0.10609279870986939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,1,power_law_1.01,0.11223679780960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,1,power_law_1.01,0.12042239904403687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,1,power_law_1.01,0.12439680099487305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,1,power_law_1.01,0.12424960136413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,8,power_law_1.01,0.036345601081848145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,1,power_law_1.01,0.13507839441299438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,8,power_law_1.01,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,1,power_law_1.01,0.14146560430526733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,8,power_law_1.01,0.035692799091339114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,1,power_law_1.01,0.15454720258712767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,8,power_law_1.01,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,1,power_law_1.01,0.19104000329971313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,1,power_law_1.01,0.21908481121063234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,1,power_law_1.01,0.2834815979003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,1,power_law_1.01,0.3403903961181641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,1,power_law_1.01,0.466099214553833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,8,power_law_1.01,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,8,power_law_1.01,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,8,power_law_1.01,0.038700801134109494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,8,power_law_1.01,0.038815999031066896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,8,power_law_1.01,0.03920640051364899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,8,power_law_1.01,0.039001598954200745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,8,power_law_1.01,0.03928320109844208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,8,power_law_1.01,0.04100480079650879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,8,power_law_1.01,0.041145598888397215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,8,power_law_1.01,0.041971200704574586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,1,power_law_1.01,0.5840511798858643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,8,power_law_1.01,0.04650880098342895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,8,power_law_1.01,0.047065600752830505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,1,power_law_1.01,0.8253631591796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,8,power_law_1.01,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,1,power_law_1.01,1.0621952056884765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,8,power_law_1.01,0.05488640069961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.01,0.06111999750137329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,1,power_law_1.01,1.5274815559387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,1,power_law_1.01,1.991596794128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.01,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,16,1,power_law_1.2,0.5284543991088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.01,0.08658559918403626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.01,0.11447679996490479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.01,0.13212159872055054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,128,power_law_1.01,0.021849599480628968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,128,power_law_1.01,0.022361600399017335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,128,power_law_1.01,0.023078399896621703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,128,power_law_1.01,0.026464000344276428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.01,0.18563200235366822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,128,power_law_1.01,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,128,power_law_1.01,0.05455999970436096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,128,power_law_1.01,0.05270400047302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,128,power_law_1.01,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.01,0.23676800727844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,128,power_law_1.01,0.05103999972343445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,128,power_law_1.01,0.052223998308181765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,128,power_law_1.01,0.05169919729232788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,128,power_law_1.01,0.05135999917984009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,128,power_law_1.01,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,128,power_law_1.01,0.045817598700523376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,128,power_law_1.01,0.04664320051670075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,128,power_law_1.01,0.04649600088596344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,128,power_law_1.01,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,128,power_law_1.01,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.01,0.07499520182609558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.01,0.08691200017929077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.01,0.1028223991394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.01,0.13363840579986572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.01,0.17426559925079346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.01,0.22766718864440919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.01,0.2991231918334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.01,0.4804800033569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.01,0.5765759944915771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,32,power_law_1.01,0.1218559980392456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,32,power_law_1.01,0.11635839939117432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,32,power_law_1.01,0.11707520484924316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,32,power_law_1.01,0.11559679508209228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,32,power_law_1.01,0.1123263955116272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,32,power_law_1.01,0.09933440089225769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,32,power_law_1.01,0.11066880226135253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,32,power_law_1.01,0.11622400283813476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,32,power_law_1.01,0.11847039461135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,32,power_law_1.01,0.11284480094909669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,32,power_law_1.01,0.1167296051979065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,32,power_law_1.01,0.11251200437545776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,32,power_law_1.01,0.115392005443573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,32,power_law_1.01,0.11754879951477051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,32,power_law_1.01,0.12133760452270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,32,power_law_1.01,0.1196287989616394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,32,power_law_1.01,0.12597119808197021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,32,power_law_1.01,0.13603199720382692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,32,power_law_1.01,0.14880000352859496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,32,power_law_1.01,0.1688639998435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,32,power_law_1.01,0.19831039905548095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,32,power_law_1.01,0.23765759468078612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,16,power_law_1.2,0.04555520117282867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,32,power_law_1.01,0.294265604019165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,16,power_law_1.2,0.042668798565864564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,16,power_law_1.2,0.040249601006507874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,32,power_law_1.01,0.3900223970413208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,16,power_law_1.2,0.04227840006351471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,32,power_law_1.01,0.4985983848571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,32,power_law_1.01,0.6483200073242188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,32,power_law_1.01,0.9115839958190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,16,power_law_1.2,0.042656001448631284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,16,power_law_1.2,0.04514560103416443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,16,power_law_1.2,0.04549759924411774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,16,power_law_1.2,0.045516800880432126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,16,power_law_1.2,0.04609920084476471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,16,power_law_1.2,0.04588159918785095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,16,power_law_1.2,0.046828800439834596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,16,power_law_1.2,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,16,power_law_1.2,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,16,power_law_1.2,0.04895359873771667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,16,power_law_1.2,0.05329279899597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,16,power_law_1.2,0.05406079888343811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,16,power_law_1.2,0.056390398740768434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,16,power_law_1.2,0.06633599996566772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,16,power_law_1.2,0.06883839964866638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,16,power_law_1.2,0.0788096010684967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,16,power_law_1.2,0.09720320105552674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,16,power_law_1.2,0.1271999955177307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,16,power_law_1.2,0.15071359872817994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,16,1,power_law_1.2,0.6807040214538574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,16,power_law_1.2,0.20701439380645753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,16,power_law_1.2,0.2758143901824951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.01,0.3529599905014038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.01,0.4311103820800781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,16,power_law_1.2,0.401196813583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,16,power_law_1.2,0.4983488082885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,32,1,balanced,0.018405333161354065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,32,1,balanced,0.017557332913080852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,32,1,balanced,0.019018666197856266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,32,1,balanced,0.018976000448067982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,32,1,balanced,0.019061333189407986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,32,1,balanced,0.01966399947802226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,32,1,balanced,0.020074666788180668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,32,1,balanced,0.02109866589307785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,32,1,balanced,0.02102400114138921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,32,1,balanced,0.021333334346612293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,32,1,balanced,0.021104000508785248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,32,1,balanced,0.02319466571013133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,32,1,balanced,0.02476266771554947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,32,1,balanced,0.025290665527184803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,32,1,balanced,0.0435146689414978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,32,1,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,4,balanced,0.04279466470082601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,4,balanced,0.043807998299598694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,4,balanced,0.04799466828505198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,4,balanced,0.057855998476346336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,4,balanced,0.08447999755541484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,4,balanced,0.09113599856694539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,4,balanced,0.0926026701927185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,4,balanced,0.0932373305161794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,4,balanced,0.09217066566149394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,4,balanced,0.09460266431172688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,4,balanced,0.09462933739026387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,4,balanced,0.09659733374913533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,4,balanced,0.09703999757766724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,4,balanced,0.10175466537475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,32,1,balanced,0.04560000201066335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,32,1,balanced,0.03979733337958654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,32,1,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,32,1,balanced,0.048485333720842995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,32,1,balanced,0.05222400029500326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,32,1,balanced,0.07181866466999054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,32,1,balanced,0.08160000046094258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,32,1,balanced,0.10971732934315999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,32,1,balanced,0.13571733236312866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,32,1,balanced,0.19089066982269287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,32,1,balanced,0.2384693423906962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,4,balanced,0.10613866647084554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,4,balanced,0.11128532886505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,4,balanced,0.11375466982523601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,4,balanced,0.12713600198427835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,4,balanced,0.13154133160909018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,4,balanced,0.16873600085576376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,4,balanced,0.18517333269119263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,4,balanced,0.2526346643765767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,4,balanced,0.27487999200820923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,4,balanced,0.3965386549631755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,4,balanced,0.48199466864267987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,4,balanced,0.6569546858469645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,4,balanced,0.8548906644185384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,64,power_law_1.2,0.02222079932689667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,64,power_law_1.2,0.01886080056428909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,64,power_law_1.2,0.018572799861431122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,64,power_law_1.2,0.018988800048828126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,64,power_law_1.2,0.02122880071401596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,64,power_law_1.2,0.025209599733352663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,64,power_law_1.2,0.02529279887676239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,64,power_law_1.2,0.038950398564338684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,64,power_law_1.2,0.03883520066738129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,64,power_law_1.2,0.03925119936466217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,4,power_law_1.01,0.09757440090179444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,4,power_law_1.01,0.14549119472503663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,64,power_law_1.2,0.037862399220466615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,4,power_law_1.01,0.24798719882965087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,4,power_law_1.01,0.2568320035934448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,64,power_law_1.2,0.03826560080051422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,4,power_law_1.01,0.3489984035491943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,4,power_law_1.01,0.397107195854187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,4,power_law_1.01,0.4372543811798096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,64,power_law_1.2,0.03767040073871612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,64,power_law_1.2,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,4,power_law_1.01,0.5173247814178467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,64,power_law_1.2,0.0350847989320755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,64,power_law_1.2,0.03852159976959228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,4,power_law_1.01,0.5404032230377197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,4,power_law_1.01,0.5369088172912597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,4,power_law_1.01,0.5555136203765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,4,power_law_1.01,0.5237247943878174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,64,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,4,power_law_1.01,0.5412543773651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,64,power_law_1.2,0.04510720074176788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.2,0.05377920269966126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,4,power_law_1.01,0.5540287971496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.2,0.059622400999069215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,4,power_law_1.01,0.5247615814208985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,4,power_law_1.01,0.5685823917388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.2,0.07120000123977661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,4,power_law_1.01,0.5769152164459228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.2,0.09029759764671326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,4,power_law_1.01,0.6218239784240722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.2,0.11374080181121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,4,power_law_1.01,0.7039872169494629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,4,power_law_1.01,0.7807680130004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,4,power_law_1.01,0.790502405166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,4,power_law_1.01,0.8981120109558105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,4,power_law_1.01,1.0291328430175781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,4,power_law_1.01,1.277843189239502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,4,power_law_1.01,1.3463552474975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,4,power_law_1.01,1.8104639053344727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.2,0.15923839807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,4,power_law_1.01,2.3002496719360352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,1,power_law_1.2,0.13049600124359131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,1,power_law_1.2,0.1865664005279541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,1,power_law_1.2,0.3219072103500366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,1,power_law_1.2,0.4911488056182861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,1,power_law_1.2,0.6532608032226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,1,power_law_1.2,1.204531192779541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.2,0.21164801120758056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,1,power_law_1.2,1.306227207183838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,1,power_law_1.2,1.3604800224304199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,1,power_law_1.2,1.4014399528503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,1,power_law_1.2,1.4643967628479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,1,power_law_1.2,1.5058176040649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,1,power_law_1.2,1.5280192375183106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,1,power_law_1.2,1.5757696151733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,1,power_law_1.2,1.6542463302612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,1,power_law_1.2,1.7104511260986328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,1,power_law_1.2,1.7524927139282227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,1,power_law_1.2,1.8309055328369142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,1,power_law_1.2,2.0175615310668946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,1,power_law_1.2,2.2131584167480467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,1,power_law_1.2,2.606809616088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,1,power_law_1.2,2.5731199264526365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,1,power_law_1.2,3.1540992736816404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,1,power_law_1.2,3.256595230102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,1,power_law_1.2,3.9889152526855467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,1,power_law_1.2,5.360262298583985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,1,power_law_1.2,6.958918762207031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,1,power_law_1.2,8.179596710205079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.2,0.2927743911743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,1,64,power_law_1.01,0.02069759964942932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,1,64,power_law_1.01,0.022265599668025972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,1,64,power_law_1.01,0.022176000475883483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,1,64,power_law_1.01,0.018387199938297273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,1,64,power_law_1.01,0.01988479942083359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,1,64,power_law_1.01,0.021887999773025513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,1,64,power_law_1.01,0.023500800132751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,1,64,power_law_1.01,0.02834559977054596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,1,64,power_law_1.01,0.02839680016040802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,1,64,power_law_1.01,0.02927359938621521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,1,64,power_law_1.01,0.039526399970054624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,1,64,power_law_1.01,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,1,64,power_law_1.01,0.03951359987258911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,1,64,power_law_1.01,0.03988479971885681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,1,64,power_law_1.01,0.041152000427246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,1,64,power_law_1.01,0.049753600358963014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,1,64,power_law_1.01,0.051744002103805545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,1,64,power_law_1.01,0.05607680082321167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,1,64,power_law_1.01,0.06618880033493042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,1,64,power_law_1.01,0.09288319945335388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,1,64,power_law_1.01,0.061292797327041626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,1,64,power_law_1.01,0.08023040294647217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,1,64,power_law_1.01,0.09775360226631165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,1,64,power_law_1.01,0.12972160577774047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,1,64,power_law_1.01,0.1717311978340149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,1,64,power_law_1.01,0.23656320571899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,1,64,power_law_1.01,0.32063999176025393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,2,power_law_1.2,0.04709759950637817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,2,power_law_1.2,0.05235840082168579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,2,power_law_1.2,0.060198402404785155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,2,power_law_1.2,0.08113920092582702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,2,power_law_1.2,0.09378560185432434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,2,power_law_1.2,0.10521600246429444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,2,power_law_1.2,0.13727999925613404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,2,power_law_1.2,0.1344256043434143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,2,power_law_1.2,0.14199680089950562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.2,0.38471040725708006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,2,power_law_1.2,0.1488703966140747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,2,power_law_1.2,0.15404160022735597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,2,power_law_1.2,0.15983999967575074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,2,power_law_1.2,0.1638335943222046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,2,power_law_1.2,0.17428480386734008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,2,power_law_1.2,0.1880511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,2,power_law_1.2,0.19946880340576173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,2,power_law_1.2,0.20178558826446533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,2,power_law_1.2,0.2350719928741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,2,power_law_1.2,0.27171199321746825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,2,power_law_1.2,0.3237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,2,power_law_1.2,0.38105599880218505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,2,power_law_1.2,0.4993919849395752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,2,power_law_1.2,0.5738687992095948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,2,power_law_1.2,0.7872384071350098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,2,power_law_1.2,1.0116543769836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,2,power_law_1.2,1.393395233154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,2,power_law_1.2,1.8385984420776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,128,power_law_1.01,0.05629439949989319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,128,power_law_1.01,0.057625597715377806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,128,power_law_1.01,0.057043200731277464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,128,power_law_1.01,0.06004480123519897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,128,power_law_1.01,0.05667200088500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,128,power_law_1.01,0.05889279842376709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,128,power_law_1.01,0.060499197244644164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,128,power_law_1.01,0.060147202014923094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,128,power_law_1.01,0.06069120168685913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,128,power_law_1.01,0.061849600076675414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,128,power_law_1.01,0.061414402723312375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,128,power_law_1.01,0.06438400149345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,128,power_law_1.01,0.06496000289916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,128,power_law_1.01,0.06535680294036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,128,power_law_1.01,0.07322239875793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,128,power_law_1.01,0.07557119727134705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,128,power_law_1.01,0.08329600095748901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,128,power_law_1.01,0.09646080136299133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,128,power_law_1.01,0.10634880065917969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,128,power_law_1.01,0.13657599687576294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,128,power_law_1.01,0.15464320182800292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,128,power_law_1.01,0.20206079483032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,128,power_law_1.01,0.21427841186523439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,128,power_law_1.01,0.2633919954299927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,128,power_law_1.01,0.33362560272216796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,128,power_law_1.01,0.44440321922302245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,128,power_law_1.01,0.5753856182098389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,32,8,balanced,0.04610133171081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,32,8,balanced,0.04394133388996124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,32,8,balanced,0.04379733403523763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,32,8,balanced,0.04607999821503957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,32,8,balanced,0.04794133206208547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,32,8,balanced,0.047653332352638245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,32,8,balanced,0.047541335225105286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,32,8,balanced,0.045893331368764244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,32,8,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,32,8,balanced,0.04810666541258494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,32,8,balanced,0.04688533147176107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,32,8,balanced,0.04779199759165446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,32,8,balanced,0.0462773342927297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,32,8,balanced,0.048298666874567665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,32,8,balanced,0.05194133520126343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,32,8,balanced,0.049957334995269775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,32,8,balanced,0.05198400219281515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,32,8,balanced,0.054858664671579994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,32,8,balanced,0.05787733197212219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,32,8,balanced,0.060191998879114784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,32,8,balanced,0.06623999774456024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,32,8,balanced,0.07764799892902374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,32,8,balanced,0.08822400371233623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,32,8,balanced,0.10709866881370544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,32,8,balanced,0.12293866276741028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,32,8,balanced,0.17097065846125284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,32,8,balanced,0.20638400316238403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,8,8,power_law_1.2,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,8,8,power_law_1.2,0.04686079919338226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,8,8,power_law_1.2,0.045440000295639035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,8,8,power_law_1.2,0.05438719987869263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,8,8,power_law_1.2,0.060313600301742556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,8,8,power_law_1.2,0.055308800935745236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,8,8,power_law_1.2,0.06048640012741089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,8,8,power_law_1.2,0.057132798433303836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,8,8,power_law_1.2,0.061459201574325564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,8,8,power_law_1.2,0.06097919940948486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,8,8,power_law_1.2,0.06425600051879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,8,8,power_law_1.2,0.06796799898147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,8,8,power_law_1.2,0.06970880031585694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,8,8,power_law_1.2,0.07278079986572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,8,8,power_law_1.2,0.07687039971351624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,8,8,power_law_1.2,0.07996799945831298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,8,8,power_law_1.2,0.08364800214767457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,8,8,power_law_1.2,0.0970304012298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,8,8,power_law_1.2,0.09821439981460571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,8,8,power_law_1.2,0.11599999666213989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,8,8,power_law_1.2,0.1307584047317505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,8,8,power_law_1.2,0.16354559659957885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,8,8,power_law_1.2,0.1994752049446106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,8,8,power_law_1.2,0.2638144016265869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,8,8,power_law_1.2,0.323142409324646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,8,8,power_law_1.2,0.4721792221069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,8,8,power_law_1.2,0.6423039913177491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,1,power_law_1.2,0.04853120148181915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,1,power_law_1.2,0.05626239776611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,1,power_law_1.2,0.07529600262641907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,1,power_law_1.2,0.0944703996181488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,1,power_law_1.2,0.11630079746246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,1,power_law_1.2,0.17086080312728882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,1,power_law_1.2,0.18231680393218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,1,power_law_1.2,0.19578880071640015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,1,power_law_1.2,0.1994752049446106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,1,power_law_1.2,0.20366079807281495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,1,power_law_1.2,0.21385600566864013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,1,power_law_1.2,0.22568960189819337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,1,power_law_1.2,0.23343360424041748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,1,power_law_1.2,0.23617920875549317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,1,power_law_1.2,0.24730238914489747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,1,power_law_1.2,0.26256000995635986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,1,power_law_1.2,0.28243839740753174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,1,power_law_1.2,0.3260672092437744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,1,power_law_1.2,0.37075200080871584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,1,power_law_1.2,0.45450239181518554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,1,balanced,0.10331733028093974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,1,power_law_1.2,0.5322624206542969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,1,power_law_1.2,0.6894336223602295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,4,power_law_1.2,0.053913599252700804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,1,power_law_1.2,0.8995967864990234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,1,balanced,0.1357439955075582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,1,balanced,0.1906399925549825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,1,balanced,0.2785173257191976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,1,power_law_1.2,1.1885120391845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,4,power_law_1.2,0.06328960061073304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,4,power_law_1.2,0.08191360235214233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,4,power_law_1.2,0.09103360176086425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,4,power_law_1.2,0.11144959926605225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,1,power_law_1.2,1.6400127410888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,4,power_law_1.2,0.1504320025444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,4,power_law_1.2,0.15808639526367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,1,balanced,0.42716801166534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,4,power_law_1.2,0.16514559984207153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,1,balanced,0.7136639753977457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,4,power_law_1.2,0.16841599941253663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,1,balanced,0.6933759848276774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,4,power_law_1.2,0.17215360403060914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,1,balanced,0.6746133168538412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,4,power_law_1.2,0.18528640270233154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,1,balanced,0.6631306807200114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,4,power_law_1.2,0.1923967957496643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,4,power_law_1.2,0.21470720767974855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,1,power_law_1.2,2.184351921081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,4,power_law_1.2,0.20176639556884765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,4,power_law_1.2,0.218720006942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,1,balanced,0.6564586559931437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,4,power_law_1.2,0.22880001068115235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,1,balanced,0.6668266455332438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,1,balanced,0.6625440120697021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,4,power_law_1.2,0.21463680267333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,1,balanced,0.6653279860814413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,4,power_law_1.2,0.23625600337982178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,1,balanced,0.6645119984944662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,1,balanced,0.6637119849522909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.2,0.34028160572052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,1,balanced,0.6768906911214193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.2,0.3362943887710571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.2,0.37341439723968506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.2,0.47585282325744627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.2,0.5247744083404541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.2,0.7417664051055908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,1,balanced,0.7008586724599203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,1,balanced,0.7245493729909261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.2,0.8522879600524902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,1,balanced,0.7602880001068115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,1,balanced,0.8218293190002441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,1,balanced,0.880890687306722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,1,balanced,1.0098666350046794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,1,balanced,1.1411306858062744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,1,balanced,1.884943962097168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,1,balanced,2.1246612866719565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.2,1.132761573791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,1,balanced,3.1503467559814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,1,balanced,3.865034739176432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.2,1.6788799285888671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,1,power_law_1.2,2.665465545654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,1,128,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,1,128,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,1,128,balanced,0.033471999069054924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,1,128,balanced,0.03550933301448822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,1,128,balanced,0.03562666724125544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,1,128,balanced,0.037674665451049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,1,128,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,1,128,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,1,128,balanced,0.037776000797748566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,1,128,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,1,128,balanced,0.03788800040880839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,1,128,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,1,128,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,1,128,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,1,128,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,1,128,balanced,0.040031999349594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,1,128,balanced,0.04223999877770742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,1,128,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,1,128,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,4,power_law_1.01,0.0929535984992981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,1,128,balanced,0.043882668018341064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,1,128,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,4,power_law_1.01,0.11975680589675904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,1,128,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,1,128,balanced,0.05064533154169718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,1,128,balanced,0.05795200169086456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,4,power_law_1.01,0.17694079875946045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,4,power_law_1.01,0.19022719860076903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,4,power_law_1.01,0.24531838893890381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,4,power_law_1.01,0.29047040939331054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,4,power_law_1.01,0.25760641098022463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,1,128,balanced,0.06272533535957336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,1,128,balanced,0.07845866680145264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,4,power_law_1.01,0.3531584024429321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,1,128,balanced,0.09453333417574565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,4,power_law_1.01,0.3956736087799072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,4,power_law_1.01,0.35564799308776857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,4,power_law_1.01,0.33055360317230226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,4,power_law_1.01,0.38586881160736086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,4,power_law_1.01,0.370412802696228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,4,power_law_1.01,0.3687488079071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,4,power_law_1.01,0.37073280811309817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,4,power_law_1.01,0.39676799774169924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,4,power_law_1.01,0.4161664009094238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,4,power_law_1.01,0.44736638069152834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,4,power_law_1.01,0.4650559902191162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,4,power_law_1.01,0.5129856109619141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,4,power_law_1.01,0.5723648071289062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,4,power_law_1.01,0.721017599105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,4,power_law_1.01,0.7804096221923829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,4,power_law_1.01,0.9833151817321777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,4,power_law_1.01,1.2561792373657226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,4,power_law_1.01,1.5779135704040528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,4,power_law_1.01,2.0909887313842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,1,power_law_1.2,0.023449599742889404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,1,power_law_1.2,0.029606398940086365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,1,power_law_1.2,0.046163201332092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,1,power_law_1.2,0.06878079771995545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,1,power_law_1.2,0.09096320271492005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,1,power_law_1.2,0.12996480464935303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,1,power_law_1.2,0.1731328010559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,1,power_law_1.2,0.18180480003356933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,1,power_law_1.2,0.18439040184020997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,1,power_law_1.2,0.193503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,1,power_law_1.2,0.20212481021881104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,1,power_law_1.2,0.21395840644836425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,1,power_law_1.2,0.22485120296478273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,1,power_law_1.2,0.2531264066696167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,1,power_law_1.2,0.2374336004257202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,1,power_law_1.2,0.24863998889923095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,1,power_law_1.2,0.2736191987991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,1,power_law_1.2,0.27548799514770506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.2,0.30728960037231445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.2,0.3272639989852905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.2,0.36798079013824464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.2,0.7912447929382325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.2,0.9224896430969238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.2,0.6986879825592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.2,0.808563232421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.2,1.0537471771240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.2,1.3011199951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,2,balanced,0.019018666197856266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,2,balanced,0.01922133316596349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,2,balanced,0.020853333175182343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,2,balanced,0.023317334552605946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,2,balanced,0.031285333136717476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,2,balanced,0.0454720010360082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,2,balanced,0.047509332497914634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,2,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,2,balanced,0.04784533381462097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,2,balanced,0.04818133513132731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,2,balanced,0.05157333115736643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,2,balanced,0.05349866549173991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,2,balanced,0.05487466851870219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,2,balanced,0.05513066550095876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,2,balanced,0.058042665322621666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,2,balanced,0.06014933188756307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,2,balanced,0.06292800108591716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,2,balanced,0.07107733190059662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,2,balanced,0.07860800127188365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,2,balanced,0.10586667060852051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,2,balanced,0.10744532942771912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,2,balanced,0.1537813345591227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,2,balanced,0.18342934052149454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,2,balanced,0.1999839941660563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,2,balanced,0.21556266148885092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,2,balanced,0.36582398414611816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,2,balanced,0.39767467975616455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,64,balanced,0.0543039987484614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,64,balanced,0.054234668612480164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,64,balanced,0.054192001620928444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,64,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,64,balanced,0.06241600215435028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,64,balanced,0.0724373310804367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,64,balanced,0.08380267024040222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,64,balanced,0.08243200182914734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,64,balanced,0.08453333377838135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,64,balanced,0.08290666838486989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,64,balanced,0.08902399738629659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,64,balanced,0.08318399886290233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,64,balanced,0.08649599552154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,64,balanced,0.08545066912968953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,64,balanced,0.09889066219329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,64,balanced,0.09362666805585225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,64,balanced,0.09922666351000468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,64,balanced,0.10797866185506184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,64,balanced,0.11945066849390666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,64,balanced,0.13396799564361572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,32,power_law_1.2,0.06228479743003845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,64,balanced,0.14524799585342407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,32,power_law_1.2,0.06079360246658325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,64,balanced,0.18082133928934732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,64,balanced,0.20973867177963257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,32,power_law_1.2,0.05850239992141724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,32,power_law_1.2,0.057183998823165896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,32,power_law_1.2,0.05738239884376526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,32,power_law_1.2,0.062003201246261595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,32,power_law_1.2,0.06232960224151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,32,power_law_1.2,0.06269440054893494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,32,power_law_1.2,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,32,power_law_1.2,0.06305279731750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,32,power_law_1.2,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,32,power_law_1.2,0.06513280272483826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,32,power_law_1.2,0.06506239771842956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,32,power_law_1.2,0.06670719981193543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,32,power_law_1.2,0.07215999960899352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,32,power_law_1.2,0.07153919935226441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,32,power_law_1.2,0.07565439939498901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,32,power_law_1.2,0.08503040075302123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,32,power_law_1.2,0.09578880071640014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,32,power_law_1.2,0.11391999721527099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,32,power_law_1.2,0.13509759902954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,32,power_law_1.2,0.16713600158691405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,32,power_law_1.2,0.19905279874801635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,64,balanced,0.2749813397725423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,64,balanced,0.3556693394978841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,64,balanced,0.47441065311431885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,64,balanced,0.613973339398702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,32,power_law_1.2,0.29270401000976565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,32,power_law_1.2,0.36450560092926027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,2,8,balanced,0.060175999999046326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,2,8,balanced,0.058933332562446594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,2,8,balanced,0.059808000922203064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,2,8,balanced,0.060191998879114784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,32,power_law_1.2,0.5450047969818115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,2,8,balanced,0.06451733410358429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,2,8,balanced,0.06122133135795593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,2,8,balanced,0.06225066880385081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,2,8,balanced,0.06299200157324474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,2,8,balanced,0.06229866544405619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,2,8,balanced,0.06186666587988535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,2,8,balanced,0.062319998939832054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,2,8,balanced,0.0576853354771932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,2,8,balanced,0.060005332032839455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,2,8,balanced,0.058592001597086586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,2,8,balanced,0.056101332108179726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,2,8,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,2,8,balanced,0.055973331133524575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,2,8,balanced,0.07204799850781758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,2,8,balanced,0.06794133285681407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,2,8,balanced,0.08854400118192036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,2,8,balanced,0.09633066256841023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,2,8,balanced,0.12533332904179892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,2,8,balanced,0.14891733725865683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,2,8,balanced,0.19795199235280356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,2,8,balanced,0.24311999479929605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,4,64,balanced,0.052842666705449425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,2,8,balanced,0.33855998516082764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,4,64,balanced,0.05239999790986379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,2,8,balanced,0.4355786641438802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,4,64,balanced,0.05023466547330221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,4,64,balanced,0.05637866755326589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,4,64,balanced,0.058730666836102806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,4,64,balanced,0.05912533402442932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,4,64,balanced,0.060271998246510826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,4,64,balanced,0.0584746648867925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,4,64,balanced,0.060602664947509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,4,64,balanced,0.05986666679382324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,4,64,balanced,0.06046399970849355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,4,64,balanced,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,4,64,balanced,0.06404800216356914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,4,64,balanced,0.0647680014371872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,4,64,balanced,0.07094933092594147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,4,64,balanced,0.07201600074768066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,4,64,balanced,0.07673599819342296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,4,64,balanced,0.09301867087682088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,4,64,balanced,0.09588799873987834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,4,64,balanced,0.12213333447774251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,4,64,balanced,0.13201600313186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,4,64,balanced,0.17550933361053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,4,64,balanced,0.1665173371632894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,4,64,balanced,0.2158506711324056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,4,64,balanced,0.23093332846959433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,4,64,balanced,0.30315200487772626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,4,64,balanced,0.3534453312555949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,32,power_law_1.2,0.7066175937652588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,1,128,balanced,0.03965866565704346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,1,128,balanced,0.03976000100374222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,1,128,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,1,128,balanced,0.040607998768488564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,1,128,balanced,0.04166933397452036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,1,128,balanced,0.043621331453323364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,1,128,balanced,0.04181866844495138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,1,128,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,1,128,balanced,0.04368533194065094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,1,128,balanced,0.04391466577847799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,1,128,balanced,0.04275199770927429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,1,128,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,1,128,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,1,128,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,1,128,balanced,0.04632533093293508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,1,128,balanced,0.04794133206208547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,1,128,balanced,0.04971200227737427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,1,128,balanced,0.05093333125114441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,1,128,balanced,0.05043200155099233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,1,128,balanced,0.056202664971351624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,1,128,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,1,128,balanced,0.07041066884994507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,1,128,balanced,0.07770133515199025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,1,128,balanced,0.10284266869227092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,1,128,balanced,0.12174399693806966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,1,128,balanced,0.15796800454457602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,1,128,balanced,0.19455466667811075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,2,balanced,0.10761599739392598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,2,balanced,0.11117866635322571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,2,balanced,0.12518399953842163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,2,balanced,0.15987733006477356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,2,balanced,0.2079733411471049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,2,balanced,0.31593600908915204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,2,balanced,0.31595200300216675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,2,balanced,0.3171893358230591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,2,balanced,0.3198026617368062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,2,balanced,0.3184853394826253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,2,balanced,0.32203733921051025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,2,balanced,0.3306079904238383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,2,balanced,0.33292800188064575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,2,balanced,0.34113065401713055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,2,balanced,0.3445386489232381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,2,balanced,0.34946131706237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,2,balanced,0.3571093479792277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,2,balanced,0.38572800159454346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,2,balanced,0.40691200892130536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,2,balanced,0.4623039960861206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,2,balanced,0.5150293509165446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,2,balanced,0.6152533292770386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,2,balanced,0.7191253503163656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,2,balanced,1.0095093250274658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,2,balanced,1.2191200256347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,1,power_law_1.2,0.0412416011095047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,1,power_law_1.2,0.045023998618125914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,1,power_law_1.2,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,1,power_law_1.2,0.05726720094680786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,1,power_law_1.2,0.05797119736671448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,1,power_law_1.2,0.05625600218772888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,1,power_law_1.2,0.057683199644088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,1,power_law_1.2,0.055955201387405396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,1,power_law_1.2,0.056934398412704465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,1,power_law_1.2,0.062015998363494876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,1,power_law_1.2,0.0565887987613678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,1,power_law_1.2,0.05904639959335327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,1,power_law_1.2,0.05971199870109558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,1,power_law_1.2,0.06104959845542908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,1,power_law_1.2,0.06757760047912598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,1,power_law_1.2,0.07080960273742676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,1,power_law_1.2,0.07777919769287109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,1,power_law_1.2,0.09015039801597595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.2,0.0995136022567749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.2,0.12300159931182861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.2,0.14149119853973388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.2,0.1857599973678589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.2,0.23252480030059813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.2,0.32467200756073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,8,power_law_1.01,0.02686080038547516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,8,power_law_1.01,0.02536959946155548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,8,power_law_1.01,0.02449920028448105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,8,power_law_1.01,0.024697600305080412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,8,power_law_1.01,0.024934400618076325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,8,power_law_1.01,0.02826879918575287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,8,power_law_1.01,0.05682560205459595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,8,power_law_1.01,0.05859839916229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,8,power_law_1.01,0.059487998485565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,8,power_law_1.01,0.05318400263786316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,8,power_law_1.01,0.04475519955158234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,8,power_law_1.01,0.08193920254707336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,8,power_law_1.01,0.07857279777526856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,8,power_law_1.01,0.060153597593307497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,8,power_law_1.01,0.0907904028892517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,8,power_law_1.01,0.07285760045051574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,8,power_law_1.01,0.09496319890022278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,8,power_law_1.01,0.1171072006225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.01,0.14265600442886353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.01,0.19983999729156493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.01,0.25654399394989014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.2,0.417523193359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.01,0.3577215909957886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.01,0.467955207824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,4,4,power_law_1.01,0.05370879769325256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,4,4,power_law_1.01,0.056403201818466184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,4,4,power_law_1.01,0.058297598361968996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,4,4,power_law_1.01,0.0673792004585266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.01,0.6733312129974365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,4,4,power_law_1.01,0.07252479791641235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,4,4,power_law_1.01,0.07574399709701538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,4,4,power_law_1.01,0.08451840281486511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,4,4,power_law_1.01,0.08537600040435792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,4,4,power_law_1.01,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,4,4,power_law_1.01,0.08784000277519226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,4,4,power_law_1.01,0.0898688018321991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,4,4,power_law_1.01,0.09345920085906982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,4,4,power_law_1.01,0.09566079974174499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,4,4,power_law_1.01,0.09723520278930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,4,4,power_law_1.01,0.10497920513153076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,4,4,power_law_1.01,0.10722559690475464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,4,4,power_law_1.01,0.11201280355453491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,4,4,power_law_1.01,0.12791680097579955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,4,4,power_law_1.01,0.13738880157470704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,4,4,power_law_1.01,0.16268800497055053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,4,4,power_law_1.01,0.18991999626159667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,4,4,power_law_1.01,0.23987839221954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,4,4,power_law_1.01,0.2930943965911865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,4,4,power_law_1.01,0.3967616081237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,4,4,power_law_1.01,0.5092095851898193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,4,4,power_law_1.01,0.7001215934753418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,4,4,power_law_1.01,0.9660032272338868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.01,0.8864895820617675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,8,balanced,0.04875733455022176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,8,balanced,0.050570666790008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,8,balanced,0.05502399802207947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,8,balanced,0.07971733311812083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,8,balanced,0.1002400020758311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,8,balanced,0.11570133765538533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,8,balanced,0.11529599626859029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,8,balanced,0.11664000153541565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,8,balanced,0.11689600348472595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,8,balanced,0.1178986628850301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,8,balanced,0.11687999963760376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,8,balanced,0.12065066893895467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,8,balanced,0.12272533774375916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,8,balanced,0.12417067090670268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,8,balanced,0.13159466783205667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,8,balanced,0.13395733634630838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,8,balanced,0.14113066593805948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,8,balanced,0.15957333644231161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,8,balanced,0.16709333658218384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,8,balanced,0.220085342725118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,8,balanced,0.2360960046450297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,8,balanced,0.3285813331604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,8,balanced,0.3232213258743286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,8,balanced,0.46533334255218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2688,1856,6,128,1,4,power_law_1.01,0.024345600605010988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,8,balanced,0.5414133469263712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2688,1856,6,128,1,4,power_law_1.01,0.035436800122261046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,8,balanced,0.7731093565622965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,8,balanced,0.9430293242136637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.2,0.5920447826385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2688,1856,6,128,1,4,power_law_1.01,0.03485440015792847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2688,1856,6,128,1,4,power_law_1.01,0.043942400813102724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2688,1856,6,128,1,4,power_law_1.01,0.04661760032176972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2688,1856,6,128,1,4,power_law_1.01,0.05326719880104065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2688,1856,6,128,1,4,power_law_1.01,0.054790401458740236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2688,1856,6,128,1,4,power_law_1.01,0.057017600536346434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2688,1856,6,128,1,4,power_law_1.01,0.05829120278358459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2688,1856,6,128,1,4,power_law_1.01,0.059033602476119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2688,1856,6,128,1,4,power_law_1.01,0.060710400342941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2688,1856,6,128,1,4,power_law_1.01,0.061459201574325564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2688,1856,6,128,1,4,power_law_1.01,0.06514559984207154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2688,1856,6,128,1,4,power_law_1.01,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2688,1856,6,128,1,4,power_law_1.01,0.06895999908447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,32,power_law_1.01,0.047891199588775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,32,power_law_1.01,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,32,power_law_1.01,0.04552960097789764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.01,1.3075200080871583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,32,power_law_1.01,0.047705599665641786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,32,power_law_1.01,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,32,power_law_1.01,0.05190399885177612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,32,power_law_1.01,0.05331839919090271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,32,power_law_1.01,0.0525056004524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,32,power_law_1.01,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,32,power_law_1.01,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,32,power_law_1.01,0.054655998945236206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,32,power_law_1.01,0.057606399059295654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,32,power_law_1.01,0.059571200609207155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,32,power_law_1.01,0.060844802856445314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.01,1.7259712219238281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,32,power_law_1.01,0.06410880088806152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,32,power_law_1.01,0.06419199705123901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,32,power_law_1.01,0.06992639899253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,32,power_law_1.01,0.0766975998878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,32,power_law_1.01,0.08721280097961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,32,power_law_1.01,0.10486400127410889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,32,power_law_1.01,0.11990400552749633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,32,power_law_1.01,0.16418559551239015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,32,power_law_1.01,0.18920960426330566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,32,power_law_1.01,0.2610687971115112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,32,power_law_1.01,0.3137984037399292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,32,power_law_1.01,0.45866241455078127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,32,power_law_1.01,0.5883584022521973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.2,0.7772031784057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,1,4,power_law_1.2,0.08750079870223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,1,4,power_law_1.2,0.10088959932327271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,1,4,power_law_1.2,0.10959359407424926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,1,4,power_law_1.2,0.12700159549713136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,1,4,power_law_1.2,0.12400000095367432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,1,4,power_law_1.2,0.12051199674606324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,1,4,power_law_1.2,0.12222080230712891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,1,4,power_law_1.2,0.11811200380325318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,1,4,power_law_1.2,0.1164736032485962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,1,4,power_law_1.2,0.11801600456237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,1,4,power_law_1.2,0.11032320261001587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,1,4,power_law_1.2,0.11936000585556031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,1,4,power_law_1.2,0.14039679765701293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,1,4,power_law_1.2,0.13334399461746216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,1,4,power_law_1.2,0.15159679651260377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,1,4,power_law_1.2,0.16330239772796631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,1,4,power_law_1.2,0.18645119667053223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,1,4,power_law_1.2,0.21276159286499025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.2,0.2812096118927002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.2,0.37500801086425783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,4,power_law_1.2,0.09349759817123413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,4,power_law_1.2,0.11930880546569825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,4,power_law_1.2,0.15247999429702758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,4,power_law_1.2,0.17989120483398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,4,power_law_1.2,0.22868480682373046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,4,power_law_1.2,0.2949120044708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,4,power_law_1.2,0.2810175895690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.2,0.4236095905303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,4,power_law_1.2,0.3609344005584717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,4,power_law_1.2,0.39187839031219485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,4,power_law_1.2,0.332806396484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,4,power_law_1.2,0.33413760662078856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,4,power_law_1.2,0.3441472053527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,4,power_law_1.2,0.37148160934448243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,4,power_law_1.2,0.38404479026794436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,4,power_law_1.2,0.39246718883514403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.2,0.5401663780212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,4,power_law_1.2,0.3873984098434448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.2,0.6974976062774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,4,power_law_1.2,0.4141439914703369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,4,power_law_1.2,0.4369408130645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,4,power_law_1.2,0.4723328113555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,4,power_law_1.2,0.5406911849975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,4,power_law_1.2,0.6063551902770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,4,power_law_1.2,0.763756799697876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,4,power_law_1.2,0.8126399993896485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,4,power_law_1.2,1.02041597366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,4,power_law_1.2,1.4414848327636718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,4,power_law_1.2,1.7668224334716798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,4,power_law_1.2,2.1303743362426757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.2,0.989356803894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.2,1.6433664321899415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.2,1.862451171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.2,2.404979133605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,8,balanced,0.05007466673851013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,8,balanced,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,8,balanced,0.054416000843048096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,8,balanced,0.05995733539263407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,8,balanced,0.06609066824118297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,8,balanced,0.07869866490364075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,8,balanced,0.08991466959317525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,8,balanced,0.09099200367927551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,8,balanced,0.09136533737182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,8,balanced,0.09290132919947307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,8,balanced,0.09264533718427022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,8,balanced,0.09693866968154907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,8,balanced,0.09849066535631816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,8,balanced,0.0981173316637675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,8,balanced,0.10501333077748616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,8,balanced,0.10802666346232097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,8,balanced,0.11762133240699768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,8,balanced,0.12924800316492716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,8,balanced,0.14235732952753702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,8,balanced,0.16711999972661337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,8,balanced,0.18058133125305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,8,balanced,0.22472000122070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,8,balanced,0.27046932776769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,8,balanced,0.35953064759572345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,8,balanced,0.48677865664164227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,8,balanced,0.6558560132980347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,8,balanced,0.8559786478678385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,4,16,power_law_1.01,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,4,16,power_law_1.01,0.03187200129032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,4,16,power_law_1.01,0.031052801012992858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,4,16,power_law_1.01,0.030131199955940248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,4,16,power_law_1.01,0.030124801397323608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,4,16,power_law_1.01,0.03261440098285675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,4,16,power_law_1.01,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,4,16,power_law_1.01,0.03230080008506775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,4,16,power_law_1.01,0.03237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,4,16,power_law_1.01,0.032742398977279666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,4,16,power_law_1.01,0.033228799700737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,4,16,power_law_1.01,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,4,16,power_law_1.01,0.03463680148124695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,4,16,power_law_1.01,0.03573760092258453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,4,16,power_law_1.01,0.03909119963645935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,4,16,power_law_1.01,0.0398719996213913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,4,16,power_law_1.01,0.043110400438308716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,4,16,power_law_1.01,0.049420800805091855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,4,16,power_law_1.01,0.048416000604629514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,4,16,power_law_1.01,0.05438719987869263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,4,16,power_law_1.01,0.060127997398376466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,4,16,power_law_1.01,0.07552000284194946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,4,16,power_law_1.01,0.08750720024108886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,4,16,power_law_1.01,0.10940159559249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,4,16,power_law_1.01,0.14239360094070436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,4,16,power_law_1.01,0.16330239772796631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,4,16,power_law_1.01,0.2147200107574463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,128,power_law_1.2,0.057792001962661745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,128,power_law_1.2,0.08412799835205079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,128,power_law_1.2,0.04958080053329468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,128,power_law_1.2,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,128,power_law_1.2,0.052236801385879515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,128,power_law_1.2,0.05107839703559876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,128,power_law_1.2,0.052160000801086424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,128,power_law_1.2,0.05294079780578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,128,power_law_1.2,0.05338240265846252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,128,power_law_1.2,0.05383679866790771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,128,power_law_1.2,0.054655998945236206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,128,power_law_1.2,0.05594879984855652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,128,power_law_1.2,0.05677440166473389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,128,power_law_1.2,0.057036799192428586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,128,power_law_1.2,0.06479359865188598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,128,power_law_1.2,0.06750079989433289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,128,power_law_1.2,0.07536640167236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,128,power_law_1.2,0.08832640051841736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.2,0.09600639939308167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.2,0.1176576018333435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.2,0.1376512050628662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.2,0.19886080026626587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.2,0.2070591926574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.2,0.3121920108795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.2,0.3870399951934814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.2,0.555187177658081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.2,0.7412864208221436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,2,32,power_law_1.2,0.0985472023487091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,2,32,power_law_1.2,0.09445760250091553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,2,32,power_law_1.2,0.09148799777030944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,2,32,power_law_1.2,0.09582719802856446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,2,32,power_law_1.2,0.0939903974533081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,2,32,power_law_1.2,0.08464639782905578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,2,32,power_law_1.2,0.08856959939002991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,2,32,power_law_1.2,0.09308800101280212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,2,32,power_law_1.2,0.09441919922828675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,2,32,power_law_1.2,0.09430400133132935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,32,4,balanced,0.04611733555793762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,32,4,balanced,0.046298667788505554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,32,4,balanced,0.04832000037034353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,32,4,balanced,0.05454400181770325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,32,4,balanced,0.06478400031725566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,32,4,balanced,0.09278399745623271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,32,4,balanced,0.09545066952705383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,32,4,balanced,0.09729066491127014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,32,4,balanced,0.09718400239944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,32,4,balanced,0.09841066598892212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,32,4,balanced,0.09871466954549153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,32,4,balanced,0.1015786627928416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,32,4,balanced,0.10360532999038696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,32,4,balanced,0.10689066847165425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,32,4,balanced,0.11083733042081197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,32,4,balanced,0.1144426663716634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,32,4,balanced,0.1185706655184428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,32,4,balanced,0.1323946714401245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,32,4,balanced,0.141375998655955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,32,4,balanced,0.1646453340848287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,32,4,balanced,0.18244266510009766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,32,4,balanced,0.2465546727180481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,32,4,balanced,0.28942400217056274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,2,32,power_law_1.2,0.0929535984992981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,32,4,balanced,0.4103093147277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,2,32,power_law_1.2,0.09566720128059387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,32,4,balanced,0.5050826470057169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,2,32,power_law_1.2,0.0959231972694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,32,4,balanced,0.7320746580759684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,2,32,power_law_1.2,0.09818239808082581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,2,32,power_law_1.2,0.10264320373535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,32,4,balanced,0.94706130027771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,2,32,power_law_1.2,0.10083199739456176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,2,32,power_law_1.2,0.1078336000442505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,2,32,power_law_1.2,0.12049920558929443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.2,0.136518394947052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.2,0.15752960443496705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.2,0.18010239601135253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.2,0.22274560928344728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.2,0.2914560079574585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.2,0.3630208015441895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.2,0.5077055931091309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.2,0.6623616218566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,1,power_law_1.2,0.023104000091552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.2,0.9551424026489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,1,power_law_1.2,0.035123199224472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,1,power_law_1.2,0.04904319941997528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,1,power_law_1.2,0.06325119733810425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,1,power_law_1.2,0.08190720081329346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,1,power_law_1.2,0.12729599475860595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,1,power_law_1.2,0.13778560161590575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,1,power_law_1.2,0.145632004737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,1,power_law_1.2,0.14903039932250978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,1,power_law_1.2,0.15291520357131957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,1,power_law_1.2,0.1579200029373169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,1,power_law_1.2,0.16570240259170532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,1,power_law_1.2,0.16981120109558107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,1,power_law_1.2,0.1650879979133606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,1,power_law_1.2,0.17836159467697144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,1,power_law_1.2,0.181004798412323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,1,power_law_1.2,0.21592319011688232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,1,power_law_1.2,0.2562432050704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.2,0.25039360523223875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.2,0.32198400497436525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.2,0.31269121170043945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.2,0.39687678813934324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.2,0.442412805557251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.2,0.5552320003509521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.2,0.6584896087646485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.2,0.8966336250305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.2,1.122879981994629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,16,4,balanced,0.04001066585381826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,16,4,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,16,4,balanced,0.04181866844495138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,16,4,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,16,4,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,16,4,balanced,0.06333866715431213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,16,4,balanced,0.06469866633415222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,16,4,balanced,0.062458669145902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,16,4,balanced,0.06156266729036967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,16,4,balanced,0.0627040018637975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,1,power_law_1.2,0.016838400065898894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,16,4,balanced,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,1,power_law_1.2,0.016531200706958772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,1,power_law_1.2,0.016806399822235106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,1,power_law_1.2,0.017382399737834932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,1,power_law_1.2,0.017875200510025023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,1,power_law_1.2,0.018041600286960603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,1,power_law_1.2,0.019308799505233766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,1,power_law_1.2,0.019392000138759614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,16,4,balanced,0.06479466458161671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,16,4,balanced,0.0672266681989034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,16,4,balanced,0.06889066596825917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,16,4,balanced,0.07721066474914551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,16,4,balanced,0.07721066474914551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,16,4,balanced,0.08099733293056488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,16,4,balanced,0.09359467029571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,16,4,balanced,0.09985066453615825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,16,4,balanced,0.12131200234095256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,16,4,balanced,0.14498666922251383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,16,4,balanced,0.19509865840276083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,1,power_law_1.2,0.02064639925956726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,16,4,balanced,0.22723199923833212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,16,4,balanced,0.32050132751464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,16,4,balanced,0.4108373324076335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,16,4,balanced,0.5880906581878662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,16,4,balanced,0.7533600330352783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,1,power_law_1.2,0.020972800254821778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,1,power_law_1.2,0.021593600511550903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,1,power_law_1.2,0.023583999276161192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,1,power_law_1.2,0.027590399980545043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,1,power_law_1.2,0.02542079985141754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,1,power_law_1.2,0.026796799898147584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,1,power_law_1.2,0.028857600688934327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,1,power_law_1.2,0.031027200818061828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,1,power_law_1.2,0.037299200892448425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,4,32,balanced,0.039247999588648476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,4,32,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,4,32,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,4,32,balanced,0.043381333351135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,4,32,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,4,32,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,4,32,balanced,0.04776533444722494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,4,32,balanced,0.04807466765244802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,4,32,balanced,0.0484799991051356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.2,0.04072319865226746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,4,32,balanced,0.050101334849993386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,4,32,balanced,0.04880533119042715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,4,32,balanced,0.05011733373006185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,4,32,balanced,0.05192000170548757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,4,32,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,4,32,balanced,0.05811200042565664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,4,32,balanced,0.057904000083605446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,4,32,balanced,0.060032000144322716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,4,32,balanced,0.06609599788983662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,4,32,balanced,0.06861333549022675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,4,32,balanced,0.07964266836643219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,4,32,balanced,0.08698667089144389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,4,32,balanced,0.11602666974067688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,4,32,balanced,0.14078933000564575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,4,32,balanced,0.18902933597564697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,4,32,balanced,0.23784534136454263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.2,0.05432959794998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,4,32,balanced,0.32628800471623737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,4,32,balanced,0.4219786723454793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.2,0.0602944016456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.2,0.0805184006690979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.2,0.09613440036773682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,16,1,power_law_1.01,0.05726079940795899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,16,1,power_law_1.01,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,16,1,power_law_1.01,0.08499839901924133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,16,1,power_law_1.01,0.11281280517578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,16,1,power_law_1.01,0.16554239988327027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,16,1,power_law_1.01,0.2143872022628784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,16,1,power_law_1.01,0.2495743989944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.2,0.12590080499649048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,16,1,power_law_1.01,0.30201599597930906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,16,1,power_law_1.01,0.31283841133117674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,16,1,power_law_1.01,0.32096641063690184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,16,1,power_law_1.01,0.3298559904098511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,16,1,power_law_1.01,0.34946560859680176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,16,1,power_law_1.01,0.35763840675354003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,16,1,power_law_1.01,0.37170560359954835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,16,1,power_law_1.01,0.38494720458984377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,16,1,power_law_1.01,0.39354240894317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,16,1,power_law_1.01,0.41354880332946775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,16,1,power_law_1.01,0.4769599914550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,16,1,power_law_1.01,0.5293759822845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,16,1,power_law_1.01,0.6319808006286621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,16,1,power_law_1.01,0.7108928203582764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,16,1,power_law_1.01,0.896889591217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,16,1,power_law_1.01,1.1030591964721679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,2,64,power_law_1.2,0.01895039975643158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,2,64,power_law_1.2,0.01854719966650009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,16,1,power_law_1.01,1.4520511627197266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,2,64,power_law_1.2,0.017689600586891174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,2,64,power_law_1.2,0.01756799966096878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,2,64,power_law_1.2,0.018067200481891633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,16,1,power_law_1.01,1.9090944290161134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,2,64,power_law_1.2,0.018873600661754607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,2,64,power_law_1.2,0.019974400103092194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,16,1,power_law_1.01,2.5820287704467773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,2,64,power_law_1.2,0.023871999979019166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,2,64,power_law_1.2,0.024710400402545928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,2,64,power_law_1.2,0.02338559925556183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,2,64,power_law_1.2,0.03400320112705231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,16,1,power_law_1.01,3.1464319229125977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,2,64,power_law_1.2,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,2,64,power_law_1.2,0.03434880077838898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,2,64,power_law_1.2,0.03443840146064758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,2,64,power_law_1.2,0.033817601203918454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,2,64,power_law_1.2,0.03399679958820343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,2,64,power_law_1.2,0.034668800234794614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,2,64,power_law_1.2,0.03711360096931458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,2,64,power_law_1.2,0.039673599600791934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,2,64,power_law_1.2,0.04649600088596344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,2,64,power_law_1.2,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,2,64,power_law_1.2,0.06967039704322815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,2,64,power_law_1.2,0.07971199750900268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,2,64,power_law_1.2,0.10479999780654907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,2,64,power_law_1.2,0.1323456048965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,2,64,power_law_1.2,0.19874559640884398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,2,64,power_law_1.2,0.24558720588684083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.2,0.15918079614639283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.2,0.2221247911453247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.2,0.2860480070114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,16,1,balanced,0.05481066803137461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,16,1,balanced,0.058864002426465355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,16,1,balanced,0.07372800012429555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,16,1,balanced,0.10686399539311726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,16,1,balanced,0.15946132938067117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,16,1,balanced,0.2616373300552368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,16,1,balanced,0.2680799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,16,1,balanced,0.2685546676317851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,16,1,balanced,0.2675253351529439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,16,1,balanced,0.271071990331014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,16,1,balanced,0.27242134014765423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,16,1,balanced,0.27985600630442303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,16,1,balanced,0.28296534220377606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,16,1,balanced,0.2905706763267517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,16,1,balanced,0.2937866648038228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,16,1,balanced,0.3031466603279114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,16,1,balanced,0.31623999277750653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,16,1,balanced,0.35655466715494794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,16,1,balanced,0.38527464866638184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,16,1,balanced,0.454309344291687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,16,1,balanced,0.5053439935048422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,8,power_law_1.2,0.034297600388526917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,8,power_law_1.2,0.03371520042419433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,8,power_law_1.2,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,8,power_law_1.2,0.03601279854774475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,8,power_law_1.2,0.03648000061511993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,8,power_law_1.2,0.03685120046138764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,8,power_law_1.2,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,8,power_law_1.2,0.036582401394844054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,8,power_law_1.2,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,8,power_law_1.2,0.0377023994922638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,16,1,balanced,0.6917333602905273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,8,power_law_1.2,0.03829759955406189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,8,power_law_1.2,0.038124799728393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,16,1,balanced,0.7866506576538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,8,power_law_1.2,0.038124799728393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,16,1,balanced,1.1342079639434814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,16,1,balanced,1.3493812878926594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,8,power_law_1.2,0.038848000764846805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,8,power_law_1.2,0.04287360012531281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,16,1,balanced,2.005056063334147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,16,1,balanced,2.576757272084554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,8,power_law_1.2,0.0455808013677597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,8,power_law_1.2,0.045414400100708005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,8,power_law_1.2,0.05107839703559876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,8,power_law_1.2,0.05589119791984558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,8,power_law_1.2,0.06718720197677612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,8,power_law_1.2,0.07219840288162231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,8,power_law_1.2,0.09594879746437072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,8,power_law_1.2,0.11199359893798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,8,power_law_1.2,0.15285120010375977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,8,power_law_1.2,0.1948032021522522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,8,power_law_1.2,0.2784640073776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,16,power_law_1.2,0.08578559756278992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,16,power_law_1.2,0.12660479545593262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,8,power_law_1.2,0.06938239932060242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,16,power_law_1.2,0.13388799428939818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,8,power_law_1.2,0.08982399702072144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,16,power_law_1.2,0.140556800365448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,16,power_law_1.2,0.13946239948272704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,16,power_law_1.2,0.14476159811019898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,8,power_law_1.2,0.09352319836616516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,8,power_law_1.2,0.1096384048461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,8,power_law_1.2,0.13152639865875243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,8,power_law_1.2,0.13341439962387086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,8,power_law_1.2,0.140774405002594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,8,power_law_1.2,0.13494399785995484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,8,power_law_1.2,0.12776960134506227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,8,power_law_1.2,0.13084800243377687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,8,power_law_1.2,0.13818880319595336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,8,power_law_1.2,0.14569599628448487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,8,power_law_1.2,0.15240960121154784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,8,power_law_1.2,0.14725120067596437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,8,power_law_1.2,0.161900794506073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,8,power_law_1.2,0.1598847985267639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,8,power_law_1.2,0.17184640169143678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,8,power_law_1.2,0.19521280527114868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,16,power_law_1.2,0.13455359935760497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,8,power_law_1.2,0.19589120149612427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,8,power_law_1.2,0.24281599521636962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,8,power_law_1.2,0.25539839267730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,8,power_law_1.2,0.3402048110961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,8,power_law_1.2,0.3962752103805542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,8,power_law_1.2,0.46201601028442385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,16,power_law_1.2,0.14142080545425414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,8,power_law_1.2,0.5976960182189941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,16,power_law_1.2,0.12367360591888428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,8,power_law_1.2,0.9163328170776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,16,power_law_1.2,0.12565120458602905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,16,power_law_1.2,0.13992960453033448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,8,power_law_1.2,1.1699968338012696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,16,power_law_1.2,0.12420480251312256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,16,power_law_1.2,0.14256639480590821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,16,power_law_1.2,0.14174720048904418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,16,power_law_1.2,0.15203839540481567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,16,power_law_1.2,0.14921599626541138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,16,power_law_1.2,0.16595200300216675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,16,power_law_1.2,0.18057600259780884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,16,power_law_1.2,0.1926144003868103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,16,power_law_1.2,0.23248639106750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,16,power_law_1.2,0.2531840085983276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,16,power_law_1.2,0.3387200117111206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,16,power_law_1.2,0.378438401222229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,16,power_law_1.2,0.5120255947113037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,16,power_law_1.2,0.595033597946167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,16,power_law_1.2,0.8836095809936524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,16,power_law_1.2,1.0959424018859862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,8,power_law_1.2,0.3606911897659302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,3072,1536,8,256,8,4,balanced,0.01952533299724261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,3072,1536,8,256,8,4,balanced,0.019007999449968338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,3072,1536,8,256,8,4,balanced,0.01937599976857503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,3072,1536,8,256,8,4,balanced,0.02195200075705846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,3072,1536,8,256,8,4,balanced,0.025562666356563568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,3072,1536,8,256,8,4,balanced,0.029317334294319153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,3072,1536,8,256,8,4,balanced,0.030671998858451843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,3072,1536,8,256,8,4,balanced,0.03133333226044973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,3072,1536,8,256,8,4,balanced,0.03258133431275686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,3072,1536,8,256,8,4,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,3072,1536,8,256,8,4,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,3072,1536,8,256,8,4,balanced,0.036677333215872444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,3072,1536,8,256,8,4,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,3072,1536,8,256,8,4,balanced,0.04286933441956838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,3072,1536,8,256,8,4,balanced,0.0446720023949941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,3072,1536,8,256,8,4,balanced,0.045567999283472695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,3072,1536,8,256,8,4,balanced,0.06356266637643178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,3072,1536,8,256,8,4,balanced,0.06401599943637848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,64,balanced,0.03183466692765554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,3072,1536,8,256,8,4,balanced,0.06719466547171275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,64,balanced,0.031530665854612984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,64,balanced,0.03166399896144867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,64,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,64,balanced,0.03544000039498011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,64,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,3072,1536,8,256,8,4,balanced,0.0757066657145818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,3072,1536,8,256,8,4,balanced,0.11003200213114421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,3072,1536,8,256,8,4,balanced,0.11937066912651062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,3072,1536,8,256,8,4,balanced,0.13537599643071493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,3072,1536,8,256,8,4,balanced,0.15758933623631796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,3072,1536,8,256,8,4,balanced,0.18331199884414673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,3072,1536,8,256,8,4,balanced,0.2847626606623332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,3072,1536,8,256,8,4,balanced,0.33168532450993854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,64,balanced,0.03579200059175491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,64,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,64,balanced,0.035802667339642845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,64,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,64,balanced,0.03737599899371465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,64,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,64,balanced,0.03782933453718821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,64,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,64,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,64,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,64,balanced,0.04814933240413666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,64,balanced,0.04944533109664917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,64,balanced,0.05493866900602976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,64,balanced,0.06038933495680491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,64,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,64,balanced,0.08075733482837677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,64,balanced,0.09282132983207703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,64,balanced,0.12176533540089925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,64,balanced,0.14808533589045206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,64,balanced,0.19420266151428223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,64,balanced,0.246394673983256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,1,power_law_1.2,0.028601598739624024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,1,power_law_1.2,0.04862079918384552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,1,power_law_1.2,0.07623040080070495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,1,power_law_1.2,0.11196800470352172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,1,power_law_1.2,0.15578880310058593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,1,power_law_1.2,0.22074239253997802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,1,power_law_1.2,0.30804479122161865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,1,power_law_1.2,0.32688639163970945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,1,power_law_1.2,0.33471360206604006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,1,power_law_1.2,0.34651520252227785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,1,power_law_1.2,0.362719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,1,power_law_1.2,0.3722496032714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,1,power_law_1.2,0.38853759765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,1,power_law_1.2,0.3928767919540405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,1,power_law_1.2,0.3985215902328491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,2,power_law_1.01,0.02531839907169342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,1,power_law_1.2,0.41580801010131835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,2,power_law_1.01,0.025484800338745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,2,power_law_1.01,0.024607999622821806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,2,power_law_1.01,0.025459200143814087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,2,power_law_1.01,0.026521599292755126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,2,power_law_1.01,0.030777600407600404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,2,power_law_1.01,0.044537600874900815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,2,power_law_1.01,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,2,power_law_1.01,0.04844799935817719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,2,power_law_1.01,0.04351359903812409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,2,power_law_1.01,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,1,power_law_1.2,0.4422143936157227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,2,power_law_1.01,0.08427519798278808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,2,power_law_1.01,0.07655680179595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,1,power_law_1.2,0.49837441444396974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,2,power_law_1.01,0.06072319746017456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.2,0.5096320152282715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,2,power_law_1.01,0.08881919980049133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,2,power_law_1.01,0.07366399765014649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.2,0.60033278465271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,2,power_law_1.01,0.09381120204925537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,2,power_law_1.01,0.11681280136108399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.2,0.6982336044311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.01,0.14207359552383422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.2,0.8905344009399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.01,0.2003391981124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.2,0.86495361328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.2,1.085811233520508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.2,1.3269248008728027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.2,1.8032192230224608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.2,2.2930816650390624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.01,0.25539839267730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.01,0.35842559337615965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.01,0.4679296016693115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.01,0.6761280059814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.01,0.8879679679870606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.01,1.308563232421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.01,1.7214143753051758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,16,balanced,0.04885333279768626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,16,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,16,balanced,0.07457066575686137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,16,balanced,0.1042133371035258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,16,balanced,0.1562933325767517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,16,balanced,0.18404799699783325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,16,balanced,0.18669867515563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,16,balanced,0.1872319976488749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,16,balanced,0.18515199422836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,16,balanced,0.18736000855763754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,16,balanced,0.18836800257364908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,16,balanced,0.18794133265813193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,16,balanced,0.18902399142583212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,16,balanced,0.19204266866048178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,16,balanced,0.19581333796183267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,16,balanced,0.19429866472880045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,16,balanced,0.19849065939585367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,16,balanced,0.21991999944051108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,16,balanced,0.20764267444610596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,16,balanced,0.25270400444666546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,16,balanced,0.2350133260091146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,16,balanced,0.3103626569112142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,16,balanced,0.2951093316078186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,16,balanced,0.4466559886932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,16,balanced,0.4503733317057292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,16,balanced,0.7423466841379801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,16,balanced,0.7609333197275797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,4,power_law_1.01,0.1365056037902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,4,power_law_1.01,0.21349759101867677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,4,power_law_1.01,0.2398655891418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,4,power_law_1.01,0.3034303903579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,4,power_law_1.01,0.36129920482635497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,4,power_law_1.01,0.36998400688171384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,4,power_law_1.01,0.4668992042541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,4,power_law_1.01,0.43894400596618655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,4,power_law_1.01,0.4618368148803711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,4,power_law_1.01,0.46455039978027346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,4,power_law_1.01,0.49251837730407716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,4,power_law_1.01,0.4735231876373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,4,power_law_1.01,0.47621760368347166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,4,power_law_1.01,0.4950399875640869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,4,power_law_1.01,0.5280447959899902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,4,power_law_1.01,0.49621758460998533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,4,power_law_1.01,0.5265664100646973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,4,power_law_1.01,0.5361408233642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.01,0.5859007835388184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.01,0.6602816104888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.01,0.709772777557373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.01,0.8464447975158691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.01,0.9827584266662598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.01,1.2378879547119142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.01,1.5412416458129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.01,2.1229759216308595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.01,2.682054328918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,1,power_law_1.2,0.10222079753875732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,1,power_law_1.2,0.1293951988220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,1,power_law_1.2,0.1736575961112976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,1,power_law_1.2,0.25870718955993655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,1,power_law_1.2,0.36097280979156493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,1,power_law_1.2,0.46579837799072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,1,power_law_1.2,0.6614272117614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,1,power_law_1.2,0.7235968112945557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,1,power_law_1.2,0.7338751792907715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,1,power_law_1.2,0.7645247936248779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,1,power_law_1.2,0.7806848049163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,1,power_law_1.2,0.793452787399292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,1,power_law_1.2,0.8218303680419922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,1,power_law_1.2,0.8612735748291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,1,power_law_1.2,0.8531328201293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,1,power_law_1.2,0.8804351806640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,1,power_law_1.2,0.9379008293151856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,1,power_law_1.2,1.0152128219604493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,1,power_law_1.2,1.1095104217529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,1,power_law_1.2,1.2429951667785644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,1,power_law_1.2,1.3157055854797364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,1,power_law_1.2,1.599891185760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,1,power_law_1.2,1.8700672149658204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,1,power_law_1.2,2.394726371765137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,4,power_law_1.01,0.025670400261878966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,4,power_law_1.01,0.024736000597476958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,4,power_law_1.01,0.02504960000514984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,4,power_law_1.01,0.029574400186538695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,1,power_law_1.2,2.9381439208984377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,4,power_law_1.01,0.03112959861755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,4,power_law_1.01,0.03544960021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,4,power_law_1.01,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,4,power_law_1.01,0.034355199337005614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,1,16,power_law_1.2,0.15834879875183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,4,power_law_1.01,0.037561601400375365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,4,power_law_1.01,0.03765760064125061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,1,16,power_law_1.2,0.17639039754867553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,4,power_law_1.01,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,1,power_law_1.2,4.022745513916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,1,16,power_law_1.2,0.18089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,1,16,power_law_1.2,0.1595072031021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,1,16,power_law_1.2,0.15134079456329347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,1,16,power_law_1.2,0.17265280485153198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,1,16,power_law_1.2,0.170796799659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,1,16,power_law_1.2,0.17872639894485473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,1,power_law_1.2,5.12174072265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,1,16,power_law_1.2,0.17201919555664064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,1,16,power_law_1.2,0.18057600259780884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,1,16,power_law_1.2,0.18819199800491332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,1,16,power_law_1.2,0.18411519527435302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,1,16,power_law_1.2,0.18803839683532714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,1,16,power_law_1.2,0.20414719581604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,4,power_law_1.01,0.03959679901599884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,4,power_law_1.01,0.039129599928855896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,4,power_law_1.01,0.04715520143508911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,4,power_law_1.01,0.05236480236053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,4,power_law_1.01,0.05541759729385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,4,power_law_1.01,0.056569600105285646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,4,power_law_1.01,0.06350079774856568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,1,16,power_law_1.2,0.21445760726928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,1,16,power_law_1.2,0.21893761157989503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,4,power_law_1.01,0.0689408004283905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,4,power_law_1.01,0.07982720136642456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,4,power_law_1.01,0.09126399755477906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,4,power_law_1.01,0.1145408034324646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,4,power_law_1.01,0.1454975962638855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,4,power_law_1.01,0.18476799726486207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,4,power_law_1.01,0.22213759422302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,4,power_law_1.01,0.3347775936126709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,1,16,power_law_1.2,0.2395456075668335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,4,power_law_1.01,0.40792322158813477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,1,16,power_law_1.2,0.27708160877227783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,1,16,power_law_1.2,0.28787200450897216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,1,16,power_law_1.2,0.4013951778411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,1,16,power_law_1.2,0.4085887908935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,1,16,power_law_1.2,0.4891007900238037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,1,16,power_law_1.2,0.5030975818634034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,1,16,power_law_1.2,0.772492790222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,1,16,power_law_1.2,0.9686847686767578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,1,16,power_law_1.2,1.3030271530151367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,1,16,power_law_1.2,1.7340288162231445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,2,balanced,0.033045334120591484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,2,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,2,balanced,0.03124266614516576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,2,balanced,0.033813332517941795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,2,balanced,0.035232000052928925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,2,balanced,0.04077333211898804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,2,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,2,balanced,0.04762666424115499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,2,balanced,0.06646933158238728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,2,balanced,0.05366933345794678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,2,balanced,0.059104000528653465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,2,balanced,0.07125866909821828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,2,balanced,0.07786133388678233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,2,balanced,0.07659199833869934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,2,balanced,0.10621333122253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,2,balanced,0.08610666791598003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,2,balanced,0.10130133231480916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,2,balanced,0.14190399646759033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,2,balanced,0.16612266500790915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,2,balanced,0.23365867137908936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,2,balanced,0.29604266087214154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,2,balanced,0.4233013391494751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,2,balanced,0.5336106618245443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,2,balanced,0.7855199972788492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,2,balanced,1.0241333643595378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,2,balanced,1.5162933667500813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,2,balanced,2.0077385902404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,16,balanced,0.04910933474699656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,16,balanced,0.05559466779232025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,16,balanced,0.07555200159549713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,16,balanced,0.10771733522415161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,16,balanced,0.14040533701578775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,16,balanced,0.1581760048866272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,16,balanced,0.1579093337059021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,16,balanced,0.16013866662979126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,16,balanced,0.1630506714185079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,16,balanced,0.1616213321685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,16,balanced,0.1602026621500651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,16,balanced,0.16263467073440552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,16,balanced,0.16194132963816324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,16,balanced,0.16474666198094687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,16,balanced,0.17018665870030722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,16,balanced,0.1707893411318461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,16,balanced,0.17659199237823486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,16,balanced,0.18886399269104004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,16,balanced,0.19792000452677408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,16,balanced,0.25252266724904376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,16,balanced,0.24252800146738687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,16,balanced,0.37412798404693604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,16,balanced,0.30843732754389447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,16,balanced,0.4452799956003825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,16,balanced,0.4482133388519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,16,balanced,0.6312479972839355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,16,balanced,0.7084426879882812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,14336,2,8,4,1,power_law_1.2,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,14336,2,8,4,1,power_law_1.2,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,14336,2,8,4,1,power_law_1.2,0.058316802978515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,14336,2,8,4,1,power_law_1.2,0.08177919983863831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,14336,2,8,4,1,power_law_1.2,0.08284800052642823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,14336,2,8,4,1,power_law_1.2,0.08498560190200806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,14336,2,8,4,1,power_law_1.2,0.08608639836311341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,14336,2,8,4,1,power_law_1.2,0.09071999788284302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,8,power_law_1.2,0.03997440040111542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,14336,2,8,4,1,power_law_1.2,0.08904320001602173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,8,power_law_1.2,0.0383296012878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,14336,2,8,4,1,power_law_1.2,0.08962560296058655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,8,power_law_1.2,0.03930239975452423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,14336,2,8,4,1,power_law_1.2,0.09144960045814514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,14336,2,8,4,1,power_law_1.2,0.09632639884948731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,14336,2,8,4,1,power_law_1.2,0.09623680114746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,14336,2,8,4,1,power_law_1.2,0.1017151951789856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,14336,2,8,4,1,power_law_1.2,0.11189119815826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,14336,2,8,4,1,power_law_1.2,0.11838079690933227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,8,power_law_1.2,0.04332799911499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,8,power_law_1.2,0.046828800439834596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,14336,2,8,4,1,power_law_1.2,0.12223360538482667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,8,power_law_1.2,0.04913919866085052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,8,power_law_1.2,0.05057280063629151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,8,power_law_1.2,0.050937598943710326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,8,power_law_1.2,0.0514303982257843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,8,power_law_1.2,0.05085440278053284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,8,power_law_1.2,0.05301759839057922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,8,power_law_1.2,0.05482239723205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,8,power_law_1.2,0.05766400098800659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,14336,2,8,4,1,power_law_1.2,0.14339840412139893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,8,power_law_1.2,0.060601598024368285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,8,power_law_1.2,0.06606720089912414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,8,power_law_1.2,0.07212799787521362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,8,power_law_1.2,0.07223039865493774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,8,power_law_1.2,0.08528000116348267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.2,0.1083840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.2,0.12881920337677003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,14336,2,8,4,1,power_law_1.2,0.15854079723358155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.2,0.13604480028152466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.2,0.1794111967086792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.2,0.21400959491729737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,14336,2,8,4,1,power_law_1.2,0.19924479722976685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,14336,2,8,4,1,power_law_1.2,0.23233280181884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.2,0.2859200000762939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.2,0.3455424070358276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,4,4,balanced,0.06865600248177846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,4,4,balanced,0.06868800024191539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,4,4,balanced,0.06895466645558675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,4,4,balanced,0.08063999811808269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,4,4,balanced,0.09148800373077393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,4,4,balanced,0.11378666758537292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,4,4,balanced,0.11281599601109822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,14336,2,8,4,1,power_law_1.2,0.3110464096069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,4,4,balanced,0.11371733744939168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,4,4,balanced,0.11576533317565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,4,4,balanced,0.11410666505495708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,4,4,balanced,0.11563733220100403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,4,4,balanced,0.12084800004959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,4,4,balanced,0.12020799517631531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,4,4,balanced,0.12154666582743327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,4,4,balanced,0.12563733259836832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,4,4,balanced,0.12485866745313008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,4,4,balanced,0.12873599926630655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,4,4,balanced,0.13926399747530618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,4,4,balanced,0.1470080018043518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,4,4,balanced,0.16859734058380127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,4,4,balanced,0.1858773430188497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,4,4,balanced,0.21891733010609946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,4,4,balanced,0.26095465819040936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,4,4,balanced,0.3773386478424072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,4,4,balanced,0.4624319871266683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,4,4,balanced,0.6623573303222656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,4,4,balanced,0.8482027053833008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,14336,2,8,4,1,power_law_1.2,0.38460800647735593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,14336,2,8,4,1,power_law_1.2,0.5197504043579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,1,power_law_1.2,0.03814400136470795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,1,power_law_1.2,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,1,power_law_1.2,0.05767040252685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,1,power_law_1.2,0.08186240196228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,1,power_law_1.2,0.08296959996223449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,1,power_law_1.2,0.08368639945983887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,1,power_law_1.2,0.08478080034255982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,1,power_law_1.2,0.08394880294799804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,1,power_law_1.2,0.08546559810638428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,1,power_law_1.2,0.08592000007629394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,1,power_law_1.2,0.08627840280532836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,1,power_law_1.2,0.09093120098114013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,1,power_law_1.2,0.09217280149459839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,1,power_law_1.2,0.09789440035820007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,1,power_law_1.2,0.10823040008544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,1,power_law_1.2,0.11264640092849731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,1,power_law_1.2,0.1134335994720459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,1,power_law_1.2,0.13281919956207275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.2,0.5353856086730957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,1,power_law_1.2,0.1410431981086731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,64,power_law_1.2,0.03212159872055054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,64,power_law_1.2,0.031737598776817325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,64,power_law_1.2,0.031001600623130798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,1,power_law_1.2,0.17824000120162964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,64,power_law_1.2,0.03212159872055054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,64,power_law_1.2,0.03247359991073608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,64,power_law_1.2,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,64,power_law_1.2,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,1,power_law_1.2,0.20421121120452881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,64,power_law_1.2,0.03421440124511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,64,power_law_1.2,0.03425279855728149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,64,power_law_1.2,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,64,power_law_1.2,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,64,power_law_1.2,0.0377920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,64,power_law_1.2,0.03880319893360138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,64,power_law_1.2,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,64,power_law_1.2,0.04474239945411682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,64,power_law_1.2,0.045484799146652224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,64,power_law_1.2,0.051692801713943484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,64,power_law_1.2,0.05820159912109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,64,power_law_1.2,0.0606656014919281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,1,power_law_1.2,0.26343679428100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,64,power_law_1.2,0.07889919877052307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,14336,2,8,4,1,power_law_1.2,0.663750410079956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,64,power_law_1.2,0.0871936023235321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.2,0.613862419128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,64,power_law_1.2,0.13445760011672975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,64,power_law_1.2,0.16154240369796752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,1,power_law_1.2,0.31626238822937014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,16,power_law_1.01,0.03375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,16,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,16,power_law_1.01,0.030995199084281923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,16,power_law_1.01,0.03284479975700379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,16,power_law_1.01,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,16,power_law_1.01,0.03427839875221252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,16,power_law_1.01,0.034995201230049136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,16,power_law_1.01,0.03580799996852875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,16,power_law_1.01,0.03589119911193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,16,power_law_1.01,0.037145599722862244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,16,power_law_1.01,0.037651199102401736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,16,power_law_1.01,0.03873279988765717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,16,power_law_1.01,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,16,power_law_1.01,0.03912320137023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,16,power_law_1.01,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,16,power_law_1.01,0.046828800439834596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,16,power_law_1.01,0.048172798752784726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,16,power_law_1.01,0.052172797918319705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,16,power_law_1.01,0.061273598670959474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,16,power_law_1.01,0.0695360004901886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,16,power_law_1.01,0.07917439937591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,16,power_law_1.01,0.1070207953453064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,16,power_law_1.01,0.14428800344467163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,16,power_law_1.01,0.21682560443878174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,16,power_law_1.01,0.24890880584716796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,64,power_law_1.2,0.2558336019515991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,16,power_law_1.01,0.35136001110076903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,16,power_law_1.01,0.505836820602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,1,power_law_1.2,0.4239935874938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,64,power_law_1.2,0.2760063886642456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,14336,2,8,4,1,power_law_1.2,0.9437055587768555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,8,16,balanced,0.049082666635513306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,8,16,balanced,0.0413973331451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,8,16,balanced,0.04171733558177948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,8,16,balanced,0.04809066653251648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,8,16,balanced,0.049728001157442726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,8,16,balanced,0.07017600039641063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,8,16,balanced,0.07252799967924754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,8,16,balanced,0.07217599948247273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,8,16,balanced,0.07209066549936931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,8,16,balanced,0.07344533503055573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,8,16,balanced,0.07254933317502339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,8,16,balanced,0.07418666779994965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,8,16,balanced,0.07484266658624013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,8,16,balanced,0.07755733529726665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,8,16,balanced,0.08095466593901317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,8,16,balanced,0.08385066191355388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,8,16,balanced,0.08449600140253703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,8,16,balanced,0.09058133761088054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,8,16,balanced,0.09497066338857015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,8,16,balanced,0.10688533385594685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,8,16,balanced,0.11892799536387126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,8,16,balanced,0.1446560025215149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,8,16,balanced,0.1686720053354899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,8,16,balanced,0.21676800648371378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,8,16,balanced,0.26873066027959186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,8,16,balanced,0.35769065221150714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,8,16,balanced,0.4508266846338908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,1,power_law_1.2,0.5402624130249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,64,power_law_1.2,0.42914562225341796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,14336,2,8,4,1,power_law_1.2,1.2217856407165528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,1,64,power_law_1.2,0.031167998909950256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,1,64,power_law_1.2,0.02842240035533905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,1,64,power_law_1.2,0.028089600801467895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,1,64,power_law_1.2,0.03000960052013397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,1,64,power_law_1.2,0.02990719974040985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,1,64,power_law_1.2,0.03162879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,1,64,power_law_1.2,0.031430399417877196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,1,64,power_law_1.2,0.031327998638153075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,1,64,power_law_1.2,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,1,64,power_law_1.2,0.0319680005311966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,1,64,power_law_1.2,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,1,64,power_law_1.2,0.03411200046539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,1,64,power_law_1.2,0.03431040048599243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,1,64,power_law_1.2,0.034918400645256045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,1,64,power_law_1.2,0.03935999870300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,1,64,power_law_1.2,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,1,64,power_law_1.2,0.04208639860153198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,64,power_law_1.2,0.521504020690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,1,64,power_law_1.2,0.046777600049972536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,1,64,power_law_1.2,0.0516543984413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,1,power_law_1.2,0.761952018737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,1,64,power_law_1.2,0.061440002918243405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,1,64,power_law_1.2,0.06532480120658875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,1,64,power_law_1.2,0.09038720130920411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,1,64,power_law_1.2,0.09895679950714112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,1,64,power_law_1.2,0.15304960012435914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,1,power_law_1.2,0.9704511642456055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,1,64,power_law_1.2,0.2039167881011963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,2,balanced,0.04817600051561991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,2,balanced,0.05019199848175049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,2,balanced,0.05535466472307841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,2,balanced,0.0650186687707901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,2,balanced,0.08205866813659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,2,balanced,0.11954133709271748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,2,balanced,0.14939199884732565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,2,balanced,0.15104533235232034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,2,balanced,0.15242666999499002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,2,balanced,0.15480533242225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,2,balanced,0.15455466508865356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,1,64,power_law_1.2,0.2751487970352173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,2,balanced,0.16033066312472025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,2,balanced,0.16477866967519125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,2,balanced,0.16681599617004395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,2,balanced,0.17843733231226602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,2,balanced,0.18222399552663168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,2,balanced,0.19305066267649332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,2,balanced,0.22404267390569052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,2,balanced,0.24664533138275146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,2,balanced,0.2993280092875163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,2,balanced,0.3399626811345418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,2,balanced,0.4405546585718791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,2,balanced,0.5677066644032797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,2,balanced,0.7645706335703532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,2,balanced,1.0379733244578044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,2,balanced,1.4772213300069172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,2,balanced,1.9138293266296387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,1,64,power_law_1.2,0.34472320079803465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,8,balanced,0.044026667873064675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,8,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,8,balanced,0.04983466863632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,8,balanced,0.07672533392906189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,8,balanced,0.10301333665847778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,8,balanced,0.11449600259462993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,8,balanced,0.116976002852122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,8,balanced,0.11726400256156921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,8,balanced,0.11618133385976155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,8,balanced,0.1179200013478597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,8,balanced,0.11950400471687317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,8,balanced,0.11938132842381795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,8,balanced,0.12014933427174886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,8,balanced,0.12242133418718974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,8,balanced,0.12545067071914673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,8,balanced,0.12914133071899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,8,balanced,0.1316480040550232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,8,balanced,0.13800000150998434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,8,balanced,0.14601066708564758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,8,balanced,0.17221866051355997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,8,balanced,0.1747466723124186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,8,balanced,0.2285226583480835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,8,balanced,0.24088533719380698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,8,balanced,0.38303999106089276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,8,balanced,0.37763198216756183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,8,balanced,0.5246400038401285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,8,balanced,0.6595840056737264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,2,8,power_law_1.2,0.09541760087013244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,2,8,power_law_1.2,0.1299456000328064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,2,8,power_law_1.2,0.11535999774932862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,2,8,power_law_1.2,0.14830080270767212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,2,8,power_law_1.2,0.1725823998451233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,2,8,power_law_1.2,0.161190402507782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,2,8,power_law_1.2,0.17511039972305298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,2,8,power_law_1.2,0.1682368040084839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,2,8,power_law_1.2,0.17522560358047484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,2,8,power_law_1.2,0.18097280263900756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,2,8,power_law_1.2,0.17517440319061278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,2,8,power_law_1.2,0.18257280588150024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,2,8,power_law_1.2,0.18522239923477174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,2,8,power_law_1.2,0.18762240409851075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,4,power_law_1.2,0.06568319797515869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,4,power_law_1.2,0.08419200181961059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,4,power_law_1.2,0.09032959938049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,4,power_law_1.2,0.12239999771118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,4,power_law_1.2,0.13708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,4,power_law_1.2,0.14707839488983154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,2,8,power_law_1.2,0.19697279930114747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,2,8,power_law_1.2,0.19800959825515746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,2,8,power_law_1.2,0.21037440299987792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,4,power_law_1.2,0.1985856056213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,4,power_law_1.2,0.1841599941253662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,4,power_law_1.2,0.18469760417938233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,4,power_law_1.2,0.19367040395736695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,4,power_law_1.2,0.1990847945213318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,2,8,power_law_1.2,0.23304319381713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,4,power_law_1.2,0.19901440143585206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,4,power_law_1.2,0.20595200061798097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,4,power_law_1.2,0.21650559902191163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,4,power_law_1.2,0.22499840259552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,4,power_law_1.2,0.2309056043624878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,4,power_law_1.2,0.2407167911529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,2,8,power_law_1.2,0.2391103982925415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,4,power_law_1.2,0.2668864011764526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,4,power_law_1.2,0.2837376117706299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,4,power_law_1.2,0.34870400428771975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,4,power_law_1.2,0.38735361099243165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,4,power_law_1.2,0.46778240203857424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,4,power_law_1.2,0.4835072040557861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,2,8,power_law_1.2,0.2924799919128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,4,power_law_1.2,0.7184192180633545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,2,8,power_law_1.2,0.3439039945602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,4,power_law_1.2,0.7905856132507324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,2,8,power_law_1.2,0.43313279151916506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,4,power_law_1.2,1.2058239936828614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,2,8,power_law_1.2,0.5375616073608398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,4,power_law_1.2,1.3988544464111328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,2,8,power_law_1.2,0.7005055904388428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,2,8,power_law_1.2,1.0064448356628417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,2,8,power_law_1.2,1.4397184371948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,2,8,power_law_1.2,1.8087615966796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,128,power_law_1.2,0.09226880073547364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,128,power_law_1.2,0.08949120044708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,128,power_law_1.2,0.09022079706192017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,128,power_law_1.2,0.09461119771003723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,128,power_law_1.2,0.09602559804916382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,128,power_law_1.2,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,128,power_law_1.2,0.10103679895401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,128,power_law_1.2,0.10030720233917237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,128,power_law_1.2,0.1025920033454895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,128,power_law_1.2,0.10049920082092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,128,power_law_1.2,0.0977728009223938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,128,power_law_1.2,0.10188159942626954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,128,power_law_1.2,0.10166399478912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,128,power_law_1.2,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,128,power_law_1.2,0.10046720504760742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,128,power_law_1.2,0.10129280090332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,128,power_law_1.2,0.10591360330581664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,128,power_law_1.2,0.1252671957015991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,128,power_law_1.2,0.12741760015487671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,128,power_law_1.2,0.1658560037612915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,128,power_law_1.2,0.19937920570373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,128,power_law_1.2,0.27930240631103515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,128,power_law_1.2,0.3404736042022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,128,power_law_1.2,0.46709117889404295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,128,power_law_1.2,0.5885248184204102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,128,power_law_1.2,0.917625617980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,128,power_law_1.2,1.2704383850097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,32,1,balanced,0.035936000446478523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,32,1,balanced,0.03585066646337509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,32,1,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,32,1,balanced,0.03977599988381068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,32,1,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,32,1,balanced,0.04119999955097834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,32,1,balanced,0.03987200061480204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,32,1,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,32,1,balanced,0.043141335248947144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,32,1,balanced,0.042362665136655174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,32,1,balanced,0.04251199960708618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,32,1,balanced,0.043525333205858864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,32,1,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,32,1,balanced,0.04643199841181437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,32,1,balanced,0.05384000142415365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,32,1,balanced,0.05593599875768026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,32,1,balanced,0.05606399973233541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,32,1,balanced,0.0682666649421056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,32,1,balanced,0.07232533395290375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,32,1,balanced,0.09477866689364116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,32,1,balanced,0.11156800389289856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,32,1,balanced,0.15044800440470377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,32,1,balanced,0.17430400848388672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,32,1,balanced,0.2440213362375895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,32,1,balanced,0.3140053351720174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,32,1,balanced,0.45154134432474774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,32,1,balanced,0.5892053445180258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,4,power_law_1.2,0.0969215989112854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,4,power_law_1.2,0.14421119689941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,4,power_law_1.2,0.1982848048210144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,4,power_law_1.2,0.254803204536438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,4,power_law_1.2,0.32442240715026854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,4,power_law_1.2,0.41245441436767577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,4,power_law_1.2,0.4500160217285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,4,power_law_1.2,0.5736512184143067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,4,power_law_1.2,0.5034111976623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,4,power_law_1.2,0.4688000202178955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,4,power_law_1.2,0.5371712207794189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,4,power_law_1.2,0.5062464237213135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,4,power_law_1.2,0.5197184085845947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,4,power_law_1.2,0.5388031959533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,4,power_law_1.2,0.5376255989074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,4,power_law_1.2,0.5607679843902588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,4,power_law_1.2,0.6133632183074951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,4,power_law_1.2,0.6520063877105713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,4,power_law_1.2,0.7149824142456055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,4,power_law_1.2,0.819155216217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,4,power_law_1.2,0.8295424461364747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,4,power_law_1.2,1.0057472229003905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,4,power_law_1.2,1.136774444580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,4,power_law_1.2,1.4348031997680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,4,power_law_1.2,1.490291213989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,4,power_law_1.2,1.8295360565185548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,4,power_law_1.2,2.5959295272827148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,8,16,power_law_1.01,0.044582399725914004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,8,16,power_law_1.01,0.04917120039463043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,8,16,power_law_1.01,0.045900800824165346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,8,16,power_law_1.01,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,8,16,power_law_1.01,0.04906879961490631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,8,16,power_law_1.01,0.04758400022983551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,8,16,power_law_1.01,0.04938240051269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,8,16,power_law_1.01,0.05025280117988586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,8,16,power_law_1.01,0.05055360198020935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,8,16,power_law_1.01,0.05084800124168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,8,16,power_law_1.01,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,8,16,power_law_1.01,0.05326719880104065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,8,16,power_law_1.01,0.05443199872970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,8,16,power_law_1.01,0.056729602813720706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,8,16,power_law_1.01,0.06225919723510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,8,16,power_law_1.01,0.06236159801483154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,8,16,power_law_1.01,0.06673920154571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,8,16,power_law_1.01,0.07453439831733703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,8,16,power_law_1.01,0.07913600206375122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,8,16,power_law_1.01,0.09695360064506531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,8,16,power_law_1.01,0.10851199626922607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,8,16,power_law_1.01,0.1422719955444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,8,16,power_law_1.01,0.17749760150909424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,8,16,power_law_1.01,0.2427072048187256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,8,1,power_law_1.2,0.05052800178527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,8,1,power_law_1.2,0.06176000237464905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,8,1,power_law_1.2,0.07678080201148987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,8,1,power_law_1.2,0.11552640199661254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,8,1,power_law_1.2,0.11520639657974244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,8,1,power_law_1.2,0.12135039567947388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,8,1,power_law_1.2,0.12624000310897826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,8,16,power_law_1.01,0.31208319664001466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,8,1,power_law_1.2,0.12101119756698608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,8,1,power_law_1.2,0.12099839448928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,8,16,power_law_1.01,0.43164801597595215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,8,1,power_law_1.2,0.12871040105819703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,8,1,power_law_1.2,0.1266816020011902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,8,1,power_law_1.2,0.13271039724349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,8,1,power_law_1.2,0.13270399570465088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,8,1,power_law_1.2,0.1410048007965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,8,1,power_law_1.2,0.14949120283126832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,8,16,power_law_1.01,0.5509759902954101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,8,1,power_law_1.2,0.16087039709091186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,8,1,power_law_1.2,0.16015360355377198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,8,1,power_law_1.2,0.186080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,8,1,power_law_1.2,0.21979520320892335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,8,1,power_law_1.2,0.25585920810699464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,8,1,power_law_1.2,0.303385591506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,2,balanced,0.023306667804718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,2,balanced,0.023423999547958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,2,balanced,0.023071999351183575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,2,balanced,0.024133334557215374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,2,balanced,0.030037333567937214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,2,balanced,0.031258667508761086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,2,balanced,0.029285334050655365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,2,balanced,0.03275199979543686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,2,balanced,0.035631999373435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,2,balanced,0.032655999064445496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,2,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,2,balanced,0.07434666653474171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,2,balanced,0.07501866420110066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,2,balanced,0.080485333998998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,2,balanced,0.061008001367251076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,2,balanced,0.05356266597906748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,2,balanced,0.06095466514428457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,2,balanced,0.07427733143170674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,2,balanced,0.09191999832789104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,2,balanced,0.1272533337275187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,8,1,power_law_1.2,0.40122241973876954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,2,balanced,0.15351466337839761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,2,balanced,0.21477866172790527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,2,balanced,0.2681279977162679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,2,balanced,0.38607998689015705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,2,balanced,0.5020800034205118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,2,balanced,0.7295200030008951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,2,balanced,0.9593973159790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,8,1,power_law_1.2,0.48549761772155764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,8,1,power_law_1.2,0.6604800224304199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,8,1,power_law_1.2,0.8356351852416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,32,8,balanced,0.03677333394686381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,32,8,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,32,8,balanced,0.03653866549332937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,32,8,balanced,0.039434666434923805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,32,8,balanced,0.03990933299064636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,32,8,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,32,8,balanced,0.03938133269548416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,32,8,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,32,8,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,32,8,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,32,8,balanced,0.03972266614437103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,32,8,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,32,8,balanced,0.04194133480389913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,32,8,balanced,0.0417546679576238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,32,8,balanced,0.044256001710891724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,32,8,balanced,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,32,8,balanced,0.04769066472848257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,32,8,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,32,8,balanced,0.053770666321118675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,32,8,balanced,0.06028266747792562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,32,8,balanced,0.06607466439406078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,16384,2,8,4,1,power_law_1.01,0.07403519749641418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,16384,2,8,4,1,power_law_1.01,0.10077439546585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,16384,2,8,4,1,power_law_1.01,0.1229375958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,16384,2,8,4,1,power_law_1.01,0.1644927978515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,16384,2,8,4,1,power_law_1.01,0.16261119842529298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,16384,2,8,4,1,power_law_1.01,0.16231679916381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,16384,2,8,4,1,power_law_1.01,0.16104320287704468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,16384,2,8,4,1,power_law_1.01,0.15697280168533326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,32,8,balanced,0.08249066770076752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,16384,2,8,4,1,power_law_1.01,0.15948159694671632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,32,8,balanced,0.0960586667060852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,32,8,balanced,0.12552000085512796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,32,8,balanced,0.15121066570281982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,32,8,balanced,0.20313066244125366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,32,8,balanced,0.25570666790008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,16384,2,8,4,1,power_law_1.01,0.15673600435256957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,16384,2,8,4,1,power_law_1.01,0.15905280113220216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,16384,2,8,4,1,power_law_1.01,0.16013439893722534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,16384,2,8,4,1,power_law_1.01,0.1658560037612915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,16384,2,8,4,1,power_law_1.01,0.17048959732055663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,16384,2,8,4,1,power_law_1.01,0.1838719964027405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,16384,2,8,4,1,power_law_1.01,0.1815616011619568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,16384,2,8,4,1,power_law_1.01,0.21189119815826415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,16384,2,8,4,1,power_law_1.01,0.23664638996124268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.01,0.2749056100845337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.01,0.3331968069076538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.01,0.41303038597106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.01,0.527846384048462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,8,1,power_law_1.2,1.181049633026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.01,0.657919979095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.01,0.9276927947998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.01,1.201260757446289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,4,1,power_law_1.2,0.031244799494743347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,4,1,power_law_1.2,0.03041279911994934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,4,1,power_law_1.2,0.03766399919986725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,32,8,power_law_1.01,0.04359039962291718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,4,1,power_law_1.2,0.044896000623703004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,32,8,power_law_1.01,0.04629760086536407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,4,1,power_law_1.2,0.05187839865684509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,32,8,power_law_1.01,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,32,8,power_law_1.01,0.04529919922351837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,32,8,power_law_1.01,0.0456959992647171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,32,8,power_law_1.01,0.046137601137161255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,32,8,power_law_1.01,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,32,8,power_law_1.01,0.04684160053730011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,32,8,power_law_1.01,0.048607999086380006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,4,1,power_law_1.2,0.07660160064697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,32,8,power_law_1.01,0.04859519898891449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,4,1,power_law_1.2,0.08034560084342957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,4,1,power_law_1.2,0.08549759984016418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,4,1,power_law_1.2,0.0868287980556488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,4,1,power_law_1.2,0.09056640267372132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,4,1,power_law_1.2,0.0948095977306366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,4,1,power_law_1.2,0.09911680221557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,4,1,power_law_1.2,0.09861119985580444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,32,8,power_law_1.01,0.04960640072822571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,4,1,power_law_1.2,0.1040768027305603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,32,8,power_law_1.01,0.050483202934265135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,32,8,power_law_1.01,0.0513152003288269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,32,8,power_law_1.01,0.052902400493621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,32,8,power_law_1.01,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,32,8,power_law_1.01,0.05730559825897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,32,8,power_law_1.01,0.062041598558425906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,32,8,power_law_1.01,0.0695680022239685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,4,1,power_law_1.2,0.10929919481277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,32,8,power_law_1.01,0.07765120267868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,4,1,power_law_1.2,0.11138559579849243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,32,8,power_law_1.01,0.09360640048980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,4,1,power_law_1.2,0.12083200216293336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,32,8,power_law_1.01,0.10581760406494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,4,1,power_law_1.2,0.14039039611816406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,32,8,power_law_1.01,0.1336959958076477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,32,8,power_law_1.01,0.17430399656295775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,32,8,power_law_1.01,0.2214207887649536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,32,8,power_law_1.01,0.3254463911056519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,32,8,power_law_1.01,0.4339776039123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,32,8,power_law_1.01,0.5556416034698486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,8,1,power_law_1.2,1.5246591567993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,4,1,power_law_1.2,0.15386879444122314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,4,1,power_law_1.2,0.18895360231399536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,4,1,power_law_1.2,0.20877439975738527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,4,1,power_law_1.2,0.27699201107025145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,4,1,power_law_1.2,0.34724481105804444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,4,1,power_law_1.2,0.4718912124633789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.01,1.694950485229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,4,1,power_law_1.2,0.6000192165374756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,4,1,power_law_1.2,0.8437760353088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,4,4,power_law_1.2,0.06046079993247986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,4,4,power_law_1.2,0.07124480009078979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,4,4,power_law_1.2,0.0880895972251892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,4,4,power_law_1.2,0.10105600357055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,4,4,power_law_1.2,0.12332799434661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,4,4,power_law_1.2,0.1270143985748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,4,4,power_law_1.2,0.13656320571899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,4,4,power_law_1.2,0.1746559977531433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,4,4,power_law_1.2,0.16308480501174927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,4,4,power_law_1.2,0.16583679914474486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,4,4,power_law_1.2,0.15889279842376708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,4,4,power_law_1.2,0.1740928053855896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,4,4,power_law_1.2,0.17927039861679078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,4,4,power_law_1.2,0.17448320388793945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,4,4,power_law_1.2,0.1879807949066162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,4,4,power_law_1.2,0.19128960371017456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,4,4,power_law_1.2,0.20280320644378663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,4,4,power_law_1.2,0.22387199401855468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,4,4,power_law_1.2,0.24618239402770997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,4,4,power_law_1.2,0.29925758838653566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,4,4,power_law_1.2,0.3346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,4,4,power_law_1.2,0.41427202224731446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,4,4,power_law_1.2,0.4875648021697998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,4,4,power_law_1.2,0.6617792129516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,4,4,power_law_1.2,0.8479680061340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,4,4,power_law_1.2,1.1542847633361817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,4,4,power_law_1.2,1.6214591979980468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.01,2.240153694152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,2,1,balanced,0.11894399921099345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,2,1,balanced,0.1869653264681498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,2,1,balanced,0.3227786620457967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,2,1,balanced,0.6001973152160645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,2,1,balanced,1.121285359064738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,2,1,balanced,1.644053300221761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,2,1,balanced,1.6458773612976074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,2,1,balanced,1.6474666595458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,2,1,balanced,1.6569066047668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,2,1,balanced,1.6609652837117512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,2,1,balanced,1.6752106348673503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,2,1,balanced,1.6857706705729167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,2,1,balanced,1.6980373064676921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,2,1,balanced,1.7192907333374023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,4,1,power_law_1.2,1.1040767669677733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,2,1,balanced,1.740645408630371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,2,1,balanced,1.7586347262064617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,2,1,balanced,1.7925653457641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,2,1,balanced,2.0138559341430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,2,1,balanced,1.9321120580037434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,2,1,balanced,2.866410573323568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,2,1,balanced,2.2386132876078286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,2,1,balanced,3.2452853520711265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,2,1,balanced,2.798645337422689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,2,1,balanced,4.113594690958659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,2,8,power_law_1.2,0.06589440107345582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,2,8,power_law_1.2,0.09057279825210571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,2,1,balanced,4.430303891499837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,2,8,power_law_1.2,0.09783679842948914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,2,8,power_law_1.2,0.10977920293807983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,2,8,power_law_1.2,0.12838399410247803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,2,8,power_law_1.2,0.14357119798660278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,2,1,balanced,7.23744519551595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,2,8,power_law_1.2,0.14321919679641723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,2,8,power_law_1.2,0.14967679977416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,2,8,power_law_1.2,0.13876479864120483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,2,8,power_law_1.2,0.1444607973098755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,2,1,balanced,7.323130925496419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,2,8,power_law_1.2,0.14067840576171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,2,8,power_law_1.2,0.1467519998550415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,2,8,power_law_1.2,0.14283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,2,8,power_law_1.2,0.14686720371246337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,2,8,power_law_1.2,0.16709760427474976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,2,8,power_law_1.2,0.1654080033302307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,2,8,power_law_1.2,0.17075200080871583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,2,8,power_law_1.2,0.19573119878768921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,2,8,power_law_1.2,0.2131903886795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,2,8,power_law_1.2,0.25719680786132815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,2,8,power_law_1.2,0.3038719892501831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,2,8,power_law_1.2,0.37816319465637205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,2,8,power_law_1.2,0.4086656093597412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,2,8,power_law_1.2,0.5807168006896972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,2,8,power_law_1.2,0.7002111911773682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,2,8,power_law_1.2,0.9643327713012695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,2,8,power_law_1.2,1.2440640449523925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,2,power_law_1.2,0.03185920119285583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,2,power_law_1.2,0.03345920145511627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,2,power_law_1.2,0.03489919900894165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,2,power_law_1.2,0.03991680145263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,2,power_law_1.2,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,2,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,2,power_law_1.2,0.055379199981689456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,2,power_law_1.2,0.05767040252685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,2,power_law_1.2,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,2,power_law_1.2,0.061900800466537474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,2,power_law_1.2,0.06306560039520263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,2,power_law_1.2,0.0679744005203247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,2,power_law_1.2,0.07394559979438782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,2,power_law_1.2,0.07294719815254211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,2,power_law_1.2,0.07914239764213563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,2,power_law_1.2,0.08635519742965699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,2,power_law_1.2,0.09477760195732117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,2,power_law_1.2,0.11208319664001465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,2,power_law_1.2,0.13232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,2,power_law_1.2,0.16517759561538697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,2,power_law_1.2,0.2037440061569214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,2,power_law_1.2,0.2684288024902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,2,power_law_1.2,0.3597759962081909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,2,power_law_1.2,0.5214591979980469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,2,power_law_1.2,0.6811776161193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,4,power_law_1.01,0.11266560554504394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,2,power_law_1.2,0.8843903541564941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,4,power_law_1.01,0.15512319803237914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,4,power_law_1.01,0.15444480180740355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,2,power_law_1.2,1.2354880332946778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,4,power_law_1.01,0.1998975992202759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,4,power_law_1.01,0.2374336004257202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,4,power_law_1.01,0.3070463895797729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,4,power_law_1.01,0.2918463945388794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,4,power_law_1.01,0.289683198928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,4,power_law_1.01,0.3005055904388428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,4,power_law_1.01,0.3156543970108032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,4,power_law_1.01,0.31393280029296877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,4,power_law_1.01,0.32629759311676027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,4,power_law_1.01,0.3284735918045044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,4,power_law_1.01,0.34384000301361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,4,power_law_1.01,0.35280001163482666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,4,power_law_1.01,0.3609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,4,power_law_1.01,0.381875205039978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,4,power_law_1.01,0.4284543991088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,4,power_law_1.01,0.4751423835754395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,4,power_law_1.01,0.5701119899749756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,4,power_law_1.01,0.6461952209472657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,4,power_law_1.01,0.8211711883544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,4,power_law_1.01,0.979366397857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,4,power_law_1.01,1.3038271903991698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,4,power_law_1.01,1.5842944145202638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,4,power_law_1.01,2.332480049133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,1,64,power_law_1.2,0.07434880137443542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,1,64,power_law_1.2,0.07272319793701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,4,power_law_1.01,2.9074304580688475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,1,64,power_law_1.2,0.07125759720802308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,1,64,power_law_1.2,0.07559679746627808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,1,64,power_law_1.2,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,1,64,power_law_1.2,0.0782912015914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,1,64,power_law_1.2,0.07733759880065919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,1,64,power_law_1.2,0.0788927972316742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,1,64,power_law_1.2,0.08284800052642823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,1,64,power_law_1.2,0.08072320222854615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,1,64,power_law_1.2,0.08196480274200439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,1,64,power_law_1.2,0.08710399866104127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,1,64,power_law_1.2,0.08841599822044373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,1,64,power_law_1.2,0.08464639782905578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,1,64,power_law_1.2,0.0972544014453888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,1,64,power_law_1.2,0.09906560182571411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,1,64,power_law_1.2,0.1035647988319397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,1,64,power_law_1.2,0.11429120302200317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,1,64,power_law_1.2,0.12906240224838256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,1,64,power_law_1.2,0.15890560150146485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,1,64,power_law_1.2,0.1877120018005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,1,64,power_law_1.2,0.2582783937454224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,1,64,power_law_1.2,0.3137216091156006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,1,64,power_law_1.2,0.45171198844909666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,1,64,power_law_1.2,0.5859392166137696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,1,64,power_law_1.2,0.8386303901672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,1,64,power_law_1.2,1.3302271842956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,1,balanced,0.14851733048756918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,1,balanced,0.19552000363667807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,1,balanced,0.28066132465998334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,1,balanced,0.4635466734568278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,1,balanced,0.8133760293324789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,1,balanced,1.1739146709442139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,1,balanced,1.1794453461964924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,1,balanced,1.1859947045644124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,1,balanced,1.1920639673868816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,1,balanced,1.196079969406128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,1,balanced,1.2081226507822673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,1,balanced,1.220410664876302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,1,balanced,1.2319786548614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,1,balanced,1.2500960032145183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,1,balanced,1.276794672012329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,1,balanced,1.2986719608306885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,1,balanced,1.3418025970458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,1,balanced,1.4402987162272136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,1,balanced,1.528773307800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,1,balanced,2.3158079783121743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,1,balanced,1.943189303080241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,1,balanced,2.794565200805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,1,balanced,3.0267041524251304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,1,balanced,4.495589256286621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,1,balanced,5.070218722025554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,256,power_law_1.01,0.02104319930076599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,256,power_law_1.01,0.020627200603485107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,256,power_law_1.01,0.0226623997092247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,256,power_law_1.01,0.028441599011421202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,1,balanced,7.765029271443685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,256,power_law_1.01,0.0514240026473999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,256,power_law_1.01,0.04286080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,256,power_law_1.01,0.04389120042324066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,256,power_law_1.01,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,256,power_law_1.01,0.04360319972038269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,256,power_law_1.01,0.044659200310707095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,1,balanced,9.442474365234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,256,power_law_1.01,0.04343680143356323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,256,power_law_1.01,0.043731200695037845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,256,power_law_1.01,0.04170880019664765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,256,power_law_1.01,0.03994239866733551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,256,power_law_1.01,0.04120959937572479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,256,power_law_1.01,0.043270400166511534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,256,power_law_1.01,0.04659200012683869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,256,power_law_1.01,0.048153600096702574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.01,0.05548800230026245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.01,0.07006080150604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.01,0.08822399973869324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.01,0.1078719973564148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.01,0.13786879777908326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.01,0.17573120594024658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.01,0.2300031900405884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.01,0.34371840953826904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.01,0.44618239402771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,1,power_law_1.2,0.056985598802566526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,1,power_law_1.2,0.06272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,1,power_law_1.2,0.07073919773101807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,1,power_law_1.2,0.09573760032653808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,1,power_law_1.2,0.1256832003593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,1,power_law_1.2,0.1552575945854187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,1,power_law_1.2,0.17468160390853882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,1,power_law_1.2,0.21904640197753905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,1,power_law_1.2,0.2270656108856201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,1,power_law_1.2,0.2333695888519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,1,power_law_1.2,0.2403968095779419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,1,power_law_1.2,0.253766393661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,1,power_law_1.2,0.26019198894500734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,1,power_law_1.2,0.27295360565185545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,1,power_law_1.2,0.28545279502868653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,1,power_law_1.2,0.2953727960586548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,1,power_law_1.2,0.3227776050567627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,1,power_law_1.2,0.3791680097579956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,1,power_law_1.2,0.41962881088256837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,1,power_law_1.2,0.5149504184722901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,1,power_law_1.2,0.6134848117828369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,1,power_law_1.2,0.7915711879730225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,1,power_law_1.2,0.9636159896850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,1,power_law_1.2,1.290022373199463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,1,power_law_1.2,1.6412416458129884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,1,power_law_1.2,2.2984575271606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,1,balanced,0.03454400102297465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,1,balanced,0.033759998778502144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,1,balanced,0.03525333354870478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,1,balanced,0.04193066557248434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,1,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,1,balanced,0.07286400099595387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,1,balanced,0.06465066472689311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,1,power_law_1.2,2.977619171142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,8,1,balanced,0.02207999924818675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,8,1,balanced,0.023669332265853882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,8,1,balanced,0.027280000348885853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,8,1,balanced,0.040074666341145836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,8,1,balanced,0.06195733447869619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,8,1,balanced,0.07176533341407776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,8,1,balanced,0.07445333401362102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,8,1,balanced,0.07889600098133087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,8,1,balanced,0.08085333307584126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,8,1,balanced,0.08397866288820903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,8,1,balanced,0.08589866757392883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,8,1,balanced,0.08960533142089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,8,1,balanced,0.0953493316968282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,8,1,balanced,0.09218666950861613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,1,balanced,0.07303466896216075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,1,balanced,0.06397333244482677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,1,balanced,0.06832000116507213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,1,balanced,0.0772213339805603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,8,1,balanced,0.09736532966295879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,1,balanced,0.07247999807198842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,1,balanced,0.0780213326215744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,1,balanced,0.08956799904505412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,1,balanced,0.1774346629778544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,1,balanced,0.2076639930407206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,1,balanced,0.2585653265317281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,1,balanced,0.13978667060534158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,1,balanced,0.16517333189646402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,1,balanced,0.23300800720850626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,1,balanced,0.2978773315747579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,1,balanced,0.42193599541982013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,8,1,balanced,0.10337066650390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,8,1,balanced,0.11172800262769063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,8,1,balanced,0.1451466679573059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,8,1,balanced,0.16049066185951233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,8,1,balanced,0.22034132480621338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,8,1,balanced,0.19815999269485474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,8,1,balanced,0.27243733406066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,8,1,balanced,0.2748746673266093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,8,1,balanced,0.439242680867513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,8,1,balanced,0.4926826556523641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,1,balanced,0.532256007194519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,8,1,balanced,0.7051946322123209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,1,balanced,0.7846240202585856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,8,1,balanced,0.9149280389149984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,1,balanced,1.0250293413798015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,1,balanced,1.516874631245931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,1,balanced,2.007749398549398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,2048,768,8,128,2,32,balanced,0.025258667767047882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,2048,768,8,128,2,32,balanced,0.023365333676338196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,2048,768,8,128,2,32,balanced,0.022656001150608063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,2048,768,8,128,2,32,balanced,0.024677333732446034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,2048,768,8,128,2,32,balanced,0.02534399926662445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,2048,768,8,128,2,32,balanced,0.02682666728893916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,2048,768,8,128,2,32,balanced,0.027322667340437572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,2048,768,8,128,2,32,balanced,0.02720000098148982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,2048,768,8,128,2,32,balanced,0.02716800073782603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,2048,768,8,128,2,32,balanced,0.02718399961789449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,2048,768,8,128,2,32,balanced,0.029135999580224354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,2048,768,8,128,2,32,balanced,0.029152000943819683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,2048,768,8,128,2,32,balanced,0.02922133356332779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,2048,768,8,128,2,32,balanced,0.031498665610949196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,2048,768,8,128,2,32,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,2048,768,8,128,2,32,balanced,0.033887999753157295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,2048,768,8,128,2,32,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,2048,768,8,128,2,32,balanced,0.037589333951473236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,2048,768,8,128,2,32,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,2048,768,8,128,2,32,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,2048,768,8,128,2,32,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,2048,768,8,128,2,32,balanced,0.04555733501911163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,2048,768,8,128,2,32,balanced,0.04595733185609182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,2048,768,8,128,2,32,balanced,0.05719466507434845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,2048,768,8,128,2,32,balanced,0.06886933247248332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,2048,768,8,128,2,32,balanced,0.08888000249862671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,2048,768,8,128,2,32,balanced,0.10129066308339436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,16,power_law_1.01,0.048793599009513855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,16,power_law_1.01,0.07901440262794494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,16,power_law_1.01,0.07592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,16,power_law_1.01,0.07711359858512878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,16,power_law_1.01,0.07624319791793824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,16,power_law_1.01,0.07306240200996399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,16,power_law_1.01,0.08124160170555114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,16,power_law_1.01,0.07687680125236511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,16,power_law_1.01,0.08097280263900757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,256,power_law_1.01,0.03782399892807007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,16,power_law_1.01,0.08080639839172363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,256,power_law_1.01,0.03937279880046844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,16,power_law_1.01,0.08249599933624267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,16,power_law_1.01,0.0837119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,16,power_law_1.01,0.09054719805717468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,16,power_law_1.01,0.09414399862289428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,16,power_law_1.01,0.09876480102539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,16,power_law_1.01,0.10644479990005493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,16,power_law_1.01,0.18250240087509156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,16,power_law_1.01,0.20940799713134767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.01,0.14988800287246704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,256,power_law_1.01,0.0382207989692688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.01,0.16248320341110228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,256,power_law_1.01,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.01,0.1842687964439392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,256,power_law_1.01,0.042438399791717527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.01,0.23452160358428956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.01,0.2782655954360962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.01,0.3786623954772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.01,0.470579195022583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.01,0.6542143821716309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.01,0.8403776168823243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,256,power_law_1.01,0.04488320052623749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,16,16,power_law_1.01,0.059468799829483034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,256,power_law_1.01,0.04479359984397888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,16,16,power_law_1.01,0.06559360027313232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,16,16,power_law_1.01,0.07342720031738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,16,16,power_law_1.01,0.06777600049972535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,16,16,power_law_1.01,0.06786559820175171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,16,16,power_law_1.01,0.06816639900207519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,16,16,power_law_1.01,0.06849920153617858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,16,16,power_law_1.01,0.06840959787368775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,16,16,power_law_1.01,0.0695039987564087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,16,16,power_law_1.01,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,16,16,power_law_1.01,0.0705024003982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,256,power_law_1.01,0.045535999536514285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,16,16,power_law_1.01,0.07184000015258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,16,16,power_law_1.01,0.07340160012245178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,16,16,power_law_1.01,0.07467520236968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,16,16,power_law_1.01,0.0807039976119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,16,16,power_law_1.01,0.08238080143928528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,16,16,power_law_1.01,0.09063680171966552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,16,16,power_law_1.01,0.10668799877166749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,16,power_law_1.2,0.016889600455760954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,16,power_law_1.2,0.01611520051956177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,16,16,power_law_1.01,0.11599359512329102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,16,power_law_1.2,0.016524800658226015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,16,16,power_law_1.01,0.14405759572982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,16,power_law_1.2,0.016652800142765045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,16,16,power_law_1.01,0.16892800331115723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,16,power_law_1.2,0.017868800461292265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,16,16,power_law_1.01,0.2288127899169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,16,power_law_1.2,0.0177279993891716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,256,power_law_1.01,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,16,power_law_1.2,0.01892479956150055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,16,16,power_law_1.01,0.25840001106262206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,256,power_law_1.01,0.04720639884471893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,16,power_law_1.2,0.018483200669288637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,16,power_law_1.2,0.019788800179958342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,16,power_law_1.2,0.020179200172424316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,16,power_law_1.2,0.02220800071954727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,16,power_law_1.2,0.022995199263095855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,16,power_law_1.2,0.02330880016088486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,16,power_law_1.2,0.03128960132598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,16,16,power_law_1.01,0.34791040420532227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,16,power_law_1.2,0.03275519907474518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,16,16,power_law_1.01,0.4267712116241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,16,power_law_1.2,0.03107840120792389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,16,power_law_1.2,0.032441601157188416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,16,16,power_law_1.01,0.6094592094421387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,16,power_law_1.2,0.035795199871063235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,16,power_law_1.2,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,16,16,power_law_1.01,0.7749184131622314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,16,power_law_1.2,0.04392319917678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,16,power_law_1.2,0.04798719882965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,256,power_law_1.01,0.05169919729232788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,256,power_law_1.01,0.05338240265846252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,256,power_law_1.01,0.048716801404953006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,256,power_law_1.01,0.05418879985809326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,256,power_law_1.01,0.05943040251731872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,256,power_law_1.01,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,16,power_law_1.2,0.05806080102920532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,16,power_law_1.2,0.07037439942359924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,256,power_law_1.01,0.06379520297050476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,256,power_law_1.01,0.07257599830627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,256,power_law_1.01,0.07682560086250305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,256,power_law_1.01,0.09434880018234253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,256,power_law_1.01,0.10773119926452637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,16,power_law_1.2,0.09216639995574952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,256,power_law_1.01,0.14224640130996705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,256,power_law_1.01,0.17198079824447632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,256,power_law_1.01,0.2504512071609497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,256,power_law_1.01,0.315500807762146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,256,power_law_1.01,0.44808320999145507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,256,power_law_1.01,0.6239039897918701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,16,power_law_1.2,0.11086080074310303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,4,1,power_law_1.2,0.0514303982257843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,4,1,power_law_1.2,0.06410239934921265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,4,1,power_law_1.2,0.07564160227775574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,4,1,power_law_1.2,0.1078719973564148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,16,power_law_1.2,0.16200319528579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,4,1,power_law_1.2,0.13953280448913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,4,1,power_law_1.2,0.16900479793548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,4,1,power_law_1.2,0.2381119966506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,4,1,power_law_1.2,0.24764800071716309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,4,1,power_law_1.2,0.257043194770813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,4,1,power_law_1.2,0.26394240856170653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,4,1,power_law_1.2,0.2695487976074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,4,1,power_law_1.2,0.2864703893661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,4,1,power_law_1.2,0.2897599935531616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,4,1,power_law_1.2,0.30500481128692625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,4,1,power_law_1.2,0.3123840093612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,32,8,balanced,0.020080000162124634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,4,1,power_law_1.2,0.3213887929916382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,4,1,power_law_1.2,0.34108800888061525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,4,1,power_law_1.2,0.376582407951355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,4,1,power_law_1.2,0.3940927982330322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,4,1,power_law_1.2,0.45829119682312014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,4,1,power_law_1.2,0.5184512138366699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,4,1,power_law_1.2,0.6453504085540771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,4,1,power_law_1.2,0.7225855827331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,4,1,power_law_1.2,0.9349760055541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,32,8,balanced,0.021087999145189922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,32,8,balanced,0.02313599983851115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,32,8,balanced,0.023562667270501454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,32,8,balanced,0.025386666258176167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,32,8,balanced,0.027802666028340656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,32,8,balanced,0.030623999734719593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,32,8,balanced,0.031583999594052635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,32,8,balanced,0.03324799984693527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,32,8,balanced,0.03369066615899404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,32,8,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,32,8,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,32,8,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,4,power_law_1.01,0.0321727991104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,32,8,balanced,0.03963200002908707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,32,8,balanced,0.041834667325019836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,4,power_law_1.01,0.04357120096683502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,4,power_law_1.01,0.046463999152183535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,4,1,power_law_1.2,1.1721471786499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,4,1,power_law_1.2,1.5946111679077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,32,8,balanced,0.04961599906285604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,32,8,balanced,0.05198400219281515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,32,8,balanced,0.06434133152167003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,32,8,balanced,0.08617599805196126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,4,power_law_1.01,0.055827200412750244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,4,power_law_1.01,0.060140800476074216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,4,power_law_1.01,0.07037439942359924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,4,power_law_1.01,0.06910079717636108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,4,power_law_1.01,0.07249280214309692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,4,power_law_1.01,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,4,power_law_1.01,0.07171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,4,1,power_law_1.2,2.0283071517944338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,4,power_law_1.01,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,32,8,balanced,0.11422399679819743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,4,power_law_1.01,0.07409279942512512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,4,power_law_1.01,0.07709439992904663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,4,power_law_1.01,0.0791104018688202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,4,power_law_1.01,0.09394559860229493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,4,power_law_1.01,0.09765120148658753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,4,power_law_1.01,0.09813119769096375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,4,power_law_1.01,0.10537600517272949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,4,power_law_1.01,0.12045439481735229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,4,power_law_1.01,0.13583359718322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,4,power_law_1.01,0.1313599944114685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,4,power_law_1.01,0.1561535954475403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,4,power_law_1.01,0.1799999952316284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,4,power_law_1.01,0.22499840259552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,4,power_law_1.01,0.2881472110748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,4,power_law_1.01,0.3789184093475342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,4,power_law_1.01,0.45751681327819826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,32,8,balanced,0.10971200466156006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,32,8,balanced,0.13904533783594766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,32,8,balanced,0.16675732533137003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,32,8,balanced,0.20866666237513223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,32,8,balanced,0.2664533257484436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,32,8,balanced,0.387445330619812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,32,8,balanced,0.4874613285064697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,16,power_law_1.2,0.19742720127105712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,2,128,power_law_1.01,0.06488959789276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,2,128,power_law_1.01,0.06390399932861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,2,128,power_law_1.01,0.06317440271377564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,2,128,power_law_1.01,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,2,128,power_law_1.01,0.0668287992477417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,2,128,power_law_1.01,0.07066239714622498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,2,128,power_law_1.01,0.06758400201797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,2,128,power_law_1.01,0.06951040029525757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,2,128,power_law_1.01,0.06751360297203064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,2,128,power_law_1.01,0.0700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,2,128,power_law_1.01,0.07077119946479797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,2,128,power_law_1.01,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,2,128,power_law_1.01,0.07137280106544494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,2,128,power_law_1.01,0.07263360023498536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,2,128,power_law_1.01,0.07561600208282471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,2,128,power_law_1.01,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,2,128,power_law_1.01,0.07959679961204529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,2,128,power_law_1.01,0.08913919925689698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,2,128,power_law_1.01,0.09961599707603455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,2,128,power_law_1.01,0.11310720443725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,2,128,power_law_1.01,0.1250496029853821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,2,128,power_law_1.01,0.15877120494842528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,2,128,power_law_1.01,0.19470080137252807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,2,128,power_law_1.01,0.2651648044586182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,2,128,power_law_1.01,0.325382399559021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,2,128,power_law_1.01,0.44946560859680174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,2,128,power_law_1.01,0.557369613647461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,16,2,balanced,0.03809066613515218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,16,2,balanced,0.04374399781227112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,16,2,balanced,0.04801600178082784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,16,2,balanced,0.046069333950678505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,16,2,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,16,2,balanced,0.04845866560935974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,16,2,balanced,0.04731733103593191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,16,2,balanced,0.04693866769472758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,16,2,balanced,0.04801600178082784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,16,2,balanced,0.048021331429481506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,16,2,balanced,0.04730666677157084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,16,2,balanced,0.04824000100294749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,16,2,balanced,0.04799466828505198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,16,2,balanced,0.047925333182017006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,16,2,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,16,2,balanced,0.052560001611709595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,16,2,balanced,0.053642665346463524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,16,2,balanced,0.06669866542021434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,16,2,balanced,0.06698133548100789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,16,2,balanced,0.08067733546098073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,16,2,balanced,0.09386666615804036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,16,2,balanced,0.12146133184432983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,16,2,balanced,0.14314666390419006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,16,2,balanced,0.19394133488337198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,16,2,balanced,0.24701333045959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,16,2,balanced,0.34690133730570477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,16,2,balanced,0.4466346502304077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,32,power_law_1.2,0.018156799674034118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,1,power_law_1.2,0.1256767988204956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,32,power_law_1.2,0.0179967999458313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,1,power_law_1.2,0.18009599447250366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,32,power_law_1.2,0.01785600036382675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,1,power_law_1.2,0.21333119869232178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,32,power_law_1.2,0.018060800433158875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,1,power_law_1.2,0.38740479946136475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,32,power_law_1.2,0.019603200256824493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,1,power_law_1.2,0.3840575933456421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,1,power_law_1.2,0.39226880073547366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,1,power_law_1.2,0.4033215999603271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,1,power_law_1.2,0.4016128063201904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,1,power_law_1.2,0.4135104179382324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,32,power_law_1.2,0.022227199375629426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,1,power_law_1.2,0.4268223762512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,32,power_law_1.2,0.03969280123710632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,32,power_law_1.2,0.03976959884166718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,1,power_law_1.2,0.4415679931640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,32,power_law_1.2,0.033766400814056394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,1,power_law_1.2,0.45824642181396485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,32,power_law_1.2,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,32,power_law_1.2,0.03363839983940124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,1,power_law_1.2,0.4514431953430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,32,power_law_1.2,0.03391999900341034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,1,power_law_1.2,0.46503682136535646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,32,power_law_1.2,0.03301759958267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,1,power_law_1.2,0.4928895950317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,32,power_law_1.2,0.03190400004386902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,32,power_law_1.2,0.034016001224517825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,1,power_law_1.2,0.5068927764892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,32,power_law_1.2,0.03441919982433319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,32,power_law_1.2,0.03761920034885406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,32,power_law_1.2,0.042182400822639465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.2,0.04517120122909546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,1,power_law_1.2,0.552288007736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.2,0.05095040202140808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,1,power_law_1.2,0.6212287902832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.2,0.058361601829528806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,1,power_law_1.2,0.7214975833892823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,1,power_law_1.2,0.9216704368591309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.2,0.08102399706840516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.2,0.09233279824256897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,1,power_law_1.2,1.105190372467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.2,0.12575360536575317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,1,power_law_1.2,1.4873663902282714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,1,power_law_1.2,1.822502326965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.2,0.15611519813537597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,1,power_law_1.2,2.61529598236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,2,64,balanced,0.04970666766166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,2,64,balanced,0.04418666660785675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,2,64,balanced,0.04595733185609182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,2,64,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,2,64,balanced,0.04845866560935974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,2,4,power_law_1.01,0.10273920297622681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.2,0.21619200706481934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,2,64,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,2,4,power_law_1.01,0.13638399839401244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,2,64,balanced,0.052005335688591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,2,64,balanced,0.053818667928377785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,2,64,balanced,0.05403199791908264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,32,1,balanced,0.05220800141493479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,2,64,balanced,0.054192001620928444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,32,1,balanced,0.05547733108202616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,2,64,balanced,0.05369600156943003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,2,64,balanced,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,2,64,balanced,0.055957332253456116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,2,64,balanced,0.05594133337338766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,2,64,balanced,0.06233066817124685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,2,64,balanced,0.0629013329744339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,2,64,balanced,0.06629866858323415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,2,64,balanced,0.07275199890136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,2,64,balanced,0.07666133344173431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,2,4,power_law_1.01,0.1396607995033264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,2,4,power_law_1.01,0.18803839683532714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,2,4,power_law_1.01,0.18303999900817872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,32,1,balanced,0.06085866689682007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,2,4,power_law_1.01,0.28333439826965334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,32,1,balanced,0.08040533463160197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,32,1,balanced,0.1127786636352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,2,4,power_law_1.01,0.2822400093078613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,2,4,power_law_1.01,0.28224639892578124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,2,4,power_law_1.01,0.2929728031158447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,2,4,power_law_1.01,0.2894079923629761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,2,4,power_law_1.01,0.2990592002868652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,2,4,power_law_1.01,0.2953279972076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,2,4,power_law_1.01,0.3014847993850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,32,1,balanced,0.16703466574350992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,32,1,balanced,0.16849599281946817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,32,1,balanced,0.1698240041732788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,32,1,balanced,0.1728960076967875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,32,1,balanced,0.17542399962743124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,4,power_law_1.01,0.18627840280532837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,32,1,balanced,0.17756799856821695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,32,1,balanced,0.18546666701634726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,32,1,balanced,0.18926932414372763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,32,1,balanced,0.19616534312566122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,32,1,balanced,0.2047520081202189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,32,1,balanced,0.21066133181254068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,32,1,balanced,0.22805333137512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,32,1,balanced,0.2693333427111308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,32,1,balanced,0.29769599437713623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,2,4,power_law_1.01,0.3236095905303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,32,1,balanced,0.3707946538925171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,2,64,balanced,0.09030399719874065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,32,1,balanced,0.42501334349314374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,2,4,power_law_1.01,0.33175039291381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,32,1,balanced,0.6356106599171957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,2,4,power_law_1.01,0.33403520584106444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,32,1,balanced,0.7282079855600992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,1,power_law_1.2,3.5190078735351564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,2,4,power_law_1.01,0.3605504035949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,32,1,balanced,1.0909813245137532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,2,4,power_law_1.01,0.40677762031555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,2,4,power_law_1.01,0.40770559310913085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,32,1,balanced,1.367952028910319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,2,4,power_law_1.01,0.49159679412841795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,32,1,balanced,1.9989013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,2,64,balanced,0.10500799616177876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,2,4,power_law_1.01,0.5635903835296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,2,64,balanced,0.13111467162768045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,2,64,balanced,0.14869333306948343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,2,64,balanced,0.19050665696461996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,2,4,power_law_1.01,0.7388864040374756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,32,1,balanced,2.635845343271891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,2,64,balanced,0.24211732546488443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,4,power_law_1.01,0.3228032112121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,2,64,balanced,0.3203253348668416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,2,4,power_law_1.01,0.6581056118011475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,2,64,balanced,0.3976000150044759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,4,power_law_1.01,0.3146048069000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,2,4,power_law_1.01,0.8171199798583985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,4,power_law_1.01,0.41605119705200194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,2,4,power_law_1.01,1.1450112342834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,4,power_law_1.01,0.5728960037231445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,2,4,power_law_1.01,1.4950336456298827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,4,power_law_1.01,0.710918378829956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,2,4,power_law_1.01,1.8149503707885741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,4,power_law_1.01,0.7219456195831299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,4,power_law_1.01,0.7169792175292968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,4,power_law_1.01,0.7286208152770997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,4,power_law_1.01,0.7291327953338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,4,power_law_1.01,0.7554048061370849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,4,power_law_1.01,0.7845695972442627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,4,power_law_1.01,0.7888319969177247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,4,power_law_1.01,0.8001472473144531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,4,power_law_1.01,0.8400128364562989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,4,power_law_1.01,0.8561152458190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,4,power_law_1.01,0.9316800117492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,4,power_law_1.01,1.049503993988037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,4,power_law_1.01,1.0327168464660645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,4,power_law_1.01,1.1974271774291991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,4,power_law_1.01,1.2652671813964844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,4,power_law_1.01,1.4844032287597657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,4,power_law_1.01,1.6392383575439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,4,power_law_1.01,1.9722944259643556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,4,power_law_1.01,2.656595230102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,4,power_law_1.01,3.334783935546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,4,power_law_1.01,3.8747840881347657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.2,0.2823296070098877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,1,power_law_1.2,5.09447021484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,1,power_law_1.2,6.581779479980469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,8,4,power_law_1.2,0.06695039868354798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,8,4,power_law_1.2,0.07655680179595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,8,4,power_law_1.2,0.0865664005279541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,8,4,power_law_1.2,0.10720640420913696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,8,4,power_law_1.2,0.12043520212173461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,8,4,power_law_1.2,0.1507840037345886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,8,4,power_law_1.2,0.14751360416412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,8,4,power_law_1.2,0.1532863974571228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,8,4,power_law_1.2,0.14684159755706788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,8,4,power_law_1.2,0.15470720529556276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,8,4,power_law_1.2,0.15631359815597534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,8,4,power_law_1.2,0.16399359703063965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,8,4,power_law_1.2,0.16787840127944947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,8,4,power_law_1.2,0.17511680126190185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,8,balanced,0.023792001108328503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,8,balanced,0.02309866746266683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,8,4,power_law_1.2,0.18391679525375365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,8,balanced,0.02149333308140437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,8,4,power_law_1.2,0.19052799940109252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,8,balanced,0.02147199958562851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,8,balanced,0.023120000958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,8,balanced,0.02535466601451238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,8,4,power_law_1.2,0.20302720069885255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,8,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,8,balanced,0.04970666766166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,8,balanced,0.043706665436426796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,8,4,power_law_1.2,0.24216320514678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,8,4,power_law_1.2,0.25886080265045164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,8,4,power_law_1.2,0.33403520584106444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,8,4,power_law_1.2,0.3994751930236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,8,4,power_law_1.2,0.5176127910614013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,8,balanced,0.041738669077555336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,8,4,power_law_1.2,0.5838335990905762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,8,4,power_law_1.2,0.8057984352111817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,8,4,power_law_1.2,1.0147583961486817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,8,balanced,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,8,balanced,0.051914667089780174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,8,balanced,0.04971200227737427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,8,balanced,0.04159466673930486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,8,balanced,0.06323733429114024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,8,4,power_law_1.2,1.4094464302062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,8,4,power_law_1.2,1.7623552322387694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,8,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,8,balanced,0.06028266747792562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,8,balanced,0.07453866799672444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,8,balanced,0.09301333626111348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,8,balanced,0.12784000237782797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,8,balanced,0.15292267004648843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,8,balanced,0.21451733509699503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,8,balanced,0.27003733317057294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,8,balanced,0.38487998644510907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,8,balanced,0.5007359981536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,8,balanced,0.7316533724466959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,8,balanced,0.9610400199890137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,1,balanced,0.05287466446558634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,1,balanced,0.06862933437029521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,1,balanced,0.09692800045013428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,1,balanced,0.15285332997639975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,1,balanced,0.2650773326555888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,1,balanced,0.3306933244069417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,1,balanced,0.34535467624664307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,1,balanced,0.35949865976969403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,1,balanced,0.3744266827901204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,1,balanced,0.38917867342631024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,1,balanced,0.4175146818161011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,32,4,balanced,0.044405331214269005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,1,balanced,0.4465493361155192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,32,4,balanced,0.046341334780057274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,1,balanced,0.39655999342600506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,32,4,balanced,0.05020266771316528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,1,balanced,0.4232586622238159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,1,balanced,0.449621319770813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,1,balanced,0.4159466822942098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,16,1,power_law_1.2,0.049369600415229795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,1,balanced,0.4464373191197713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,16,1,power_law_1.2,0.05304960012435913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,32,4,balanced,0.0561706672112147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,16,1,power_law_1.2,0.05518720149993896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,32,4,balanced,0.06481066842873891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,32,4,balanced,0.09359467029571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,16,1,power_law_1.2,0.06954240202903747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,32,4,balanced,0.12062399586041768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,16,1,power_law_1.2,0.08275840282440186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,32,4,balanced,0.12343999743461609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,32,4,balanced,0.12037866314252217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,16,1,power_law_1.2,0.09633920192718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,32,4,balanced,0.12355732917785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,32,4,balanced,0.11738133430480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,32,4,balanced,0.12071466445922852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,32,4,balanced,0.12199466427167256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,32,4,balanced,0.12613333264986673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,32,4,balanced,0.13074666261672974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,32,4,balanced,0.13447466492652893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,32,4,balanced,0.14028799533843994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,32,4,balanced,0.15496533115704855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,32,4,balanced,0.16661866505940756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,32,4,balanced,0.1916053295135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,16,1,power_law_1.2,0.11779839992523193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,1,balanced,0.5032693147659302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,16,1,power_law_1.2,0.12156800031661988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,1,balanced,0.48842668533325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,16,1,power_law_1.2,0.12847360372543334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,1,balanced,1.0815040270487468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,16,1,power_law_1.2,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,16,1,power_law_1.2,0.13555840253829957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,1,balanced,1.2266080379486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,16,1,power_law_1.2,0.13875839710235596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,1,balanced,0.8554933071136475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,16,1,power_law_1.2,0.14421119689941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,1,balanced,0.84115203221639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,16,1,power_law_1.2,0.14866559505462645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,16,1,power_law_1.2,0.15581439733505248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,1,8,balanced,0.06492800017197926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,1,8,balanced,0.06373866895834605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,1,8,balanced,0.06238933404286703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,1,8,balanced,0.06646400193373363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,1,8,balanced,0.06885333359241486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,1,8,balanced,0.06853333115577698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,1,8,balanced,0.07017600039641063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,32,4,balanced,0.2182613412539164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,32,4,balanced,0.2728479901949565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,32,4,balanced,0.34426132837931317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,32,4,balanced,0.45119468371073407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,32,4,balanced,0.586464007695516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,1,balanced,1.5123200416564941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,16,1,power_law_1.2,0.16028800010681152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,32,4,balanced,0.8535733222961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,1,balanced,1.594922701517741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,16,1,power_law_1.2,0.18533120155334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,32,4,balanced,1.093295971552531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,16,1,power_law_1.2,0.21378560066223146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,1,8,balanced,0.07023466626803081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,4,16,power_law_1.01,0.03726719915866852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,16,1,power_law_1.2,0.23990399837493898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,1,balanced,2.3644960721333823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,4,16,power_law_1.01,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,16,1,power_law_1.2,0.29580159187316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,4,16,power_law_1.01,0.03121280074119568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,16,1,power_law_1.2,0.3462591886520386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,4,16,power_law_1.01,0.0326335996389389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,1,balanced,3.0223681131998696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,16,1,power_law_1.2,0.4401792049407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,4,16,power_law_1.01,0.03255679905414581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,4,16,power_law_1.01,0.03447040021419525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,16,1,power_law_1.2,0.543839979171753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,4,16,power_law_1.01,0.035180801153182985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,1,8,balanced,0.07419200241565704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,1,8,balanced,0.07464533547560374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,1,8,balanced,0.07474133372306824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,4,16,power_law_1.01,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,16,1,power_law_1.2,0.7519743919372559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,4,16,power_law_1.01,0.03671039938926697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,4,16,power_law_1.01,0.03766399919986725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,16,1,power_law_1.2,0.9710463523864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,16,1,power_law_1.2,1.360211181640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,16,1,power_law_1.2,1.7536767959594726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,1,8,balanced,0.07735466460386912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,1,8,balanced,0.07653866708278656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,4,16,power_law_1.01,0.03943040072917938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,4,16,power_law_1.01,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,4,16,power_law_1.01,0.040992000699043275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,4,16,power_law_1.01,0.04273920059204102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,4,16,power_law_1.01,0.04717440009117126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,4,16,power_law_1.01,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,4,16,power_law_1.01,0.05120639801025391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,4,16,power_law_1.01,0.06027519702911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,4,16,power_law_1.01,0.06594560146331788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,1,8,balanced,0.08290666838486989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,4,16,power_law_1.01,0.07746559977531434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,4,16,power_law_1.01,0.09827839732170104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,1,8,balanced,0.09071466326713562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,4,16,power_law_1.01,0.132096004486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,4,16,power_law_1.01,0.15344640016555786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,1,8,balanced,0.10521599650382996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,1,8,balanced,0.09192533294359843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,1,8,balanced,0.11921599507331848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,4,16,power_law_1.01,0.22206718921661378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,1,8,balanced,0.1113866666952769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,1,8,balanced,0.1442400018374125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,1,8,balanced,0.16370133558909097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,1,8,balanced,0.20974934101104736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,1,8,balanced,0.27213333050409955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,1,8,balanced,0.36314666271209717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,1,8,balanced,0.46591468652089435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,1,8,balanced,0.6796853542327881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,4,16,power_law_1.01,0.2750591993331909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,4,16,power_law_1.01,0.3852096080780029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,1,8,balanced,0.8719093004862467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,4,16,power_law_1.01,0.4651328086853027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,16,4,power_law_1.2,0.031673601269721983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,16,4,power_law_1.2,0.03216640055179596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,16,4,power_law_1.2,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,16,4,power_law_1.2,0.03573119938373566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,16,4,power_law_1.2,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,16,4,power_law_1.2,0.04197759926319122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,16,4,power_law_1.2,0.04287360012531281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,16,4,power_law_1.2,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,16,4,power_law_1.2,0.04599039852619171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,16,4,power_law_1.2,0.046214398741722104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,16,4,power_law_1.2,0.05066879987716675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,16,4,power_law_1.2,0.053401601314544675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,16,4,power_law_1.2,0.05583360195159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,16,4,power_law_1.2,0.0541055977344513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,16,4,power_law_1.2,0.06367999911308289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,16,4,power_law_1.2,0.06783360242843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,16,4,power_law_1.2,0.07065600156784058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,16,4,power_law_1.2,0.08118399977684021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,16,4,power_law_1.2,0.09372159838676453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,16,4,power_law_1.2,0.12024960517883301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,16,4,power_law_1.2,0.14763519763946534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,16,4,power_law_1.2,0.18961280584335327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,16,4,power_law_1.2,0.28185598850250243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,16,4,power_law_1.2,0.36446080207824705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,16,4,power_law_1.2,0.45229439735412597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,16,4,power_law_1.2,0.720198392868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,4,1,power_law_1.01,0.052288001775741576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,4,1,power_law_1.01,0.05914239883422852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,16,4,power_law_1.2,0.8744640350341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,4,1,power_law_1.01,0.07427840232849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,4,1,power_law_1.01,0.09723520278930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,4,1,power_law_1.01,0.09717760086059571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,4,1,power_law_1.01,0.09873279929161072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,4,1,power_law_1.01,0.09911680221557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,4,1,power_law_1.01,0.10034559965133667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,4,1,power_law_1.01,0.09996160268783569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,4,1,power_law_1.01,0.10030080080032348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,4,1,power_law_1.01,0.10081280469894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,4,1,power_law_1.01,0.10375679731369018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,4,1,power_law_1.01,0.10551680326461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,4,1,power_law_1.01,0.1181439995765686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,4,1,power_law_1.01,0.12616959810256959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,4,1,power_law_1.01,0.12929279804229737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,128,balanced,0.05300800005594889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,128,balanced,0.052000001072883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,128,balanced,0.052042668064435325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,128,balanced,0.05622399846712748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,128,balanced,0.05400000015894572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,128,balanced,0.06228266656398773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,4,1,power_law_1.01,0.133241605758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,128,balanced,0.07043200234572093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,128,balanced,0.07259200016657512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,4,1,power_law_1.01,0.16267520189285278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.01,0.18961280584335327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,128,balanced,0.07751999795436859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,128,balanced,0.07765333354473114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,128,balanced,0.07258666555086772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.01,0.25319039821624756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,128,balanced,0.07467733323574066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,128,balanced,0.08222400148709615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,128,balanced,0.07852266728878021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,128,balanced,0.09270933270454407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,128,balanced,0.09276266892751057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,128,balanced,0.09530666470527649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,128,balanced,0.10538666447003682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,128,balanced,0.11158399780591328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,128,balanced,0.1234933336575826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.01,0.30328960418701173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,128,balanced,0.13755733768145242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,128,balanced,0.1705120007197062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,128,balanced,0.19512534141540527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,128,balanced,0.24865599473317465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,128,balanced,0.2978293299674988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,128,balanced,0.40378665924072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,128,balanced,0.5155200163523356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.01,0.41458559036254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.01,0.5207424163818359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.01,0.7628799915313721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.01,0.9733759880065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,2,power_law_1.01,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,2,power_law_1.01,0.0927616000175476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,2,power_law_1.01,0.11103999614715576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,2,power_law_1.01,0.1398848056793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,2,power_law_1.01,0.18394240140914916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,2,power_law_1.01,0.27231359481811523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.01,1.400051212310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,2,power_law_1.01,0.26703999042510984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,2,power_law_1.01,0.27653119564056394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,2,power_law_1.01,0.2814784049987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,2,power_law_1.01,0.29655680656433103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,2,power_law_1.01,0.29818239212036135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,2,power_law_1.01,0.310534405708313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,2,power_law_1.01,0.31084160804748534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,2,power_law_1.01,0.3162751913070679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,2,power_law_1.01,0.33870720863342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,2,power_law_1.01,0.34450559616088866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,2,power_law_1.01,0.3568320035934448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,2,power_law_1.01,0.3838720083236694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,2,power_law_1.01,0.416761589050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,2,balanced,0.021344001094500225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,2,balanced,0.021386665602525074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,2,power_law_1.01,0.498854398727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,2,balanced,0.022976001103719074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,2,balanced,0.03137599925200144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,2,power_law_1.01,0.5594560146331787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,2,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,2,balanced,0.048826664686203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,2,power_law_1.01,0.698252820968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,2,balanced,0.051274667183558144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,2,balanced,0.052442664901415505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,2,power_law_1.01,0.7502079963684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,2,balanced,0.055973331133524575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,2688,1856,6,128,1,4,power_law_1.2,0.06115840077400207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,2,balanced,0.05540800094604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,2,balanced,0.0572213331858317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,2,power_law_1.01,0.9372735977172851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,2688,1856,6,128,1,4,power_law_1.2,0.06280959844589233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,2,balanced,0.06232533355553945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,2688,1856,6,128,1,4,power_law_1.2,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,2688,1856,6,128,1,4,power_law_1.2,0.07428479790687562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,2688,1856,6,128,1,4,power_law_1.2,0.09975680112838745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,2688,1856,6,128,1,4,power_law_1.2,0.10025600194931031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,2688,1856,6,128,1,4,power_law_1.2,0.10558719635009765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,2688,1856,6,128,1,4,power_law_1.2,0.10807679891586304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,2,balanced,0.06392533580462138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,2,balanced,0.06419733166694641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,2,balanced,0.08169599870840709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,2,balanced,0.08135466774304707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,2,balanced,0.08238933483759563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,2,balanced,0.11716266473134358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,2,balanced,0.12541866302490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,2,balanced,0.14682132999102274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,2,balanced,0.15413332978884378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,2,balanced,0.16351999839146933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,2,balanced,0.17521599928538004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,2,balanced,0.284335990746816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,2,balanced,0.30209600925445557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,2688,1856,6,128,1,4,power_law_1.2,0.10039039850234985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,2,balanced,0.43317333857218426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,2688,1856,6,128,1,4,power_law_1.2,0.10517120361328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,2,balanced,0.5660853385925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,2688,1856,6,128,1,4,power_law_1.2,0.10766719579696656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,2688,1856,6,128,1,4,power_law_1.2,0.11200640201568604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,2688,1856,6,128,1,4,power_law_1.2,0.10992640256881714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,2688,1856,6,128,1,4,power_law_1.2,0.10792959928512573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,2688,1856,6,128,1,4,power_law_1.2,0.1173632025718689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,2688,1856,6,128,1,4,power_law_1.2,0.11697920560836791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,2688,1856,6,128,1,4,power_law_1.2,0.12186880111694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,2,power_law_1.01,1.2663552284240722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,2688,1856,6,128,1,4,power_law_1.2,0.13033599853515626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,2688,1856,6,128,1,4,power_law_1.2,0.14521600008010865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,2688,1856,6,128,1,4,power_law_1.2,0.1689344048500061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,2,power_law_1.01,1.722822380065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,2688,1856,6,128,1,4,power_law_1.2,0.19900799989700318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,2688,1856,6,128,1,4,power_law_1.2,0.2541248083114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,2688,1856,6,128,1,4,power_law_1.2,0.317248010635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,2,power_law_1.01,2.177964782714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,2688,1856,6,128,1,4,power_law_1.2,0.4039936065673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,2688,1856,6,128,1,4,power_law_1.2,0.5021759986877441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,384,1,2,power_law_1.01,0.09586560130119323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,384,1,2,power_law_1.01,0.12947200536727904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,2688,1856,6,128,1,4,power_law_1.2,0.7614463806152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,384,1,2,power_law_1.01,0.18651520013809203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,1,power_law_1.01,0.018988800048828126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,384,1,2,power_law_1.01,0.2267967939376831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,1,power_law_1.01,0.021747200191020964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.01,1.8659072875976563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,384,1,2,power_law_1.01,0.3376384019851685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,384,1,2,power_law_1.01,0.4181056022644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,384,1,2,power_law_1.01,0.47718400955200196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,2688,1856,6,128,1,4,power_law_1.2,0.8953472137451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,384,1,2,power_law_1.01,0.629030418395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,1,power_law_1.01,0.028697600960731505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,1,power_law_1.01,0.036575999855995175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,384,1,2,power_law_1.01,0.5915647983551026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,1,power_law_1.01,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,1,power_law_1.01,0.061440002918243405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,1,power_law_1.01,0.06379520297050476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,1,power_law_1.01,0.06746879816055298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,1,power_law_1.01,0.06746240258216858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,1,power_law_1.01,0.07088639736175537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,384,1,2,power_law_1.01,0.6608255863189697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,1,power_law_1.01,0.07564799785614014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,1,power_law_1.01,0.07463679909706115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,384,1,2,power_law_1.01,0.6302015781402588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,1,power_law_1.01,0.07858560085296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,1,power_law_1.01,0.08193280100822449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,384,1,2,power_law_1.01,0.6529151916503906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,384,1,2,power_law_1.01,0.6607999801635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,1,power_law_1.01,0.08158079981803894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,384,1,2,power_law_1.01,0.6714176177978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,1,power_law_1.01,0.08249599933624267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,384,1,2,power_law_1.01,0.6871168136596679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,1,power_law_1.01,0.09210879802703857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,1,power_law_1.01,0.09619200229644775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,384,1,2,power_law_1.01,0.7344639778137207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.01,0.10661120414733886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.01,0.20358400344848632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,384,1,2,power_law_1.01,0.7354623794555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.01,0.2370687961578369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.01,0.20230400562286377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.01,0.23048319816589355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.01,0.2883519887924194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.01,0.3557503938674927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.01,0.48630399703979493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.01,0.6201600074768067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,384,1,2,power_law_1.01,0.7906176090240479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,384,1,2,power_law_1.01,0.8575424194335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,384,1,2,power_law_1.01,0.9284928321838379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,384,1,2,power_law_1.01,1.0472384452819825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,384,1,2,power_law_1.01,1.241049575805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,384,1,2,power_law_1.01,1.327552032470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,384,1,2,power_law_1.01,1.6945856094360352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,384,1,2,power_law_1.01,2.0467071533203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,384,1,2,power_law_1.01,2.8304447174072265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,384,1,2,power_law_1.01,3.315225601196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,32,balanced,0.02372266600529353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,32,balanced,0.023205332458019257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,32,balanced,0.023178666830062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,32,balanced,0.023002666731675465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,32,balanced,0.023056000471115112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,32,balanced,0.023434666295846302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,32,balanced,0.028010666370391846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,32,balanced,0.028880000114440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,32,balanced,0.0552106648683548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,32,balanced,0.0539626677831014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,32,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,32,balanced,0.0470773329337438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,32,balanced,0.04600533346335093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,32,balanced,0.047685335079828896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,32,balanced,0.05146133402983347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,32,balanced,0.054048001766204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,32,balanced,0.053904001911481224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,32,balanced,0.060415998101234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,32,balanced,0.06915199756622314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,32,balanced,0.08242666721343994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,32,balanced,0.09690666198730469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,32,balanced,0.1322773297627767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,32,balanced,0.15985600153605142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,32,balanced,0.22036266326904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,32,balanced,0.283680001894633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,32,balanced,0.4111573298772176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,32,balanced,0.5424373149871826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,2048,768,8,128,2,128,power_law_1.01,0.026624000072479247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,2048,768,8,128,2,128,power_law_1.01,0.026694399118423463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,2048,768,8,128,2,128,power_law_1.01,0.02648960053920746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,2048,768,8,128,2,128,power_law_1.01,0.028832000494003297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,2048,768,8,128,2,128,power_law_1.01,0.029177600145339967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,2048,768,8,128,2,128,power_law_1.01,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,2048,768,8,128,2,128,power_law_1.01,0.030585598945617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,2048,768,8,128,2,128,power_law_1.01,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,2048,768,8,128,2,128,power_law_1.01,0.0310591995716095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,2048,768,8,128,2,128,power_law_1.01,0.030752000212669373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,2048,768,8,128,2,128,power_law_1.01,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,2048,768,8,128,2,128,power_law_1.01,0.03225600123405457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,2048,768,8,128,2,128,power_law_1.01,0.03331199884414673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,2048,768,8,128,2,128,power_law_1.01,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,2048,768,8,128,2,128,power_law_1.01,0.03840000033378601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,2048,768,8,128,2,128,power_law_1.01,0.03948160111904144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,2048,768,8,128,2,128,power_law_1.01,0.04344319999217987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,2048,768,8,128,2,128,power_law_1.01,0.04737280011177063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,2048,768,8,128,2,128,power_law_1.01,0.055174398422241214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,2048,768,8,128,2,128,power_law_1.01,0.06436480283737182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,2048,768,8,128,2,128,power_law_1.01,0.07822719812393189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,2048,768,8,128,2,128,power_law_1.01,0.1093951940536499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,2048,768,8,128,2,128,power_law_1.01,0.135315203666687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,2048,768,8,128,2,128,power_law_1.01,0.18606719970703126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,2048,768,8,128,2,128,power_law_1.01,0.24085121154785155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,2048,768,8,128,2,128,power_law_1.01,0.3432255983352661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,2048,768,8,128,2,128,power_law_1.01,0.4746687889099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,8,power_law_1.2,0.033011201024055484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,8,power_law_1.2,0.030451199412345885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,64,balanced,0.02826133370399475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,64,balanced,0.02757866680622101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,64,balanced,0.02589333305756251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,64,balanced,0.027514666318893433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,8,power_law_1.2,0.031167998909950256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,64,balanced,0.031285333136717476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,8,power_law_1.2,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,8,power_law_1.2,0.032051199674606325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,8,power_law_1.2,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,8,power_law_1.2,0.03500159978866577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,8,power_law_1.2,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,8,power_law_1.2,0.036236798763275145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,8,power_law_1.2,0.03654400110244751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,64,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,8,power_law_1.2,0.03919999897480011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,8,power_law_1.2,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,8,power_law_1.2,0.042419201135635375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,8,power_law_1.2,0.04287999868392944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,64,balanced,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,8,power_law_1.2,0.046323201060295104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,64,balanced,0.05789866546789805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,8,power_law_1.2,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,64,balanced,0.06771199901898702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,8,power_law_1.2,0.05143679976463318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,64,balanced,0.06589866677920024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,8,power_law_1.2,0.058316802978515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,64,balanced,0.06419733166694641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,8,power_law_1.2,0.06643199920654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,64,balanced,0.06188266475995382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,64,balanced,0.05527999997138977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,8,power_law_1.2,0.07827200293540955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,64,balanced,0.04790399968624115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,64,balanced,0.0832479993502299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,8,power_law_1.2,0.0933247983455658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,64,balanced,0.08496000369389851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,64,balanced,0.06629866858323415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,8,power_law_1.2,0.12613120079040527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,64,balanced,0.08468266328175862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,64,balanced,0.10354133447011311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,8,power_law_1.2,0.14536319971084594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,2,power_law_1.01,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,2,power_law_1.01,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,2,power_law_1.01,0.06199679970741272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,64,balanced,0.1384106675783793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,64,balanced,0.16528000434239706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,2,power_law_1.01,0.07114880084991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,64,balanced,0.23814932505289713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,2,power_law_1.01,0.08712319731712341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,2,power_law_1.01,0.10549759864807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,2,power_law_1.01,0.11445120573043824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,2,power_law_1.01,0.1284991979598999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,2,power_law_1.01,0.13293440341949464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,4,balanced,0.08131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,64,balanced,0.3034506638844808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,2,power_law_1.01,0.13404159545898436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,64,balanced,0.43879465262095135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,2,power_law_1.01,0.1422976016998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,2,power_law_1.01,0.14625920057296754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,2,power_law_1.01,0.15534720420837403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,2,power_law_1.01,0.16028800010681152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,4,balanced,0.08896533648173015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,4,balanced,0.10129066308339436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,4,balanced,0.1328053375085195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,64,balanced,0.5799466768900553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,8,power_law_1.2,0.21989119052886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,64,balanced,0.8614772955576578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,8,16,power_law_1.01,0.06224640011787415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,64,balanced,1.1401173273722331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,8,16,power_law_1.01,0.07486720085144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,8,power_law_1.2,0.2832191944122314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,8,16,power_law_1.01,0.05653759837150574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,2,power_law_1.01,0.17640960216522217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,8,16,power_law_1.01,0.058534401655197146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,2,power_law_1.01,0.1870911955833435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,4,balanced,0.1833546757698059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,8,16,power_law_1.01,0.05882880091667175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,4,balanced,0.28074665864308673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,8,16,power_law_1.01,0.058316802978515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,8,16,power_law_1.01,0.06100479960441589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,8,16,power_law_1.01,0.06133120059967041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,8,16,power_law_1.01,0.06267520189285278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,8,16,power_law_1.01,0.06258559823036194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,8,16,power_law_1.01,0.06331520080566407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,2,power_law_1.01,0.20818560123443602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,8,16,power_law_1.01,0.06475520133972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,4,balanced,0.28036266565322876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,2,power_law_1.01,0.25401599407196046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,8,16,power_law_1.01,0.06522240042686463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,4,balanced,0.28309865792592365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,8,16,power_law_1.01,0.06702719926834107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,2,power_law_1.01,0.2436608076095581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,8,16,power_law_1.01,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,2,power_law_1.01,0.32216320037841795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,8,16,power_law_1.01,0.07767040133476258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,8,16,power_law_1.01,0.08457599878311158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,2,power_law_1.01,0.37830400466918945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,8,16,power_law_1.01,0.09649919867515563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,2,power_law_1.01,0.4928895950317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.01,0.11022080183029175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.01,0.12877440452575684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,2,power_law_1.01,0.611513614654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,14336,2,8,8,4,power_law_1.2,0.033081600069999696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,4,balanced,0.2844853401184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.01,0.15593600273132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,4,balanced,0.28278932968775433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,2,power_law_1.01,0.8558848381042481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,4,balanced,0.2858293255170186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,4,balanced,0.2911786635716756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,2,power_law_1.01,1.0707776069641113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,2,power_law_1.01,1.5162943840026855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,14336,2,8,8,4,power_law_1.2,0.03357439935207367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,14336,2,8,8,4,power_law_1.2,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.01,0.19411840438842773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,14336,2,8,8,4,power_law_1.2,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,14336,2,8,8,4,power_law_1.2,0.03466239869594574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.01,0.22552320957183838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,14336,2,8,8,4,power_law_1.2,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,14336,2,8,8,4,power_law_1.2,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,14336,2,8,8,4,power_law_1.2,0.0359360009431839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,14336,2,8,8,4,power_law_1.2,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,14336,2,8,8,4,power_law_1.2,0.03754239976406097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,14336,2,8,8,4,power_law_1.2,0.03779839873313904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,14336,2,8,8,4,power_law_1.2,0.03953920006752014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,2,power_law_1.01,1.9331775665283204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,14336,2,8,8,4,power_law_1.2,0.039526399970054624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,4,balanced,0.29178667068481445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.01,0.3073472023010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,4,balanced,0.29596267143885296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,4,balanced,0.302181343237559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.01,0.38903040885925294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,4,balanced,0.3025226593017578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,4,balanced,0.308351993560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,4,balanced,0.3243359923362732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,4,balanced,0.3349119822184245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,4,balanced,0.3684533437093099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,14336,2,8,8,4,power_law_1.2,0.04270080029964447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,14336,2,8,8,4,power_law_1.2,0.050425601005554196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,14336,2,8,8,4,power_law_1.2,0.051718401908874514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.01,0.5372096061706543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,14336,2,8,8,4,power_law_1.2,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,4,balanced,0.3987893263498942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,4,balanced,0.46085333824157715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,4,balanced,0.5272213220596313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,14336,2,8,8,4,power_law_1.2,0.06193280220031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,4,balanced,0.7380533218383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.01,0.710374402999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,4,balanced,0.8622879981994629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.2,0.0654528021812439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.2,0.08150399923324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.2,0.10698879957199096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,8,power_law_1.2,0.40274558067321775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.2,0.12354559898376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.2,0.15483520030975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.2,0.20615038871765137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,8,power_law_1.2,0.4786367893218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.2,0.27915520668029786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.2,0.3654207944869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,2,8,power_law_1.01,0.06092159748077393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,2,8,power_law_1.01,0.07709439992904663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,2,8,power_law_1.01,0.08243200182914734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,2,8,power_law_1.01,0.08732799887657165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,2,8,power_law_1.01,0.08963840007781983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,2,8,power_law_1.01,0.09238399863243103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,2,8,power_law_1.01,0.09709439873695373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,2,8,power_law_1.01,0.09834880232810975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,2,8,power_law_1.01,0.10105600357055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,2,8,power_law_1.01,0.09784960150718688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,2,8,power_law_1.01,0.10158720016479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,2,8,power_law_1.01,0.10076800584793091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,2,8,power_law_1.01,0.10165760517120362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,2,8,power_law_1.01,0.10542720556259155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,2,8,power_law_1.01,0.10729600191116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,2,8,power_law_1.01,0.10905599594116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,2,8,power_law_1.01,0.1128383994102478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,2,8,power_law_1.01,0.12175999879837036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,2,8,power_law_1.01,0.13167359828948974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,2,8,power_law_1.01,0.1548735976219177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,2,8,power_law_1.01,0.1790336012840271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,2,8,power_law_1.01,0.21822080612182618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,2,8,power_law_1.01,0.2629568099975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,2,8,power_law_1.01,0.35368320941925047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,2,8,power_law_1.01,0.4615808010101318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,2,8,power_law_1.01,0.7115520000457763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.2,0.5687808036804199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,2,8,power_law_1.01,0.9131199836730957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1,4096,2688,22,512,4,64,power_law_1.01,0.05416319966316223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2,4096,2688,22,512,4,64,power_law_1.01,0.05390080213546753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4,4096,2688,22,512,4,64,power_law_1.01,0.05219200253486633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8,4096,2688,22,512,4,64,power_law_1.01,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16,4096,2688,22,512,4,64,power_law_1.01,0.05413119792938233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,32,4096,2688,22,512,4,64,power_law_1.01,0.05596799850463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,48,4096,2688,22,512,4,64,power_law_1.01,0.05563520193099976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,64,4096,2688,22,512,4,64,power_law_1.01,0.056454402208328244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,80,4096,2688,22,512,4,64,power_law_1.01,0.056441599130630495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,96,4096,2688,22,512,4,64,power_law_1.01,0.05720319747924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,128,4096,2688,22,512,4,64,power_law_1.01,0.057843202352523805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,160,4096,2688,22,512,4,64,power_law_1.01,0.061875200271606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,192,4096,2688,22,512,4,64,power_law_1.01,0.06309120059013366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,256,4096,2688,22,512,4,64,power_law_1.01,0.0636672019958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,320,4096,2688,22,512,4,64,power_law_1.01,0.07272319793701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,384,4096,2688,22,512,4,64,power_law_1.01,0.07255039811134338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,512,4096,2688,22,512,4,64,power_law_1.01,0.07921919822692872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,768,4096,2688,22,512,4,64,power_law_1.01,0.09540479779243469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1024,4096,2688,22,512,4,64,power_law_1.01,0.1014847993850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,1536,4096,2688,22,512,4,64,power_law_1.01,0.1295359969139099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,2048,4096,2688,22,512,4,64,power_law_1.01,0.15289599895477296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,3072,4096,2688,22,512,4,64,power_law_1.01,0.2075711965560913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,4096,4096,2688,22,512,4,64,power_law_1.01,0.2158911943435669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,6144,4096,2688,22,512,4,64,power_law_1.01,0.2832576036453247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,8192,4096,2688,22,512,4,64,power_law_1.01,0.37177600860595705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,12288,4096,2688,22,512,4,64,power_law_1.01,0.4895167827606201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,fp8,16384,4096,2688,22,512,4,64,power_law_1.01,0.5805759906768799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,1,balanced,0.029461334149042766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,1,balanced,0.03162666658560435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,1,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,1,balanced,0.03745600084463755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,1,balanced,0.037104000647862755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,1,balanced,0.03800000001986822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,1,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,1,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,1,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,1,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,1,balanced,0.042277331153551735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,1,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,1,balanced,0.043525333205858864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,1,balanced,0.05096533397833506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,1,balanced,0.05931200087070465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,1,balanced,0.05900266766548157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,1,balanced,0.06433600187301636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,1,balanced,0.07828799883524577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,1,balanced,0.08214400211970012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,1,balanced,0.10366933544476827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,1,balanced,0.11382399996121724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,1,balanced,0.15727999806404114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,1,balanced,0.16965333620707193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,1,balanced,0.23863999048868814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,1,balanced,0.2997066577275594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,1,balanced,0.4291306734085083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,1,balanced,0.537338654200236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,1,power_law_1.01,0.020960000157356263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,1,power_law_1.01,0.022176000475883483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,1,power_law_1.01,0.02956799864768982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,1,power_law_1.01,0.04488320052623749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,1,power_law_1.01,0.06114559769630432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,1,power_law_1.01,0.08010879755020142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,1,power_law_1.01,0.0945088028907776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,1,power_law_1.01,0.10083839893341065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,1,power_law_1.01,0.10805759429931641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,1,power_law_1.01,0.1081663966178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,1,power_law_1.01,0.11229439973831176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,1,power_law_1.01,0.11526399850845337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,1,power_law_1.01,0.12023040056228637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,1,power_law_1.01,0.12323839664459228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,1,power_law_1.01,0.12775039672851562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,1,power_law_1.01,0.13271039724349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,16,power_law_1.01,0.05011839866638183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,16,power_law_1.01,0.06007680296897888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,1,power_law_1.01,0.13912960290908813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,1,power_law_1.01,0.1602687954902649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,1,power_law_1.01,0.17323520183563232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,1,power_law_1.01,0.208351993560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,1,power_law_1.01,0.24915199279785155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,16,power_law_1.01,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,16,power_law_1.01,0.049004799127578734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,1,power_law_1.01,0.3242111921310425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,16,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,1,power_law_1.01,0.33791360855102537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,1,power_law_1.01,0.45528321266174315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,1,power_law_1.01,0.5620160102844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,1,power_law_1.01,0.7948416233062744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,16,power_law_1.01,0.050547200441360476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,1,power_law_1.01,1.0211711883544923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,16,power_law_1.01,0.05294719934463501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,16,power_law_1.01,0.0532800018787384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,16,power_law_1.01,0.05394560098648071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,16,power_law_1.01,0.05443840026855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,16,power_law_1.01,0.056729602813720706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,16,power_law_1.01,0.05812479853630066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,16,power_law_1.01,0.06101120114326477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,16,power_law_1.01,0.0648959994316101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,16,power_law_1.01,0.07503359913825988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,16,power_law_1.01,0.0738431990146637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,16,power_law_1.01,0.07745919823646545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,16,power_law_1.01,0.0930176019668579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,16,power_law_1.01,0.10161919593811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,16,power_law_1.01,0.12743040323257446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,384,4,16,power_law_1.2,0.08453119993209839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,16,power_law_1.01,0.1476415991783142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,16,power_law_1.01,0.2032831907272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,384,4,16,power_law_1.2,0.09760000109672547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,16,power_law_1.01,0.24746880531311036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,384,4,16,power_law_1.2,0.10744960308074951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,16,power_law_1.01,0.32665600776672366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,384,4,16,power_law_1.2,0.11153919696807861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,384,4,16,power_law_1.2,0.11223679780960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,384,4,16,power_law_1.2,0.11401599645614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,384,4,16,power_law_1.2,0.11036159992218017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,16,power_law_1.01,0.42131838798522947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,384,4,16,power_law_1.2,0.11022720336914063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,16,power_law_1.01,0.6354112148284912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,384,4,16,power_law_1.2,0.10633599758148193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,16,power_law_1.01,0.8040255546569824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,384,4,16,power_law_1.2,0.10887680053710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,384,4,16,power_law_1.2,0.10895999670028686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,384,4,16,power_law_1.2,0.11265920400619507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,384,4,16,power_law_1.2,0.11215360164642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,384,4,16,power_law_1.2,0.11433600187301636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,384,4,16,power_law_1.2,0.12702080011367797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,384,4,16,power_law_1.2,0.1221951961517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,384,4,16,power_law_1.2,0.13292160034179687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,384,4,16,power_law_1.2,0.1487552046775818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,384,4,16,power_law_1.2,0.1674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,384,4,16,power_law_1.2,0.1988927960395813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,384,4,16,power_law_1.2,0.22656641006469727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,384,4,16,power_law_1.2,0.2907520055770874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,384,4,16,power_law_1.2,0.3751487970352173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,384,4,16,power_law_1.2,0.5075263977050781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,384,4,16,power_law_1.2,0.6252927780151367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,384,4,16,power_law_1.2,1.0881728172302245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,384,4,16,power_law_1.2,1.2382399559020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,256,power_law_1.01,0.020326399803161622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,256,power_law_1.01,0.020294399559497835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,256,power_law_1.01,0.021254399418830873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,256,power_law_1.01,0.025862398743629455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,256,power_law_1.01,0.04750080108642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,256,power_law_1.01,0.04229759871959686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,256,power_law_1.01,0.042124798893928526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,256,power_law_1.01,0.042284798622131345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,256,power_law_1.01,0.04245760142803192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,256,power_law_1.01,0.04229120016098022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,256,power_law_1.01,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,256,power_law_1.01,0.04177280068397522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,256,power_law_1.01,0.041536000370979306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,256,power_law_1.01,0.03976320028305054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,256,power_law_1.01,0.041459199786186215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,256,power_law_1.01,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,256,power_law_1.01,0.04433920085430145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,256,power_law_1.01,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,256,power_law_1.01,0.048870399594306946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,256,power_law_1.01,0.060601598024368285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,256,power_law_1.01,0.07592319846153259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,256,power_law_1.01,0.10446720123291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,256,power_law_1.01,0.12015360593795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,256,power_law_1.01,0.1675968050956726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,256,power_law_1.01,0.21064960956573486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,256,power_law_1.01,0.31563520431518555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,256,power_law_1.01,0.401964807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,8,2,balanced,0.048437332113583885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,8,2,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,8,2,balanced,0.05376533170541128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,8,2,balanced,0.07865599791208903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,8,2,balanced,0.10525332887967427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,8,2,balanced,0.1594986617565155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,8,2,balanced,0.16230400403340658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,8,2,balanced,0.15576000014940897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,8,2,balanced,0.15759999553362528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,8,2,balanced,0.1588640014330546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,8,2,balanced,0.16030933459599814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,8,2,balanced,0.164410670598348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,8,2,balanced,0.16897600889205933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,8,2,balanced,0.1730239987373352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,8,2,balanced,0.1826080083847046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,8,2,balanced,0.186191995938619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,8,2,balanced,0.2025973399480184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,8,2,balanced,0.22902933756510416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,8,2,balanced,0.24914133548736572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,8,2,balanced,0.2917813261349996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,8,2,balanced,0.3453813393910726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,8,2,balanced,0.4363360007603963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,8,2,balanced,0.5225973526636759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,8,2,balanced,0.7423413594563802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,8,2,balanced,0.920362631479899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,8,4,power_law_1.2,0.0471231997013092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,8,2,balanced,1.3216959635416667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,8,4,power_law_1.2,0.05007359981536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,8,2,balanced,1.7187093098958333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,8,4,power_law_1.2,0.05648639798164368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,8,4,power_law_1.2,0.06553599834442139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,8,4,power_law_1.2,0.07505919933319091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,8,4,power_law_1.2,0.08364160060882568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,8,4,power_law_1.2,0.08647040128707886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,8,4,power_law_1.2,0.08769279718399048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,8,4,power_law_1.2,0.09174399971961975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,8,4,power_law_1.2,0.09237759709358215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,8,4,power_law_1.2,0.09927039742469787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,8,4,power_law_1.2,0.10107519626617431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,8,4,power_law_1.2,0.10483200550079345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,8,4,power_law_1.2,0.10711040496826171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,8,4,power_law_1.2,0.11464320421218872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,8,4,power_law_1.2,0.11843199729919433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,8,4,power_law_1.2,0.1295359969139099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,8,4,power_law_1.2,0.14800000190734863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,8,4,power_law_1.2,0.17367039918899535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,8,4,power_law_1.2,0.21043200492858888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,8,4,power_law_1.2,0.24431359767913818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,8,4,power_law_1.2,0.3316927909851074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,8,4,power_law_1.2,0.45282559394836425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,8,4,power_law_1.2,0.5824384212493896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,8,4,power_law_1.2,0.7472000122070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,8,4,power_law_1.2,1.0098496437072755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,8,4,power_law_1.2,1.3387328147888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,4,4,balanced,0.06032533446947733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,4,4,balanced,0.06443733473618825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,4,4,balanced,0.0718560020128886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,4,4,balanced,0.10125333070755005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,4,4,balanced,0.13524799545605978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,4,4,balanced,0.20427199204762778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,4,4,balanced,0.20804800589879355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,4,4,balanced,0.19464532534281412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,4,4,balanced,0.18968532482783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,4,4,balanced,0.19103999932607016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,4,4,balanced,0.18765334288279215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,4,4,balanced,0.18954133987426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,4,4,balanced,0.18563199043273926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,4,4,balanced,0.19338667392730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,4,4,balanced,0.19400533040364584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,4,4,balanced,0.19315199057261148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,4,4,balanced,0.20009599129358926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,4,4,balanced,0.2118133306503296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,4,4,balanced,0.22193600734074911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,4,4,balanced,0.24687999486923218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,4,4,balanced,0.2656960090001424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,4,4,balanced,0.31152000029881793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,4,4,balanced,0.36081600189208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,4,4,balanced,0.5459359884262085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,4,4,balanced,0.6334666808446249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,4,4,balanced,0.8907039960225424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,4,4,balanced,1.1489919821421306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,4,power_law_1.2,0.04601599872112274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,4,power_law_1.2,0.04679040014743805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,4,power_law_1.2,0.044627198576927186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,4,power_law_1.2,0.04748800098896026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,4,power_law_1.2,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,4,power_law_1.2,0.05307520031929016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,4,power_law_1.2,0.07439360022544861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,4,power_law_1.2,0.06681600213050842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,4,power_law_1.2,0.08786560297012329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,4,power_law_1.2,0.07948799729347229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,4,power_law_1.2,0.07505919933319091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,4,power_law_1.2,0.13129600286483764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,4,power_law_1.2,0.12444800138473511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,4,power_law_1.2,0.10026240348815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,4,power_law_1.2,0.144652795791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,4,power_law_1.2,0.1248128056526184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,4,power_law_1.2,0.1610368013381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,4,power_law_1.2,0.21048319339752197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.2,0.2723007917404175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.2,0.3874495983123779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.2,0.5192383766174317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.2,0.7451200008392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.2,0.9907456398010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,3072,1536,8,256,1,64,power_law_1.2,0.08739839792251587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.2,1.458137607574463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,3072,1536,8,256,1,64,power_law_1.2,0.059334397315979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,3072,1536,8,256,1,64,power_law_1.2,0.05498239994049072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,3072,1536,8,256,1,64,power_law_1.2,0.06116480231285095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,3072,1536,8,256,1,64,power_law_1.2,0.05502079725265503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,3072,1536,8,256,1,64,power_law_1.2,0.05395200252532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,3072,1536,8,256,1,64,power_law_1.2,0.055289602279663085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,3072,1536,8,256,1,64,power_law_1.2,0.059571200609207155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,3072,1536,8,256,1,64,power_law_1.2,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,16,2,power_law_1.2,0.04921599924564361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,3072,1536,8,256,1,64,power_law_1.2,0.059539198875427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,16,2,power_law_1.2,0.05167359709739685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,16,2,power_law_1.2,0.06446719765663148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,16,2,power_law_1.2,0.07249280214309692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,16,2,power_law_1.2,0.08621439933776856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,16,2,power_law_1.2,0.11232000589370728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,16,2,power_law_1.2,0.11804159879684448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,16,2,power_law_1.2,0.12383999824523925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,3072,1536,8,256,1,64,power_law_1.2,0.06506879925727845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,16,2,power_law_1.2,0.12502399682998658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,3072,1536,8,256,1,64,power_law_1.2,0.06657919883728028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,16,2,power_law_1.2,0.12910079956054688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,3072,1536,8,256,1,64,power_law_1.2,0.06780800223350525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,16,2,power_law_1.2,0.13427200317382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,3072,1536,8,256,1,64,power_law_1.2,0.07210239768028259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,16,2,power_law_1.2,0.13852159976959227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,16,2,power_law_1.2,0.1415168046951294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,3072,1536,8,256,1,64,power_law_1.2,0.07912319898605347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,16,2,power_law_1.2,0.15210239887237548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,3072,1536,8,256,1,64,power_law_1.2,0.08166400194168091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,16,2,power_law_1.2,0.15955840349197387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,16,2,power_law_1.2,0.16476800441741943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,16,2,power_law_1.2,0.17592320442199708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,16,2,power_law_1.2,0.2028287887573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,16,2,power_law_1.2,0.22028160095214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,16,2,power_law_1.2,0.28596479892730714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,16,2,power_law_1.2,0.3482687950134277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,16,2,power_law_1.2,0.42529921531677245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,16,2,power_law_1.2,0.5333439826965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,3072,1536,8,256,1,64,power_law_1.2,0.07828480005264282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,16,2,power_law_1.2,0.7107776165008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,3072,1536,8,256,1,64,power_law_1.2,0.09141119718551635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,3072,1536,8,256,1,64,power_law_1.2,0.09345920085906982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,16,2,power_law_1.2,0.9956095695495606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,3072,1536,8,256,1,64,power_law_1.2,0.11032960414886475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,3072,1536,8,256,1,64,power_law_1.2,0.12060799598693847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,3072,1536,8,256,1,64,power_law_1.2,0.15255039930343628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,3072,1536,8,256,1,64,power_law_1.2,0.17996159791946412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,3072,1536,8,256,1,64,power_law_1.2,0.25315840244293214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,3072,1536,8,256,1,64,power_law_1.2,0.2917248010635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,3072,1536,8,256,1,64,power_law_1.2,0.41726078987121584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,3072,1536,8,256,1,64,power_law_1.2,0.541868782043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,16384,2,8,4,8,power_law_1.2,0.03865599930286408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,16384,2,8,4,8,power_law_1.2,0.037062400579452516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,16384,2,8,4,8,power_law_1.2,0.03729279935359955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,16384,2,8,4,8,power_law_1.2,0.03997440040111542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,16384,2,8,4,8,power_law_1.2,0.040166398882865904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,16384,2,8,4,8,power_law_1.2,0.040908798575401306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,16384,2,8,4,8,power_law_1.2,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,16384,2,8,4,8,power_law_1.2,0.04416640102863312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,1,2,balanced,0.045781334241231285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,1,2,balanced,0.05858133236567179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,1,2,balanced,0.08245866497357686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,1,2,balanced,0.12618666887283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,1,2,balanced,0.22114133834838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,1,2,balanced,0.4038933515548706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,1,2,balanced,0.4138880173365275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,1,2,balanced,0.42548267046610516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,1,2,balanced,0.436901330947876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,1,2,balanced,0.444976011912028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,1,2,balanced,0.4673333168029785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,1,2,balanced,0.4459520181020101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,1,2,balanced,0.4543573458989461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,1,2,balanced,0.47421332200368244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,1,2,balanced,0.5007253487904867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,1,2,balanced,0.4979786475499471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,16384,2,8,4,8,power_law_1.2,0.04883840084075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,1,2,balanced,0.5041280190149943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,16,2,power_law_1.2,1.3771136283874512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,16384,2,8,4,8,power_law_1.2,0.0491456001996994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,1,2,balanced,0.6908426284790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,16384,2,8,4,8,power_law_1.2,0.050483202934265135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,1,2,balanced,0.6531466643015543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,1,2,balanced,1.1872159639994304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,1,2,balanced,1.2923893133799236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,1,2,balanced,1.0611733595530193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,1,2,balanced,1.0936960379282634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,1,2,balanced,1.105631987253825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,1,2,balanced,1.1752533117930095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,1,2,balanced,2.1186134020487466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,16384,2,8,4,8,power_law_1.2,0.0606719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.2,1.918342399597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,1,2,balanced,2.1572799682617188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,16384,2,8,4,8,power_law_1.2,0.06149759888648987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,16384,2,8,4,8,power_law_1.2,0.06993920207023621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,16384,2,8,4,8,power_law_1.2,0.08674560189247131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,16384,2,8,4,8,power_law_1.2,0.08030080199241638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,16384,2,8,4,8,power_law_1.2,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.2,2.874720001220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,16384,2,8,4,8,power_law_1.2,0.13477760553359985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,16384,2,8,4,8,power_law_1.2,0.149619197845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,16384,2,8,4,8,power_law_1.2,0.18664319515228273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,16384,2,8,4,8,power_law_1.2,0.26666879653930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,16,2,power_law_1.2,1.643270492553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,16384,2,8,4,8,power_law_1.2,0.35900800228118895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,16384,2,8,4,8,power_law_1.2,0.5162623882293701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2560,8,160,1,16,balanced,0.045994664231936135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2560,8,160,1,16,balanced,0.04595200220743815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2560,8,160,1,16,balanced,0.049839998284975685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2560,8,160,1,16,balanced,0.07830933233102162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2560,8,160,1,16,balanced,0.10487467050552368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2560,8,160,1,16,balanced,0.11912000179290771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2560,8,160,1,16,balanced,0.11950400471687317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2560,8,160,1,16,balanced,0.11932800213495891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2560,8,160,1,16,balanced,0.12080533305803935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2560,8,160,1,16,balanced,0.12155200044314067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2560,8,160,1,16,balanced,0.12230400244394939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2560,8,160,1,16,balanced,0.12468799948692322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2560,8,160,1,16,balanced,0.12785067160924277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2560,8,160,1,16,balanced,0.12916266918182373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2560,8,160,1,16,balanced,0.13427733381589255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2560,8,160,1,16,balanced,0.1360640029112498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2560,8,160,1,16,balanced,0.13826133807500204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2560,8,160,1,16,balanced,0.1486026644706726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2560,8,160,1,16,balanced,0.15779200196266174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2560,8,160,1,16,balanced,0.20249066750208536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2560,8,160,1,16,balanced,0.18773333231608072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2560,8,160,1,16,balanced,0.2432159980138143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2560,8,160,1,16,balanced,0.25990400711695355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2560,8,160,1,16,balanced,0.3744800090789795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2560,8,160,1,16,balanced,0.4310826857884725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2560,8,160,1,16,balanced,0.6486239830652872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2560,8,160,1,16,balanced,0.750495990117391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,16384,2,8,4,8,power_law_1.2,0.7674816131591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,2688,1856,6,128,2,1,power_law_1.01,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,2688,1856,6,128,2,1,power_law_1.01,0.04853120148181915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,2688,1856,6,128,2,1,power_law_1.01,0.06327679753303528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,2688,1856,6,128,2,1,power_law_1.01,0.0938431978225708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,2688,1856,6,128,2,1,power_law_1.01,0.11716480255126953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,2688,1856,6,128,2,1,power_law_1.01,0.17588479518890382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,2688,1856,6,128,2,1,power_law_1.01,0.18481279611587526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,2688,1856,6,128,2,1,power_law_1.01,0.18723200559616088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,2688,1856,6,128,2,1,power_law_1.01,0.19619200229644776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,2688,1856,6,128,2,1,power_law_1.01,0.19467519521713256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.2,3.8036094665527345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,2688,1856,6,128,2,1,power_law_1.01,0.20734078884124757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,2688,1856,6,128,2,1,power_law_1.01,0.20778241157531738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,2688,1856,6,128,2,1,power_law_1.01,0.20623359680175782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,2688,1856,6,128,2,1,power_law_1.01,0.2160576105117798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,2688,1856,6,128,2,1,power_law_1.01,0.22284159660339356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,2688,1856,6,128,2,1,power_law_1.01,0.23054718971252441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,2688,1856,6,128,2,1,power_law_1.01,0.2395263910293579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,2688,1856,6,128,2,1,power_law_1.01,0.27204480171203616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,2688,1856,6,128,2,1,power_law_1.01,0.2720767974853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,2688,1856,6,128,2,1,power_law_1.01,0.31843841075897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,2688,1856,6,128,2,1,power_law_1.01,0.34615039825439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,2688,1856,6,128,2,1,power_law_1.01,0.4163072109222412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,2688,1856,6,128,2,1,power_law_1.01,0.4559807777404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,2688,1856,6,128,2,1,power_law_1.01,0.5812416076660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,2688,1856,6,128,2,1,power_law_1.01,0.7662271976470947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,2688,1856,6,128,2,1,power_law_1.01,1.0234496116638183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,384,32,1,power_law_1.2,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,384,32,1,power_law_1.2,0.062054401636123656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,384,32,1,power_law_1.2,0.06675840020179749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,384,32,1,power_law_1.2,0.08706560134887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,384,32,1,power_law_1.2,0.11556479930877686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,384,32,1,power_law_1.2,0.13692159652709962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,2688,1856,6,128,2,1,power_law_1.01,1.2096896171569824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,384,32,1,power_law_1.2,0.15607680082321168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,384,32,1,power_law_1.2,0.18906240463256835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,384,32,1,power_law_1.2,0.19716479778289794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,16384,2,8,4,8,power_law_1.2,1.0014847755432128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,384,32,1,power_law_1.2,0.20144639015197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,384,32,1,power_law_1.2,0.21147520542144777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,384,32,1,power_law_1.2,0.22960638999938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,384,32,1,power_law_1.2,0.24165759086608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,384,32,1,power_law_1.2,0.25897600650787356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,384,32,1,power_law_1.2,0.2797631978988647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,384,32,1,power_law_1.2,0.29637761116027833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,384,32,1,power_law_1.2,0.32797439098358155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,384,32,1,power_law_1.2,0.4136191844940186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,384,32,1,power_law_1.2,0.4073919773101807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,384,32,1,power_law_1.2,0.5084735870361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,384,32,1,power_law_1.2,0.5920447826385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,384,32,1,power_law_1.2,0.7806528091430665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,384,32,1,power_law_1.2,0.9890175819396972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,384,32,1,power_law_1.2,1.3450559616088866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,384,32,1,power_law_1.2,1.8386880874633789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,384,32,1,power_law_1.2,2.531078338623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,384,32,1,power_law_1.2,3.105190467834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,16384,2,8,4,8,power_law_1.2,1.4946111679077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,16384,2,8,4,8,power_law_1.2,1.8095935821533202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,8,power_law_1.2,0.017843200266361235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,8,power_law_1.2,0.01793919950723648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,8,power_law_1.2,0.018118399381637573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,8,power_law_1.2,0.018169599771499633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,8,power_law_1.2,0.018412800133228303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,8,power_law_1.2,0.01889919936656952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,8,power_law_1.2,0.02222079932689667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,8,power_law_1.2,0.020985600352287293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,8,power_law_1.2,0.02271360009908676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,8,power_law_1.2,0.024396799504756927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,8,power_law_1.2,0.023104000091552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,8,power_law_1.2,0.025094398856163026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,8,power_law_1.2,0.02598400115966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,8,power_law_1.2,0.0251008003950119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,8,power_law_1.2,0.026476800441741943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,8,power_law_1.2,0.027699199318885804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,8,power_law_1.2,0.040115201473236085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,8,power_law_1.2,0.04596480131149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,8,power_law_1.2,0.04967679977416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,8,power_law_1.2,0.05934720039367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,8,power_law_1.2,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,8,power_law_1.2,0.06432639956474304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,2,balanced,0.03309866786003113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,2,balanced,0.033728001018365227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,2,balanced,0.050714666644732155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,2,balanced,0.05384000142415365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,2,balanced,0.05403199791908264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,2,balanced,0.05584533512592316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,2,balanced,0.05659199754397074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,2,balanced,0.05533333122730255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,2,balanced,0.05685866872469584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,2,balanced,0.05649599929650625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,2,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,2,balanced,0.058176000912984215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,2,balanced,0.058431997895240784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,8,power_law_1.2,0.07801600098609925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,2,balanced,0.06195199986298879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,2,balanced,0.0683786670366923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,2,balanced,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,2,balanced,0.0736053337653478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,2,balanced,0.08074133098125458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,2,balanced,0.08667199810345967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,2,balanced,0.10326400399208069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,2,balanced,0.12148267030715942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,2,balanced,0.1567466656366984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,2,balanced,0.19401599963506064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,2,balanced,0.2546186645825704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,8,power_law_1.2,0.10169600248336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,2,balanced,0.30612800518671673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,2,balanced,0.42954134941101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,2,balanced,0.5280266602834066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,8,power_law_1.2,0.12338559627532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,32,balanced,0.03579200059175491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,32,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,32,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,32,balanced,0.04011733333269755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,32,balanced,0.04764266808827718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,32,balanced,0.04632000128428141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,32,balanced,0.04574400186538696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,32,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,32,balanced,0.047925333182017006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,32,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,32,balanced,0.048010667165120445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,32,balanced,0.04736533264319102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,32,balanced,0.04971200227737427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,32,balanced,0.04965866605440775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,32,balanced,0.05197333296140035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,32,balanced,0.0537120004494985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,32,balanced,0.053914666175842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,32,balanced,0.055306668082873024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,32,balanced,0.056143999099731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,32,balanced,0.0620959997177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,32,balanced,0.06619733572006226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,32,balanced,0.0848533312479655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,32,balanced,0.09527466694513957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,32,balanced,0.12069867054621379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,32,balanced,0.1434719959894816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,32,balanced,0.19081066052118936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,32,balanced,0.2402133345603943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,8,power_law_1.2,0.17148799896240235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,8,power_law_1.2,0.22001919746398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,2,64,balanced,0.0622026671965917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,2,64,balanced,0.045791998505592346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,2,64,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,2,64,balanced,0.050069332122802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,2,64,balanced,0.05170666674772898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,2,64,balanced,0.07439466814200084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,2,64,balanced,0.09268266956011455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,2,64,balanced,0.08877866466840108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,2,64,balanced,0.08829333384831746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,2,64,balanced,0.08886933326721191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,2,64,balanced,0.09075199564297994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,2,64,balanced,0.09134399890899658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,2,64,balanced,0.09273067116737366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,2,64,balanced,0.09572266538937886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,2,64,balanced,0.09949333469072978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,2,64,balanced,0.10088533163070679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,2,64,balanced,0.10496000448862712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,2,64,balanced,0.11115733782450359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,2,64,balanced,0.11551466584205627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,2,64,balanced,0.12773332993189493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,2,64,balanced,0.13944533467292786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,2,64,balanced,0.16457600394884744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,2,64,balanced,0.18954133987426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,2,64,balanced,0.2364693284034729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,2,64,balanced,0.28994133075078327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,2,64,balanced,0.38384532928466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,2,64,balanced,0.48475201924641925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,2,4,power_law_1.01,0.04933120012283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,2,4,power_law_1.01,0.06039680242538452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,2,4,power_law_1.01,0.06535680294036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,2,4,power_law_1.01,0.078847998380661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,2,4,power_law_1.01,0.0899071991443634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,2,4,power_law_1.01,0.09924479722976684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,2,4,power_law_1.01,0.10727039575576783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,2,4,power_law_1.01,0.10503040552139283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,2,4,power_law_1.01,0.10546560287475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,3072,1536,8,256,1,16,power_law_1.01,0.06643199920654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,3072,1536,8,256,1,16,power_law_1.01,0.08769279718399048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,3072,1536,8,256,1,16,power_law_1.01,0.08567039966583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,3072,1536,8,256,1,16,power_law_1.01,0.08821120262145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,3072,1536,8,256,1,16,power_law_1.01,0.08698239922523499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,2,4,power_law_1.01,0.10790400505065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,3072,1536,8,256,1,16,power_law_1.01,0.08896639943122864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,2,4,power_law_1.01,0.11696640253067017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,3072,1536,8,256,1,16,power_law_1.01,0.0913919985294342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,2,4,power_law_1.01,0.11767040491104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,2,4,power_law_1.01,0.12003840208053589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,3072,1536,8,256,1,16,power_law_1.01,0.09304320216178893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,2,4,power_law_1.01,0.12302720546722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,3072,1536,8,256,1,16,power_law_1.01,0.09232640266418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,2,4,power_law_1.01,0.12992639541625978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,2,4,power_law_1.01,0.13235199451446533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,2,4,power_law_1.01,0.13946239948272704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,2,4,power_law_1.01,0.16313600540161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,2,4,power_law_1.01,0.16874239444732667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,2,4,power_law_1.01,0.20034561157226563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,3072,1536,8,256,1,16,power_law_1.01,0.09178239703178406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,2,4,power_law_1.01,0.221612811088562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,3072,1536,8,256,1,16,power_law_1.01,0.09340159893035889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,3072,1536,8,256,1,16,power_law_1.01,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,3072,1536,8,256,1,16,power_law_1.01,0.09512320160865784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,3072,1536,8,256,1,16,power_law_1.01,0.09708160161972046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,3072,1536,8,256,1,16,power_law_1.01,0.10049920082092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,3072,1536,8,256,1,16,power_law_1.01,0.10174720287322998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,2,4,power_law_1.01,0.27752320766448973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,3072,1536,8,256,1,16,power_law_1.01,0.10409599542617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,2,4,power_law_1.01,0.32829439640045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,2,4,power_law_1.01,0.44235520362854003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,2,4,power_law_1.01,0.593555212020874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,3072,1536,8,256,1,16,power_law_1.01,0.11470719575881957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,2,4,power_law_1.01,0.8117888450622559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,3072,1536,8,256,1,16,power_law_1.01,0.1258239984512329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,3072,1536,8,256,1,16,power_law_1.01,0.14905600547790526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,2,4,power_law_1.01,1.049177646636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,3072,1536,8,256,1,16,power_law_1.01,0.17200640439987183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,3072,1536,8,256,1,16,power_law_1.01,0.20856959819793702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,3072,1536,8,256,1,16,power_law_1.01,0.2462143898010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,3072,1536,8,256,1,16,power_law_1.01,0.3392640113830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,3072,1536,8,256,1,16,power_law_1.01,0.4276224136352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,3072,1536,8,256,1,16,power_law_1.01,0.5937471866607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,3072,1536,8,256,1,16,power_law_1.01,0.8056384086608886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,4,power_law_1.2,0.09598079919815064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,4,power_law_1.2,0.14298880100250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,4,power_law_1.2,0.16714240312576295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,4,power_law_1.2,0.21183359622955322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,4,power_law_1.2,0.23048319816589355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,4,power_law_1.2,0.3118655920028687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,4,power_law_1.2,0.3400896072387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,4,power_law_1.2,0.37772159576416015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,4,power_law_1.2,0.4041344165802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,4,power_law_1.2,0.38933119773864744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,4,power_law_1.2,0.38852479457855227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,4,power_law_1.2,0.4068416118621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,4,power_law_1.2,0.3872191905975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,4,power_law_1.2,0.41313920021057127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,4,power_law_1.2,0.43174400329589846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,4,power_law_1.2,0.45873279571533204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,4,power_law_1.2,0.4849599838256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,4,power_law_1.2,0.5537536144256592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,4,power_law_1.2,0.5682240009307862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,4,power_law_1.2,0.6624959945678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,4,power_law_1.2,0.6534592151641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,4,power_law_1.2,0.7848639965057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,4,power_law_1.2,0.925164794921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,4,power_law_1.2,1.239129638671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,4,power_law_1.2,1.4506879806518556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,4,power_law_1.2,1.9922943115234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,4,power_law_1.2,2.760915184020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,8,power_law_1.01,0.032576000690460204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,8,power_law_1.01,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,8,power_law_1.01,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,8,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,8,power_law_1.01,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,8,power_law_1.01,0.04012799859046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,8,power_law_1.01,0.04776319861412048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,8,power_law_1.01,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,8,power_law_1.01,0.04824959933757782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,8,power_law_1.01,0.04938240051269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,8,power_law_1.01,0.04940159916877747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,8,power_law_1.01,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,8,power_law_1.01,0.056364798545837404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,8,power_law_1.01,0.06033279895782471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,8,power_law_1.01,0.06704000234603882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,8,power_law_1.01,0.07527679800987244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,8,power_law_1.01,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,8,power_law_1.01,0.1047551989555359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,8,power_law_1.01,0.10962560176849365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,8,power_law_1.01,0.14824960231781006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,8,power_law_1.01,0.18353279829025268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,8,power_law_1.01,0.24053120613098145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,2,8,power_law_1.01,0.06616960167884826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,2,8,power_law_1.01,0.08901119828224183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,2,8,power_law_1.01,0.0927295982837677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,8,power_law_1.01,0.31004159450531005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,2,8,power_law_1.01,0.11555839776992798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,2,8,power_law_1.01,0.12041599750518799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,2,8,power_law_1.01,0.11781120300292969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,2,8,power_law_1.01,0.1284991979598999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,2,8,power_law_1.01,0.12817280292510985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,2,8,power_law_1.01,0.1307136058807373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,2,8,power_law_1.01,0.12865279912948607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,2,8,power_law_1.01,0.12819199562072753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,2,8,power_law_1.01,0.13419519662857055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,2,8,power_law_1.01,0.1372032046318054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,2,8,power_law_1.01,0.14195200204849243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,2,8,power_law_1.01,0.14851839542388917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,2,8,power_law_1.01,0.14746880531311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,2,8,power_law_1.01,0.16060800552368165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,2,8,power_law_1.01,0.18347519636154175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,2,8,power_law_1.01,0.19112319946289064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,2,8,power_law_1.01,0.23386878967285157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,2,8,power_law_1.01,0.2598144054412842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,2,8,power_law_1.01,0.32293119430541994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,2,8,power_law_1.01,0.3772480010986328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,2,8,power_law_1.01,0.5176256179809571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,8,power_law_1.01,0.4487423896789551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,2,8,power_law_1.01,0.6382271766662597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,2,8,power_law_1.01,0.8696703910827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,8,power_law_1.01,0.4741312026977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,2,8,power_law_1.01,1.1819328308105468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,16,power_law_1.01,0.05281919836997986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,16,power_law_1.01,0.07141119837760926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,16,power_law_1.01,0.06926079988479614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,16,power_law_1.01,0.0726144015789032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,16,power_law_1.01,0.0744704008102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,16,power_law_1.01,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,16,power_law_1.01,0.0742143988609314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,16,power_law_1.01,0.07388799786567687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,16,power_law_1.01,0.07461760044097901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,16,power_law_1.01,0.07715200185775757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,16,power_law_1.01,0.07677440047264099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,16,power_law_1.01,0.07799040079116822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,16,power_law_1.01,0.08321279883384705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,16,power_law_1.01,0.08699520230293274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,16,power_law_1.01,0.09095680117607116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,16,power_law_1.01,0.10088959932327271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,16,power_law_1.01,0.10620160102844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,16,power_law_1.01,0.11987199783325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,16,power_law_1.01,0.1295359969139099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,16,power_law_1.01,0.15838719606399537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,16,power_law_1.01,0.17797759771347046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,16,power_law_1.01,0.22547199726104736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,16,power_law_1.01,0.28244481086730955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,16,power_law_1.01,0.38355839252471924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,16,4,power_law_1.2,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,16,power_law_1.01,0.5009600162506104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,16,4,power_law_1.2,0.04963200092315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,16,power_law_1.01,0.6885759830474854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,16,4,power_law_1.2,0.05101439952850342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,16,4,power_law_1.2,0.05966079831123352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,16,power_law_1.01,0.9459648132324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,16,4,power_law_1.2,0.06360960006713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,16,4,power_law_1.2,0.06917120218276977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,16,4,power_law_1.2,0.08449280261993408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,16,4,power_law_1.2,0.08451840281486511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,16,4,power_law_1.2,0.08387200236320495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,16,4,power_law_1.2,0.0881600022315979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,16,4,power_law_1.2,0.09089279770851136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,16,4,power_law_1.2,0.09312639832496643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,16,4,power_law_1.2,0.10206719636917114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,16,4,power_law_1.2,0.10472320318222046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,16,4,power_law_1.2,0.11422079801559448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,16,4,power_law_1.2,0.11939840316772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,16,4,power_law_1.2,0.12475520372390747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,16,4,power_law_1.2,0.14438400268554688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,16,4,power_law_1.2,0.15889279842376708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,16,4,power_law_1.2,0.1947648048400879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,16,4,power_law_1.2,0.22385919094085693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,16,4,power_law_1.2,0.2935296058654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,16,4,power_law_1.2,0.37496318817138674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,16,4,power_law_1.2,0.5107840061187744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,16,4,power_law_1.2,0.6644608020782471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,16,4,power_law_1.2,0.9927424430847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,16,4,power_law_1.2,1.2469696044921874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,8,power_law_1.01,0.7889152050018311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,8,power_law_1.01,0.9303487777709961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,1,8,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,1,8,balanced,0.07042133311430614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,1,8,balanced,0.093941330909729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,1,8,balanced,0.14949867129325867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,1,8,balanced,0.24648000796635947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,1,8,balanced,0.4465493361155192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,1,8,balanced,0.6471999883651733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,1,8,balanced,0.6430399815241495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,1,8,balanced,0.6415093342463175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,1,8,balanced,0.6478773355484009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,1,8,balanced,0.6454559961954752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,1,8,balanced,0.6496373414993286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,1,8,balanced,0.6497600078582764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,1,8,balanced,0.650218685468038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,1,8,balanced,0.6565706729888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,1,8,balanced,0.6643733183542887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,1,8,balanced,0.6675946712493896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,1,8,balanced,0.687274694442749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,1,8,balanced,0.6884799798329672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,1,8,balanced,0.7132053375244141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,1,8,balanced,0.7402079900105795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,1,8,balanced,0.793071985244751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,1,8,balanced,0.840778668721517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,1,8,balanced,0.9131733576456705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,1,8,balanced,1.0591413180033367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,1,8,balanced,1.217146635055542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,1,8,balanced,1.654863993326823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,32,power_law_1.2,0.07773439884185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,32,power_law_1.2,0.07402880191802978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,32,power_law_1.2,0.07246720194816589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,32,power_law_1.2,0.07769600152969361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,32,power_law_1.2,0.07335039973258972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,32,power_law_1.2,0.07417600154876709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,32,power_law_1.2,0.07891200184822082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,32,power_law_1.2,0.08041599988937378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,32,power_law_1.2,0.07978879809379577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,32,power_law_1.2,0.08047360181808472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,32,power_law_1.2,0.08201599717140198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,32,power_law_1.2,0.08570880293846131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,32,power_law_1.2,0.0832319974899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,32,power_law_1.2,0.08723840117454529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,32,power_law_1.2,0.09853439927101135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,32,power_law_1.2,0.09994879961013795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,32,power_law_1.2,0.10105600357055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,32,power_law_1.2,0.11251200437545776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,32,power_law_1.2,0.12662400007247926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,32,power_law_1.2,0.14966399669647218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,32,power_law_1.2,0.1817023992538452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,32,power_law_1.2,0.22686080932617186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,4,32,power_law_1.2,0.031999999284744264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,32,power_law_1.2,0.27790079116821287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,4,32,power_law_1.2,0.031007999181747438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,4,32,power_law_1.2,0.030687999725341798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,4,32,power_law_1.2,0.03338240087032318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,4,32,power_law_1.2,0.033670398592948916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,4,32,power_law_1.2,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,4,32,power_law_1.2,0.035097599029541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,4,32,power_law_1.2,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,32,power_law_1.2,0.40478081703186036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,32,power_law_1.2,0.5614719867706299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,2,32,power_law_1.2,0.8223679542541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,4,32,power_law_1.2,0.03639039993286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,2,32,power_law_1.2,1.1399423599243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,4,32,power_law_1.2,0.03633280098438263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,4,32,power_law_1.2,0.03758719861507416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,4,32,power_law_1.2,0.03856639862060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,4,32,power_law_1.2,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,4,32,power_law_1.2,0.04056319892406464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,4,32,power_law_1.2,0.04560000002384186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,4,32,power_law_1.2,0.04766719937324524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,4,32,power_law_1.2,0.054995197057724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,4,32,power_law_1.2,0.05868160128593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,4,32,power_law_1.2,0.0732479989528656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,4,32,power_law_1.2,0.08917760252952575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,4,32,power_law_1.2,0.11729919910430908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,4,32,power_law_1.2,0.14727040529251098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,4,32,power_law_1.2,0.18906240463256835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,4,32,power_law_1.2,0.29776639938354493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,4,32,power_law_1.2,0.3701440095901489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,4,8,power_law_1.01,0.05025920271873474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,4,8,power_law_1.01,0.06997759938240052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,4,8,power_law_1.01,0.08033279776573181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,4,8,power_law_1.01,0.08757759928703308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,4,8,power_law_1.01,0.09405440092086792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,4,8,power_law_1.01,0.08419839739799499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,4,8,power_law_1.01,0.1000704050064087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,4,8,power_law_1.01,0.09735680222511292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,4,8,power_law_1.01,0.10360959768295289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,4,8,power_law_1.01,0.10355839729309083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,4,8,power_law_1.01,0.10561280250549317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,4,8,power_law_1.01,0.10224640369415283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,4,8,power_law_1.01,0.11112960577011108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,4,8,power_law_1.01,0.11514879465103149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,4,8,power_law_1.01,0.12116479873657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,4,8,power_law_1.01,0.1245311975479126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,4,8,power_law_1.01,0.13477760553359985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,4,8,power_law_1.01,0.1515071988105774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,4,8,power_law_1.01,0.16510080099105834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,4,8,power_law_1.01,0.1991871953010559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,4,8,power_law_1.01,0.22740480899810792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,4,8,power_law_1.01,0.2976191997528076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,4,8,power_law_1.01,0.3578815937042236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,4,8,power_law_1.01,0.48416638374328613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,4,8,power_law_1.01,0.595692777633667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,4,8,power_law_1.01,0.8297727584838868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,4,8,power_law_1.01,1.1227904319763184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,4,32,power_law_1.2,0.5909952163696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,384,4,1,power_law_1.2,0.07831040024757385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,384,4,1,power_law_1.2,0.10398080348968505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,384,4,1,power_law_1.2,0.1472000002861023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,384,4,1,power_law_1.2,0.24800000190734864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,384,4,1,power_law_1.2,0.3716799974441528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,384,4,1,power_law_1.2,0.4993408203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,2,32,power_law_1.01,0.04583680033683777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,2,32,power_law_1.01,0.04547199904918671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,2,32,power_law_1.01,0.04394879937171936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,2,32,power_law_1.01,0.04480000138282776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,2,32,power_law_1.01,0.042752000689506534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,2,32,power_law_1.01,0.04620800018310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,2,32,power_law_1.01,0.0455808013677597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,2,32,power_law_1.01,0.04628480076789856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,384,4,1,power_law_1.2,0.593779182434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,2,32,power_law_1.01,0.045900800824165346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,2,32,power_law_1.01,0.04715520143508911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,2,32,power_law_1.01,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,384,4,1,power_law_1.2,0.8118528366088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,2,32,power_law_1.01,0.04772480130195618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,2,32,power_law_1.01,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,384,4,1,power_law_1.2,0.8467776298522949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,2,32,power_law_1.01,0.04888960123062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,2,32,power_law_1.01,0.0516543984413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,2,32,power_law_1.01,0.05258880257606506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,2,32,power_law_1.01,0.05504639744758606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,2,32,power_law_1.01,0.06109439730644226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.01,0.06668800115585327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.01,0.07882239818572997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.01,0.09017599821090698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,384,4,1,power_law_1.2,0.8884032249450684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.01,0.11191040277481079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.01,0.12844159603118896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,384,4,1,power_law_1.2,0.9296640396118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.01,0.1835263967514038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.01,0.22176640033721923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,384,4,1,power_law_1.2,0.9673151969909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.01,0.3148927927017212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,384,4,1,power_law_1.2,0.9844287872314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,384,4,1,power_law_1.2,1.0423487663269042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,384,4,1,power_law_1.2,1.0679807662963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.01,0.46892800331115725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,384,4,1,power_law_1.2,1.0921600341796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,384,4,1,power_law_1.2,1.1676287651062012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,384,4,1,power_law_1.2,1.2522560119628907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,4,32,power_law_1.2,0.844115161895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,384,4,1,power_law_1.2,1.3521856307983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,384,4,1,power_law_1.2,1.5606080055236817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,384,4,1,power_law_1.2,1.5936960220336913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,384,4,1,power_law_1.2,1.9572479248046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,384,4,1,power_law_1.2,2.187513542175293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,384,4,1,power_law_1.2,2.780166435241699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,384,4,1,power_law_1.2,2.9998783111572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,4,balanced,0.07057600220044453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2560,8,160,4,4,balanced,0.07460799813270569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2560,8,160,4,4,balanced,0.08659199873606364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2560,8,160,4,4,balanced,0.10542399684588115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2560,8,160,4,4,balanced,0.14089066783587137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,384,4,1,power_law_1.2,3.8436862945556642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,384,4,1,power_law_1.2,5.176812744140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2560,8,160,4,4,balanced,0.15405866503715515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2560,8,160,4,4,balanced,0.15424000223477682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2560,8,160,4,4,balanced,0.15296533703804016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2560,8,160,4,4,balanced,0.1557973325252533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2560,8,160,4,4,balanced,0.1543839971224467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2560,8,160,4,4,balanced,0.1572373310724894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2560,8,160,4,4,balanced,0.1612160007158915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2560,8,160,4,4,balanced,0.16198399662971497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2560,8,160,4,4,balanced,0.16602133711179098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2560,8,160,4,4,balanced,0.17227200667063394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2560,8,160,4,4,balanced,0.17303999265034994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2560,8,160,4,4,balanced,0.17735467354456583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2560,8,160,4,4,balanced,0.1906720002492269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2560,8,160,4,4,balanced,0.20018666982650757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2560,8,160,4,4,balanced,0.22957332928975424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2560,8,160,4,4,balanced,0.257749338944753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2560,8,160,4,4,balanced,0.360426664352417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2560,8,160,4,4,balanced,0.4228266477584839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2560,8,160,4,4,balanced,0.574399987856547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2560,8,160,4,4,balanced,0.7536640167236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2560,8,160,4,4,balanced,1.0349866549173992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2560,8,160,4,4,balanced,1.3686134020487468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2048,8,256,4,1,power_law_1.01,0.06888960003852844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2048,8,256,4,1,power_law_1.01,0.0917248010635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2048,8,256,4,1,power_law_1.01,0.13065600395202637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2048,8,256,4,1,power_law_1.01,0.2167680025100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2048,8,256,4,1,power_law_1.01,0.3035007953643799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2048,8,256,4,1,power_law_1.01,0.3996031999588013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2048,8,256,4,1,power_law_1.01,0.5431168079376221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2048,8,256,4,1,power_law_1.01,0.5631743907928467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2048,8,256,4,1,power_law_1.01,0.5910912036895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2048,8,256,4,1,power_law_1.01,0.6028416156768799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2048,8,256,4,1,power_law_1.01,0.6269760131835938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2048,8,256,4,1,power_law_1.01,0.6287551879882812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2048,8,256,4,1,power_law_1.01,0.6486783981323242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2048,8,256,4,1,power_law_1.01,0.6597951889038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2048,8,256,4,1,power_law_1.01,0.6857984066009521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2048,8,256,4,1,power_law_1.01,0.7110015869140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2048,8,256,4,1,power_law_1.01,0.7461631774902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2048,8,256,4,1,power_law_1.01,0.8339455604553223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2048,8,256,4,1,power_law_1.01,0.8616512298583985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2048,8,256,4,1,power_law_1.01,0.9927040100097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2048,8,256,4,1,power_law_1.01,1.0300864219665526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2048,8,256,4,1,power_law_1.01,1.2063232421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2048,8,256,4,1,power_law_1.01,1.4089280128479005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2048,8,256,4,1,power_law_1.01,1.760576057434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2048,8,256,4,1,power_law_1.01,2.137228775024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2048,8,256,4,1,power_law_1.01,2.8947391510009766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2048,8,256,4,1,power_law_1.01,3.638739013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,4,8,power_law_1.2,0.07068799734115601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,4,8,power_law_1.2,0.08697599768638611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,4,8,power_law_1.2,0.08191999793052673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,4,8,power_law_1.2,0.10268160104751586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,4,8,power_law_1.2,0.10920959711074829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,4,8,power_law_1.2,0.11030399799346924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,4,8,power_law_1.2,0.1149183988571167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,4,8,power_law_1.2,0.11085439920425415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,4,8,power_law_1.2,0.11767040491104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,4,8,power_law_1.2,0.11520639657974244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,4,8,power_law_1.2,0.1162943959236145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,4,8,power_law_1.2,0.11729279756546021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,4,8,power_law_1.2,0.11868159770965576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,4,8,power_law_1.2,0.12303999662399293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,4,8,power_law_1.2,0.1329599976539612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,4,8,power_law_1.2,0.13313280344009398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,4,8,power_law_1.2,0.14069119691848755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,4,8,power_law_1.2,0.1625856041908264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,4,8,power_law_1.2,0.17560960054397584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,4,8,power_law_1.2,0.20869119167327882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,4,8,power_law_1.2,0.2325439929962158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,4,8,power_law_1.2,0.3149696111679077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,4,8,power_law_1.2,0.3938368082046509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,4,8,power_law_1.2,0.5794623851776123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,4,8,power_law_1.2,0.7345600128173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,6144,2048,8,256,4,8,power_law_1.2,1.0950464248657226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,6144,2048,8,256,4,8,power_law_1.2,1.5114944458007813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,8,2,power_law_1.2,0.032646399736404416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,8,2,power_law_1.2,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,8,2,power_law_1.2,0.038099199533462524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,8,2,power_law_1.2,0.04584319889545441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,8,2,power_law_1.2,0.05023360252380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,8,2,power_law_1.2,0.06551039814949036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,8,2,power_law_1.2,0.06833279728889466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,8,2,power_law_1.2,0.07039999961853027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,8,2,power_law_1.2,0.07448319792747497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,8,2,power_law_1.2,0.07534719705581665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,8,2,power_law_1.2,0.07360000014305115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,8,2,power_law_1.2,0.07993599772453308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,8,2,power_law_1.2,0.08254719972610473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,8,2,power_law_1.2,0.0850495994091034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,8,2,power_law_1.2,0.09404159784317016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,8,2,power_law_1.2,0.10160000324249267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,8,2,power_law_1.2,0.10885759592056274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,8,2,power_law_1.2,0.12237440347671509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,8,2,power_law_1.2,0.14577920436859132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,8,2,power_law_1.2,0.1753216028213501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,8,2,power_law_1.2,0.21624319553375243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,8,2,power_law_1.2,0.2894144058227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,8,2,power_law_1.2,0.38125441074371336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,8,2,power_law_1.2,0.5441472053527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,8,2,power_law_1.2,0.6715328216552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,8,2,power_law_1.2,1.0012224197387696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,4,1,power_law_1.2,0.05271040201187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,4,1,power_law_1.2,0.06487680077552796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,4,1,power_law_1.2,0.07848960161209106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,4,1,power_law_1.2,0.1147968053817749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,4,1,power_law_1.2,0.11767040491104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,4,1,power_law_1.2,0.11786240339279175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,4,1,power_law_1.2,0.12374399900436402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,4,1,power_law_1.2,0.12480640411376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,4,1,power_law_1.2,0.12817920446395875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,4,1,power_law_1.2,0.13210879564285277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,4,1,power_law_1.2,0.13300479650497438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,4,1,power_law_1.2,0.14079999923706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,4,1,power_law_1.2,0.14013439416885376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,4,1,power_law_1.2,0.14592000246047973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,4,1,power_law_1.2,0.15976959466934204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,8,2,power_law_1.2,1.400153636932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,4,1,power_law_1.2,0.16483839750289916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,4,1,power_law_1.2,0.17454080581665038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,4,1,power_law_1.2,0.20371201038360595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,4,4,balanced,0.05508266886075338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,4,4,balanced,0.053354665637016296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,4,4,balanced,0.08004266520341237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,4,4,balanced,0.07864533364772797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,4,4,balanced,0.07872533301512401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,4,4,balanced,0.08038400113582611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,4,4,balanced,0.07830933233102162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,4,4,balanced,0.08000533282756805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,4,4,balanced,0.08084799846013387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,4,4,balanced,0.08101333181063335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,4,4,balanced,0.08290666838486989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,4,4,balanced,0.08247999846935272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,4,4,balanced,0.08208000163237254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,4,4,balanced,0.08562666177749634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,4,4,balanced,0.09265066186587016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,4,1,power_law_1.2,0.23354239463806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,4,4,balanced,0.09160533547401428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,4,4,balanced,0.10269332925478618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,4,4,balanced,0.11142933368682861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,4,4,balanced,0.12211199601491292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,4,4,balanced,0.16645333170890808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,4,4,balanced,0.1654933293660482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,4,4,balanced,0.21175465981165567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,4,4,balanced,0.2505333423614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,4,4,balanced,0.3368320067723592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,4,4,balanced,0.42001601060231525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,4,4,balanced,0.6049066781997681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,4,4,balanced,0.7717066605885824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,4,1,power_law_1.2,0.2885504007339478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,2048,8,256,1,64,balanced,0.0606826643149058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,2048,8,256,1,64,balanced,0.05861333509286245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,2048,8,256,1,64,balanced,0.059215997656186424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,2048,8,256,1,64,balanced,0.06404266754786174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,2048,8,256,1,64,balanced,0.06572266419728597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,2048,8,256,1,64,balanced,0.08316799998283386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,2048,8,256,1,64,balanced,0.08452266454696655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,2048,8,256,1,64,balanced,0.08854933579762776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,2048,8,256,1,64,balanced,0.08265066643555959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,2048,8,256,1,64,balanced,0.08299200236797333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,2048,8,256,1,64,balanced,0.09676266709963481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,2048,8,256,1,64,balanced,0.08943999807039897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,2048,8,256,1,64,balanced,0.08988799651463826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,2048,8,256,1,64,balanced,0.09291733304659526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,2048,8,256,1,64,balanced,0.09870933492978413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,2048,8,256,1,64,balanced,0.0990133285522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,2048,8,256,1,64,balanced,0.10244266192118327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,2048,8,256,1,64,balanced,0.11343466242154439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,2048,8,256,1,64,balanced,0.11793067057927449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,2048,8,256,1,64,balanced,0.13371733824412027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,2048,8,256,1,64,balanced,0.14825600385665894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,2048,8,256,1,64,balanced,0.18413867553075156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,2048,8,256,1,64,balanced,0.216154674688975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,2048,8,256,1,64,balanced,0.2914026578267415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,2048,8,256,1,64,balanced,0.3564586639404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,2048,8,256,1,64,balanced,0.49595733483632404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,2048,8,256,1,64,balanced,0.63591468334198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,4,1,power_law_1.2,0.34664320945739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,4,1,power_law_1.2,0.4505727767944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,2,power_law_1.2,0.0646336019039154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,2,power_law_1.2,0.06970239877700805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,2,power_law_1.2,0.07538560032844543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,4,1,power_law_1.2,0.5570367813110352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,2,power_law_1.2,0.08673920035362244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,2,power_law_1.2,0.10126719474792481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,2,power_law_1.2,0.11787519454956055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,2,power_law_1.2,0.1357375979423523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,2,power_law_1.2,0.14300800561904908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,2,power_law_1.2,0.1410815954208374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,2,power_law_1.2,0.14850560426712037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,2,power_law_1.2,0.14697599411010742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,2,power_law_1.2,0.15071359872817994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,2,power_law_1.2,0.15575040578842164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,2,power_law_1.2,0.16087039709091186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,2,power_law_1.2,0.16897280216217042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,2,power_law_1.2,0.17659519910812377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,2,power_law_1.2,0.18495999574661254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,2,power_law_1.2,0.21142399311065674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,2,power_law_1.2,0.23679358959198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,2,power_law_1.2,0.2797823905944824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,2,power_law_1.2,0.33780479431152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,2,power_law_1.2,0.4549568176269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,2,power_law_1.2,0.5576000213623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,2,power_law_1.2,0.7640063762664795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,2,power_law_1.2,0.9570303916931152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,2,power_law_1.2,1.4008064270019531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,2,power_law_1.2,1.8823680877685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,4,1,power_law_1.2,0.7871679782867431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,4,1,power_law_1.2,0.9871552467346192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,1536,8,128,1,64,power_law_1.2,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,1536,8,128,1,64,power_law_1.2,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,1536,8,128,1,64,power_law_1.2,0.03198719918727875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,1536,8,128,1,64,power_law_1.2,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,1536,8,128,1,64,power_law_1.2,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,1536,8,128,1,64,power_law_1.2,0.03524479866027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,1536,8,128,1,64,power_law_1.2,0.03523840010166168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,1536,8,128,1,64,power_law_1.2,0.03553279936313629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,1536,8,128,1,64,power_law_1.2,0.03537920117378235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,1536,8,128,1,64,power_law_1.2,0.03856000006198883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,1536,8,128,1,64,power_law_1.2,0.041875201463699344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,1536,8,128,1,64,power_law_1.2,0.042956799268722534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,1536,8,128,1,64,power_law_1.2,0.04707199931144714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,1536,8,128,1,64,power_law_1.2,0.050387197732925416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,1536,8,128,1,64,power_law_1.2,0.05447040200233459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,1536,8,128,1,64,power_law_1.2,0.06295679807662964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,1536,8,128,1,64,power_law_1.2,0.06044800281524658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,1536,8,128,1,64,power_law_1.2,0.07283200025558471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,1536,8,128,1,64,power_law_1.2,0.07691519856452941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,1536,8,128,1,64,power_law_1.2,0.0930176019668579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,1536,8,128,1,64,power_law_1.2,0.11786880493164062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,1536,8,128,1,64,power_law_1.2,0.14949120283126832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,1536,8,128,1,64,power_law_1.2,0.17077759504318238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,1536,8,128,1,64,power_law_1.2,0.2810431957244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,4,1,power_law_1.2,1.4083776473999023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,1536,8,128,1,64,power_law_1.2,0.277728009223938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,1536,8,128,1,64,power_law_1.2,0.3543936014175415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,4,1,power_law_1.2,1.8348031997680665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,1536,8,128,1,64,power_law_1.2,0.55032320022583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,2,4,power_law_1.2,0.05504639744758606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,2,4,power_law_1.2,0.060627198219299315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,2,4,power_law_1.2,0.07578240036964416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,2,4,power_law_1.2,0.08048639893531799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,2,4,power_law_1.2,0.08022400140762329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,2,4,power_law_1.2,0.08424320220947265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,2,4,power_law_1.2,0.08883200287818908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,2,4,power_law_1.2,0.08504319787025452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,2,4,power_law_1.2,0.09015039801597595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,2,4,power_law_1.2,0.08883839845657349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,2,4,power_law_1.2,0.09217280149459839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,2,4,power_law_1.2,0.10001920461654663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,2,4,power_law_1.2,0.10343040227890014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,2,4,power_law_1.2,0.10597120523452759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,2,4,power_law_1.2,0.13422720432281493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,2,4,power_law_1.2,0.14421119689941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,2,4,power_law_1.2,0.1380288004875183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,2,4,power_law_1.2,0.1622015953063965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,2,4,power_law_1.2,0.2114880084991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,2,4,power_law_1.2,0.2611327886581421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,2,4,power_law_1.2,0.34035840034484866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,2,4,power_law_1.2,0.4546495914459229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,2,4,power_law_1.2,0.5296768188476563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2560,8,160,8,16,power_law_1.2,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2560,8,160,8,16,power_law_1.2,0.07139840126037597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2560,8,160,8,16,power_law_1.2,0.07066239714622498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2560,8,160,8,16,power_law_1.2,0.06947199702262878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2560,8,160,8,16,power_law_1.2,0.06649600267410279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2560,8,160,8,16,power_law_1.2,0.0715071976184845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2560,8,160,8,16,power_law_1.2,0.0713599979877472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2560,8,160,8,16,power_law_1.2,0.07300480008125305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2560,8,160,8,16,power_law_1.2,0.07210879921913146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2560,8,160,8,16,power_law_1.2,0.07149440050125122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2560,8,160,8,16,power_law_1.2,0.07276800274848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2560,8,160,8,16,power_law_1.2,0.07441920042037964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2560,8,160,8,16,power_law_1.2,0.0748416006565094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2560,8,160,8,16,power_law_1.2,0.07590399980545044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2560,8,160,8,16,power_law_1.2,0.0791104018688202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2560,8,160,8,16,power_law_1.2,0.08457599878311158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2560,8,160,8,16,power_law_1.2,0.08849279880523682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2560,8,160,8,16,power_law_1.2,0.0986303985118866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2560,8,160,8,16,power_law_1.2,0.11249279975891113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2560,8,160,8,16,power_law_1.2,0.1424064040184021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2560,8,160,8,16,power_law_1.2,0.15267839431762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2560,8,160,8,16,power_law_1.2,0.19136639833450317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2560,8,160,8,16,power_law_1.2,0.25729920864105227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2560,8,160,8,16,power_law_1.2,0.3680255889892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2560,8,160,8,16,power_law_1.2,0.428115177154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,2,4,power_law_1.2,0.9288576126098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2560,8,160,8,16,power_law_1.2,0.6262976169586182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,2,4,power_law_1.2,1.2151552200317384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2560,8,160,8,16,power_law_1.2,0.7695680141448975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,14336,2,8,16,2,power_law_1.2,0.030483201146125793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,14336,2,8,16,2,power_law_1.2,0.031065601110458373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,14336,2,8,16,2,power_law_1.2,0.03229439854621887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,14336,2,8,16,2,power_law_1.2,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,14336,2,8,16,2,power_law_1.2,0.0350271999835968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,14336,2,8,16,2,power_law_1.2,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,14336,2,8,16,2,power_law_1.2,0.036339199542999266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,14336,2,8,16,2,power_law_1.2,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,14336,2,8,16,2,power_law_1.2,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,14336,2,8,16,2,power_law_1.2,0.03653759956359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,14336,2,8,16,2,power_law_1.2,0.036652800440788266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,14336,2,8,16,2,power_law_1.2,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,14336,2,8,16,2,power_law_1.2,0.04159359931945801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,14336,2,8,16,2,power_law_1.2,0.044486400485038755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,14336,2,8,16,2,power_law_1.2,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,14336,2,8,16,2,power_law_1.2,0.057004797458648684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,14336,2,8,16,2,power_law_1.2,0.0615231990814209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,14336,2,8,16,2,power_law_1.2,0.07354879975318909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,32,2,balanced,0.03156266609827677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,32,2,balanced,0.02961066613594691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,32,2,balanced,0.03164266546567281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,32,2,balanced,0.03357866654793421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,32,2,balanced,0.03373866776625315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,32,2,balanced,0.03453333427508672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,32,2,balanced,0.03373866776625315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,32,2,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,32,2,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,32,2,balanced,0.03485333422819773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,32,2,balanced,0.03572800010442734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,32,2,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,32,2,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,32,2,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,32,2,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,32,2,balanced,0.04604266583919525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,32,2,balanced,0.04597333570321401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,32,2,balanced,0.05091733237107595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,32,2,balanced,0.055573334296544395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,32,2,balanced,0.07039466500282288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,32,2,balanced,0.0746559997399648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,32,2,balanced,0.10543466607729594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,32,2,balanced,0.11301333705584209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,32,2,balanced,0.1564959983030955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,32,2,balanced,0.17898666858673096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,32,2,balanced,0.2526879906654358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,32,2,balanced,0.32628800471623737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,14336,2,8,16,2,power_law_1.2,0.08737919926643371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,14336,2,8,16,2,power_law_1.2,0.11585279703140258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,16,4,power_law_1.01,0.047040000557899475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,14336,2,8,16,2,power_law_1.2,0.14561280012130737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,16,4,power_law_1.01,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,16,4,power_law_1.01,0.056531202793121335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,16,4,power_law_1.01,0.06401919722557067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,16,4,power_law_1.01,0.07022079825401306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,16,4,power_law_1.01,0.07802240252494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,16,4,power_law_1.01,0.08915839791297912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,16,4,power_law_1.01,0.08765439987182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,16,4,power_law_1.01,0.09093120098114013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,16,4,power_law_1.01,0.09367039799690247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,16,4,power_law_1.01,0.09313279986381531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,16,4,power_law_1.01,0.0998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,16,4,power_law_1.01,0.10735360383987427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,16,4,power_law_1.01,0.11296000480651855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,16,4,power_law_1.01,0.12239359617233277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,16,4,power_law_1.01,0.12365440130233765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,16,4,power_law_1.01,0.13269120454788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,16,4,power_law_1.01,0.15306240320205688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,14336,2,8,16,2,power_law_1.2,0.1820479989051819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,16,4,power_law_1.01,0.1692415952682495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,16,4,power_law_1.01,0.21220479011535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,16,4,power_law_1.01,0.24946560859680175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,16,4,power_law_1.01,0.33518080711364745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,16,4,power_law_1.01,0.4013823986053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,16,4,power_law_1.01,0.5599232196807862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,2,4,power_law_1.2,1.666854476928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,16,4,power_law_1.01,0.7194496154785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,16,4,power_law_1.01,1.0297344207763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,16,4,power_law_1.01,1.3273983955383302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,14336,2,8,16,2,power_law_1.2,0.21116158962249756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,14336,2,8,16,2,power_law_1.2,0.35520639419555666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,2,4,power_law_1.2,1.9964736938476562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,14336,2,8,16,2,power_law_1.2,0.42383999824523927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,14336,2,8,16,2,power_law_1.2,0.6911935806274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,2,32,power_law_1.01,0.03406080007553101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,2,32,power_law_1.01,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,2,32,power_law_1.01,0.03252480030059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,14336,2,8,16,2,power_law_1.2,0.9643775939941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,2,32,power_law_1.01,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,2,32,power_law_1.01,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,2,32,power_law_1.01,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,2,32,power_law_1.01,0.03675520122051239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,2,32,power_law_1.01,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,2,32,power_law_1.01,0.03864319920539856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,2,32,power_law_1.01,0.03860479891300202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,2,32,power_law_1.01,0.03909760117530823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,2,32,power_law_1.01,0.04170880019664765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,2,32,power_law_1.01,0.04199039936065674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,2,32,power_law_1.01,0.04537599980831146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,2,32,power_law_1.01,0.05214080214500427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,2,32,power_law_1.01,0.048716801404953006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,2,32,power_law_1.01,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,2,32,power_law_1.01,0.06417919993400574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,2,32,power_law_1.01,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,2,32,power_law_1.01,0.08726400136947632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,2,32,power_law_1.01,0.10056320428848267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,2,32,power_law_1.01,0.13084800243377687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,2,32,power_law_1.01,0.16959359645843505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,2,32,power_law_1.01,0.22872319221496581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,2,32,power_law_1.01,0.31605119705200196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,2,32,power_law_1.01,0.5094592094421386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,2,32,power_law_1.01,0.6530240058898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,4,power_law_1.01,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,4,power_law_1.01,0.07084800004959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,4,power_law_1.01,0.08578559756278992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,4,power_law_1.01,0.11341439485549927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,4,power_law_1.01,0.12487679719924927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,4,power_law_1.01,0.14822399616241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,4,power_law_1.01,0.16116479635238648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,4,power_law_1.01,0.15640319585800172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,4,power_law_1.01,0.16669440269470215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,4,power_law_1.01,0.1700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,4,power_law_1.01,0.18273919820785522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,4,power_law_1.01,0.19434239864349365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,4,power_law_1.01,0.206329607963562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,4,power_law_1.01,0.19113600254058838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,4,power_law_1.01,0.20803840160369874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,4,power_law_1.01,0.21985280513763428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,4,power_law_1.01,0.20592639446258545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,4,power_law_1.01,0.22441599369049073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.01,0.3164031982421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.01,0.32515840530395507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.01,0.3637887954711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.01,0.4104640007019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.01,0.5111680030822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.01,0.6536320209503174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.01,0.7976384162902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.01,1.092518424987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.01,1.441055965423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,3072,1536,8,256,1,4,power_law_1.01,0.05904639959335327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,3072,1536,8,256,1,4,power_law_1.01,0.07756800055503846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,3072,1536,8,256,1,4,power_law_1.01,0.0875328004360199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,3072,1536,8,256,1,4,power_law_1.01,0.10797439813613892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,3072,1536,8,256,1,4,power_law_1.01,0.12717440128326415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,3072,1536,8,256,1,4,power_law_1.01,0.13451520204544068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,3072,1536,8,256,1,4,power_law_1.01,0.16066559553146362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,3072,1536,8,256,1,4,power_law_1.01,0.14974720478057862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,3072,1536,8,256,1,4,power_law_1.01,0.1620352029800415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,3072,1536,8,256,1,4,power_law_1.01,0.1657920002937317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,3072,1536,8,256,1,4,power_law_1.01,0.1650431990623474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,3072,1536,8,256,1,4,power_law_1.01,0.16393599510192872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,3072,1536,8,256,1,4,power_law_1.01,0.1718016028404236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,3072,1536,8,256,1,4,power_law_1.01,0.17265280485153198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,3072,1536,8,256,1,4,power_law_1.01,0.1877120018005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,3072,1536,8,256,1,4,power_law_1.01,0.18955520391464234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,3072,1536,8,256,1,4,power_law_1.01,0.20297598838806152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,3072,1536,8,256,1,4,power_law_1.01,0.22247040271759033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,3072,1536,8,256,1,4,power_law_1.01,0.22751998901367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,3072,1536,8,256,1,4,power_law_1.01,0.26871678829193113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,3072,1536,8,256,1,4,power_law_1.01,0.30134398937225343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,3072,1536,8,256,1,4,power_law_1.01,0.3874815940856934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,3072,1536,8,256,1,4,power_law_1.01,0.4040128231048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,3072,1536,8,256,1,4,power_law_1.01,0.5473728179931641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,3072,1536,8,256,1,4,power_law_1.01,0.6567872047424317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,3072,1536,8,256,1,4,power_law_1.01,0.9530495643615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,3072,1536,8,256,1,4,power_law_1.01,1.1818623542785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1,4096,2688,22,512,1,32,balanced,0.04877333343029022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2,4096,2688,22,512,1,32,balanced,0.066021333138148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4,4096,2688,22,512,1,32,balanced,0.08242666721343994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8,4096,2688,22,512,1,32,balanced,0.12495467066764832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16,4096,2688,22,512,1,32,balanced,0.1550986667474111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,32,4096,2688,22,512,1,32,balanced,0.15707199772198996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,48,4096,2688,22,512,1,32,balanced,0.15954666336377463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,64,4096,2688,22,512,1,32,balanced,0.15712533394495645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,80,4096,2688,22,512,1,32,balanced,0.16087999939918518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,96,4096,2688,22,512,1,32,balanced,0.1599573294321696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,128,4096,2688,22,512,1,32,balanced,0.1593119998772939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,160,4096,2688,22,512,1,32,balanced,0.16479466358820596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,192,4096,2688,22,512,1,32,balanced,0.16506133476893106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,256,4096,2688,22,512,1,32,balanced,0.1642453372478485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,320,4096,2688,22,512,1,32,balanced,0.16852267583211264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,384,4096,2688,22,512,1,32,balanced,0.1689280072848002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,512,4096,2688,22,512,1,32,balanced,0.1743519902229309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,768,4096,2688,22,512,1,32,balanced,0.20339733362197876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1024,4096,2688,22,512,1,32,balanced,0.19315199057261148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,1536,4096,2688,22,512,1,32,balanced,0.2423200011253357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,2048,4096,2688,22,512,1,32,balanced,0.23483733336130777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,3072,4096,2688,22,512,1,32,balanced,0.3629386822382609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,4096,4096,2688,22,512,1,32,balanced,0.2877440055211385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,6144,4096,2688,22,512,1,32,balanced,0.40026132265726727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,8192,4096,2688,22,512,1,32,balanced,0.4090293248494466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,12288,4096,2688,22,512,1,32,balanced,0.6295093297958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,float16,16384,4096,2688,22,512,1,32,balanced,0.635317325592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,16,1,power_law_1.01,0.09669119715690613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,16,1,power_law_1.01,0.10631040334701539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,16,1,power_law_1.01,0.12714879512786864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,16,1,power_law_1.01,0.1684864044189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,16,1,power_law_1.01,0.20824320316314698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,16,1,power_law_1.01,0.24588160514831542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,16,1,power_law_1.01,0.29828479290008547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,16,1,power_law_1.01,0.31965439319610595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,16,1,power_law_1.01,0.32534399032592776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,16,1,power_law_1.01,0.33115520477294924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,16,1,power_law_1.01,0.33914239406585694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,16,1,power_law_1.01,0.35219199657440187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,16,1,power_law_1.01,0.34664320945739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,16,1,power_law_1.01,0.3635328054428101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,16,1,power_law_1.01,0.37512960433959963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,16,1,power_law_1.01,0.37795200347900393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,16,1,power_law_1.01,0.4026175975799561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,16,1,power_law_1.01,0.45166721343994143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,4096,1536,8,128,8,1,power_law_1.2,0.052275198698043826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,4096,1536,8,128,8,1,power_law_1.2,0.05511040091514587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,4096,1536,8,128,8,1,power_law_1.2,0.06157439947128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,4096,1536,8,128,8,1,power_law_1.2,0.0741375982761383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,4096,1536,8,128,8,1,power_law_1.2,0.084934401512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,4096,1536,8,128,8,1,power_law_1.2,0.11848959922790528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,4096,1536,8,128,8,1,power_law_1.2,0.12350720167160034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,4096,1536,8,128,8,1,power_law_1.2,0.12709759473800658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,4096,1536,8,128,8,1,power_law_1.2,0.12922879457473754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,4096,1536,8,128,8,1,power_law_1.2,0.12917120456695558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,16,1,power_law_1.01,0.5005695819854736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,4096,1536,8,128,8,1,power_law_1.2,0.13720959424972534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,4096,1536,8,128,8,1,power_law_1.2,0.14024959802627562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,16,1,power_law_1.01,0.5868800163269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,4096,1536,8,128,8,1,power_law_1.2,0.14218879938125611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,4096,1536,8,128,8,1,power_law_1.2,0.145798397064209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,16,1,power_law_1.01,0.671347188949585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,4096,1536,8,128,8,1,power_law_1.2,0.15645439624786378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,4096,1536,8,128,8,1,power_law_1.2,0.1617792010307312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,16,1,power_law_1.01,0.8672639846801757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,4096,1536,8,128,8,1,power_law_1.2,0.17732479572296142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,4096,1536,8,128,8,1,power_law_1.2,0.21020159721374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,16,1,power_law_1.01,1.0446463584899903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.2,0.2419071912765503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,16,1,power_law_1.01,1.4173055648803712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.2,0.30869760513305666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.2,0.3723839998245239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,16,1,power_law_1.01,1.7899648666381835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,16,1,power_law_1.01,2.5187328338623045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.2,0.5075456142425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,16,1,power_law_1.01,3.253510284423828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.2,0.637113618850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.2,0.9035776138305665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.2,1.1760704040527343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,4,8,power_law_1.01,0.06983680129051209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,4,8,power_law_1.01,0.09192960262298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,4,8,power_law_1.01,0.09561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,4,8,power_law_1.01,0.11854079961776734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,4,8,power_law_1.01,0.1232192039489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,4,8,power_law_1.01,0.12668800354003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.2,1.695884895324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,4,8,power_law_1.01,0.13806719779968263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,4,8,power_law_1.01,0.13193600177764891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,4,8,power_law_1.01,0.133951997756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,4,8,power_law_1.01,0.13786879777908326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,4,8,power_law_1.01,0.14256639480590821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,4096,14336,2,8,8,8,balanced,0.035487999518712364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,4096,14336,2,8,8,8,balanced,0.03403199960788091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,4096,14336,2,8,8,8,balanced,0.02992533395687739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,4096,14336,2,8,8,8,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,4096,14336,2,8,8,8,balanced,0.03159466634194056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,4096,14336,2,8,8,8,balanced,0.03336533407370249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,4096,14336,2,8,8,8,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,4096,14336,2,8,8,8,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,4096,14336,2,8,8,8,balanced,0.0331839993596077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,4096,14336,2,8,8,8,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,4096,14336,2,8,8,8,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,4096,14336,2,8,8,8,balanced,0.0352906659245491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,4096,14336,2,8,8,8,balanced,0.03357866654793421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,4096,14336,2,8,8,8,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,4096,14336,2,8,8,8,balanced,0.04145599901676178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,4,8,power_law_1.01,0.14583040475845338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,4,8,power_law_1.01,0.14755840301513673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,4,8,power_law_1.01,0.14752639532089235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,4,8,power_law_1.01,0.1537279963493347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,4,8,power_law_1.01,0.15806080102920533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,4,8,power_law_1.01,0.1713152050971985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,4,8,power_law_1.01,0.1934592008590698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,4096,14336,2,8,8,8,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,4096,14336,2,8,8,8,balanced,0.04154133299986521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,4096,14336,2,8,8,8,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,4096,14336,2,8,8,8,balanced,0.04570133487383524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,4096,14336,2,8,8,8,balanced,0.05797866483529409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,4096,14336,2,8,8,8,balanced,0.056186666091283165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,4096,14336,2,8,8,8,balanced,0.07629866898059845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,4096,14336,2,8,8,8,balanced,0.0792799989382426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,4096,14336,2,8,8,8,balanced,0.11069867014884949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,4096,14336,2,8,8,8,balanced,0.11886933445930481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,4096,14336,2,8,8,8,balanced,0.16007467110951742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,4096,14336,2,8,8,8,balanced,0.2014346718788147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,4,8,power_law_1.01,0.1967296004295349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,4,8,power_law_1.01,0.22772479057312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,4,8,power_law_1.01,0.2524415969848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,4,8,power_law_1.01,0.30785279273986815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,4,8,power_law_1.01,0.35583999156951907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,4,8,power_law_1.01,0.4591807842254639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,4,8,power_law_1.01,0.533574390411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,4,8,power_law_1.01,0.769708776473999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,4,8,power_law_1.01,0.9812095642089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.2,2.2058431625366213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,1,128,power_law_1.01,0.07317119836807251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,1,128,power_law_1.01,0.07155200242996215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,1,128,power_law_1.01,0.07155200242996215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,1,128,power_law_1.01,0.072953599691391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,1,128,power_law_1.01,0.07038080096244811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,1,128,power_law_1.01,0.07427840232849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,1,128,power_law_1.01,0.07530879974365234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,1,128,power_law_1.01,0.07739520072937012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,1,128,power_law_1.01,0.07406719923019409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,1,128,power_law_1.01,0.07438719868659974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,1,128,power_law_1.01,0.07529600262641907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,1,128,power_law_1.01,0.077702397108078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,1,128,power_law_1.01,0.0762943983078003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,1,128,power_law_1.01,0.08095359802246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,1,128,power_law_1.01,0.08560640215873719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,1,128,power_law_1.01,0.08556159734725952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,1,128,power_law_1.01,0.089683198928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,1,128,power_law_1.01,0.10492160320281982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,1,128,power_law_1.01,0.11786880493164062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,1,128,power_law_1.01,0.14739199876785278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,1,128,power_law_1.01,0.16631679534912108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,1,128,power_law_1.01,0.2119231939315796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,1,128,power_law_1.01,0.262604808807373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,1,128,power_law_1.01,0.35852799415588377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,1,128,power_law_1.01,0.4667200088500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,1,128,power_law_1.01,0.6659840106964111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,1,128,power_law_1.01,0.8432512283325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2560,8,160,4,2,balanced,0.09520000219345093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,6144,2048,8,256,1,1,power_law_1.01,0.10286719799041748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,6144,2048,8,256,1,1,power_law_1.01,0.13034240007400513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,6144,2048,8,256,1,1,power_law_1.01,0.1738368034362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,6144,2048,8,256,1,1,power_law_1.01,0.28472959995269775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,6144,2048,8,256,1,1,power_law_1.01,0.39984641075134275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,6144,2048,8,256,1,1,power_law_1.01,0.4960000038146973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,6144,2048,8,256,1,1,power_law_1.01,0.6749695777893067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,6144,2048,8,256,1,1,power_law_1.01,0.7177536010742187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,6144,2048,8,256,1,1,power_law_1.01,0.7288512229919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,6144,2048,8,256,1,1,power_law_1.01,0.7432256221771241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,6144,2048,8,256,1,1,power_law_1.01,0.7863232135772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,6144,2048,8,256,1,1,power_law_1.01,0.7939648151397705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,6144,2048,8,256,1,1,power_law_1.01,0.7982719898223877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,6144,2048,8,256,1,1,power_law_1.01,0.835200023651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,6144,2048,8,256,1,1,power_law_1.01,0.8545856475830078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,6144,2048,8,256,1,1,power_law_1.01,0.8803008079528809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,6144,2048,8,256,1,1,power_law_1.01,0.9118335723876954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,6144,2048,8,256,1,1,power_law_1.01,0.9775232315063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,6144,2048,8,256,1,1,power_law_1.01,1.049120044708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,6144,2048,8,256,1,1,power_law_1.01,1.2030400276184081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,6144,2048,8,256,1,1,power_law_1.01,1.277939224243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,6144,2048,8,256,1,1,power_law_1.01,1.5370176315307618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,6144,2048,8,256,1,1,power_law_1.01,1.8077056884765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,6144,2048,8,256,1,1,power_law_1.01,2.333087921142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,6144,2048,8,256,1,1,power_law_1.01,2.8799615859985352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,6144,2048,8,256,1,1,power_law_1.01,3.950592041015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,6144,2048,8,256,1,1,power_law_1.01,5.040671920776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,4,8,power_law_1.01,0.08016639947891235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,4,8,power_law_1.01,0.09668480157852173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,4,8,power_law_1.01,0.1078976035118103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,4,8,power_law_1.01,0.1189568042755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,4,8,power_law_1.01,0.12328319549560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,4,8,power_law_1.01,0.1247871994972229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,4,8,power_law_1.01,0.12777600288391114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,4,8,power_law_1.01,0.12989439964294433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,4,8,power_law_1.01,0.12830079793930055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,4,8,power_law_1.01,0.12933119535446166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,4,8,power_law_1.01,0.12974079847335815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,4,8,power_law_1.01,0.13181439638137818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,4,8,power_law_1.01,0.1369279980659485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,4,8,power_law_1.01,0.13736319541931152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,4,8,power_law_1.01,0.1477120041847229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,4,8,power_law_1.01,0.14536319971084594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,4,8,power_law_1.01,0.14903680086135865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,4,8,power_law_1.01,0.17296639680862427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,4,8,power_law_1.01,0.1839679956436157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,4,8,power_law_1.01,0.2097536087036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,4,8,power_law_1.01,0.24479999542236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,4,8,power_law_1.01,0.316646409034729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,4,8,power_law_1.01,0.3985856056213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,4,8,power_law_1.01,0.5290431976318359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,4,8,power_law_1.01,0.6606592178344727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,4,8,power_law_1.01,1.0259391784667968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,4,8,power_law_1.01,1.3258048057556153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,128,power_law_1.01,0.031411200761795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,128,power_law_1.01,0.030105599761009218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,128,power_law_1.01,0.030220800638198854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,128,power_law_1.01,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,128,power_law_1.01,0.033580800890922545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,128,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,128,power_law_1.01,0.035692799091339114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,128,power_law_1.01,0.036524799466133115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,128,power_law_1.01,0.03759360015392303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,128,power_law_1.01,0.03891839981079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,128,power_law_1.01,0.0412992000579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,4,4,power_law_1.01,0.035462400317192076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,128,power_law_1.01,0.04286080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,4,4,power_law_1.01,0.030553600192070006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,4,4,power_law_1.01,0.031167998909950256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,4,4,power_law_1.01,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,4,4,power_law_1.01,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,4,4,power_law_1.01,0.03701759874820709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,128,power_law_1.01,0.04123519957065582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,128,power_law_1.01,0.044352000951766966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,128,power_law_1.01,0.05010560154914856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,128,power_law_1.01,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,4,4,power_law_1.01,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,128,power_law_1.01,0.054681599140167236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,4,4,power_law_1.01,0.037939199805259706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,128,power_law_1.01,0.06151679754257202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,4,4,power_law_1.01,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,128,power_law_1.01,0.06939520239830017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,4,4,power_law_1.01,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,4,4,power_law_1.01,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,4,4,power_law_1.01,0.04251520037651062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,4,4,power_law_1.01,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,4,4,power_law_1.01,0.04650880098342895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,4,4,power_law_1.01,0.05025280117988586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,4,4,power_law_1.01,0.052172797918319705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,128,power_law_1.01,0.08579840064048767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,4,4,power_law_1.01,0.05724800229072571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,128,power_law_1.01,0.10756479501724243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,4,4,power_law_1.01,0.06470400094985962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,128,power_law_1.01,0.1523967981338501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,4,4,power_law_1.01,0.07393280267715455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,4,4,power_law_1.01,0.09227520227432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,128,power_law_1.01,0.19600000381469726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,4,4,power_law_1.01,0.10120960474014282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,128,power_law_1.01,0.2664128065109253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,4,4,power_law_1.01,0.14123519659042358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,4,4,power_law_1.01,0.16941440105438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,128,power_law_1.01,0.34444799423217776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,4,4,power_law_1.01,0.24770560264587402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,128,power_law_1.01,0.5953855991363526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,128,power_law_1.01,0.9119104385375977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,4,4,power_law_1.01,0.29368319511413576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,4,4,power_law_1.01,0.41101441383361814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,4,4,power_law_1.01,0.6136576175689697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,16384,2,8,16,8,power_law_1.2,0.0482367992401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,16384,2,8,16,8,power_law_1.2,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,16384,2,8,16,8,power_law_1.2,0.048179200291633605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,16384,2,8,16,8,power_law_1.2,0.049420800805091855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,16384,2,8,16,8,power_law_1.2,0.04880000054836273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,16384,2,8,16,8,power_law_1.2,0.04787200093269348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,16384,2,8,16,8,power_law_1.2,0.04686720073223114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,16384,2,8,16,8,power_law_1.2,0.04723840057849884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,16384,2,8,16,8,power_law_1.2,0.04659200012683869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,16384,2,8,16,8,power_law_1.2,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,16384,2,8,16,8,power_law_1.2,0.04612480103969574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,16384,2,8,16,8,power_law_1.2,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,16384,2,8,16,8,power_law_1.2,0.04917120039463043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,7168,2048,8,256,1,16,power_law_1.01,0.14729599952697753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,16384,2,8,16,8,power_law_1.2,0.04893440008163452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,7168,2048,8,256,1,16,power_law_1.01,0.19783040285110473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,16384,2,8,16,8,power_law_1.2,0.05305600166320801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,7168,2048,8,256,1,16,power_law_1.01,0.21081600189208985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,16384,2,8,16,8,power_law_1.2,0.0541055977344513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,7168,2048,8,256,1,16,power_law_1.01,0.19125759601593018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,16384,2,8,16,8,power_law_1.2,0.06083199977874756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,7168,2048,8,256,1,16,power_law_1.01,0.21396479606628419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,7168,2048,8,256,1,16,power_law_1.01,0.16088320016860963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,16384,2,8,16,8,power_law_1.2,0.06866559982299805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,7168,2048,8,256,1,16,power_law_1.01,0.19073280096054077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,7168,2048,8,256,1,16,power_law_1.01,0.1799936056137085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,7168,2048,8,256,1,16,power_law_1.01,0.17730560302734374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,7168,2048,8,256,1,16,power_law_1.01,0.17871999740600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,7168,2048,8,256,1,16,power_law_1.01,0.17674239873886108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,7168,2048,8,256,1,16,power_law_1.01,0.1868607997894287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,7168,2048,8,256,1,16,power_law_1.01,0.1889407992362976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,7168,2048,8,256,1,16,power_law_1.01,0.18614399433135986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,7168,2048,8,256,1,16,power_law_1.01,0.19594240188598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,7168,2048,8,256,1,16,power_law_1.01,0.18894720077514648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,16384,2,8,16,8,power_law_1.2,0.07801600098609925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,7168,2048,8,256,1,16,power_law_1.01,0.20168960094451904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,7168,2048,8,256,1,16,power_law_1.01,0.21639680862426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,16384,2,8,16,8,power_law_1.2,0.10013439655303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.01,0.22744319438934327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.01,0.2597759962081909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.01,0.3017535924911499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.01,0.36757760047912597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.01,0.4123392105102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.01,0.5411263942718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.01,0.6988736152648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.01,0.973043155670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.01,1.2622783660888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,16384,2,8,16,8,power_law_1.2,0.12131839990615845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,16384,2,8,16,8,power_law_1.2,0.1544319987297058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,16384,2,8,16,8,power_law_1.2,0.18037760257720947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,1,balanced,0.027727998793125153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,1,balanced,0.02741333345572154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,1,balanced,0.02722666660944621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,1,balanced,0.03316800047953924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,1,balanced,0.04589866598447164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,1,balanced,0.06250666578610738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,1,balanced,0.053674668073654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,1,balanced,0.06222933530807495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,1,balanced,0.05347733199596405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,1,balanced,0.057349334160486855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,1,balanced,0.06666666766007741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,1,balanced,0.06111466884613037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,1,balanced,0.06600533425807953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,1,balanced,0.0782293329636256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,1,balanced,0.16124266386032104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,1,balanced,0.18480000893274942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,1,balanced,0.22020800908406576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,1,balanced,0.11992533008257548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,1,balanced,0.14551466703414917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,1,balanced,0.20387200514475504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,1,balanced,0.2609120011329651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,1,balanced,0.3627573251724243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,1,balanced,0.4694720109303792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,1,balanced,0.687274694442749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,1,balanced,0.8899040222167969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,1,balanced,1.3095893065134685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,1,balanced,1.7294507026672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2048,8,256,8,32,power_law_1.2,0.024063999950885772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2048,8,256,8,32,power_law_1.2,0.02033279985189438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2048,8,256,8,32,power_law_1.2,0.020262399315834047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2048,8,256,8,32,power_law_1.2,0.02021760046482086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2048,8,256,8,32,power_law_1.2,0.020678399503231047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2048,8,256,8,32,power_law_1.2,0.021798400580883025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2048,8,256,8,32,power_law_1.2,0.022771200537681578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2048,8,256,8,32,power_law_1.2,0.022176000475883483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2048,8,256,8,32,power_law_1.2,0.023423999547958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2048,8,256,8,32,power_law_1.2,0.023379200696945192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2048,8,256,8,32,power_law_1.2,0.027616000175476073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2048,8,256,8,32,power_law_1.2,0.027807998657226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2048,8,256,8,32,power_law_1.2,0.028384000062942505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2048,8,256,8,32,power_law_1.2,0.0441536009311676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2048,8,256,8,32,power_law_1.2,0.04344319999217987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2048,8,256,8,32,power_law_1.2,0.04320639967918396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2048,8,256,8,32,power_law_1.2,0.04529280066490173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2048,8,256,8,32,power_law_1.2,0.04673919975757599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2048,8,256,8,32,power_law_1.2,0.053427201509475705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2048,8,256,8,32,power_law_1.2,0.06421120166778564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2048,8,256,8,32,power_law_1.2,0.0726527988910675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2048,8,256,8,32,power_law_1.2,0.09326080083847046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2048,8,256,8,32,power_law_1.2,0.11942399740219116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2048,8,256,8,32,power_law_1.2,0.15752320289611815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2048,8,256,8,32,power_law_1.2,0.19390079975128174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2048,8,256,8,32,power_law_1.2,0.2681407928466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2048,8,256,8,32,power_law_1.2,0.3532288074493408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,16384,2,8,16,8,power_law_1.2,0.2887423992156982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,16384,2,8,16,8,power_law_1.2,0.34567039012908934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,6144,2560,8,160,4,4,balanced,0.04647466540336609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,6144,2560,8,160,4,4,balanced,0.050341332952181496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,6144,2560,8,160,4,4,balanced,0.07119999825954437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,6144,2560,8,160,4,4,balanced,0.10965866843859355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,6144,2560,8,160,4,4,balanced,0.15653333067893982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,6144,2560,8,160,4,4,balanced,0.18332266807556152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,6144,2560,8,160,4,4,balanced,0.18290666739145914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,6144,2560,8,160,4,4,balanced,0.1855413317680359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,6144,2560,8,160,4,4,balanced,0.18753600120544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,6144,2560,8,160,4,4,balanced,0.18780799706776938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,6144,2560,8,160,4,4,balanced,0.18756266434987387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,6144,2560,8,160,4,4,balanced,0.18849066893259683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,6144,2560,8,160,4,4,balanced,0.19011199474334717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,6144,2560,8,160,4,4,balanced,0.19520533084869385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,6144,2560,8,160,4,4,balanced,0.19924799601236978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,6144,2560,8,160,4,4,balanced,0.20578134059906006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,6144,2560,8,160,4,4,balanced,0.20784533023834229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,6144,2560,8,160,4,4,balanced,0.22295467058817545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,6144,2560,8,160,4,4,balanced,0.22884267568588257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,6144,2560,8,160,4,4,balanced,0.2691146731376648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,6144,2560,8,160,4,4,balanced,0.27638934055964154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,6144,2560,8,160,4,4,balanced,0.3447306553522746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,6144,2560,8,160,4,4,balanced,0.3766186634699504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,6144,2560,8,160,4,4,balanced,0.5239520072937012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,6144,2560,8,160,4,4,balanced,0.6155680020650228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,2560,8,160,4,4,balanced,0.8502079645792643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,2560,8,160,4,4,balanced,1.0723946889241536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,6144,16384,2,8,16,8,power_law_1.2,0.5063231945037842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,2048,768,8,128,2,8,power_law_1.01,0.03421440124511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,2048,768,8,128,2,8,power_law_1.01,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,2048,768,8,128,2,8,power_law_1.01,0.030988800525665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,2048,768,8,128,2,8,power_law_1.01,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,2048,768,8,128,2,8,power_law_1.01,0.03407999873161316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,2048,768,8,128,2,8,power_law_1.01,0.03492479920387268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,2048,768,8,128,2,8,power_law_1.01,0.03542400002479553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,2048,768,8,128,2,8,power_law_1.01,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,2048,768,8,128,2,8,power_law_1.01,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,2048,768,8,128,2,8,power_law_1.01,0.0365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,2048,768,8,128,2,8,power_law_1.01,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,2048,768,8,128,2,8,power_law_1.01,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,2048,768,8,128,2,8,power_law_1.01,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,2048,768,8,128,2,8,power_law_1.01,0.0423552006483078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,2048,768,8,128,2,8,power_law_1.01,0.04688639938831329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,2048,768,8,128,2,8,power_law_1.01,0.04873600006103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,2048,768,8,128,2,8,power_law_1.01,0.04981760084629059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,2048,768,8,128,2,8,power_law_1.01,0.056704002618789676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,2048,768,8,128,2,8,power_law_1.01,0.06448000073432922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,2048,768,8,128,2,8,power_law_1.01,0.07626240253448487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,2048,768,8,128,2,8,power_law_1.01,0.0851967990398407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,2048,768,8,128,2,8,power_law_1.01,0.1177791953086853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,2048,768,8,128,2,8,power_law_1.01,0.14061440229415895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,2048,768,8,128,2,8,power_law_1.01,0.1727231979370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,6144,16384,2,8,16,8,power_law_1.2,0.6499648094177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,2048,768,8,128,2,8,power_law_1.01,0.22999680042266846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,2048,768,8,128,2,8,power_law_1.01,0.3563584089279175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,2048,768,8,128,2,8,power_law_1.01,0.45829758644104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,7168,2048,8,256,2,1,balanced,0.07268799841403961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,7168,2048,8,256,2,1,balanced,0.09823466340700786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,7168,2048,8,256,2,1,balanced,0.14736533164978027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,7168,2048,8,256,2,1,balanced,0.2520906726519267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,7168,2048,8,256,2,1,balanced,0.44757866859436035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,7168,2048,8,256,2,1,balanced,0.8450400034586588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,7168,2048,8,256,2,1,balanced,0.8500693639119467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,7168,2048,8,256,2,1,balanced,0.849232037862142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,7168,2048,8,256,2,1,balanced,0.8571893374125162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,7168,2048,8,256,2,1,balanced,0.8595200379689535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,7168,2048,8,256,2,1,balanced,0.864853302637736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,7168,2048,8,256,2,1,balanced,0.8744853337605795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,7168,2048,8,256,2,1,balanced,0.881114641825358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,7168,2048,8,256,2,1,balanced,0.8930026690165201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,7168,2048,8,256,2,1,balanced,0.9116960366566976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,7168,2048,8,256,2,1,balanced,0.9213759899139404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,7168,2048,8,256,2,1,balanced,0.9434666633605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,7168,2048,8,256,2,1,balanced,0.9965706666310629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,7168,2048,8,256,2,1,balanced,1.0350293318430583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,7168,2048,8,256,2,1,balanced,1.111733357111613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,7168,2048,8,256,2,1,balanced,1.2135519981384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,7168,2048,8,256,2,1,balanced,1.3742079734802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,7168,2048,8,256,2,1,balanced,1.5490026473999023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,7168,2048,8,256,2,1,balanced,2.1464106241861978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,7168,2048,8,256,2,1,balanced,2.350282669067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,7168,2048,8,256,2,1,balanced,3.439258575439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,7168,2048,8,256,2,1,balanced,4.222837448120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1,7168,2048,8,256,1,128,balanced,0.08282666901747386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2,7168,2048,8,256,1,128,balanced,0.08228800197442372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4,7168,2048,8,256,1,128,balanced,0.08229866623878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8,7168,2048,8,256,1,128,balanced,0.08476266264915466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16,7168,2048,8,256,1,128,balanced,0.08434133728345235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,32,7168,2048,8,256,1,128,balanced,0.09867200255393982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,48,7168,2048,8,256,1,128,balanced,0.10095999638239543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,64,7168,2048,8,256,1,128,balanced,0.10053867101669312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,80,7168,2048,8,256,1,128,balanced,0.10362133383750916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,96,7168,2048,8,256,1,128,balanced,0.10681066910425822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,128,7168,2048,8,256,1,128,balanced,0.10630933443705241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,160,7168,2048,8,256,1,128,balanced,0.1002506713072459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,192,7168,2048,8,256,1,128,balanced,0.11572800079981486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,256,7168,2048,8,256,1,128,balanced,0.10333866874376933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,320,7168,2048,8,256,1,128,balanced,0.10900266965230306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,384,7168,2048,8,256,1,128,balanced,0.10674132903416951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,512,7168,2048,8,256,1,128,balanced,0.1074133316675822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,768,7168,2048,8,256,1,128,balanced,0.11691199739774068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1024,7168,2048,8,256,1,128,balanced,0.11897066235542297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,1536,7168,2048,8,256,1,128,balanced,0.12524267037709555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,2048,7168,2048,8,256,1,128,balanced,0.13580800096193948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,3072,7168,2048,8,256,1,128,balanced,0.15651200215021768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,4096,7168,2048,8,256,1,128,balanced,0.17878933747609457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,6144,7168,2048,8,256,1,128,balanced,0.22755199670791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,8192,7168,2048,8,256,1,128,balanced,0.26420799891153973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,12288,7168,2048,8,256,1,128,balanced,0.3704799811045329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,float16,16384,7168,2048,8,256,1,128,balanced,0.44968533515930176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,4096,1536,8,128,1,4,power_law_1.01,0.07416319847106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,4096,1536,8,128,1,4,power_law_1.01,0.0825215995311737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,4096,1536,8,128,1,4,power_law_1.01,0.09032959938049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,4096,1536,8,128,1,4,power_law_1.01,0.10870399475097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,4096,1536,8,128,1,4,power_law_1.01,0.12762240171432496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,4096,1536,8,128,1,4,power_law_1.01,0.14603519439697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,4096,1536,8,128,1,4,power_law_1.01,0.14852479696273804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,4096,1536,8,128,1,4,power_law_1.01,0.15050239562988282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,4096,1536,8,128,1,4,power_law_1.01,0.15017600059509278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,4096,1536,8,128,1,4,power_law_1.01,0.15155199766159058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,4096,1536,8,128,1,4,power_law_1.01,0.15674879550933837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,4096,1536,8,128,1,4,power_law_1.01,0.15886720418930053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,4096,1536,8,128,1,4,power_law_1.01,0.15941120386123658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,4096,1536,8,128,1,4,power_law_1.01,0.16436480283737182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,4096,1536,8,128,1,4,power_law_1.01,0.16775039434432984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,4096,1536,8,128,1,4,power_law_1.01,0.17204480171203612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,4096,1536,8,128,1,4,power_law_1.01,0.17663999795913696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,4096,1536,8,128,1,4,power_law_1.01,0.1959488034248352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,4,power_law_1.01,0.05863040089607239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,4096,1536,8,128,1,4,power_law_1.01,0.2130176067352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,4,power_law_1.01,0.07079039812088013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,4,power_law_1.01,0.08076800107955932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,4,power_law_1.01,0.09226880073547364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,4,power_law_1.01,0.10819840431213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,4096,1536,8,128,1,4,power_law_1.01,0.25946879386901855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,4,power_law_1.01,0.11492480039596557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,4,power_law_1.01,0.13030400276184081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,4,power_law_1.01,0.13715200424194335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,4,power_law_1.01,0.12858879566192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,4,power_law_1.01,0.1314303994178772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,4,power_law_1.01,0.13355519771575927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,4,power_law_1.01,0.13443200588226317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,4096,1536,8,128,1,4,power_law_1.01,0.2759936094284058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,4096,1536,8,128,1,4,power_law_1.01,0.38607358932495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,4096,1536,8,128,1,4,power_law_1.01,0.47588481903076174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,4096,1536,8,128,1,4,power_law_1.01,0.7328127861022949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,4,power_law_1.01,0.1401535987854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,4,power_law_1.01,0.1444991946220398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,4096,1536,8,128,1,4,power_law_1.01,0.8346303939819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,4,power_law_1.01,0.15696640014648439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,4,power_law_1.01,0.15670399665832518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,4096,1536,8,128,1,4,power_law_1.01,1.2363967895507812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,4,power_law_1.01,0.16430720090866088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,4,power_law_1.01,0.17857919931411742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,4,power_law_1.01,0.19336960315704346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,4096,1536,8,128,1,4,power_law_1.01,1.6865728378295899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,4,power_law_1.01,0.2344896078109741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,4,power_law_1.01,0.25583999156951903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,4,power_law_1.01,0.3212224006652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,4,power_law_1.01,0.378656005859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,4,power_law_1.01,0.5113279819488525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,4,power_law_1.01,0.6209727764129639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,4,power_law_1.01,0.8874943733215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,4,power_law_1.01,1.1659775733947755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,4096,1536,8,128,2,8,balanced,0.030213333666324615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,4096,1536,8,128,2,8,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,4096,1536,8,128,2,8,balanced,0.03336533407370249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,4096,1536,8,128,2,8,balanced,0.03745066622893015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,4096,1536,8,128,2,8,balanced,0.05997333427270254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,4096,1536,8,128,2,8,balanced,0.060191998879114784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,4096,1536,8,128,2,8,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,4096,1536,8,128,2,8,balanced,0.06141866743564606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,4096,1536,8,128,2,8,balanced,0.06191466748714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,4096,1536,8,128,2,8,balanced,0.06089599927266439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,4096,1536,8,128,2,8,balanced,0.06344000001748402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,4096,1536,8,128,2,8,balanced,0.06329600016276042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,4096,1536,8,128,2,8,balanced,0.06473599870999654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,4096,1536,8,128,2,8,balanced,0.0653599997361501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,4096,1536,8,128,2,8,balanced,0.070592001080513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,4096,1536,8,128,2,8,balanced,0.07363200187683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,4096,1536,8,128,2,8,balanced,0.07533866663773854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,4096,1536,8,128,2,8,balanced,0.08111999928951263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,4096,1536,8,128,2,8,balanced,0.09437867005666097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,4096,1536,8,128,2,8,balanced,0.10700800021489461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,4096,1536,8,128,2,8,balanced,0.11400533715883891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,4096,1536,8,128,2,8,balanced,0.14306666453679404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,4096,1536,8,128,2,8,balanced,0.1710933248202006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,4096,1536,8,128,2,8,balanced,0.23123733202616373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,4096,1536,8,128,2,8,balanced,0.2894879976908366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,4096,1536,8,128,2,8,balanced,0.39900267124176025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,4096,1536,8,128,2,8,balanced,0.5030399958292643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,6144,2048,8,256,2,4,power_law_1.2,0.09815040230751038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,6144,2048,8,256,2,4,power_law_1.2,0.13427200317382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,6144,2048,8,256,2,4,power_law_1.2,0.13150080442428588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,6144,2048,8,256,2,4,power_law_1.2,0.15629440546035767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,6144,2048,8,256,2,4,power_law_1.2,0.19459840059280395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,6144,2048,8,256,2,4,power_law_1.2,0.20190720558166503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,6144,2048,8,256,2,4,power_law_1.2,0.25055999755859376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,6144,2048,8,256,2,4,power_law_1.2,0.2597440004348755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,6144,2048,8,256,2,4,power_law_1.2,0.24430079460144044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,6144,2048,8,256,2,4,power_law_1.2,0.27302401065826415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,6144,2048,8,256,2,4,power_law_1.2,0.25834879875183103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,6144,2048,8,256,2,4,power_law_1.2,0.2664383888244629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,6144,2048,8,256,2,4,power_law_1.2,0.2713855981826782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,6144,2048,8,256,2,4,power_law_1.2,0.2808383941650391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,6144,2048,8,256,2,4,power_law_1.2,0.2926464080810547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,6144,2048,8,256,2,4,power_law_1.2,0.29437439441680907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,6144,2048,8,256,2,4,power_law_1.2,0.30959999561309814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,6144,2048,8,256,2,4,power_law_1.2,0.3219520092010498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,6144,2048,8,256,2,4,power_law_1.2,0.3532095909118652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,6144,2048,8,256,2,4,power_law_1.2,0.39950079917907716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,6144,2048,8,256,2,4,power_law_1.2,0.46381440162658694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,6144,2048,8,256,2,4,power_law_1.2,0.5981056213378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,6144,2048,8,256,2,4,power_law_1.2,0.7016064167022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,6144,2048,8,256,2,4,power_law_1.2,0.9336511611938476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,6144,2048,8,256,2,4,power_law_1.2,1.1872384071350097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1,3072,1536,8,256,1,16,power_law_1.01,0.07761920094490052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2,3072,1536,8,256,1,16,power_law_1.01,0.10087679624557495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4,3072,1536,8,256,1,16,power_law_1.01,0.10143359899520873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8,3072,1536,8,256,1,16,power_law_1.01,0.10123519897460938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16,3072,1536,8,256,1,16,power_law_1.01,0.0992576003074646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,32,3072,1536,8,256,1,16,power_law_1.01,0.0834559977054596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,48,3072,1536,8,256,1,16,power_law_1.01,0.10358400344848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,64,3072,1536,8,256,1,16,power_law_1.01,0.09747840166091919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,80,3072,1536,8,256,1,16,power_law_1.01,0.09385600090026855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,96,3072,1536,8,256,1,16,power_law_1.01,0.09403520226478576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,128,3072,1536,8,256,1,16,power_law_1.01,0.09610880017280579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,160,3072,1536,8,256,1,16,power_law_1.01,0.09899520277976989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,192,3072,1536,8,256,1,16,power_law_1.01,0.10014079809188843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,256,3072,1536,8,256,1,16,power_law_1.01,0.10839680433273316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,320,3072,1536,8,256,1,16,power_law_1.01,0.10584959983825684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,384,3072,1536,8,256,1,16,power_law_1.01,0.10705280303955078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,512,3072,1536,8,256,1,16,power_law_1.01,0.1098688006401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,768,3072,1536,8,256,1,16,power_law_1.01,0.11962239742279053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1024,3072,1536,8,256,1,16,power_law_1.01,0.1257599949836731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,1536,3072,1536,8,256,1,16,power_law_1.01,0.1429311990737915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,2048,3072,1536,8,256,1,16,power_law_1.01,0.1635200023651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,3072,3072,1536,8,256,1,16,power_law_1.01,0.20160000324249266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,4096,3072,1536,8,256,1,16,power_law_1.01,0.2464128017425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,6144,3072,1536,8,256,1,16,power_law_1.01,0.3019903898239136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,8192,3072,1536,8,256,1,16,power_law_1.01,0.3722496032714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,12288,3072,1536,8,256,1,16,power_law_1.01,0.5269184112548828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,nvfp4,16384,3072,1536,8,256,1,16,power_law_1.01,0.6894207954406738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1,7168,2048,8,256,1,32,power_law_1.2,0.1158784031867981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2,7168,2048,8,256,1,32,power_law_1.2,0.11136000156402588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4,7168,2048,8,256,1,32,power_law_1.2,0.10894080400466918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8,7168,2048,8,256,1,32,power_law_1.2,0.11401599645614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16,7168,2048,8,256,1,32,power_law_1.2,0.11408640146255493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,32,7168,2048,8,256,1,32,power_law_1.2,0.10658559799194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,48,7168,2048,8,256,1,32,power_law_1.2,0.11441279649734497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,64,7168,2048,8,256,1,32,power_law_1.2,0.11449600458145141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1,6144,16384,2,8,1,2,power_law_1.01,0.0995199978351593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2,6144,16384,2,8,1,2,power_law_1.01,0.11644799709320068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,80,7168,2048,8,256,1,32,power_law_1.2,0.11523840427398682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4,6144,16384,2,8,1,2,power_law_1.01,0.14145920276641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,96,7168,2048,8,256,1,32,power_law_1.2,0.11754239797592163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8,6144,16384,2,8,1,2,power_law_1.01,0.215065598487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,128,7168,2048,8,256,1,32,power_law_1.2,0.11582080125808716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16,6144,16384,2,8,1,2,power_law_1.01,0.2143615961074829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,32,6144,16384,2,8,1,2,power_law_1.01,0.21510400772094726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,48,6144,16384,2,8,1,2,power_law_1.01,0.21918718814849852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,64,6144,16384,2,8,1,2,power_law_1.01,0.22731521129608154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,80,6144,16384,2,8,1,2,power_law_1.01,0.23274240493774415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,96,6144,16384,2,8,1,2,power_law_1.01,0.23517439365386963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,128,6144,16384,2,8,1,2,power_law_1.01,0.23593599796295167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,160,6144,16384,2,8,1,2,power_law_1.01,0.24424319267272948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,160,7168,2048,8,256,1,32,power_law_1.2,0.12346880435943604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,192,6144,16384,2,8,1,2,power_law_1.01,0.2588992118835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,192,7168,2048,8,256,1,32,power_law_1.2,0.1232192039489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,256,6144,16384,2,8,1,2,power_law_1.01,0.26582400798797606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,256,7168,2048,8,256,1,32,power_law_1.2,0.12676479816436767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,320,6144,16384,2,8,1,2,power_law_1.01,0.2920896053314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,384,6144,16384,2,8,1,2,power_law_1.01,0.30586240291595457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,512,6144,16384,2,8,1,2,power_law_1.01,0.33085439205169676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,768,6144,16384,2,8,1,2,power_law_1.01,0.38224000930786134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,320,7168,2048,8,256,1,32,power_law_1.2,0.14097919464111328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,384,7168,2048,8,256,1,32,power_law_1.2,0.14586880207061767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,512,7168,2048,8,256,1,32,power_law_1.2,0.1541440010070801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,768,7168,2048,8,256,1,32,power_law_1.2,0.1659775972366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1024,6144,16384,2,8,1,2,power_law_1.01,0.46030077934265134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1024,7168,2048,8,256,1,32,power_law_1.2,0.17581440210342408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,1536,6144,16384,2,8,1,2,power_law_1.01,0.5334527969360352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,1536,7168,2048,8,256,1,32,power_law_1.2,0.201363205909729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,2048,7168,2048,8,256,1,32,power_law_1.2,0.2495487928390503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,3072,7168,2048,8,256,1,32,power_law_1.2,0.34586880207061765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,4096,7168,2048,8,256,1,32,power_law_1.2,0.3958336114883423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,2048,6144,16384,2,8,1,2,power_law_1.01,0.7293695926666259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,6144,7168,2048,8,256,1,32,power_law_1.2,0.5576255798339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,8192,7168,2048,8,256,1,32,power_law_1.2,0.740505599975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,12288,7168,2048,8,256,1,32,power_law_1.2,1.0414848327636719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,3072,6144,16384,2,8,1,2,power_law_1.01,1.099123191833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,deepgemm,fp8_block,16384,7168,2048,8,256,1,32,power_law_1.2,1.7196352005004882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,4096,6144,16384,2,8,1,2,power_law_1.01,1.0987135887145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,6144,6144,16384,2,8,1,2,power_law_1.01,1.9098112106323242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1,4096,2688,22,512,1,4,balanced,0.0706826647122701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2,4096,2688,22,512,1,4,balanced,0.08715200424194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4,4096,2688,22,512,1,4,balanced,0.11405332883199056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8,4096,2688,22,512,1,4,balanced,0.17076265811920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16,4096,2688,22,512,1,4,balanced,0.2544320027033488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,32,4096,2688,22,512,1,4,balanced,0.33081066608428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,48,4096,2688,22,512,1,4,balanced,0.3362720012664795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,64,4096,2688,22,512,1,4,balanced,0.339136004447937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,80,4096,2688,22,512,1,4,balanced,0.341813325881958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,96,4096,2688,22,512,1,4,balanced,0.3442240158716838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,128,4096,2688,22,512,1,4,balanced,0.34806398550669354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,160,4096,2688,22,512,1,4,balanced,0.3532266616821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,192,4096,2688,22,512,1,4,balanced,0.35497065385182697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,256,4096,2688,22,512,1,4,balanced,0.36159467697143555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,320,4096,2688,22,512,1,4,balanced,0.3706933259963989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,384,4096,2688,22,512,1,4,balanced,0.3752319812774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,512,4096,2688,22,512,1,4,balanced,0.38893334070841473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,768,4096,2688,22,512,1,4,balanced,0.4177600145339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1024,4096,2688,22,512,1,4,balanced,0.4416853189468384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,1536,4096,2688,22,512,1,4,balanced,0.6113866567611694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,2048,4096,2688,22,512,1,4,balanced,0.5755413373311361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,3072,4096,2688,22,512,1,4,balanced,0.8047573566436768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,4096,4096,2688,22,512,1,4,balanced,0.8534826437632242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,6144,4096,2688,22,512,1,4,balanced,1.2337653636932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,8192,4096,2688,22,512,1,4,balanced,1.3875519434611003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,12288,4096,2688,22,512,1,4,balanced,2.060981273651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_nongated,nvfp4,16384,4096,2688,22,512,1,4,balanced,2.505407969156901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,8192,6144,16384,2,8,1,2,power_law_1.01,2.299622344970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,12288,6144,16384,2,8,1,2,power_law_1.01,3.453388977050781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_cutlass,fp8,16384,6144,16384,2,8,1,2,power_law_1.01,4.706975936889648
